Description
Conflict management is the process of limiting the negative aspects of conflict while increasing the positive aspects of conflict.
ABSTRACT Title of Dissertation: ATTRIBUTIONAL PROCESSES IN ACCOUNTING FOR CONFLICT BEHAVIORS Shuo Yao, Doctor of Philosophy, 2009 Dissertation Directed By: Professors Deborah A. Cai & Edward L. Fink Department of Communication
One simple way to handle interpersonal conflict is to use accounts to explain one’s behaviors. Although accounts play a significant role in managing conflict, relatively little research has explored the processes offenders use to determine the accounts selected in conflict situations. The goal of this dissertation is to investigate the attributional processes offenders use that determine their accounts in conflict. Ten hypotheses were proposed about how the severity of the conflict outcome and the closeness between the parties involved in the conflict influence offenders’ choice of accounts. A structural equation model was developed and tested based on the proposed hypotheses. An experiment was conducted, with two levels of outcome severity and three levels of relational closeness. Offenders’ attributions (i.e., the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability), anticipated consequences (i.e., expected responsibility and expected anger), and offenders’ expected choice of accounts (i.e., the likelihood of selecting concessions, justifications, excuses, and refusals) were measured. Two hundred thirty-eight participants were recruited and randomly assigned to one of the six experimental conditions. Participants read a hypothetical conflict scenario,
imagined that they were the offender in the scenario, and completed a questionnaire that had the dependent measures. Results indicated that outcome severity influenced offenders’ choice of accounts directly and indirectly. Offenders tended to choose a more defensive account when they perceived the outcome to be severe than when the outcome was not severe. The influence of outcome severity on offenders’ choice of accounts was also mediated by the attributions offenders made, the responsibility expected to be assigned to offenders, and anger expected to be felt by victims. When offenders perceived the outcome to be severe, offenders made more attributions, expected more responsibility to be assigned to them, and expected that victims felt angrier about offenders’ behavior than when the outcome was not severe. Consequently, when offenders expected more anger from victims, they tended to be less defensive. Interpretations and implications of results, the limitations of the study, and future directions were discussed.
ATTRIBUTIONAL PROCESSES IN ACCOUNTING FOR CONFLICT BEHAVIORS
by Shuo Yao
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2009
Advisory Committee: Professor Deborah A. Cai, Co-Chair Professor Edward L. Fink, Co-Chair Professor Torsten O. Reimer Professor Harold Sigall Professor Monique M. Turner
© Copyright by Shuo Yao 2009
ii
Dedication
This dissertation is dedicated to my parents and my sister, for their love and support.
iii
Acknowledgement
I would like to express my gratitude to all those who gave me the possibility to complete this dissertation. I am deeply indebted to my two advisors, Dr. Deborah A. Cai and Dr. Edward L. Fink. Dr. Cai always makes herself available when I need her. I cannot recall how many times we sat in her office and talked about research ideas or problems in my life. She always inspires me and guides me on my way. Two years ago I was in the most difficult time of my life. If it was not Dr. Cai who listened to me patiently and encouraged me to concentrate on my work, I probably would not be able to finish the program. My other advisor, Dr. Fink, is the most intelligent and diligent scholars I have ever encountered. I remember my fellow students and I used to say that Dr. Fink is a “walking encyclopedia.” No matter what questions we have, we can always get some answers from him. Dr. Fink met with me extensively to work on this dissertation. Even when I am away to teach at another university, he told me to call him anytime when I have a question. Without Dr. Fink’s help, this dissertation would be impossible to complete. Being able to be one of his students is the luckiest thing that has ever happened to me. My appreciation to my two advisors cannot be expressed fully by words. I also want to thank my other committee members: Drs. Monique M. Turner, Torsten O. Reimer, and Harold Sigall. I am so lucky that I could be in Dr. Turner’s research team during my first year at Maryland. The experience taught me way much than what I could have learned from books. Dr. Turner also sets a role model for me not only in doing research but also in teaching. Dr. Reimer is always polite and helpful. When I worked with him on one of his research ideas, he always surprised me with new
iv ideas or perspectives to conduct the study. I am also grateful to have Dr. Sigall as one of my committee members. I would like to thank Dr. Timothy J Ng. When I was about to defend my dissertation prospectus, Dr. Sigall was sick and could not attend the defense. On a short notice, Dr. Ng still generously agreed to replace Dr. Sigall and attended the defense. My former colleagues from the Department of Communication, University of Maryland, College Park, supported me in my research work. I want to thank them all for their help, support, interest, and valuable ideas. Especially, I am obliged to Drs. Qi Wang, Xiaoying Xie, and Bing Han. Qi was my roommate for four years and we share so many happy moments. Qi is also the person that I will first contact if I have any question about my research and teaching. I arrived at Maryland the same year as Bing and Xiaoying. We took the same classes together and I am so lucky to have these two smart people as my cohort. I also want to thank my friend, Dr. Jennifer Vardeman, who helped me in various aspects of life, especially when I was struggling two years ago. My friend, Stephanie Materese George, can always find a way to put a smile on my face. Dr. Lan Ni talks with me over the phone and helps me go through the transition stage from a student to a professor. Hua Jiang, Hongmei Shen, Ai Zhang, Sanja Sipek, Sabine Chai, Ioana Cionea, Sejal R. Patel, John Banas, Elena Bessarabova Banas, Eirca Lamm, Timothy Barney, Art Herbig, Ahnlee Jang, Yi Luo, I thank you for your help and support. And, especially, I would like to thank Jiangming Teng, who accompanied me during the darkest time of my life. His support and understanding enabled me to complete this work.
v
Table of Contents
Dedication ...................................................................................................................... i Acknowledgements ....................................................................................................... ii Table of Contents .......................................................................................................... v List of Tables ............................................................................................................... vi List of Figures ............................................................................................................. vii CHAPTER I: Introduction ............................................................................................ 1 CHAPTER II: Attributional Processes in Accounting for Conflict Behaviors ............ 6 Interpreonal Conflict and Conflict management....................................................... 6 Accounts ................................................................................................................... 8 Outcome Severity.................................................................................................... 16 Relational Closeness ............................................................................................... 20 Attribution Theory .................................................................................................. 22 Weiner’s Attribution Theory of Motivation and Emotion .................................. 23 CHAPTER III: Method ............................................................................................... 31 Overall Design, Stimulus Message, and Instruments ............................................. 31 Overall Design .................................................................................................... 31 Hypothetical Interpersonal Conflict Scenario ..................................................... 32 Role Playing in a Hypothetical Conflict Scenario .............................................. 33 Naming the Hypothetical Other Party................................................................. 34 Instruments .......................................................................................................... 34 Summary ............................................................................................................. 37 Pilot Studies ............................................................................................................ 37 Pilot Study 1: Collecting Conflict Scenarios ...................................................... 38 Pilot Study 2: Developing a Hypothetical Conflict Scenario ............................. 40 Pilot Study 3: Manipulation Checks ................................................................... 42 Pilot Study 4: Creating the Measure of Accounts ............................................... 49 Pilot Study 5: Testing the Defensiveness Level of Each Account...................... 50 Pilot Study 6: Pilot Testing Dependent Measures .............................................. 65 The Formal Study ................................................................................................... 72 Analytical Strategies ........................................................................................... 72 Design ................................................................................................................. 74 Sample................................................................................................................. 74 Procedure ............................................................................................................ 75 CHAPTER IV: Results ............................................................................................... 78 Preliminary Analyses .............................................................................................. 78 Data Entry Checks .............................................................................................. 78 Missing Data ....................................................................................................... 78 Data Transformation ........................................................................................... 79 Instruments .......................................................................................................... 80 Manipulation Checks .......................................................................................... 96 Model Assessment and Hypothesis Testing ........................................................... 97 Model Assessment .............................................................................................. 97 Hypothesis Testing............................................................................................ 101
vi Post-Hoc Analyses ................................................................................................ 106 Hypothesis Testing............................................................................................ 106 Modification Indices ......................................................................................... 111 Model Revising ................................................................................................. 113 Testing the Revised Model ............................................................................... 113 CHAPTER V: Discussion ........................................................................................ 120 Summary of the Study .......................................................................................... 120 Summary of the Results ........................................................................................ 122 Hypothesis 1: Influence of Outcome Severity on Expected Responsibility ..... 122 Research Question 1: Effect of Outcome Severity on the Choice of Accounts 122 Hypothesis 2: Influence of Relational Closeness on the Choice of Accounts .. 123 Hypothesis 3: Interaction and the Choice of Accounts ..................................... 124 Hypothesis 4: Attributional Processes .............................................................. 124 Hypothesis 5: Internal Attribution and Anticipated Consequences .................. 125 Hypothesis 6: Controllability and Anticipated Consequences.......................... 125 Hypothesis 7: External Attribution and Anticipated Consequences ................. 126 Hypothesis 8: Uncontrollability and Anticipated Consequences...................... 126 Hypothesis 9: Expected Responsibility and Expected Anger ........................... 127 Hypothesis 10: Expected Anger and the Choice of Accounts .......................... 127 How do People Decide the Accounts in Conflict Situations ............................ 128 Limitations of the Study and the Direction for Future Research .......................... 129 Limitation Regarding the Conceptulization of Accounts ................................. 129 Limitation Regarding the Dependent Measures ............................................... 130 Limitation Regarding the Hypothetical Conflict Scenario ............................... 131 Future Direction: Cultural Differences in Accounts ......................................... 132 Significance of the Study and Conclusion ............................................................ 134 Appendix A Instruments for Manipulation Checks in the Formal Study ................. 139 Appendix B Instruments for Attributions in the Formal Study ................................ 140 Appendix C Instruments for Anticipated Consequences in the Formal Study ......... 143 Appendix D Instruments for the Likelihood of Selecting an Account in the Formal Study ......................................................................................................................... 145 Appendix E Pilot Study 1: Questionnaire for Collecting Conflict Scenarios ........... 148 Appendix F Pilot Study 2: Questionnaire for Sorting Task ...................................... 153 Appendix G Conflict Incidents Reported in Pilot Study 1 ....................................... 159 Appendix H Hypothetical Conflict Scenarios Used in the Formal Study ................ 161 Appendix I Pilot Study 3: Questionnaire for Manipulation Checks ......................... 166 Appendix J Instruments for Defensiveness of Accounts in Pilot Study 5 ................ 171 Appendix K List of Open-Ended Questions Used in Pilot Study 6 .......................... 172 Appendix L Descriptives of the Indicatiors in the Formal Study after Transformation ................................................................................................................................... 173 Appendix L Covariance Matrix in LISREL for the Proposed Model in Figure 2 .... 177 Appendix M LISREL Syntax for Testing the Proposed Model in Figure 2 ............. 185 Appendix N Covariance Matrix in LISREL for the Revised Model in Figure 6 ...... 186 Appendix O LISREL Syntax for Testing the Revised Model in Figure 6 ................ 192 References ................................................................................................................. 193
vii
List of Tables
Table 1. Table 2. Major Types of Accounts, Sources, and Definitions. Coding Scheme for Four Types of Accounts, Adapted from Fritsche’s (2002) Review (pp. 389-394). Table 3. Table 4. 51 12
Intercorrelations between Items for Defensiveness of Each Account. 55 The Cronbach’s Alpha for the Sum of Items that Measured Defensiveness Level of Each Account (N = 22). 62
Table 5.
Eigenvalues of the First Principal Components of Endogenous Variables with the Proportions of the Variance Explained. 88
Table 6.
The Cronbach’s Alpha for the Sum of Measurement Items for All Endogenous variables in the Formal Study. 89
Table 7.
Manipulation Checks, Indicator Loadings, and Measurement Model Indices. 91
Table 8.
Attributional Processes Variables, Indicator Loadings, and Measurement Model Indices. 93
Table 9.
Expected Consequences Variables, Indicator Loadings, and Measurement Model Indices. 95 98 99 107
Table 10. Table 11. Table 12. Table 13.
Structural Equations of the Model Proposed by the Hypotheses. The Unstandardized Structural Coefficients for the Model. Intercorrelations between the Five Exogenous Variables.
Structural Equations of the Revised Model Based on Modification Indices and Exploratory Analyses. 115 116
Table 14.
Unstandardized Structural Coefficients for the Revised Model.
viii
List of Figures
Figure 1. Types of accounts on the defensiveness continuum, as proposed by Schönbach (1990). Figure 2. Figure 3. Structural relations based on the hypotheses. Defensiveness level of concessions, justifications, excuses, and refusals. Figure 4. Structural relations proposed in the hypotheses with unstandardized structural coefficients. Figure 5. Figure 6. Structural relations based on post-hoc analysis. The revised model with unstandardized structural coefficients. 100 114 117 64 16 29
1 CHAPTER I Introduction Person A: You are supposed to cover this training session with me together but you never showed up! How many times are you going to leave me to work this group by myself? Person B: My other job has had me working sixty hours a week for the past few weeks to close out an important project! (Pause) Person B: But it will end soon. By that time I can be back to work with you regularly. I understand where you are coming from. (Extract from a self-report conflict scenario in Pilot Study 1 of the current research). Despite growing interest in behavioral strategies for conflict management, little attention has been given to a simple way to handle conflicts: the use of accounts. Accounts are statements made by offenders to explain their unanticipated or normcontradictive behaviors (Scott & Lyman, 1968). As shown in the above example, when an interpersonal conflict occurs, communication plays a significant role in the conflict process (Putnam & Poole, 1987). People often use accounts to try to influence the affected party’s perception of the conflict, to avoid further escalation of the conflict or to maintain their relationship with the affected party. Verbal explanation is especially important for offenders whose behavior has a negative influence on victims in conflict situations. What offenders say to explain their behaviors directly influences a victim’s perceptions of (1) the responsibility assigned to the offender, (2) the offenders’ intentions, and (3) the way that the conflict can be handled (Sitkin & Bies, 1993).
2 Studying communication that occurs in conflict situations, especially accounts offenders use to explain their behaviors, provides a foundation for conflict management research because accounts often set the tone for how a conflict can be resolved. Further, examining offenders’ choice of accounts in a conflict situation can provide insight to the account literature. Past research on accounts has focused on developing different types of accounts and examining their functions in interpersonal situations (e.g., Jellison, 1990; Tedeschi & Reiss, 1981). Relatively little work has explored how people construct their accounts (for exceptions see Read, 1992; Rest, 1984). Read argued that offenders construct accounts based on a typical form, such as information about the goals of offenders, factors that stimulate those goals, behaviors or strategies to achieve those goals, outcomes (e.g., whether those goals are satisfied), and “the physical and social situations in which the action occurred” (Read, 1992, p. 6). Rest looked into how offenders reconstruct the definition of situations to mitigate the consequences of their actions. Rest pointed out that the defensive reinterpretation, a process to reframe offenders’ understanding about the causes of their actions and their judgments of responsibility, was used by offenders to protect them in the course of a negative event. Read’s (1992) and Rest’s (1984) studies are examples of research examining the cognitive processes offenders use to determine their accounts. McLaughlin, Cody, and Read (1992) suggested that more attention should be given to understanding how people create accounts. The current research answers the call from McLaughlin et al. (1992) by focusing on one specific process, the attributional process, to understand the accounts that offenders use in conflict situations. The question addressed by this research is: What attributional processes do offenders use to determine
3 their accounts in a conflict? Attribution theories provide a framework for understanding how people make sense of their world (e.g., Weiner, 1986, 1992). Causal inferences organize people’s perceptions of reality and make the experiential world more understandable and predicable (Kelley, 1971). By using attribution theories, this study can test the cognitive and emotional processes that offenders engage in after their behavior leads to an interpersonal conflict. Weiner’s (1986, 1995, 2006) attribution theory of motivation and emotion provides the theoretical background for this research. Weiner argued that to explain a norm-contradictive behavior, offenders often start by examining antecedent conditions that lead to the behavior. In a conflict situation, the antecedent conditions could be the relationship between the offender and the victim or the severity of the situation. According to Weiner, once offenders have analyzed the antecedent conditions, they will make causal inferences about why the behavior has occurred. The behavior could have happened because of environmental reasons or because of the offenders’ personal traits. Weiner further argued that based on the causal inferences made, offenders will have expectations about consequences if such inferences were presented to the victim. Expected consequences refer to the level of responsibility to be assigned to offenders and the level of the victim’s anger. Offenders then will decide the accounts to be conveyed to victims based on the expected consequences. The current research examined offenders’ account-giving processes by focusing on two antecedent conditions—outcome severity and relational closeness—and causal inferences associated with the offending behavior. Outcome severity refers to the level of
4 severity in consequences caused by offenders’ behavior. Attribution studies have suggested that people are “motivate[d] to be highly engaged in an attributional search in response to more threatening or unusual events because there is a greater need to explain such events” than events that have a less severe outcome (Robbenolt, 2000, p. 2576). Research also indicated that different levels of outcome severity lead to a different amount of responsibility to be assigned to offenders (Walster, 1960). When the outcome of a conflict is severe, offenders expect victims to assign more responsibility to offenders. To reduce offenders’ negative feelings about themselves, offenders may blame themselves and use an account to comfort victims, such as by offering compensation or showing regret. When the outcome of a conflict is not severe, offenders expect victims to assign less responsibility to offenders; thus, offenders will be more likely to defend themselves, compared to the situation in which the outcome is severe. The current research also posits that people’s perception about conflict and managing conflict varies, depending on with whom they are in conflict. Relational closeness, the level of closeness between the offender and the victim, influences the choice of accounts. Fitzpatrick and Winke (1979) pointed out that when a conflict occurs between close friends, people perceive that they have a lower risk of relationship termination than if the conflict is between people who are not close. Offenders may then take a defensive stand to handle the conflict. If the conflict occurs between acquaintances, given the desire to maintain the relationship with the victim, offenders often will be mitigating to cope with the conflict. However, when a conflict occurs between strangers, with a low possibility of future interaction, offenders will try to defend themselves and, as a result, will be aggressive.
5 The current research focuses on these two contextual variables (i.e., outcome severity and relational closeness between offenders and victims) of interpersonal conflicts and examines the cognitive and emotional processes involved in the provision of accounts. An experiment that manipulated three levels of relational closeness (high vs. moderate vs. low) and two levels of outcome severity (high vs. low) in a hypothetical conflict scenario was conducted to further our understanding of the offenders’ choice of accounts in a conflict situation.
6 CHAPTER II Attributional Processes in Accounting for Conflict Behaviors Interpersonal Conflict and Conflict Management Although no definition of conflict is universally agreed upon, the central characteristic of conflict is generally viewed as incompatibility of goals (Roloff & Soule, 2002). For example, Mack and Snyder (1957) defined conflict as a “particular kind of social interaction process or ‘interaction relationship’ between parties who have mutually exclusive or incompatible values” (cited in Fink, 1968, p. 432). Similarly, Wilmot and Hocker (2005) defined conflict as “an expressed struggle between at least two interdependent parties who perceive incompatible goals, scarce resources, and interference from others in achieving their goals” (p. 9). Conflicts can occur under different circumstances, such as when two parties’ behavioral preferences are inconsistent, when there are not enough resources to fulfill each party’s desire, or when there is a conflict of values or goals (Rahim, 1992; Rahim & Bonoma, 1979). The primary focus of conflict research has been the examination of people’s preference for conflict management styles (Thomas, 1992). Rahim’s (1983) typology of conflict management styles has been widely studied. These styles are dominating, integrating, compromising, obliging, and avoiding. Although different versions of Rahim’s five conflict management styles have been used in conflict research (e.g., Ohbuchi, Fukushima & Tedeschi, 1999), a majority of conflict management styles can be categorized based on the dual-concern model (Carnevale & Pruitt, 1992; Pruitt & Rubin, 1986). Based on Blake and Mouton’s (1964) and Thomas’ (1976) research on conflict, Pruitt and Rubin (1986) developed the dual-concern model to represent different conflict
7 styles. The dual concerns refer to concerns for one’s own outcomes and for the outcome of others involved in the conflict. The dual-concern model suggests that dominating involves a high concern for one’s own outcome and a low concern for the other’s outcome; integrating reflects a high concern for both own and other’s outcome; compromising represents a moderate concern for one’s own and the other’s outcome; obliging involves a low concern for one’s own outcome but a high concern for the other’s outcome; and avoiding reflects a low concern for both own and other’s outcome. Although conflict management style preferences have been widely studied (e.g., Ohbuchi et al., 1999), relatively little work has explicated the communication aspects of conflict management—accounts offenders use to explain the behaviors that have led to a conflict. Conflict management styles have been studied as a general way to handle conflict; for example, Cai and Fink (2002) found that compromising and integrating are preferred more by collectivists than by individualists. This result suggested that someone high on collectivism is more likely to use compromising or integrating to handle a conflict than those high on individualism. Accounts are different from conflict management styles in that accounts depend on specific contexts. For example, McLaughlin, O’Hair, and Cody (1983) studied contextual determinants of accounts and found that when there was a reproach, when the offense was severe, and when the offender felt guilty, offenders were more likely to use concessions to explain their behaviors. Contexts play an important role in determining the accounts that will be used in conflict situations. The use of a particular account is not necessarily connected to individuals’ dispositional characteristics and not necessarily correlated with individuals’ conflict management styles.
8 Accounts Accounts have been recognized as one impression management strategy. For example, Goffman (1971) discussed accounts as a remedial strategy to protect an offender from attacks to his or her social identity (Goffman, 1971). This theoretical orientation is also found in Scott and Lyman’s (1968) typology of accounts. According to Scott and Lyman, an account is “a statement made by a social actor to explain unanticipated or untoward behavior” (p. 46). Offenders use accounts to influence victims’ perception of themselves or of the causes of the conduct in a specific incident (Mills, 1940). As argued by Mills, “When an agent vocalizes or imputes motives, he is not trying to describe his experienced social action. He is not merely stating ‘reasons.’ He is influencing others—and himself” (p. 907). Most early studies of accounts were descriptive and focused on the development of account taxonomies (e.g., Scott & Lyman, 1968; Tedeschi & Reiss, 1981). For example, in a classic study on accounts (Sykes & Matza, 1957), the authors examined the techniques (i.e., techniques of neutralization) that juvenile delinquents used to justify their deviant behaviors and to protect themselves from self-blame or blame from others. Sykes and Matza (1957) classified the techniques of neutralization that juvenile delinquents used to provide explanations for their behaviors into five categories: denial of responsibility (i.e., a claim that something happened due to uncontrollable forces), denial of injury (i.e., a claim that no harm has been caused), denial of victim (i.e., a claim that the victim deserved the treatment given), condemnation of the condemners (i.e., a claim that condemners are reacting out of personal spite), or appeal to higher loyalties (i.e., a claim that loyalty to important others requires such behaviors). By using the above
9 techniques, juvenile delinquents could adjust their understanding of social norms to be consistent with their behaviors, thus avoiding self-blame and disregarding blame from others. Inspired by Sykes and Matza’s (1957) research, Scott and Lyman (1968) distinguished two types of accounts: excuses and justifications. An excuse refers to an explanation in which people admit the inappropriateness of the behavior in question but cite some external reasons to deny their full responsibility. In the example cited at the beginning of the current research, Person B used having another job as an excuse for why he or she did not fulfill the commitment to Person A. By doing so, Person B acknowledged that the behavior (i.e., not showing up to the training session) was not appropriate; however, the other job for an important project was seemingly more timeconsuming and demanding, which excused Person B’s misconduct (i.e., not showing up to the training session). Thus, Person B would not take full responsibility for missing the training session. The other type of account, justification, is defined as an explanation that denies the behavior in question is wrong but admits responsibility for it. Often people who use this account claim that their conduct fulfills a superordinate goal. Scott and Lyman (1968) offered the following example of a justification: A soldier could assert that killing in a war is for a superordinate goal—fighting for the cause of freedom. Therefore, although the solider should take the responsibility for killing, he or she does not think the behavior is inappropriate because killing in a war is necessary and perhaps even desirable. Scholars have extended Scott and Lyman’s (1968) account taxonomy (e.g., Bies,
10 1987; Schlenker, 1980; Schönbach, 1990; Sitkin & Bies, 1993; Tedeschi & Reiss, 1981). Much work has focused on clarifying the distinction between excuses and justifications (e.g., Schlenker, 1980). Schönbach expanded the account taxonomy by developing two additional categories: concession and refusal. Concession is defined as an explanation that admits the inappropriateness of the behavior in question and also admits partial or total responsibility for the behavior. Concessions are often accompanied with a confession, expression of regret or guilt, or an offer of restitution (Fincham, 1992). In the example cited at the beginning of the current research, if a concession were to be used, Person B would probably say something like this: “I am sorry for being absent from all those training sessions. It’s my fault. Maybe I can cover the next two sessions by myself so you can have some time to do your work.” In this concession, Person B admits that his or her behavior (i.e., not showing up to the training session) was not appropriate and offered compensation. On the other hand, when people choose refusals, they deny guilt or responsibility, claim that the behavior in question has not occurred, deny personal involvement with the behavior in question, or attribute guilt to others, even to the accuser (Konovsky & Jaster, 1989; Schönbach, 1980). If a refusal would be used in the above example, Person B probably would argue like this: “You did not tell me I had to be there for the training session. If you wanted me to be there, you should have let me know ahead of time, rather than complaining afterwards.” By attributing the guilt to Person A (i.e., not informing Person B about his or her duty), Person B used a refusal to deny his or her responsibility. Although a few more types of accounts have been discussed in the account literature, the majority of the new types of accounts are similar to Scott and Lyman’s
11 (1968) and Schönbach’s (1980) conceptualizations (e.g., Bies, 1987; Hamilton & Hagiwara, 1992; McLaughlin, O’Hair et al., 1983). See Table 1 for definitions of major types of accounts as proposed by various authors. For example, Bies (1987) described four different types of accounts: causal accounts, ideological accounts, referential accounts, and penitential accounts. Bie’s framework is relatively similar to Schönbach’s taxonomy. Causal accounts are in essence excuses; ideological accounts are similar to justifications; and penitential accounts consist of apology and remorse, which often are used in concessions. The only difference between Bies’ framework and Schönbach’s taxonomy is that Bies de-emphasized refusals but highlighted referential accounts (i.e., when social comparison and consensus information are used to explain the behavior in question). Scholars perceived apology as another type of account (e.g., Hamilton & Hagiwara, 1992). Apology admits that harm was done but does not necessarily remove or reduce responsibility. Using an apology assuages negative emotions aroused by the behavior in question. Apologies can be identified as a part of concessions (Fincham, 1992). McLaughlin, O’Hair et al. (1983) argued that under certain circumstances (for instance, when the embarrassment associated with the behavior in question is severe) the best choice to deal with the conflict may be silence. Silence is defined as avoiding any reference to or verbal explanation about the untoward behavior. Although silence or avoidance sometimes may be the best way to prevent interpersonal conflicts from escalating, it is not a type of accounts because it does not provide any verbal explanation about the behaviors in question. Because concessions, justifications, excuses, and refusals overlap with most types of accounts discussed in the account literature, these four types
12 Table 1 Major Types of Accounts, Sources, and Definitions Types of accounts Excuse Sources Definition Explanations in which people admit the inappropriateness of the behavior in question but cite external reasons to deny their full responsibility. Justification Scott & Lyman (1968) Explanations in which people admit full responsibility for what has happened but indicate that the behavior in question is to fulfill a superordinate goal; thus, the behavior is not inappropriate. Concession Schönbach (1980) Explanations in which people admit partial or total responsibility or guilt, express regret concerning the behavior in question, and offer compensation. Refusal Schönbach (1980) Explanations in which people deny guilt or responsibility, claim that the behavior in question has not occurred, and attribute guilt to others. Causal Bies (1987) Explanations in which people use a reason to mitigate offenders’ responsibility, or explanations that are more commonly
Scott & Lyman (1968)
13 referred to as excuses. Ideological Bies (1987) Explanations in which the actor attempts to invoke a superordinate goal or to place a positive label on the deed. Referential Bies (1987) Explanations in which social comparison and consensus information is presented. Penitential Bies (1987) Explanations that consist of apology and remorse. Consensus Hamilton & Hagiwara (1992) External excuse Hamilton & Hagiwara (1992) Explanations that use social comparison and declare that “everybody’s doing it.” Explanations that include a broad range of circumstances and interventions by human or natural causes. Internal excuse Hamilton & Hagiwara (1992) Explanations that refers to something inside the person that caused the outcome, whether or not it was controllable by that person. Apology Hamilton & Hagiwara (1992) Explanations that admit the wrong-doing and the responsibility; an apology is expected to assuage feelings.
14 of accounts will be examined in the current research. More than developing account taxonomies, McLaughlin, Cody, and Rosenstein (1983) and Schönbach (1990) advanced the study of accounts by using a continuum to lay out the underlying dimension for the four types of accounts (i.e., concessions, justifications, excuses, and refusals). McLaughlin, Cody et al. used an aggravatingmitigating continuum whereas Schönbach investigated each type of accounts in terms of its defensiveness, that is, to the extent to which using each type of accounts helps offenders to defend themselves. Regardless of the terminology, the continua proposed by these two groups of researchers are similar. According to Schönbach (1990), refusals, which deny the connection between offenders and the behavior in question, are most defensive. Further, excuses are less defensive than refusals because when excuses are used, offenders admit that their behavior is wrong and point to external reasons to reduce some responsibility. Schönbach argued that justifications are less defensive than excuses because offenders admit partial or full responsibility for their behavior, although they claim their behaviors are not inappropriate. According to Schönbach, concessions, which admit offenders’ responsibility or guilt for the behavior in question, are the least defensive type of accounts. On McLaughlin, Cody et al.’s aggravating-mitigating continuum, concessions and excuses are mitigating types of accounts whereas refusals and justifications are aggravating types of accounts. In particular, McLaughlin, Cody et al. argued that concessions are most mitigating and excuses are next to most mitigating type of accounts. Refusals are most aggravating and justifications are less aggravating, compared to refusals.
15 McLaughlin, Cody et al.’s (1983) assumed that aggravating accounts and mitigating accounts are different and exclusive: When one type of accounts is mitigating, it will not be considered as an aggravating type of accounts, and vice versa. For example, according to McLaughlin, Cody et al., excuses are a type of mitigating accounts, not an aggravating account. And justifications are a type of aggravating accounts, not a mitigating account. However, the definitions of these two types of accounts do not necessarily suggest this interpretation. By using a justification, an offender could mitigate the conflict to some extent because a justification admits the offender’s responsibility for what has happened. On the other hand, an excuse could aggregate the conflict because a victim may regard an offender’s using excuses as being dishonest and therefore think that the offenders is not willing to take his or her responsibility. A type of account may be both mitigating and aggravating at the same time. There was no empirical support to both the mitigating-aggravating continuum proposed by McLaughlin, Cody et al. and the defensiveness continuum proposed by Schönbach. Given the unclear distinction between mitigating and aggravating accounts, the current research decided to follow Schönbach’s defensiveness continuum to study accounts. Figure 1 indicates the possible ranking of four types of accounts on a defensiveness continuum as proposed by Schönbach. The use of accounts is associated with the personal responsibility assigned to offenders (McLaughlin, Cody, & French, 1990). For example, justifications and excuses are more likely to be connected with personal responsibility than concessions because justifications and excuses often “call attention to the character and motives of the defendant” (McLaughlin et al., 1990, p. 252). One factor that influences people’s judgment of responsibility is outcome severity.
16 Low Defensiveness High Defensiveness
Concession
Justification
Excuse
Refusal
Figure 1. Types of accounts on the defensiveness continuum, as proposed by Schönbach (1990). Outcome Severity Outcome severity determines the amount of responsibility assigned to a potentially responsible actor (Tennen & Affleck, 1990). For example, in Walster’s (1966) study, participants listened to a story about a car accident and rated the responsibility of offenders. Walster found a positive relationship between outcome severity and responsibility assigned to the offender; that is, when the outcome was severe, participants attributed more responsibility to the offender than when the outcome was not severe. However, later research provided mixed support for this proposition (Robbennolt, 2000). A number of studies have replicated Walters’s findings (e.g., Gleason & Harris, 1976; Tennen & Affleck, 1990; Wilson & Jonah, 1988), but other studies have either failed to find a relationship between outcome severity and expected responsibility (e.g., Thomas & Parpal, 1987) or have found their relationship to be negative (e.g., Shaver, 1970). Robbennolt (2000) conducted a meta-analysis to further understand the relationship between outcome severity and judgment of responsibility. Seventy-five studies were collected for the meta-analysis. Although the strength of correlation varied in terms of types of judgment that participants were asked to make (e.g., the correlation was smallest for liability judgments and largest for assessments of punishment), Robbenolt’s meta-analysis supported Walster's hypothesis that, overall, people assign
17 more responsibility to offenders when an outcome of a negative event is more severe than when the outcome is less severe. Robbenolt reported that the effect sizes of all studies that she examined ranged from -.26 (Shaver, 1970) to .75 (Mitchell & Kalb, 1981). The overall effect size was positive and significantly different from zero (r = .08, p < .01).1 The majority of studies on outcome severity and responsibility focus on the observer’s perspective: having a third person assign responsibility to the offender. Relatively little research has been conducted from the offender’s perspective. The question is whether offenders understand the violation the same way that observers do. When the outcome of a conflict is severe, does the offender expect more responsibility to be assigned to themselves than when the outcome is not severe? McLaughlin et al. (1990) pointed out that severe outcomes often include distinct damage to victims, such as bodily harm or loss of a large sum of money. With evidence of damage, those offenses are less deniable and offenders perceive a greater chance that they would be held responsible by others. Thus, it is hypothesized that offenders will expect more responsibility assigned to them in a severe outcome situation than a not severe outcome situation. H1: Offenders expect more responsibility to be assigned to them when the outcome is severe than when the outcome is not severe. McLaughlin et al. (1990) argued that in a severe outcome situation, an offender holds out little hope that he or she will not be found guilty for his or her behaviors. An offender will then “direct all of his or her efforts toward mitigation” (McLaughlin et al., 1990, p. 266). A less defensive account is more likely to be used by the offender to
According to Robbennolt (2000), the results for each study she examined “were converted into correlation coefficients (r) describing the relationship between outcome severity and the responsibility-related judgment” (p. 2586). A weighted average based on sample size was calculated if more than one effect between outcome severity and responsibility judgment was reported.
1
18 achieve his or her goal of mitigation when the outcome is severe than when the outcome is not severe. McLaughlin, O’Hair et al.’s (1983) study provided support for this hypothesis. McLaughlin, O’Hair et al. had participants report an incident in which their behavior led to a negative event. Participants provided the actual accounts they used in the situation and rated the severity of the negative event. The results suggested that offenders were likely to use concessions when the negative event was severe than when it was not severe. However, McLaughlin, O’Hair et al.’s (1983) results may only provide one side of the story. Schönbach (1990) argued that the self-reported incidents in McLaughlin, O’Hair et al.’s study were incidents that only had minor consequences, such as being late for a meeting. Schönbach argued that if the offense were extremely severe, such as causing bodily harm, the pressure on offenders for taking full responsibility becomes unbearable. To direct some responsibility to factors other than themselves, offenders are expected to use defensive accounts, such as justifications. Schönbach cited studies from criminal justice to support this idea. For example, Henderson and Hewstone (1984) coded explanations provided by inmates about their current convictions. They found offenders overwhelmingly pointed to external reasons, such as characteristics of victims or something about the environment, to explain their criminal behaviors. Justifications and excuses were identified as the most frequently used types of accounts in those inmates’ explanations than concessions. In the case of a severe outcome, there is a possibility that offenders will use a more defensive strategy to take some responsibility off their shoulders than when the outcome is not severe. Schönbach (1990) argued that there are two possibilities about the influence of
19 outcome severity on the defensiveness of accounts. On one hand, defensiveness of the chosen account may be negatively associated with outcome severity. McLaughlin, O’Hair et al.’s (1983) study provided support for this idea. When the outcome of a conflict is severe, offenders expect negative responses from victims: Victims will be angry and will assign more responsibility to offenders, compared to when the outcome of the conflict is not severe. To reduce those negative responses, offenders are more likely to use a less defensive account to amend the possible damages offenders’ behavior has brought to victims. On the other hand, the increase in outcome severity leads to a higher level of defensiveness, which results in a more defensive account. When the outcome is severe, taking full responsibility becomes a greater burden for offenders. For example, admitting the wrong-doing fully to one self should damage offenders’ self image and make offenders look immoral. Offenders will try to defend themselves against the damages to their face by attributing their behavior to situational rather than dispositional causes, or justify their behavior by linking it to a superordinate goal. Results from Henderson and Hewstone’s (1984) study provide support to the idea that offenders use defensive accounts to reduce the responsibility assigned to offenders when the outcome is severe. Schönbach called for more work to examine the influence of outcome severity on the level of defensiveness for a chosen account. Thus, a research question is proposed in the current research: RQ1: How does outcome severity influence the defensiveness of an account chosen by offenders? The direction of the effect of outcome severity on accounts is unclear and may be clarified by understanding the relational closeness between offenders and victims
20 (McLaughlin, O’Hair et al. 1983). Relational Closeness Research on aggression and the relationship between friends provides indirect support for the influence of relational closeness on choice of accounts. Richardson and Green (2006) found that in a conflict, people tend to use direct aggression in response to anger from their romantic partners but use indirect aggression toward their friends, who are less close to them than their romantic partners. According to Richardson and Green, direct aggression “confronts the target face to face, delivering harm through direct verbal or physical means” (p. 2493) whereas indirect aggression refers to when people involved in a conflict “[deliver] harm circuitously, through another person or object” (p. 2493). Richardson and Green’s finding is consistent with Fitzpatrick and Winke’s (1979) argument that in a close relationship offenders are less concerned with the threat of relationship termination than when the relationship is not close. As a result, offenders are likely to express their opinions directly to their close friends and defend their own behaviors, believing that their close friends will not be offended or they can compensate to their close friends at a later time. Following this logic, if a relationship is not so close, offenders need to be cautious when they choose an account to explain their behavior because a defensive account may escalate the conflict and ultimately lead to the dissolution of the relationship. For example, when a conflict occurs between acquaintances, offenders realize that their relationship with victims is much less stable and more likely to be dissolved than if the conflict were with close friends. When a conflict occurs between parties that have a moderately close relationship, compared to the conflict that occurs between close friends, offenders will use a less defensive account,
21 hoping it will reduce the possibility for relationship termination. On the other hand, if there is no established relationship, such as when a conflict occurs between strangers, offenders will be likely to defend or justify their behaviors by using a defensive account because they do not perceive a high possibility of future interaction with the victim. Based on the above arguments, it is hypothesized that: H2: Relational closeness has a curvilinear effect on the defensiveness of accounts used in a conflict situation. In particular, less defensive accounts are used when a conflict occurs between two parties that have a moderately close relationship and more defensive accounts are used when a conflict occurs between parties that are more distant or are closer. When discussing the influence of relational closeness on conflict strategies, neither Richard and Green (2006) nor Fitzpatrick and Winke (1979) took outcome severity into consideration. The findings from Richard and Green’s study may be valid only when the outcome of a conflict is not very severe. When the outcome of a conflict between close friends is not severe, offenders may feel that they can compensate their victims later. Thus, at the moment of the conflict, offenders could justify their behaviors or defend themselves. However, if the outcome is severe enough to possibly damage the relationship, offenders will be likely to admit responsibility and use a less defensive account than when the outcome is not severe. When offenders and victims are acquaintances, the risk of relationship termination and the desire to preserve a good image causes offenders to be willing to use a less defensive account regardless of severity of the outcome. Furthermore, when facing a stranger, with whom one has a small chance of future interaction, offenders will be more likely to use a defensive account if the
22 outcome is severe than if the outcome is not severe. Thus, the current research predicts the following relationship: H3: The relational closeness and the severity of outcome interact to influence the defensiveness of the account used in a conflict situation. (H3a) When a severe conflict occurs between close friends, offenders are more likely to use a less defensive account than when the conflict is not severe. (H3b) When the conflict is between parties that are moderately close, offenders are more likely to use a less defensive account than when the conflict is between parties that are more distant or closer, regardless of the severity of outcome. (H3c) When a severe conflict occurs between strangers, offenders are more likely to use a more defensive account than when the conflict is not severe. The above hypothesis predicts that the defensiveness of the account used in a conflict situation depends on outcome severity and relational closeness. But the hypothesis does not explain the process by which offenders choose certain accounts to explain their behaviors. Weiner’s (1986, 1992, 2006) attribution theory of motivation and emotions provides the framework to examine the cognitive and emotional processes that offenders engage in to determine the accounts used in conflict situations. Attribution Theory Research on accounts is closely related to attribution theory (Fiske & Taylor, 1991; Harvey & Weary, 1984; Heider, 1958; Jones & Davis, 1965). Attribution theory has focused on the process by which people create explanations for their own or others’ actions (Heider, 1958; Jones & Davis, 1965; Kelley, 1986). These explanations allow offenders in a conflict situation to interpret their behaviors and decide an account to use
23 (Orbuch, 1997). Offenders can attribute their behaviors in terms of locus (internal vs. external) and controllability (controllable vs. uncontrollable). Causal inferences influence what tactics people choose to handle conflict. For example, Sillar (1980) found that when people attributed the causes of conflict to others (e.g., their roommates), they used more distributive acts (e.g., explicitly expressing negative evaluations of the other party) to handle conflict than integrative acts (e.g., using neutral or positive statements to describe the other party involved in conflict). How offenders make attributions can also influence their choice of accounts. Weiner’s Attribution Theory of Motivation and Emotion Weiner (1986, 2000) identified three dimensions of causality—locus, stability, and controllability—to explain attribution processes. Locus refers to whether a cause is internal or external to the person. Stability refers to whether a cause is relatively enduring across situations or is situation-specific. Controllability refers to whether a cause can be subject to volitional influence. For example, if a meeting is cancelled because of a hurricane, the cause of the scheduling conflict is considered external, unstable, and uncontrollable. Weiner’s (1986, 1992, 2006) work has focused on using locus and controllability to explain people’s account-giving processes. According to Weiner (1995), when a social contract has been broken, offenders often first consider the antecedent conditions about what happened and why the social contract was broken. Furthermore, offenders make causal inferences, such as whether the reasons for the violation of the social contract are related to offenders or related to environmental factors. Weiner argued that based on the causal inferences, offenders imagine the consequences if the reasons were communicated to victims. Perceptions of
24 the causes of the behavior, such as offenders’ ability, effort, and luck, lead to inferences about responsibility, which in turn generate emotional responses, such as anger, from victims.2 With the anticipated consequences in their mind, offenders will decide on the explanations that they want to communicate to victims or to other people. There could be many antecedent conditions that lead to a conflict. Here the focus is on two contextual factors: outcome severity and relational closeness. People tend to attribute their own success or positive behaviors to internal reasons and failure or negative behaviors to external reasons (i.e., the ego-protective bias; Kelley, 1986). When offenders’ behavior leads to a conflict, offenders are likely to make external attributions about their behavior in question. The more severe the outcome of the conflict is, the more likely offenders will tend to explain their behaviors by pointing to factors outside of them or factors that are not in their control. This attribution may change depending on how close offenders are with the victim. Specifically, when offenders and victims are relationally close, offenders understand the importance of keeping the relationship with victims and also offenders understand that victims may know them very well to figure out why the behavior in question has occurred. When the outcome of the conflict between close friends is severe, offenders tend to attribute their behavior to internal and controllable reasons, as a sign to show sincerity and honesty to their friends and as a sign to show their desire to maintain the relationship. When a conflict occurs between strangers, the concern for the relationship is minimal. Offenders are likely to experience ego-protective bias, especially when the outcome of conflict is severe. Thus, the following prediction is made:
Weiner (1986, 2000) discussed two emotions: anger and sympathy. The current research will only examine anger because the study is placed in a conflict context, within which the arousal of negative emotions is more likely.
2
25 H4: The relational closeness and outcome severity interact to affect offenders’ attributions. When a severe conflict occurs between close friends, offenders will make more internal attribution (H4a) and more controllable attribution (H4b) than when the conflict is not severe. When a severe conflict occurs between strangers, offenders will make more external attribution (H4c) and more uncontrollable attribution (H4d) than when the conflict is not severe. Perceptions of causes, such as internal or external causes and controllable or uncontrollable causes, lead to judgment of responsibility, which in turn generates emotional responses such as anger from victims. When causes are perceived as internal or controllable, offenders often are judged personally responsible for their behavior because they could have done otherwise to prevent the negative event. On the contrary, when causes are perceived to be external or uncontrollable, offenders often are judged as having little or no personal responsibility for their behavior because there was not much they could have done to prevent what had happened. The judgment of responsibility is associated with expected anger. When offenders expect more responsibility to be assigned to them, they will expect victims to feel angrier about the behavior in question. Weiner and his associates studied the influence of attributions on expected responsibility assigned to offenders and expected anger felt by victims (Weiner, Amirkhan, Folkes, & Verette, 1987; Weiner, Figueroa-Muñoz, & Kakihara, 1991; Weiner & Handel, 1985; Yirmiya & Weiner, 1986). For example, Graham, Weiner, and Benesh-Weiner (1995) found that when victims perceived causes of a negative event as controllable, victims held offenders more personally responsible for their behavior and more anger was elicited, as compared to when victims perceived the causes as
26 uncontrollable. When victims assign more responsibility to offenders and feel angrier about offenders’ behavior, they tend to evaluate offenders negatively and show little interest in future interaction with the offenders (Weiner et al., 1987). Weiner’s research suggested that attributing causes of actions to external and uncontrollable reasons benefits offenders because victims are likely to forgive offenders’ behavior and assign less responsibility to them and victims also will feel less angry about the offenders’ behavior, compared to the situation when offenders attribute causes of actions to internal and controllable reasons. Although Weiner’s (1986, 2000) attribution theory of motivation and emotions was mainly applied to the excuse-giving process, his framework has provided insights into understanding offenders’ choice of other types of accounts, such as concession, refusal, and justification. When offenders perceive the causes of their actions as internal or controllable, they expect that victims assign more responsibility to them. A less defensive account acknowledges offenders’ wrong-doing and their desire to take full or partial responsibility, which helps to make victims believe that the offenders’ behaviors were not intentional, and consequently, decreases the possible punishment directed towards offenders. Thus, a less defensive type of accounts, such as concessions, is a better choice than a more defensive type of accounts, such as refusals. On the other hand, when causes are perceived as external or uncontrollable, there is not much that offenders can do to prevent what has already occurred; offenders will be judged as having little or no personal responsibility for their behaviors (Liu & Yao, 2007). Thus, offenders are more likely to use a more defensive account when they perceive the causes of conflict as external or uncontrollable than when they perceive the causes of conflict to be internal or
27 controllable. According to Weiner (1995), assignment of responsibility is associated with emotional reactions toward the behavior in question. Specifically, the assignment of personal responsibility is associated with anger and a tendency to attack or punish. Inspired by Weiner’s theory, Allred, Mallozzi, Matsui, and Raia (1997) investigated the influence of anger and compassion on negotiation performance. Their findings supported Weiner’s (1995) argument: Participants who attributed a higher level of responsibility to their partners reported feeling more angry than participants who attributed a lower level of responsibility to their partners. Liu and Yao’s (2007) study of cultural differences in the association between emotional arousal and attribution of personal responsibility had similar findings. Averill’s (1983) study on anger provided further support to the proposition regarding the relationship between attributions and emotional response. Averill asked participants to report recent events that made them angry. Nearly 80% of the reported incidents that elicited anger involved controllable behaviors. On the contrary, when victims perceived that the causes of a conflict were not controllable, victims were more likely to forgive offenders’ behaviors because victims perceived that the behaviors in question were not intentional. In this case, responsibility was less likely to be assigned to offenders and less anger was aroused than when the causes were believed to be controllable. Based on the above arguments, the following hypotheses are proposed: H5: When offenders make more internal attributions about their behavior in a conflict situation, offenders expect more responsibility to be assigned to them (H5a) and expect more anger to be felt by victims (H5b).
28 H6: When offenders perceive the causes of a conflict as more controllable, offenders expect more responsibility to be assigned to them (H6a) and expect more anger to be felt by victims (H6b). H7: When offenders make more external attributions in a conflict situation, offenders expect less responsibility to be assigned to them (H7a) and expect less anger to be felt by victims (H7b). H8: When offenders perceive the causes of a conflict as more uncontrollable, offenders expect less responsibility to be assigned to them (H8a) and expect less anger to be felt by victims (H8b). H9: When offenders expect more responsibility to be assigned to them, they expect more anger to be felt by victims. H10: When more anger is expected, a less defensive account is used. A causal model is established to explain the attributional processes offenders use to determine the accounts selected in conflict situations (see Figure 2). The rationale of all links has been presented in the research question and the ten hypotheses discussed in this chapter.
29
OS (?1)
POS (?1)
RC (?2)
PRC (?2)
Inter1 (?3) LDA (?9)
IA (?3) Inter2 (?4) CON (?4) Eresp (?7)
EA (?5) Inter3 (?5) Uncon (?6)
Eanger (?8)
Figure 2. Structural relations based on the hypotheses. Note. All exogenous variables are allowed to covary and ?s are also allowed to covary (errors in equations). OS (i.e., ?1) refers to outcome severity manipulation (low outcome severity was coded as -1 and high outcome severity was coded 1) and RC (i.e., ?2) is relational closeness manipulation (quadratic coding: low relational closeness was coded as 1, moderate relational closeness was coded as -2, and high relational closeness was coded as 1). Inter1 (i.e., ?3), Inter2 (i.e., ?4), and Inter 3 (i.e., ?5) are the interactions between outcome severity and relational closeness. In the model testing, ?3 was calculated by multiplying the outcome severity manipulation (coded as 1, -1) and relational closeness manipulation (linear coding: low relational closeness was coded as -
30 1, moderate relational closeness was coded as 0, and high relational closeness was coded as 1). ?4 refers to the coding of the high severity and close relationship condition (coded as 1) versus all other five conditions (coded as 0). ?5 refers to the coding of the interaction between outcome severity and relational closeness: The high severity and low relational closeness condition is coded as 1 and other five conditions are coded as 0. POS (i.e., ?1) is perceived outcome severity, with six indicators. PRC (i.e., ?2) refers to perceived quadratic relation closeness, the manipulation check for the quadratic coding of relational closeness. ?2 was created by squaring the first component score of the six items that measured relational closeness. ?2 only has one indicator in the model. IA (i.e., ?3) is the degree of internal attribution, measured by four of seven items (three items were deleted based on confirmatory factor analysis and principal components analysis results; see details in the result section for the formal study). CON (i.e., ? 4) is controllability, with five indicators. EA (i.e., ?5) is the degree of external attribution, measured by seven items. Uncon (i.e., ?6) is uncontrollability, with five indicators. Eresp (i.e., ?7) refers to expected responsibility assigned to the offender, measured by six items. Eanger (i.e., ?8) is expected anger, measured by six items. And LDA (i.e., ?9) refers to the measure of the likelihood of selecting a defensive account. ? 9 only has one indicator and the detail for obtaining the composite measure for ?9 was discussed in Pilot Study 5 and the formal study.
31 CHAPTER III Method This chapter describes the proposed stimulus message and instruments, the pilot studies that tested these proposed materials, and the methods of the formal study. Scales used in the current research were created by the researcher, and pilot studies were conducted to evaluate the effectiveness of the inductions and the measurements to be used in the final study. The formal study tested the causal model (see Figure 2) and the hypotheses. The pilot studies and the formal study were approved by the University of Maryland Institutional Review Board on May 28, 2007 (for the first two pilot studies) and on February 12, 2008 (Pilot Study 3 to Pilot Study 6 and the formal study). Data for Pilot Study 5 were collected at Villanova University and all other studies were conducted at University of Maryland. Approvals for conducting Pilot Study 5 at Villanova University were obtained from the University of Maryland and the Villanova University Institutional Review Boards on April 10 and April 15, 2008, respectively. Data were collected during the summer of 2007 and between March and June, 2008. Overall Design, Stimulus Message, and Instruments Overall Design An experiment was conducted to test the causal model, research question, and the hypotheses. Outcome severity and relational closeness were manipulated in a hypothetical scenario that described an interpersonal conflict in a group project at school. Two levels of outcome severity and three levels of relational closeness (high vs. moderate vs. low) were created in the hypothetical scenario. Six experimental conditions were developed. Attributions (the degree of internal attribution, the degree of external
32 attribution, controllability, and uncontrollability), anticipated consequences (expected responsibility assigned to offenders and expected anger from victims), the defensiveness level of each type of accounts (concession, justification, excuse, and refusal), and the likelihood of selecting each account were measured.3 Participants were randomly assigned to one of six conditions. They read a hypothetical conflict scenario, were told to imagine they were the offender in the scenario, and answered questions afterwards from the offender’s perspective. Hypothetical Interpersonal Conflict Scenario Two criteria need to be fulfilled to create a hypothetical conflict scenario. First, the conflict incident used in the scenario should be familiar to participants so that participants can easily relate to the situation and answer questions without difficulty. Second, the current research focuses on the accounts offenders use to explain their behaviors. However, not every incident invites an explanation. Participants may use avoidance to handle conflicts, but avoidance is not the focus of the current research. To ensure an account is offered, the need to explain a wrong-doing should be clearly demonstrated in the scenario. Pilots Studies 1 and 2 were used to create a hypothetical scenario that meet the above two criteria. Details are provided in the section on pilot studies. Based on the results of Pilot Studies 1 and 2, a scenario about a conflict between group members for a class project was chosen to be developed as the stimulus message in the proposed study. Group work is a familiar practice for college students. Because college students were the
Pilot Study 5 measured the defensiveness level of each account and the formal study measured the likelihood of selecting each account. Results from Pilot Study 5 and the formal study were combined to create the composite measure of choosing accounts as the dependent measure in model testing. Details can be found in the discussion of Pilot Study 5.
3
33 major source of participants in the studies, using a hypothetical scenario that involves a conflict between group members is easily understood by participants. To ensure that an account would be necessary, the hypothetical conflict scenario clearly specified an offense conducted by the offender. In particular, the hypothetical scenario described a group project between two classmates: writing a review paper together. One wrote his or her portion of the review paper based on the wrong readings, which reduced the grade for both group members. Participants were told to take the role of the offender and that an explanation is necessary when the scenario stated that the offender’s behavior brought negative consequences to the victim. The details of the hypothetical conflict scenario can be found in the section on Pilot Studies 1 and 2. Role Playing in a Hypothetical Conflict Scenario A hypothetical conflict scenario was used for several reasons. First, Martin (2006) argued that a hypothetical scenario allows the researcher to systematically manipulate variables of interest, minimizing the influence of possible extraneous variables on dependent variables. Using a hypothetical conflict scenario, different levels of outcome severity and relational closeness may be manipulated and their influence on offenders’ choice of accounts can be examined. Second, a hypothetical scenario can present respondents with situations that they may actually encounter, allowing researchers to "tap into the expectations and reactions which they (respondents) would have in similar social circumstances" (Gerber, 1994, p. 4). Third, as argued by Converse and Presser (1986), a hypothetical scenario can keep respondents’ focus on the variables intended by researchers. Given the above reasons, a hypothetical scenario was used in the current research and participants were asked to take the role of the offender in the scenario and
34 answer questions from the offender’s perspective. Naming the Hypothetical Other Party Participants were told to imagine that they were the offender in the group project described in the hypothetical scenario; the scenario used “you” to refer to the participants. The teammate (i.e., the victim) in the scenario needs a name to ensure the flow of the story. Although participants may find it easier to relate to scenarios that use actual names (Wang, 2006), actual names often are associated with specific gender and racial information, which may influence how participants process the stimulus message. For example, a labor market study by Bertrand and Mullainathan (2003) found that a very White-sounding name on a resume received significantly more callbacks for interviews than a very African-American-sounding name. Moreover, El-Sheikh, Buckhalt, and Reiter (2000) discovered that different degrees of emotional arousal are associated with gender in hypothetical scenarios. Men perceived more emotional arousal with a male hypothetical character than women perceived with a female counterpart in a hypothetical situation (cited in Wang, 2006). To ensure that participants will not be influenced by the gender and racial information attached with actual names, the current research decided to use “Person B” instead of an actual name for the victim in the scenario. Instruments Perceived outcome severity. The hypothetical scenario varied two levels of outcome severity (high severity vs. low severity). A list of measures was developed to assess whether the manipulation of outcome severity was successful (see Appendix A). An example item for outcome severity is, “how severely does your mistake affect Person B?” Magnitude scales were used in the study where 0 indicates not severe at all and 100
35 indicates a moderate level of severity. The items were tested and revised based on Pilot Study 3. Perceived relational closeness. Three levels of relational closeness were manipulated in the hypothetical scenario. A list of measures was developed to examine whether the manipulation of relational closeness was successful (see Appendix A). An example question is, “how familiar are you with Person B?” Magnitude scales were used in the study where 0 indicates not familiar at all and 100 indicates a moderate level of familiarity. The items were tested and revised based on Pilot Study 3. Attributions. Although studies (e.g., Ross, 1982) have examined attributions in terms of causal (i.e., the degree of internal attribution is at one end of the continuum and the degree of external attribution is at the other end) and controllability (i.e., controllability and uncontrollability are at the two ends of the continuum) dimensions, whether internal attribution is opposite to external attribution and controllability is opposite to uncontrollability remains questionable (Harvey & Weary, 1984). The current research does not want to assume the relationships among the four types of attribution. Thus, these four types of attributions were measured separately: the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability. Different attributions regarding the offender’s behavior were listed and participants indicated their belief about how likely each reason influenced his or her behavior (see Appendix B). Again, magnitude scales were used. Example questions include “how likely is that what has happened was due to your own carelessness?” for the degree of internal attribution; “how likely is that what has happened was due to the professor not giving
36 clear instructions for the assignment?” for the degree of external attribution; “how likely is that you could have prevented this mistake?” for controllability; and “how likely is that what has happened was out of your control?” for uncontrollability. In all magnitude scales, 0 = not likely at all, and 100 indicates a moderate level of likelihood. Anticipated consequences. Anticipated consequences referred to two variables: expected responsibility assigned to the offender and expected anger felt by victims. For expected responsibility, participants were asked to indicate their expectation about how much responsibility would be assigned to them. A magnitude scale was used with 0 meaning not expected to be responsible at all and 100 indicating a moderate level of expected responsibility. A sample item is, “how much responsibility would Person B assign to you for what has happened?” For expected anger, participants indicated the level of expected anger felt by victims by a magnitude scale, where 0 means not angry at all and 100 indicates a moderate level of expected anger. An example item is, “how annoyed do you expect Person B to be?” The complete list of items that measure both expected responsibility and expected anger is included in Appendix C. Expected choice of accounts. Unlike traditional accounts research in which coders code actual accounts that participants provide to explain their behaviors into different types of accounts, a list of twenty accounts (5 concessions, 5 justifications, 5 excuses, and 5 refusals) was developed in Pilot Studies 3 and 4. In the formal study participants were asked to indicate the likelihood that they would choose each account in the hypothetical conflict situation by using a magnitude scale (see Appendix D for the complete list of measures on choice of accounts). On that scale, zero means not likely at all and 100
37 indicates a moderate level of likelihood. The hypotheses proposed in the current research predicted that the influence of outcome severity and relational closeness was not only on the likelihood of selecting accounts but also on the defensiveness level of accounts. For example, Hypothesis 3a states that, “when a severe conflict occurs between close friends, offenders are more likely to use a less defensive account than when the conflict is not severe.” Only using the likelihood of selecting accounts is not enough to test the hypotheses. To create a measure that incorporates both the likelihood and the defensiveness level of each account, Pilot Study 5 was conducted to determine the defensiveness of each type of accounts and the results of Pilot Study 5 were used with the results in the formal study to create the composite measure needed for hypothesis testing (i.e., LDA: the likelihood of selecting a defensive account). The details of creating the composite measure were presented in the discussion of Pilot Study 5 and the formal study. Summary. This section has introduced the overall experimental design and the operationalizations of the relevant constructs proposed in the study: the hypothetical scenario, the name choices in the hypothetical scenario, and the instruments—for manipulation checks, the attributional processes, anticipated consequences, and the choice of accounts. The realism of the experimental inductions and the effectiveness of the instruments proposed for the current research were tested in six pilot studies, which are described below. Pilot Studies Six pilot studies were conducted in the summer term (June to July, 2007) and
38 between March and June, 2008. The majority of participants in each pilot study were undergraduate students enrolled at the University of Maryland (except that Pilot Study 5 was conducted at Villanova University). Participants received a small amount of extra credit from their course instructors in exchange for their participation. Students who participated in the summer term were informed not to participate in any other portions of this research that would take place later. Because all the participants had to sign up for the studies through an online participation pool program, students’ online identification numbers were used to ensure that they could only participate in one of the studies. All participants were debriefed in detail after they completed their participation. Pilot Study 1: Collecting Conflict Scenarios Purpose. The purpose of this pilot study was to collect self-reported conflict incidents to provide information for developing a hypothetical scenario that could be used in the formal study. Participants. Self-reported conflict scenarios were collected in the United States and China.4 Forty-nine participants were recruited from undergraduate communication courses at the University of Maryland. The age range for American respondents was 18
National cultures (i.e., China vs. the U. S.) were considered as one independent variable in the prospectus for the current research. Data were collected in both the United States and China in Pilot Studies 1, 2, and 3. Before the researcher could continue collecting data for the rest of the pilot studies and the formal study, two issues arose that interrupted the data collection process. First, it became apparent that the incentives were different for participants in each culture. Although obtaining extra credit in exchange of the participation is a common practice at the American university where the study was conducted, Chinese instructors would not give permission for their students to receive extra credit. Chinese students participated in the study voluntarily. Differences in incentives may be a factor that contributes to the differences in the results, making the conclusions problematic. Second, massive protests against media bias occurred in China after collecting data for Pilot Study 3, with college students being the major protesters. Given that this study measures perceptions in a hypothetical conflict situation, there is a concern that Chinese perceptions about conflict might be influenced by these protests. Consequently, the validity of the results may be questionable. Therefore, the researcher decided not to include national culture as a variable in the formal study. Results for Pilot Studies 1 and 2 still included Chinese data to create a hypothetical conflict scenario that could be used in data collection in China in the future. Pilot Study 3 only analyzed data collected in the United States because the formal study would only recruited American participants; results based on Chinese data were not relevant.
4
39 to 39 (M = 21.73, SD = 3.38, Mdn = 21.00). Seventy-six percent of participants were female (n = 37). Forty-four Chinese participants were recruited from undergraduate and graduate courses in two Chinese northern cities (i.e., Beijing, the capital of China and Changchun, the capital of Ji Lin Province). The age range for Chinese participants was 19 to 44 (M = 24.10, SD = 5.08, Mdn = 23.00). Fifty-seven percent of Chinese participants were female (n = 25). Self-reported conflicts were collected in both China and the United States to develop a hypothetical conflict scenario that could be used in different cultures for the future studies. Procedure. This pilot study was conducted as an online survey. American participants completed the survey through an online participation pool program and Chinese participants completed the survey through an online survey Web site. Information about the purpose of the study was provided, followed by a consent form. Participants were asked to read the consent form and provide their electronic signature if they agreed to participate in the study. Only after they agreed, participants were allowed to continue with the questionnaire. The questionnaire had two sections. The first section asked participants to recall a situation in which a conflict occurred recently between them and another person. Specifically, participants were asked to recall a conflict that happened due to their behavior. Participants described the conflict in detail; for example, participants were asked to provide information about the relationship between them and the other person involved in the conflict, the cause of the conflict from participants’ perspective and from the other person’s perspective, the consequences of the conflict, the emotions participants experienced and they expected the other party to have experienced at the time of the
40 conflict, etc.. In the second section, participants were instructed to recall in detail their explanation about the behavior that led to the conflict they described. All the questions were open-ended, allowing participants to provide details of a recent conflict they experienced. See Appendix E for the complete questionnaire. Results. A total of 93 different conflict scenarios were generated. A sorting task (Pilot Study 2) was conducted to sort the reported incidents based on the outcome severity and the relational closeness between the two parties involved in the conflict. Pilot Study 2: Developing a Hypothetical Conflict Scenario Purpose. The purpose of Pilot Study 2 was to develop a hypothetical conflict scenario based on the qualitative responses collected in Pilot Study 1. Participants. Seven Americans and six Chinese were recruited to participate in the second pilot study: sorting ninety-three conflict scenarios that were collected in Pilot Study 1. The ninety-three conflict scenarios included some incidents collected in China. Therefore, Chinese participants were recruited to ensure the sorting would be not biased by having only American participants. The age range for American sorters was 17 to 21 (M = 19.14, SD = 1.35, Mdn = 19.00), whereas Chinese sorters’ age range was 26 to 33 (M = 29.33, SD = 2.50, Mdn = 29.00). Six of the American sorters were female and five of the Chinese sorters were female. American sorters were undergraduate students enrolled in communication courses at the University of Maryland. Due to the difficulty of locating Chinese natives in undergraduate courses, Chinese graduate students at the same university were asked to participate in the sorting task. Procedure. Sorters arrived at a pre-arranged classroom alone. The researcher explained the purpose of the study and obtained the consent from participants that agreed
41 to be part of the study. Each sorter was provided with ninety-three numbered index cards. On each index card, there was one conflict scenario collected in Pilot Study 1. Sorters were instructed to first read over all ninety-three conflicts carefully and ask questions if they had difficulty understanding any scenario. After the researcher made sure that sorters fully understood the 93 scenarios, they were asked to sort the scenarios based on its outcome severity and relational closeness. Outcome severity was defined in the instructions as how serious the consequence of the conflict might be and relational closeness was the extent to which the people involved in the conflict were close to each other. Based on the hypotheses, a sorting table with two levels of outcome severity (high vs. low) and three levels of relationship closeness (high vs. moderate vs. low) was presented to the sorters. Sorters were instructed to put the number of the card in the most appropriate cell in the sorting table. They were told that each conflict could be put in one and only one cell in the table. See Appendix F for the complete instructions and the sorting table used in the Pilot Study 2. Results. Looking at frequency of participants’ sorting of ninety-three scenarios indicated that five incidents were evenly sorted into six cells (see Appendix G for these five incidents reported by participants in Pilot Study 1), which allowed the researcher to revise the incidents easily based on the independent variables. Among all five incidents, one scenario about a conflict in school projects was chosen for the hypothetical conflict scenario to be used in the formal study because participants in the formal study were to be mainly undergraduate students. Working on a group project for a class is a familiar practice to the participants. The selected incident was revised to reflect the variations of two independent
42 variables (outcome severity and relational closeness). Six versions of the hypothetical conflict scenario were generated, with two levels of outcome severity and three levels of relational closeness. All six versions had 12 sentences, with an average of 17 words per sentence and a Fresch reading ease rating of 74.5 (Microsoft Office, 2003). All scenarios started by defining relational closeness between an offender and a victim, followed by the description of a conflict between the offender and the victim and the consequence of the conflict. For example, in the high relational closeness conditions, the relationship between the offender and the victim reads: “You and Person B grew up together and attended the same university. The two of you have been best friends for many years.” The moderate relational closeness condition says “You and Person B were in couple of classes together before. Although the two of you know each other, you only see each other in class.” When the relational closeness between the offender and the victim is low, the scenario reads “You and Person B do not really know each other. You happened to sit next to each other on the first day of class.” Similarly, the two levels of outcome severity were manipulated in the scenarios. The details of the manipulations are discussed in the method section of the formal study; see Appendix H for all six versions of the hypothetical conflict scenario. Pilot Study 3: Manipulation Checks Purpose. The purpose of this pilot study was to test whether the manipulations of the two independent variables were effective. Although the hypothetical conflict scenario was created in a way to be suitable in both China and the United States, only data collected from Americans participants were analyzed because the formal study would not be conducted outside of the United States.
43 Participants. Participants (N = 80) were recruited from undergraduate communication courses at the University of Maryland. The age range was 18 to 25 (M = 20.38, SD = 1.58, Mdn = 20.00). Eighty-one percent of participants were female (n = 65). Sixty-five percent of participants identified themselves as Caucasian. Study design and procedure. A 3 (relational closeness: high vs. moderate vs. low) X 2 (the severity of an outcome: high severity vs. low severity) independent groups experiment was conducted. Participants were recruited through a departmental participation pool program, which connected participants to an external online survey Web site to participate in the study. Participants were randomly assigned to read one of six hypothetical conflict scenarios and they were asked to imagine they were the offender in the scenario. After reading the stimulus message, each participant was first asked to provide an explanation about the behavior described in the scenario (i.e., wrote their portion of paper based on the wrong readings) from the perspective of the offender. After that, two questions that were not relevant to the hypothetical scenario were provided to participants for them to practice using magnitude scales before they answered the questions. Those two questions were: “How much do you like the food offered at the dining hall of your school?” and “How much do you like the service at the main library of your school?” Participants were told to continue with the questionnaire only when they felt ready to use the scale. The questionnaire had measures for perceived outcome severity and perceived relational closeness (see Appendix I for the instructions and measures used in Pilot Study 3). After participants completed the questionnaire, they were debriefed. The same two practice questions were used in Pilot Studies 5 and 6 and the formal study.
44 Data coding and data transformation. Outcome severity has two levels (high and low) and relational closeness has three levels (high, moderate, and low). When coding the data, high severity is coded as 1 and low severity is coded as 0. Three levels of relationship closeness (high, moderate, and low) were coded as 1, 0, and -1, respectively. Participants used any nonnegative number to indicate their views about the hypothetical conflict. Data collected in Pilot Study 3 were submitted to a descriptive analysis; the skewness and kurtosis were examined to see whether a transformation was necessary. The ratio of the skewness over the standard error for each item was calculated and if the ratio was greater than 1.96, the data were significantly positively skewed and if the ratio was smaller than -1.96, the data were significantly negatively skewed (Frey, Botan, & Kreps, 2000). Using the (-1.96, 1.96) cutoff rule, all eight indicators (four for perceived relational closeness and four for perceived outcome severity) resulted into positively skewed data. A data transformation was needed. A power transformation is considered one strategy to transform positively skewed data resulting from magnitude scales (Bauer & Fink, 1983). A transformation can also helps with the assumption of homescedasticity that was used for testing the manipulation checks. When the variance of the residuals is not uniform across all levels of the independent variables, the homoscedasticity assumption is violated. Kline (2005) argued that a transformation “may remedy heteroscedasticity due to nonnormality” (p. 5). Before transformation, the data set was first trimmed at the 90th percentile to eliminate the extreme values (e.g., the 90th percentile for one item to measure outcome severity was 500; thus, any score larger than 500 was converted to 500). The data after trimming were submitted to a power transformation using the following formula:
45 New indicator = (original indicator) ?, different values of ? were tried in the power transformation and when ? = 0.40, the skewness of each indicator was not significant (p > .05). Therefore, 0.40 was used as the value of ? in this pilot study. For the power transformations conducted in the following pilot studies and the formal study, the same process was used: Different values of ? were tried and the one that ensured the non-significant skewness of all or the majority of indicators was selected. Instruments. A 4-item magnitude scale was used to assess the effectiveness of outcome severity manipulation (see Appendix I). An example item is “How much influence does your mistake have on Person B’s scholarship for the next year?” Magnitude scales were used in this pilot study where 0 indicated not severe at all and 100 indicated a moderate level of outcome severity. Confirmative factor analyses (CFAs) were conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results indicated that the scale for perceived outcome severity was not unidimensional, ?²(2, N = 80) = 7.01, p < .05; NFI = 0.97; CFI = 0.98; SRMR = 0.03. However, when a principal components analysis with no rotation was conducted on the four items for perceived outcome severity, only one eigenvalue was larger than 1 (eigenvalue = 3.10), which explained 76% of the variance in the items. The first component score was saved as the manipulation check. The Cronbach’s alpha for the sum of all four items testing perceived outcome severity was .71. Another 4-item magnitude scale measured the perceived relational closeness. Participants were asked questions such as “How familiar are you with Person B?” Magnitude scales were used in this pilot study where 0 indicated not close at all and 100
46 indicated a moderate level of perceived relational closeness. Confirmative factor analyses (CFAs) were conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results supported a one-factor structure for the perceived relational closeness, ?²(2, N = 90) = 1.73, p > .05; NFI = 1.00; CFI = 1.00; SRMR = 0.00. A principal components analysis without rotation was conducted to obtain the first component score for perceived relational closeness, which was used as the manipulation check for relational closeness. The Cronbach’s alpha for the sum of all four items testing perceived relational closeness was .96.5 Results. Two-way ANOVAs were performed to assess whether the manipulations of outcome severity and relational closeness were successful. In the first ANOVA, the dependent variable was perceived outcome severity and the independent variables were the manipulations of outcome severity and relational closeness. Results indicated that (a) the manipulation of outcome severity influenced perceived outcome severity significantly, F(1, 74) = 12.09, p < .01, ?² = .14, (b) relational closeness did not have a significant influence on perceived outcome severity, F(2, 74) = 0.60, p = .55, ?² = .02, and (c) no significant interaction between the manipulations of outcome severity and relational closeness was found, F(2, 74) = 0.11, p = .90, ?² = .00. Participants in the high outcome-severity condition (M = .38, SD = 1.12) regarded the outcome to be more severe than those in the low outcome-severity condition (M = -.36, SD = .71). The manipulation
In Pilot Study 3, both the scales for perceived outcome severity and perceived relational closeness had four items. The reliability for the scale of the perceived outcome severity was acceptable but not very satisfactory. Additionally, CFA results did not yield a unidimensional factor structure for perceived outcome severity. Because only four items were used, dropping items from the scale was not appropriate (a saturated measurement model would result if any item were dropped from the perceived outcome severity scale). To enhance the reliability of the scales for the formal study, additional items need to be added (Wrench, Thomas-Maddox, Richmond, & McCroskey, 2008), which also allows the researcher to drop one or more items if the factor structure were not unidimensional. Two items were added to the scale for perceived outcome severity, which were phrased similar to the other items. Two additional items were added to the scale for perceived relational closeness as well.
5
47 of outcome severity was successful. The second ANOVA had perceived relational closeness as the dependent variable and the manipulations of outcome severity and relational closeness as the independent variables. The results suggested that (a) relational closeness was manipulated effectively, F(2, 74) = 72.58, p < .01, ?² = .60, (b) the manipulation of outcome severity did not have a significant effect on perceived relationship, F(1, 74) = 0.47, p = .50, ?² = .01, and (c) no significant interaction was found between the manipulations of outcome severity and relational closeness, F(2, 74) = 0.33, p = .72, ?² = .01. Participants in the high relational closeness condition (M = 1.19, SD = .86) reported being closest to Person B than those in the moderate closeness condition (M = -.43, SD = .47) and in the low relational closeness condition (M = -.64, SD = .34). A linear relationship between the three levels of relational closeness was established, F(1, 79) = 104.13, p < .01, ?² = .56. Limitation and revision. The effect size for perceived outcome severity was only .14. According to Cohen (1988), the threshold for a small effect size is .20. Therefore, the hypothetical scenario needed to be revised to get a larger effect size for the outcome severity manipulation. To revise the scenario, five undergraduate students were interviewed on campus (4 were female). Two versions of the hypothetical scenarios (high vs. low outcome severity with relational closeness controlled) were read by these five interviewees. The researcher then interviewed the participants with questions such as whether the story was realistic, what the researcher intended to vary in two scenarios, and how to enlarge the difference between low and high level of outcome severity. Based on interviewees’ suggestions, appropriate revisions were made to the hypothetical conflict scenario and a
48 second test was conducted to specifically assess the manipulation of outcome severity. Two versions of the hypothetical conflict scenarios were used in the second test with relational closeness controlled at the high closeness level. In the second test, sixteen participants were recruited from an introductory communication course at the University of Maryland. Their age range was 18 to 19 (M = 18.35, SD = 0.49, Mdn = 18.00). Seventy-one percent of participants were male (n = 12), and fifty-nine percent of participants identified themselves as Caucasian. Participants were randomly assigned to read one of two versions of the hypothetical conflict scenario (high vs. low outcome severity). Participants were not asked to provide explanations but simply completed the manipulation checks, imagining they were the offender in the scenario; the same four-item scale in the first test of Pilot Study 3 was used. The data were positively skewed and a power transformation with ? = 0.50 was conducted to meet the assumption of normality for the analyses of the variance. A CFA was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The Chi-square value suggested a unidimensional factor structure although the fit indices were not satisfactory, ?²(2, N = 16) = 3.74, p > .05; NFI = 0.90; CFI = 0.93; SRMR = 0.05. The Cronbach’s alpha for the sum of all four items testing perceived outcome severity was .83. A principal components analysis without rotation was conducted and the first principal component score was saved for the analysis of the variance. In the ANOVA, the dependent variable was perceived outcome severity and the independent variable was the manipulation of outcome severity. Results indicated a significant main effect of outcome severity manipulation, F(1, 14) = 6.66, p < .05, ?² = .32. Participants in the high outcome-severity condition (M = -.12, SD = .20) perceived the consequence of the
49 conflict to be more severe than participants in the low outcome-severity condition (M = -.36, SD = .17). Based on these results, the induction of outcome severity was successful. The revised scenarios were used in the later studies. Pilot Study 4: Creating the Measure of Accounts Purpose. Pilot Study 4 was conducted to develop a measure for accounts. Coders and procedure. In Pilot Study 3, participants were asked to imagine they were the offender and to provide an explanation about the behavior described in the hypothetical scenario presented to them. Eighty explanations were collected from participants. Two trained coders coded the eighty explanations into four types of accounts examined in the current research (i.e., concession, justification, excuse, and refusal), and coders were also asked to select the five most frequently used explanations for each type of accounts. Coders used a coding scheme adapted from Fritsche's (2002) review of accounts (see Table 2). Results. Two undergraduate student coders (both females) were trained on the list of the twenty actual explanations collected in Pilot Study 3 until they obtained acceptable inter-coder reliability. They were instructed to code each explanation in terms of four types of accounts (i.e., concession, justification, excuse, and refusal). The overall initial agreement among two coders was 76% and intercoder reliability was calculated using Cohen’s (1960) kappa. The kappa equaled .68. Differences in coding were then discussed and resolved by the two coders. The coders also listed the five most frequently used explanations in each type of accounts, which created a list of twenty statements—five statements for each account type. Those statements were developed into a measure of accounts used in Pilot Studies 5 and 6 and in the formal study (Appendix D presents the
50 measure of accounts). Pilot Study 5: Testing the Defensiveness of Each Account Purpose. Schönbach (1990) argued that each type of accounts is associated with a certain defensiveness level. He arbitrarily placed the four types of accounts on a continuum of defensiveness (see Figure 1 in the section of literature review). However, little empirical research examined the location of each type of accounts on the defensiveness continuum. Pilot Study 5 served two purposes. First, the results would provide empirical evidence to support or disconfirm Schönbach’s (1990) ranking of the four types of accounts on the defensiveness continuum. Second, in the formal study, participants indicated their likelihood of using each type of accounts in a conflict situation. However, in the hypotheses, the likelihood of using a less or more defensive account was predicted. To be consistent with the hypotheses, a composite score (LDA: the likelihood of selecting a defensive account) was developed by using the results on defensiveness of each type of accounts from Pilot Study 5 and the results on likelihood of selecting each type of accounts from the formal study. LDA, was used for the hypothesis testing. The formula to calculate LDA is as follows: LDA = ? (Di • Li) / ? Li where Di refers to the defensiveness level of each type of accounts and Li refers to the likelihood of selecting each type of accounts, where i = 4 to represent the four types of accounts (i.e., concession, justification, excuse, and refusal). Participants. Twenty-two participants were recruited from undergraduate communication students at Villanova University. Their age range was 20 to 21 (M = 20.59, SD = 0.50, Mdn = 21.00). Eighty-two percent of participants were female (n = 18),
51 Table 2 Coding Scheme for Four Types of Accounts, Adapted from Fritsche’s (2002) Review (pp. 389-394) Types of accounts Refusal Descriptions An offender responds to a reproach that the behavior has been norm contradictive by questioning the legitimacy of the reproach. This is made possible by explicit refusal to deal with the reproach, by rejecting the validity of the salient social norm, by denying the behavior or relevance of the salience norm, or by attacking the reproaching instance. Excuse An offender responds to a reproach that the behavior has been norm contradictive by denying the connection between self and the critical behavior. The offender points to a restricted freedom of decision, which can emerge from a lack of control, an absence of planning or missing an opportunity to plan the behavior, as well as from external constraints and incentives that are not under the offender’s control. Justification An offender responds to a reproach that the behavior has been norm contradictive by denying that there is a connection between the behavior and the salient norm. The offender disputes that he or she has violated a salient norm. Concession An offender neither denies the connection to the deed, nor
52 its norm-contradictive character, nor bypassing it. Instead the guilt is admitted and the behavior is regretted.
and eighty-two percent of participants identified themselves as Caucasian (n = 18). Procedure. Participants completed the study through an online survey Web site. The link to the study was distributed to participants by an e-mail. Each participant was asked to provide an electronic signature, indicating their agreement to participate in the study, before completing the questionnaire. The questionnaire had two parts. The first part introduced a conflict scenario to participants. The purpose of this pilot study was to examine the defensiveness level of each type of accounts, not to test the influence of outcome severity and relational closeness on how people account for their behaviors in conflict situations. Therefore, outcome severity and relational closeness were not varied in the hypothetical conflict scenario presented in this pilot study. The second part of the questionnaire had four types of accounts generated in Pilot Study 4. Each type (i.e., concession, justification, excuse, and refusal) had five accounts. Twenty accounts were listed in total. For example, one of the refusals was, “I know the readings I read were the right material for the paper. I did not find anything wrong to what I contributed to the paper.” Participants indicated the defensiveness level of each account by using a six-item magnitude scale, which generated a total of one hundred and twenty questions (6 X 20) in the questionnaire. The six items for each account were the same. An example question reads, “To what extent were you being self-protective?” (Appendix J lists the scale measuring the defensiveness level of the accounts). To familiarize participants with using magnitude scales, the two practice questions in Pilot
53 Study 3 were used. Data transformation. Participants used nonnegative numbers to indicate their views about the defensiveness level of each account provided by the researcher. The data were submitted to the descriptive analysis for the skewness and kurtosis to see whether a transformation was needed. A (-1.96, 1.96) cutoff rule was used to determine whether the data were approximately normal (Kline, 2005). Based on the cutoff rule, fourteen of the one hundred and twenty items were approximately normal and the others were positively skewed. A data transformation was necessary to meet the assumption of normality for the later analyses. The data set was trimmed at the 90th percentile to eliminate the extreme values (e.g., the 90th percentile for one item to measure defensiveness was 400; thus, any score larger than 400 was converted to 400). A power transformation was used on the data set, with 0.40 selected as the value of ? that ensured the skewness of the majority of items after transformation was not significant (p > .05). Instruments. A list of twenty accounts was provided to participants. For each account, a 6-item magnitude scale was used to measure the defensiveness level of that account. For example, after reading one concession (e.g., “I am sorry that I misunderstood the instructions. Is there any way that I can make it up to you?”), participants were asked: “To what extent were you being self-protective?”, “To what extent were you guarded?”, “To what extent were you sticking out for yourself?”, “To what extent were you being assertive?”, “To what extent were you shifting responsibility?”, and “To what extent were you being defensive?” Correlations between the six items for each account were calculated (see Table 3). Four of the six items correlated with each other significantly across all twenty accounts (i.e., being self-
54 protective, being guarded, sticking out for one self, and being defensive). Therefore, the other two items (i.e., being assertive and shifting responsibility) were dropped from the scales. Twenty reliabilities were calculated with the remaining four items of the defensiveness scales. The Cronbach’s alphas for the sum of the four items measuring the defensiveness of all twenty accounts were acceptable, ranging from .68 to .92 (see the complete list of reliability coefficients in Table 4). The averages of the four-item scales after transformation were calculated for the following analysis. Results. An ANOVA with a repeated measure was conducted with four factors (i.e., the defensiveness level for concessions, justifications, excuses, and refusals). The
55 Table 3 Intercorrelations between Items for Defensiveness of Each Account Items 1 2 3 4 5 6
Concession 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Concession 2 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Concession 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 1 .47* 1 .33** .26 1 .39 .12 .82** 1 .28 .45* .19 .17 1 .37 .49* .28 .09 .56** 1 .67** 1 .64** .73** 1 .20 .45* .70** 1 .34 .08 .05 -.02 1 .46* .54** .67** .34 .25 1 1 .84** 1 .40 .54** 1 .35 .52** .62** 1 .19 .37 .56** .35 1 .36 .55** .59** .42 .59** 1
56 6. Defensive Concession 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Concession 5 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 2 (n = 22) 1 .33 1 .85** .24 1 .04 -.39 .21 1 -.22 .24 -.13 .22 1 .41 .16 .42* .30 .23 1 1 .21 1 .62** .61** 1 .61** .14 .38 1 .37 .38 .61** .12 1 .19 .81** .60** .11 .47* 1 1 .79** 1 .67** .57** 1 .11 .46* .44* 1 .38 .49 .52 .20 1 .30 .37 .56** .56** .23 1 1
57 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 5 (n = 22) 1. Self-protective 2. Guarded 1 .32 1 .88** .13 .23 -.18 .17 .82** .85** .23 1 .44* 1 .95** .45* 1 .48* -.04 .27 1 .56** .20 .55** .42 1 .86** .48* .82** .44* .57** 1 1 .34 1 .91** .17 1 .35 .21 .35 1 -.04 .55** -.18 .28 1 .84** .28 .86** .30 -.04 1 1 .35 1 .93** .33 1 .00 -.00 .02 1 -.08 .28 -.09 -.02 1 .60** .14 .59** -.01 -.01 1
58 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 2 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 1 .86** 1 .71** .50* 1 .46* .41 .65** 1 .26 .39 .29 .28 .46* .75** .37 .38 1 .86** 1 .74** .69* 1 .15 .28 .24 1 .74** .76** .76** .05 1 .55* .54* .74** .20 .71** 1 1 .55** 1 .66** .22 1 .23 .03 .27 1 .45* .13 .43* .12 1 .75** .44* .59** .35 .60** 1 1 .32 1 -.03 -.11 1 .75** .25 .08 1
59 5. Shifting responsibility 6. Defensive Excuse 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 5 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive 1 .53* 1 .53* .23 1 .05 -.32 .19 1 .17 .30 .16 .07 1 .52* .63** .70** .05 .22 1 1 .77** 1 .41 .59** 1 .49* .53* .34 1 .51* .34 .43* .14 1 .57** .47* -.10 .60** .11 1 1 .59** 1 .71** .50* 1 .46* .41 .65** 1 .26 .39 .29 .28 1 .46* .75** .37 .38 .14 1 1 .14 1
60 Refusal 2 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 5 (n = 22) 1. Self-protective 1 .58** .56** -.02 .80** .84** 1 .42 1 .50* .48* 1 .42* .21 .60** 1 .68** .41 .78** .86** 1 .56** .44* .90** .53* .70** 1 1 .83** 1 .76** .76** 1 .62** .49* .74** 1 .69** .55** .79** .87** 1 .71** .56** .86** .90** .94** 1 1 .39 1 .82** .23 1 .78** .15 .94** 1 .88** .37 .77** .78** 1 .51* .65** .53* .50* .42 1
61 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive * p < .05; ** p < .01. Note. Each number refers to a single item that measures the defensiveness level of the accounts. See Appendix D for the complete list of accounts and labels of accounts presented to participants. 1 .50* 1 -.20 .21 1 .61** .79** .08 1 .53* .59** -.01 .89** 1
62 Table 4 The Cronbach’s Alpha for the Sum of Items that Measured Defensiveness Level of Each Account (N = 22) Defensiveness Concession 1 Concession 2 Concession 3 Concession 4 Concession 5 Justification 1 Justification 2 Justification 3 Justification 4 Justification 5 Excuse 1 Excuse 2 Excuse 3 Excuse 4 Excuse 5 Refusal 1 Refusal 2 Refusal 3 Refusal 4 Cronbach’s ? .82 .87 .70 .81 .80 .68 .76 .82 .86 .77 .81 .84 .88 .83 .78 .77 .79 .92 .81
63 Refusal 5 .85
Note. See Appendix D for the complete list of the accounts and labels of the accounts presented to participants.
type of accounts was the between-subject variable and the within-subject variable was the defensiveness level of the accounts. A significant within-subject effect was found, F(1, 21) = 30.14, p < .01, ?² = .59. In addition, the results indicated a significant linear and a quadratic effect of the defensiveness level of the accounts, F(1, 21) = 33.42, p < .01, ?² = .61 for the linear effect and F(1, 21) = 6.99, p < .05, ?² = .25 for the quadratic effect (Figure 3 pictures the defensiveness level of four types of accounts). Consistent with Schönbach’s (1990) argument, concessions had the lowest defensiveness level (M = 4.58, SD = 1.77) and refusals’ defensiveness level was highest (M = 8.29, SD = 2.32). On the defensiveness continuum proposed by Schönbach, justifications and excuses had moderate levels of defensiveness and excuses had a higher defensiveness level than justifications. The results indicated otherwise: Justifications’ defensiveness level (M = 7.63, SD = 1.95) was higher than the defensiveness level of excuses (M = 6.22, SD = 1.47). The difference was significant, t(21) = 4.53, p < .01. As discussed in Pilot Study 4, the defensiveness level of the accounts would be used with the likelihood of selecting each account in the formal study to constitute the composite measure (LDA) in the formal study. The average defensiveness level of each type of accounts was used as the Di in the formula to calculate LDA (LDA = ? [Di • Li] / ? Li). Specifically, Di for concession, justification, excuse, and refusals was 4.58, 7.63, 6.22, and 8.29, respectively.
64
9
8
Defensiveness Level
7
6
5
4 Concession Justification Excuse Refusal
Accounts
Figure 3. Defensiveness level of concessions, justifications, excuses, and refusals.
65 Pilot Study 6: Pilot Testing Dependent Measures Purpose. Pilot Study 6 was conducted to test the measures of all the dependent variables. Participants. Sixty-three undergraduate students at the University of Maryland participated in this pilot study. The age range of participants was 18 to 22 (M = 19.40, SD = 1.06, Mdn = 19.00). Sixty percent of participants were female (n = 38). And 60% of participants identified themselves as Caucasian (n = 38). Procedure. When participants arrived, the researcher explained the purpose of the study and asked them to sign the study’s consent form before they completed the questionnaire. The questionnaire had three parts: the stimulus message, all the dependent measures, and open-ended questions about the clarity of the dependent measures. Only two versions of the hypothetical conflict scenario were used in this pilot study: the scenario that induces a high outcome severity and a high level of relational closeness and the scenario that induces a low outcome severity and a low level of relational closeness. After reading the scenario, two practice questions were provided to participants to learn how to use magnitude scales. These two questions were the same ones used in Pilot Studies 3 and 5. After participants felt ready to use magnitude scales, they were asked to continue with the second part of the questionnaire: questions for the degree of internal attribution (five indicators), the degree of external attribution (five indicators), controllability (five indicators), uncontrollability (five indicators), expected responsibility (five indicators), expected anger (six indicators), likelihood of using concessions (five indicators), likelihood of using justifications (five indicators), likelihood of using excuses (five indicators), and likelihood of using refusals (five indicators). A total of fifty-one
66 indicators were used. All the measures used magnitude scales. The third part of the questionnaire listed open-ended questions regarding participants’ opinions about the dependent measures. For example, participants were asked to explain any confusion in the questionnaire. Appendix K listed all questions regarding the clarity of the questions used in the Pilot Study 6. Data transformation. Participants used nonnegative numbers to answer the questions regarding the hypothetical conflict. The data were submitted to descriptive analysis to see whether a transformation was needed. Following the (-1.96, 1.96) cutoff rule, all fifty-one indicators were significantly positively skewed. A transformation was necessary. The data were trimmed at the 90th percentile to eliminate the extreme values (e.g., the 90th percentile for one item to measure defensiveness was 400; thus, any score larger than 400 was converted to 400). Power transformations were used with ? ranges from 0.30 to 0.50 for different variables. Instruments and results. The pilot study measured the following variables: the degree of internal attribution, the degree of external attribution, controllability, uncontrollability, expected responsibility, expected anger, and likelihood of selecting each account (four types of accounts were used and each type had five accounts; in total twenty accounts were listed in the questionnaire). Degree of internal attribution. To measure the degree of internal attribution, a 5item magnitude scale was used (e.g., “How likely is it that your mistake was due to your own carelessness?”). A CFA was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results supported a one-factor structure although the fit indices were not satisfactory, ?²(5, N = 63) = 10.24, p > .05; NFI = 0.87; CFI = 0.93; SRMR = 0.08. A
67 principal components analysis without rotation was conducted. The results indicated two components with eigenvalues larger than 1. The first component had an eigenvalue of 2.27 and explained 45% of the variance. The second component’s eigenvalue equaled 1.05 and 21% of the variance was explained by the second component. The Cronbach’s alpha for the sum of the five items was .68. The degree of external attributions was measured by a 5-item magnitude scale (e.g., one item was, “How likely is it that your mistake was due to the professor not giving clear instructions for the assignment?”). CFA results did not support a one-factor structure, ?²(5, N = 63) = 14.22, p < .05; NFI = 0.89; CFI = 0.92; SRMR = 0.07. However, a principal components analysis yielded only one component with an eigenvalue greater than 1 (eigenvalue = 2.79). Fifty-five percent of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .79. A 5-item magnitude scale was used to measure controllability (e.g., one item was, “how likely is it that you could have prevented this mistake?”). CFA results supported a one-factor structure, ?²(5, N = 63) = 9.89, p > .05; NFI = 0.96; CFI = 0.98; SRMR = 0.03. A principal components analysis without rotation yielded two components with eigenvalues larger than 1. The eigenvalue of the first component was 2.96, explaining 59% of the variance and the second component’s eigenvalue equaled 1.00, explaining 20% of the variance. The Cronbach’s alpha for the sum of the five items was .89. Uncontrollability was measured by a 5-item magnitude scale (e.g., one item was, “How likely is that what happened was out of your control?”). CFA results indicated a one-factor structure, ?²(5, N = 63) = 7.93, p > .05; NFI = 0.95; CFI = 0.98; SRMR = 0.05. A principal components analysis with no rotation indicated that only one component had
68 an eigenvalue larger than 1 (eigenvalue = 2.92) and 58% of the variance explained by this component. The Cronbach’s alpha for the sum of the five items was .83. Anticipated consequences. To measure participants’ expected responsibility, a 5item magnitude scale was used (e.g., one item was, “how much responsibility do you have if Person B did not achieve his or her goal of being awarded the fellowship?”). CFA results did not support a one-factor structure, ?²(5, N = 63) = 38.67, p < .05; NFI = 0.66; CFI = 0.67; SRMR = 0.13. A principal components analysis without rotation found two components with eigenvalue larger than 1. The first component had an eigenvalue of 2.91 and explained 58% of the variance, and the second component had an eigenvalue of 1.32. Twenty-six percent of the variance was explained by the second component. The Cronbach’s alpha for the sum of the five items was .82. A 6-item magnitude scale was used to measure participants’ expected anger (e.g., “How mad do you expect Person B to be?”). CFA results did not support a one-factor structure, ?²(9, N = 63) = 45.64, p < .05; NFI = 0.90; CFI = 0.91; SRMR = 0.04. However, a principal components analysis without rotation indicated that only one component had an eigenvalue larger than 1. The first component had an eigenvalue of 5.12, explaining 85% of the variance. The Cronbach’s alpha for the sum of the five items was .97. Likelihood of selecting each account. Five accounts of concessions (developed in Pilot Study 4) were provided to participants. Participants were asked to rate the likelihood of using each concession when explaining their behavior in the conflict scenario. A magnitude scale was used. CFA results did not support a one-factor structure, ?²(5, N = 63) = 14.59, p < .05; NFI = 0.91; CFI = 0.94; SRMR = 0.07; however, the principal components analysis without rotation found only one component with an eigenvalue
69 greater than 1 (eigenvalue = 3.00), and 60% of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .81. Participants were asked to rate the likelihood of using each of the five justifications when explaining their behavior in the conflict scenario by using a magnitude scale. A one-factor structure was not supported by CFA results, ?²(5, N = 63) = 27.57, p < .05; NFI = 0.88; CFI = 0.89; SRMR = 0.08. However, the principal components analysis without rotation found only one component with an eigenvalue larger than 1 (eigenvalue = 3.35) and 67% of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .87. Participants also rated the likelihood of using each of the five excuses by using a magnitude scale. CFA results supported a one-factor structure marginally, ?²(5, N = 63) = 11.04, p = .05; NFI = 1.00; CFI = 1.00; SRMR = 0.06. A principal components analysis with no rotation found only one component with an eigenvalue larger than 1 (eigenvalue = 2.80), explaining 56% of the variance. The Cronbach’s alpha for the sum of the five items was .78. Five refusals were presented to participants to indicate their likelihood of using each refusal in the conflict situation. CFA results did not support a one-factor structure, ?²(5, N = 63) = 27.21, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.12. However, a principal components analysis without rotation found only one component with an eigenvalue larger than 1. The eigenvalue of the first component equaled 2.89, and 58% of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .79. Revisions of the measures. Participants answered six open-ended questions
70 regarding the clarity of the questionnaire (see Appendix K). For example, participants were asked "What do you think the researchers want to measure?" Other questions examined the clarity and difficult levels of the questions. The researcher also interviewed sixteen participants about their perceptions of the items on the questionnaire. The majority of participants were able to identify three or four variables the researchers intended to measure, and most questions were perceived to be clearly stated and easy to answer. Revisions were made based on participants’ comments. For example, to measure expected responsibility of the offender, one question was revised from "How much responsibility do you have for what has happened" to "How much responsibility would Person B assign to you for what has happened?" The Cronbach’s alphas for all dependent measures were acceptable, ranging from .68 to .97. The degree of internal attribution had the lowest reliability coefficient (? = .68). A principal components analysis with no rotation indicated two components with eigenvalue larger than 1 for the degree of internal attribution. Two items (i.e., item 4: "How likely is it that your mistake was because you did not care about what grade you and Person B could get for this assignment?" and item 5: "To what extent do you believe what has happened was because you did not spend enough time on the assignment?") loaded highly on both the first and second components (.54 and .63 for item 4 and .50 and .47 for item 5). When eliminating both items, the Cronbach’s alpha for the sum of the remaining three items increased to .73; however, if these items were eliminated, the measurement model would be saturated and therefore fit perfectly, which is problematic. To ensure the measurement model for the degree of internal attribution was overidentified, three new items, phrased similarly as the remaining three items, were added to
71 the scale, and all six items were used in the formal study. For the scale of expected responsibility, two items loaded high on both first and second components (.63 and .71 for item 4, and .70 and .64 for item 5). These two items were "How much responsibility can be assigned to others (e.g., the professor or Person B) if Person B did not achieve his or her goal?" (item 4), and "How accountable were others (e.g., the professor or Person B) if Person B did not achieve his or her goal?" (item 5). After eliminating the two items from the scale, the Cronbach’s alpha for the sum of the remaining three items increased to .87. The measurement model for the expected responsibility with the remaining three items was saturated. Three new items were added to the scale and all six items were used in the formal study. CFA results indicated that the scales for the degree of external attribution, likelihood of using concessions, likelihood of using justifications, and likelihood of using refusals were not unidimensional. However, the principal components analyses found only one component with an eigenvalue greater than 1 for all four scales. The scales measuring the likelihood of using concessions, justifications, and refusals remained the same. However, based on participants’ comments, two more items were added to the scale for degree of external attribution. Based on the results of Pilot Study 3 and Pilot Study 6, in the formal study outcome severity and relational closeness manipulation checks were each measured by 6 items. The degree of internal attribution was measured by a 7-item magnitude scale, as was the degree of external attribution. A 5-item scale was used to measure controllability. Uncontrollability also was measured by a 5-item scale. Expected responsibility and expected anger were each measured by a 6-item scale. And five accounts were listed
72 under each type of accounts (i.e., concession, justification, excuse, and refusal). In total, 20 accounts were provided to participants to investigate their likelihood of using those accounts. The Formal Study Data collection for the formal study was conducted at the University of Maryland between May 6th and June 18th, 2008. This section presents the analytical strategy, the determination of the sample size and the sample description, the variables of interest, the experimental design, and the data collection procedure. Analytical Strategy The hypotheses and the proposed model were tested with analyses of variance and with structural equation modeling. Analyses of variance (ANOVAs). ANOVAs were used mainly to examine the success of the manipulations. In particular, the statistical difference between two levels of outcome severity (high vs. low) and the statistical difference between three levels of relational closeness (high vs. moderate vs. low) were tested. Structural equation modeling (SEM). SEM was used to test the causal relationships proposed by the hypotheses (see Figure 2). Severity of outcome (?1; high outcome severity was coded as 1 and low outcome severity was coded as -1), relational closeness (?2; quadratic coding: high relational closeness was coded as -1, moderate relational closeness as 2, and low relational closeness as -1), and the interactions (?3, ?4, and ?5) between outcome severity and relational closeness were the independent variables; manipulation checks for outcome severity (?1) and relational closeness (?2) were mediating variables; and the degree of internal attribution (?3), controllability (?4),
73 the degree of external attributions (?5), uncontrollability (?6), expected responsibility (?7), expected anger (?8), and LDA (?9) were outcome variables. See Figure 2 for labels and coding of each variable used in model testing. Two stages were used to test the structural equation model. First, confirmation factor analyses were conducted on each variable to determine the underlying dimensions. Second, the theoretical and the measurement model were tested together using LISREL 8.70 (Jöreskog & Sörbom, 2004). The influence of the outcome severity manipulation on expected responsibility and LDA was mediated by the outcome severity manipulation check (i.e., perceived outcome severity). The relational closeness manipulation (quadratic coding) and relational closeness manipulation check (i.e., perceived quadratic relational closeness) were used to test the curvilinear relationship between relational closeness and offenders’ choice of accounts. A direct influence of the interaction between outcome severity (linear coding) and relational closeness (linear coding) on LDA was proposed by Hypothesis 3. According to Hypothesis 4, the interaction between outcome severity and relational closeness also indirectly influences the defensiveness of the accounts through the attributional process variables (e.g., degree of internal and external attribution). Both direct and indirect paths were tested using LISREL8.70 (Jöreskog & Sörbom, 2004). Confirmatory factor analyses (CFAs) were performed using LISREL 8.70 (Jöreskog & Sörbom, 2004) to test measurement model fit. The goodness-of-fit criteria was based mainly on Hu and Bentler’s (1998, 1999) recommendations. Hu and Bentler (1999) found variation of performance among the fit indices. They suggested that to distinguish good models from poor ones, a two-index presentation strategy needs to be used that the maximum likelihood (ML) based SRMR is supplemented with another ML-
74 based fit index, such as TLI, RNI, CFI, Gamma Hat, Mc, or RMSEA. Hu and Bentler (1998) specified that a cutoff value close to .95 for CFI, a cutoff value close to .90 for Mc, a cutoff value close to .08 for SRMR, and a cutoff value close to .06 for RMSEA could “result in lower Type II error rates (with acceptable costs of Type-I error rates)” (p. 27). Hu and Bentler’s criteria have been largely accepted (e.g., Hancock, 2006). The above goodness-of-fit criteria also were used to show how well the data fit the proposed model. Testing total effects of the independent variables on the outcome variables indicated whether the hypotheses were supported. Significant coefficients from the independent variables to the outcome variables indicated the extent to which the independent variables influenced the dependent variables. Design A 2 (Outcome severity: severe vs. not severe) X 3 (Relational closeness: high vs. moderate vs. low) independent groups experimental design was employed. A scenario describing an interpersonal conflict in a course project at school was used to induce different levels of outcome severity and relational closeness. All participants were randomly assigned to one of six conditions formed by severity of an outcome and closeness of relationship. The cognitive (the degree of internal attribution, the degree of external attribution, controllability, uncontrollability, and expected responsibility) and emotional (expected anger) processes participants engaged in and the likelihood of selection of different accounts (concessions, justifications, excuses, and refusals) were measured. Sample Proposed sample size. Structural equation modeling (SEM) was used in data
75 analysis. Different researchers recommend different ways to determine sample size (e.g., Bentler & Chou, 1987; Jackson, 2003). The current research adopted the recommendation from Gagné and Hancock (2006). Gagné and Hancock took into consideration the ratio of the number of indicators per factor (p/f) and their loading magnitudes when deciding whether a sample size is acceptable. The ratio of the number of indicators per factor for the current research was 3.86. Based on Gagné and Hancock’s guideline, an acceptable estimation of sample size for the current research was between 200 and 400. Sample. Participants were 238 students; they were undergraduates (32.8% freshmen, 25.6% sophomores, 28.2% juniors, 13.0% seniors, and 0.4% others) recruited from the University of Maryland. Their age range was 18 to 37 (M = 19.76, SD = 1.78, Mdn = 20.00). One hundred and twenty-nine of the participants were female (54.2%). The majority were Caucasian (58.8%), 17.6% were Asian Americans, 13.4% were African American, and 4.6% were Hispanic. Approximately 5.5% of the participants checked the category “other.” Participants received extra credit in exchange of their participation in the study. Procedure Participants arrived at a pre-assigned classroom, in a group of five or six. The researcher informed the participants the purpose and the procedure of the study and asked them to sign a consent form before they participated in the study. All people coming to the study agreed to participate. After the signed consent forms were collected, a package including a single conflict scenario and a questionnaire was distributed. Participants were randomly assigned to receive one of the six versions of the hypothetical conflict scenario (as mentioned before, Appendix H lists all six versions of the hypothetical conflict
76 scenario). Participants were told to imagine that the hypothetical situation happened between them and another person (i.e., “Person B”). The scenario described a class project between participants (“you”) and Person B. The participants’ behavior in the scenario (i.e., writing his or her portion of a review paper based on the wrong readings) resulted in a poor grade for both the offender (i.e., the role the participant took) and the victim (i.e., Person B in the scenario), which, they were told, influenced the awarding of Person B’s scholarship for the next year. To induce relational closeness, participants and Person B’s relationship was described as follow: (a) They grew up together and have been best friends for many years (high level of closeness), (b) they were in a couple of classes together before, and they know each other but only see each other in class (moderate level of closeness), or (3) they don’t really know each other and they happened to sit together on the first day of class (low level of closeness). Outcome severity was manipulated in several ways. First, in the high outcomeseverity condition, the class project (i.e., writing a review paper) constituted a large portion of the class grade (i.e., 40%); the project was described as only counting for 10% of the grade in the low outcome-severity condition. Second, the influence of the class grade on Person B’s scholarship was different. In high outcome-severity condition, participants were told that “if the grade for this course was not good, his or her (i.e., Person B) scholarship for the next year would definitely be in jeopardy.” In the low outcome-severity condition, Person B told participants that “if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be
77 influenced.” Third, the consequence of the conflict was different. In the high outcomeseverity condition, after obtaining a poor grade for the class project, “Person B was very upset because such a low grade resulted in him or her not getting next year's scholarship,” compared to “Person B was not happy but he or she did great on other assignments and believed it would not influence his or her getting next year's scholarship” (low outcomeseverity). Each participant read one version of the hypothetical conflict scenario. Participants were asked to complete a questionnaire by taking the role of the offender. The questionnaire had the manipulation checks (i.e., for perceived outcome severity and perceived relational closeness), measures of attributions (i.e., the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability), measures of anticipated consequences (i.e., expected responsibility and expected anger), and measures of accounts (i.e., likelihood of selecting each account). Participants’ demographic information was collected at the end of the questionnaire. After completing the questionnaire, participants were debriefed in detail.
78 CHAPTER IV Results This chapter consists of two parts. The first part presents data preparation for the primary analyses, reliability and confirmatory factor analyses for measurement models, and manipulation checks. The second part presents the primary data analyses and tests of hypotheses. Preliminary Analyses Data Entry Checks Frequencies of all the variables were examined to minimize data entry error. All frequencies fell in the preset range defined by the lower boundary and the upper boundary for each item. Furthermore, all the responses in the questionnaires were checked against the data entries. Out of 238 participants’ data (each participant answered 72 question including demographic information), 6 errors were found, yielding an error rate of 6 ÷ (238 x 72) = .0004. The six incorrect inputs were corrected. Missing Data The completed questionnaires were numbered in the order of when they were turned in. Two participants had incomplete data. The missing data in both questionnaires did not show a systematic pattern. Kline (2005) suggested that non-systematic missing data can be ignored. He further argued that in statistical analyses such as principal components analyses and analyses of variance, pairwise deletion should be used to maintain the statistical power (Kline, 2005). However, pairwise deletion may pose a problem for confirmatory factor analyses and SEM; Kline suggested that when the number of cases with missing data is small, listwise deletion is preferable to pairwise
79 deletion for CFA and SEM. Thus, pairwise deletion was used when ANOVA was conducted for the manipulation checks and listwise deletion was used when CFA and SEM were conducted for testing the proposed model. Data Transformation Magnitude scales were used on the questionnaire for all dependent measures and manipulation checks. Participants used nonnegative numbers to answer the questions. Based on the results of the six pilot studies, perceived outcome severity, perceived relational closeness, expected responsibility, and expected anger each had six indicators; the degree of internal attribution and the degree of external attribution each had seven indicators; controllability and uncontrollability each had five indicators; and likelihood of selecting an account had five indicators for each of four types of accounts (20 items in total). Therefore, the total number of indicators for these measures was sixty-eight. The sixty-eight indicators were submitted to a descriptive analysis to see whether any data transformation would be needed. The skewness and kurtosis were examined. The ratio of skewness to its standard error was calculated, and an indicator’s distribution was considered approaching normality when the ratio ranged from -1.96 to 1.96 (see Frey et al., 2000). Following the |1.96| cutoff rule, all sixty-eight indicators were positively skewed. A data transformation was necessary. Before transformation, to eliminate extreme values, all data were trimmed at the 90th percentile (e.g., if 500 was the 90th percentile for an item, any score larger than 500 was converted to 500). Power transformation was used to treat the skewed data resulting from magnitude scales (Bauer & Fink, 1983): New indicator = (original indicator) ?,
80 the researcher tried different values of ? and used the ones that ensured the majority of indicators had non-significant statistics in skewness. The values of ? ranged from 0.28 to 0.55 for the formal study data. Transformed data were used in confirmatory factor analyses, principal components analyses, ANOVAs for the manipulation checks, and model testing. Appendix L provides descriptive statistics for the indicators used in the formal study after transformation. Instruments As mentioned above, Appendices A to D listed all the questions that measured the following variables. Manipulation checks. Manipulation checks involved two variables: perceived outcome severity and perceived relational closeness. Perceived outcome severity was measured by a 6-item magnitude scale, where 0 indicated not severe at all and 100 indicated a moderate level of severity. An example question is, “How severely does your mistake affect Person B?” A CFA was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results did not support a one-factor structure, ?²(9, N = 236) = 71.16, p < .05; NFI = 0.96; CFI = 0.97; SRMR = 0.03. When a principal components analysis without rotation was conducted, the results indicated only one component had an eigenvalue larger than 1 (eigenvalue = 4.86), explaining 81% of the variance. The Cronbach’s alpha for the sum of the six items was .95. The first component was used as the manipulation check. For model testing, all six items after transformation were used in testing the proposed model. Perceived relational closeness was also measured by a 6-item magnitude scale, where 0 indicated not close at all and 100 indicated a moderate level of relational
81 closeness. An example item was, “How familiar are you with Person B?” CFA results did not support a one-factor structure, ?²(9, N = 236) = 88.63, p < .05; NFI = 0.95; CFI = 0.95; SRMR = 0.03. A principal components analysis without rotation was conducted to examine the underlying components. The results indicated that only one component had an eigenvalue larger than 1 (eigenvalue = 5.49), explaining 92% of the variance. The Cronbach’s alpha for the sum of the six items was .98. The first component was obtained for two purposes. First, the first component score of the perceived relational closeness was used in ANOVA for assessing whether the manipulation of relational closeness (linear coding) was successful. Second, the first component score was used to create a manipulation check for quadratic relational closeness. A curvilinear relationship between relational closeness and selection of accounts was predicted in Hypothesis 2. To test this curvilinear relationship in the model, a manipulation check for the quadratic relational closeness needed to be included to mediate the relationship between relational closeness (quadratic coding) and selection of accounts. The following formula was used to create the perceived quadratic relational closeness: Perceived quadratic relational closeness = (first component score – mean)², where the mean for the first component score equals zero. Thus, the perceived quadratic relational closeness simply was created by squaring the first component score of the six items that measured the perceived relational closeness. In the model testing, only one indicator (i.e., perceived quadratic relational closeness) was included to represent this variable. Attributions. The current research did not assume the relationships between the degree of internal attribution and the degree of external attribution and between
82 controllability and uncontrollability. Four types of attributions were measured separately; they are the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability. Different reasons regarding the offender’s behavior were listed and participants indicated their perception of how likely each reason influenced his or her behavior, using magnitude scales where 0 = not likely at all and 100 indicated a moderate level of likelihood. The degree of internal attribution was measured by a 7-item magnitude scale. For example, participants were asked, “How likely is that what has happened was due to your own carelessness?” CFA results did not support a one-factor structure, ?²(14, N = 236) = 123.79, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.11. A principal components analysis without rotation was conducted to examine the underlying components. The results indicated two components had eigenvalues larger than 1. The first component’s eigenvalue was 2.74, explaining 39% of the variance. The second component’s eigenvalue equaled 1.37, explaining 20% of the variance. Item 2 (i.e., “How likely is that what has happened was due to internal traits, such as your personality?”), item 4 (i.e., “How likely is that what has happened was because you did not care about what grade you and Person B could get for this assignment?”), and item 7 (i.e., “To what extent do you think your personality determines what has happened?”) loaded high on both first and second components. After deleting these three items, a CFA with the remaining four items had good fit indices but the chi-square value was significant, ?²(2, N = 236) = 7.80, p < .05; NFI = 0.97; CFI = 0.98; SRMR = 0.04. A principal components analysis with no rotation was conducted on the remaining four items, indicating only one component had an eigenvalue larger than 1 (eigenvalue = 2.26) and explaining 56% of the variance. The
83 Cronbach’s alpha for the sum of these four items was .73. These remaining four items were used for the model testing. The first component score for the remaining four items was saved to examine the relationship between the degree of internal attribution and the degree of external attribution. The degree of external attribution was measured by a 7-item magnitude scale. An example question was, “How likely is that what has happened was due to the professor not giving clear instructions for the assignment?” CFA results suggested a multidimensional factor structure, ?²(14, N = 236) = 58.17, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.06. A principal components analysis without rotation found that only one component had an eigenvalue larger than 1 (eigenvalue = 2.90), explaining 41% of the variance. The Cronbach’s alpha for the sum of the seven items was .75. All seven items were used in the model testing. The first component score for all seven items was saved to investigate the relationship between the degree of internal attribution and the degree of external attribution. The correlation between the degree of internal attribution and the degree of external attribution was significant, r = .33, p < .01. Controllability was measured by a 5-item magnitude scale. For example, participants were asked to indicate “How much control do you have over what has happened?” Confirmatory factor analysis was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). Results did not indicate a unidimensional factor structure, ?²(5, N = 236) = 27.41, p < .05; NFI = 0.96; CFI = 0.97; SRMR = 0.04. However, a principal components analysis without rotation found only one component with an eigenvalue larger than 1. The first component’s eigenvalue equaled 3.27, explaining 65% of the variance. All five items were used in the model testing. The Cronbach’s alpha for the sum
84 of the five items was .85. The first component score for the five items was saved to test the relationship between controllability and uncontrollability. A 5-item magnitude scale was used to measure uncontrollability. An example question was, “To what extent do you believe there is not much you could do to prevent what has happened?” The CFA did not support a one-factor structure, ?²(5, N = 236) = 28.72, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.05. The principal components analysis with no rotation found only one component that had an eigenvalue larger than 1. The first component’s eigenvalue equaled 3.12, explaining 62% of the variance. All five items were kept for the model testing. The Cronbach’s alpha for the sum of the five items was .85. The first component score was used to test the relationship between controllability and uncontrollability. The correlation was significant, r = -.13, p < .05. Anticipated consequences. Anticipated consequences were represented by two variables: expected responsibility assigned to the offender and expected anger felt by victims. For expected responsibility, participants were asked to indicate their expectation about how much responsibility would be assigned to them. A 6-item magnitude scale was used with 0 indicated not responsible at all and 100 indicated a moderate level of expected responsibility. For example, one item is, “How much responsibility would Person B assign to you for what has happened?” CFA results did not support a one-factor structure, ?²(9, N = 236) = 34.60, p < .05; NFI = 0.99; CFI = 0.99; SRMR = 0.02. A principal components analysis with no rotation found only one component with an eigenvalue larger than 1 (eigenvalue = 5.24), explaining 87% of the variance. All six items were used in the model testing. The Cronbach’s alpha for the sum of the six items was .97.
85 For expected anger, participants indicated the expected anger felt by victims by a 6-item magnitude scale. For example, an item is, “How annoyed do you expect Person B to be?” One-factor structure was not supported by CFA results, ?²(9, N = 236) = 196.93, p < .05; NFI = 0.91; CFI = 0.91; SRMR = 0.04. However, only one component with an eigenvalue larger than 1 was found in a principal components analysis with no rotation. The eigenvalue of the first component was 5.17, explaining 86% of the variance. All six items were used for the model testing. The Cronbach’s alpha for the sum of the six items was .97. Likelihood of selecting each account. A list of 20 accounts (5 statements for each type of accounts) was provided to participants to indicate the likelihood of selecting each account by using a magnitude scale, where 0 indicated not likely at all and 100 indicated a moderate level of likelihood. For example, a concession stated, “I am sorry that I misunderstood the instruction. Is there any way that I can make it up to you?” An example of a justification was like the following, “I am sorry that the grade influenced your scholarship. But I worked very hard on this assignment. We should have gotten together to edit the paper earlier.” An example excuse was, “I am sorry that I misunderstood the instruction. But the wording in the direction was really confusing and that may be the reason why I wrote the wrong portion of the paper.” An example refusal was, “I know the readings I read were the right material for the paper. I did not find anything wrong to what I contributed to the paper.” Participants were asked to rate their likelihood of using each account. To convert the results into a data set for likelihood (i.e., with a range from 0% to 100%), the original data were first trimmed at the 90th percentile and a power transformation was conducted,
86 with ? = 0.45. After transformation, the sum of all transformed likelihood data were calculated (named the Ltotal in the following formula) and used as the base for the likelihood of each account. The likelihood of each account was calculated by dividing the transformed data by Ltotal. For example, the likelihood of selecting a concession (named the Lc in the following formula) was calculated as: Lci = Ci / Ltotal, where C refers to the likelihood of selecting a concession and i ranges from 1 to 5. When the likelihood of selecting the five concessions was obtained (i.e., Lc1, Lc2, Lc3, Lc4, and Lc5), the average of Lc1 to Lc5 was calculated, resulting the likelihood of selecting concessions (Lc). The likelihood of selecting justifications (Lj), excuses (Le) and refusals (Lr) were calculated the same way. As discussed above, participants indicated how likely it was that they would select each account. However, the dissertation’s hypotheses refer to both likelihood of selecting each account and the defensiveness level of each account. Only using the data collected in the formal study would not be sufficient and not appropriate to test the hypotheses. To create an appropriate measure for accounts used in the hypotheses testing, the defensiveness level of each type of accounts resulted from Pilot Study 5 was combined with the likelihood of selecting a defensive account (i.e., LDA) resulted from the formal study. The formula to calculate LDA is as follows: LDA = ? (Di • Li) / ? Li, where Di refers to the defensiveness level of each type of accounts and Li refers to the likelihood of using each type of accounts (i.e., Lc, Lj, Le, and Lr). Based on the results from Pilot Study 5, Di for concessions, justifications, excuses, and refusals were 4.58,
87 7.63, 6.22, and 8.29, respectively. Therefore, LDA = (4.58 • Lc + 7.63 • Lj + 6.22 • Le + 8.29 • Lr) / (Lc + Lj + Le + Lr), the LDA value indicates the likelihood of selecting a defensive account. The higher the value of LDA, the more likely a more defensive account will be selected. Overall, for manipulation checks, the first components of perceived outcome severity and perceived relational closeness were submitted to ANOVA. For hypothesis and model testing, perceived outcome severity had six indicators. Perceived quadratic relational closeness had one indicator. Four indicators were used for the degree of internal attribution where the degree of external attribution had seven indicators. Controllability and uncontrollability each had five indicators. Expected responsibility assigned to the offender had six indicators and expected anger had six indicators. The composite measure of accounts (LDA) had one indicator. The model had five exogenous variables and nine endogenous variables, with forty-six indicators in all. Table 5 summarizes the results of principal components analyses of all instruments except LDA. Table 6 summarizes the Cronbach’s alpha for all instruments except LDA. Table 7 through Table 9 present each indicator’s factor loading from CFA and the measurement model fit indices. Loadings and fit indices for perceived outcome severity and perceived relational closeness are included in Table 7, loadings and fit indices for attributions are included in Table 8, and loadings and fit indices for expected responsibility and expected anger are presented in Table 9.
88 Table 5 Eigenvalues of the First Principal Components of Endogenous Variables with the Proportions of the Variance Explaineda Variable Perceived Outcome Severity Perceived Relational Closenessb Degree of Internal Attributionc Degree of External Attribution Controllability Uncontrollability Expected Responsibility Expected Anger Eigenvalue 4.86 5.49 2.26 2.90 3.27 3.12 5.24 5.17 % Variance Explained 81.04% 91.54% 56.49% 41.38% 65.43% 62.49% 87.25% 86.23%
Note. a. The table listed all endogenous variables except LDA (i.e., the likelihood of selecting a defensive account) because LDA only had one indicator and no principal components analysis was conducted on a single-indicator variable. b. The principal components analysis results reported here are based on the six items that measured the perceived relational closeness for the linear relation manipulation check. c. Degree of internal attribution was measured originally by a 7-item magnitude scale. The principal components analyses indicated three items loaded highly on both the first and second components (the eigenvalues for both components were larger than 1). These three items were deleted from the scale. The statistics reported in this table represent the results after deletion.
89 Table 6 The Cronbach’s Alpha for the Sum of Measurement Items for All Endogenous Variables in the Formal Studya Variables Perceived outcome Severity Perceived Relational Closenessb Degree of Internal Attributionc Degree of External Attribution Controllability Uncontrollability Expected Responsibility Expected Anger Likelihood of Concessions Likelihood of Justifications Likelihood of Excuses Likelihood of Refusals Valid N 237 237 237 237 237 237 238 238 237 238 238 238 Cronbach’s ? .95 .98 .73 .75 .85 .85 .97 .97 .80 .90 .78 .87 Number of items 6 6 4 7 5 5 6 6 5 5 5 5
Note. The indicators for each variable are found in Appendices A to D. a. Although model testing used LDA for the choice of accounts, the formal study used four scales to measure the likelihood of selecting each type of accounts. Therefore, the table also listed the reliability coefficients for these four measurement scales. b. The reliability results reported here was based on the six items that measured the perceived relational closeness for the linear manipulation check. c. Degree of internal attribution was measured originally by a 7-item magnitude
90 scale. The principal components analyses indicated three items loaded highly on both the first and second components (the eigenvalues for both components were larger than 1). The three items were deleted from the scale. The statistics reported in this table represent the results after deletion.
91 Table 7 Manipulation Checks, Indicator Loadings, and Measurement Model Indices Manipulation Checks and Indicators Unstandardized Loadings (Standardized) Outcome Severity To what extent does your mistake influence Person B? How severe is the outcome of this situation for Person B? How severely does your mistake affect Person B? How much influence does your mistake have on Person B’s scholarship for the next year? How much difficulty does your mistake have on Person B’s obtaining his or her scholarship for the next year? To what extent do you think the outcome is severe? 1.56 (.60)** 1.67 (.61)** 1.00 (.73)** 1.78 (.55)** 1.57 (.56)** 2.04 (.61)**
?² (9, N = 236) = 71.16, p < .05, RMSEA = .17, SRMR = .03, CFI = .97. Relational Closenessa To what extent do you think Person B is a good friend of yours? 1.00 (.58)**
To what extent do you think you and Person B are close to each other? 1.08 (.56)** How much knowledge do you have of Person B? How familiar are you with Person B? How close are you with Person B? How well do you know Person B? 0.95 (.55)** 0.94 (.55)** 1.07 (.54)** 1.01 (.52)**
?² (9, N = 236) = 88.63, p < .05, RMSEA = .19, SRMR = .03, CFI = .95. * p < .05; ** p < .01.
92 Note. Bold values represent fixed unstandardized loadings for reference indicators. a. The confirmatory factor analysis results reported here are based on the six items that measured the perceived relational closeness for the linear manipulation check.
93 Table 8 Attributional Processes Variables, Indicator Loadings, and Measurement Model Indices Attributional Processes and Indicators Unstandardized Loadings (Standardized) Degree of Internal Attributionsa How likely…due to your own carelessness? How likely...due to you not paying enough attention? How likely…because you did not spend enough time? To what extent…because of you? 1.00 (.74)** 1.06 (.96)** 0.24 (.43)** 0.96 (.93)**
?² (2, N = 236) = 7.80, p < .05, RMSEA = .11, SRMR = .04, CFI = .98. Degree of External Attributions How likely…due to the professor? How likely…caused by other people? How likely....because the wording was confusing? How likely…due to some external reasons? How likely…because of Person B? To what extent…because of other people? To what extent…because of environmental reasons? 1.00 (.78)** 0.86 (.76)** 0.75 (.86)** 0.55 (.90)** 0.78 (.91)** 0.78 (.77)** 0.50 (.91)**
?² (14, N = 236) = 58.17, p < .05, RMSEA = .00, SRMR = .06, CFI = 1.00. Controllability How likely...was avoidable? How much control do you have…? How likely…you could have prevented the mistake? 1.00 (.73)** 0.86 (.84)** 1.14 (.57)**
94 To what extent…you could have done otherwise? To what extent…you could have done differently? 0.91 (.66)** 0.69 (.74)**
?² (5, N = 236) = 27.41, p < .05, RMSEA = .14, SRMR = .04, CFI = .97. Uncontrollability How likely...was out of your control? To what extent…there is not much you could do? How likely…you have no control? To what extent…you cannot prevent…? To what extent…was inevitable? 1.00 (.81)** 1.04 (.78)** 1.39 (.67)** 1.46 (.63)** 1.13 (.77)**
?² (5, N = 236) = 28.72, p < .05, RMSEA = .00, SRMR = .05, CFI = 1.00. * p < .05; ** p < .01. Note. Bold values represent fixed unstandardized loadings for reference indicators. See the complete list of items for attributions in Appendix B. a. Degree of internal attribution was measured originally by a 7-item magnitude scale. The principal components analyses indicated three items loaded highly on both the first and second components (the eigenvalues for both components were larger than 1). The three items were deleted from the scale. The statistics reported in this table represent the results after deletion.
95 Table 9 Expected Consequences Variables, Indicator Loadings, and Measurement Model Indices Expected Consequences and Indicators Unstandardized Loadings (Standardized) Expected Anger How angry do you expect Person B to be? How irritated do you expect Person B to be? How furious do you expect Person B to be? How mad do you expect Person B to be? How fuming do you expect Person B to be? How annoyed do you expect Person B to be? 1.00 (.58)** 0.92 (.62)** 1.07 (.65)** 1.01 (.61)** 1.03 (.69)** 0.88 (.65)**
?² (9, N = 236) = 196.93, p < .05, RMSEA = .30, SRMR = .04, CFI = .91. Expected Responsibility How much responsibility…assigned to you? To what extent…Person B think you are responsible? To what extent…Person B think you are chargeable? To what extent…Person B think you are accountable? To what extent…should be attributed to you? 1.00 (.63)** 0.98 (.61)** 0.98 (.61)** 0.97 (.61)** 0.87 (.75)**
?² (9, N = 236) = 34.60, p < .05, RMSEA = .11, SRMR = .02, CFI = .99. * p < .05; ** p < .01. Note. Bold values represent fixed unstandardized loadings for reference indicators. See the complete list of items for anticipated consequences in Appendix C.
96 Manipulation Checks Two-way ANOVAs were conducted to see whether manipulations of outcome severity and relational closeness were successful. In the first ANOVA, the dependent variable was perceived outcome severity and the independent variables were the manipulations of outcome severity (high vs. low, coded as 1, -1) and relational closeness (high vs. moderate vs. low, coded as 1, 0, -1). Results indicated that (a) outcome severity had a significant influence on perceived outcome severity, F(1, 231) = 160.35, p < .01, ?² = .41, (b) relational closeness did not have a significant influence on perceived outcome severity, F(2, 231) = 0.13, p = .88, ?² = .00, and (c) no significant interaction between outcome severity and relational closeness was found, F(2, 231) = 1.28, p = .28, ?² = .01. Participants in the high severity conditions (M = .62, SD = .84) regarded the outcome to be more severe than those in the low severity conditions (M = -.66, SD = .70). The manipulation of outcome severity was successful. The second ANOVA had perceived relational closeness as the dependent variables and the manipulations of outcome severity (high vs. low, coded as 1, -1) and relational closeness (high vs. moderate vs. low, coded as 1, 0, -1) as the independent variables. The results suggested that (a) the manipulation of relational closeness significantly influenced perceived relational closeness, F(2, 231) = 150.54, p < .01, ?² = .57, (b) the manipulation of outcome severity did not have a significant effect on perceived relational closeness, F(1, 231) = 3.58, p = .06, ?² = .02, and (c) no significant interaction was found between the manipulations of outcome severity and relational closeness, F(2, 231) = 0.87, p = .42, ?² = .01. Participants in high relational closeness conditions (M = 1.00, SD = .77) reported being closest to Person B than those in the
97 moderate closeness conditions (M = -.35, SD = .49) and in the low relational closeness condition (M = -.72, SD = .69). A linear relationship between three levels of relational closeness was established, F(1, 226) = 238.877, p < .01, ?² = .50. The manipulation of relational closeness was successful. Model Assessment and Hypothesis Testing The hypotheses were tested with structural equation modeling using LISREL 8.70 (Jöreskog & Sörbom, 2004). The model consists of structural and measurement components. The model was tested to see the effect of outcome severity and relational closeness on individuals’ account selection processes, mediated by manipulation checks (i.e., perceived outcome severity and perceived quadratic relational closeness). Model Assessment The model (Figure 2) shows all links that were proposed by the hypotheses. Model specifications are listed in Table 10 and the unstandardized structural coefficients are reported in Table 11 and Figure 4. The covariance matrix analyzed in LISREL is presented in Appendix M. LISREL syntax for model testing can be found in Appendix N. Figure 2 only illustrates the structural model. The measurement models are not presented in the figure but they were included in the model testing. Overall model fit and model comparison. Chi-square results did not indicate a good model fit, ?²(966, N = 236) = 2475.89, p < .05, RMSEA = .08 with a 90% CI of (.08, .09), SRMR = .21, NFI = .91, GFI = .94. A comparison with the null model was conducted. In the null model, no covariance was assumed between any pair of constructs. For the null model, ?²(1,035, N = 236) = 30724.93, p < .05. Comparing both models resulted in ??²(69, N = 236) = 28249.04, p < .01. The tested model was a significant
98 Table 10 Structural Equations of the Model Proposed by the Hypotheses ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = ?91 ?92 ?71 ?73 ?83 ?74 ?84 ?75 ?85 ?76 ?86 ?87 ?98 ?93 ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 ?11 ?22 ?34 ?44 ?55 ?65 ?2 ?3 ?4 ?5 ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9
Note. ?1 and ?2 represent the manipulation checks (i.e., perceived outcome severity and perceived quadratic relational closeness). ?3 through ?6 are the degree of internal attribution, controllability, the degree of external attribution, and uncontrollability. ?7 is expected responsibility assigned to the offender and ?8 is expected anger felt by victims. ?9 refers to LDA (the composite measure for the likelihood of selecting a defensive account). ?1 through ?5 are outcome severity and relational closeness manipulations and the interactions variables (see Figure 2 for the explanation of the interaction variables).
99 Table 11 The Unstandardized Structural Coefficients for the Model ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = 0.10** -0.02 0.36** 0.09** 0.26** 0.04** -0.12** -0.01 0.26** -0.01 0.13** 0.57** -0.12** -0.02 ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 0.74** 0.31** 0.94 0.30 -0.89 0.03 ?2 ?3 ?4 ?5
** p < .01. Note. See Figure 2 for the explanation of each label.
100 0.74** OS POS
0.10** RC 0.31** PRC 0.36** -0.02 -0.02 Inter1 0.09** IA 0.94 Inter2 0.30 CON -0.01 0.26** 0.26** EA Inter3 -0.89 0.03 Uncon Figure 4. Structural relations proposed in the hypotheses with unstandardized structural coefficients. ** p < .01. See Figure 2 for the explanation of each label. -0.12** 0.04** -0.01 Eanger 0.13** 0.57** Eresp -0.12** LDA
101 improvement over the null model. Although the ?² value was used as an indication of model fit, some scholars argue that the ?² is not an appropriate measure of model fit with a sample larger than 200 (Kline, 2005). Kline proposed that the ratio of ?²/df as a more effective measure of fit when N > 200; Kline suggested a ratio less than 3.00 was acceptable. For the current model, ?²/df = 2.56. The proposed model still could be considered as a good fit based on Kline’s suggestion. SEM scholars (e.g., Bollen, 1989; Hu & Bentler, 1998, 1999) also have recommended that multiple fit indices should be considered to evaluate the fit of a model, given the advantages and disadvantages of each fit index. The goodness-of-fit criteria used here are based on Hu and Bentler’s (1998, 1999) recommendation of a twoindex presentation strategy, using SRMR supplemented with other indices such as CFI, RNI, and RMSEA. Hu and Bentler (1999) specified that a cutoff value close to .95 for CFI, a cutoff value close to .08 for SRMR, a cutoff value close to .95 for RNI, and a cutoff value close to .06 for RMSEA. The proposed model had following fit indices: SRMR = .21, CFI = .94 and RMSEA = .08 with a 90% CI of (.08, .09). Based on Hu and Bentler’s criteria, the proposed model did not have a good fit. Hypotheses Testing Ten hypotheses were proposed regarding the attributional processes individuals engage in to determine their accounts. The effects of the exogenous variables on their relevant endogenous variables were examined. H1 predicts that outcome severity influences the expected responsibility assigned to offenders. In particular, when the outcome is severe, offenders expect more responsibility assigned to them than when the outcome is not severe. The effect of
102 outcome severity on the expected responsibility was significant, ? = 0.36, z = 6.50, p < .01. H1 was supported. The positive sign of the coefficient indicated that more responsibility was expected to be assigned to offenders when the outcome was severe than when the outcome was not severe. One research question was proposed regarding the influence of outcome severity on the selection of an account. The effect of outcome severity on the selection of an account was significant, ? = 0.10, z = 3.45, p < .01. As discussed in the construction of LDA (the composite measure of the likelihood of selecting a defensive account), the higher the value of LDA, a more defensive account will be more likely to be selected. The positive sign of the ? coefficient indicates that when the outcome was severe, offenders were more likely to use a more defensive account than when the outcome was not severe. The results are consistent with Schönbach’s (1990) idea. Schönbach argued that when facing severe outcome, offenders often fear the pressure to take full responsibility for their offenses. To shift some responsibility to reasons other than themselves, the offenders tend to take a more defensive stand than when the outcome of their behavior is not severe. H2 proposed that relational closeness influences people’s choice of accounts curvilinearly. It is predicted that a defensive account is less likely to be used when two parties involved in a conflict have a moderate level of relational closeness than when two parties are strangers or when two parties are close friends. The results were not significant, ? = -0.02, z = -30.77, p > .05. The negative sign of the coefficient implies that when conflict occurs between two parties with a moderate level of relational closeness, offenders are more likely to use a defensive account than when the two parties are close
103 friends or strangers. The direction was opposite to the hypothesis. H2 was not supported. H3 predicts that outcome severity and relational closeness interact to influence the offenders’ selection of accounts in a conflict situation. It is hypothesized that when a severe conflict occurs between two close friends, a less defensive account is more likely to be used than when the conflict is not severe. However, when a severe conflict happens between two strangers, offenders are more likely to use a more defensive account than when the conflict is not severe. When two parties involved in a conflict have a moderate level of relational closeness, a less defensive account is more likely to be used than when the conflict occurs between parties that are distant or close, regardless of outcome severity. The independent variable was the interaction variable (?3). H3 was not supported, r = -0.02, z = -0.44, p > .05. There was no interaction of outcome severity and relational closeness on offenders’ selection of accounts. H4 had four subsets of hypotheses regarding the attributional processes that offenders engage in after a conflict has occurred. It was predicted that when a severe conflict occurs between close friends, internal (H4a) and controllable attributions (H4b) are more likely than in all other conditions; however, when a severe conflict occurs between strangers, offenders tend to attribute their own negative behaviors to external (H4c) and uncontrollable reasons (H4d) more than in all other conditions. No significant results were found, r = 0.94, z = 1.87, p > .05 for internal attribution; ? = 0.30, z = 1.09, p > .05 for controllability; ? = -0.89, z = -1.17, p > .05 for external attribution; and ? = 0.03, z = 0.10, p > .05 for uncontrollability. H4 was not supported. Although the results were not significant, the signs of the coefficients suggest that when a severe conflict occurs between close friends, internal and controllable attributions are more likely to be made by
104 offenders than other conditions. However, when a severe conflict happens between strangers, offenders is less likely to make external attribution and more likely to make uncontrollable attribution than other conditions. H5 to H8 examined the relationship between attributions and expected responsibility and expected anger. Specifically, H5a specified a positive influence of internal attribution on expected responsibility, and H5b predicted a positive influence of internal attribution on expected anger. H5a was supported, ? = 0.09, z = 3.72, p < .01. The positive sign of the ? coefficient indicates that when offenders attribute their behaviors more to their internal traits, they expect more responsibility to be assigned to them. H5b was not supported, ? = -0.01, z = -0.32, p > .05. The negative sign of the coefficient suggests an opposite direction to what was predicted in H5b: When offenders attribute their behaviors more to their internal characteristics, they expect less anger would be felt by the victim. H5 was partially supported. H6 predicted that controllability positively influences expected responsibility to be assigned to offenders and expected anger felt by the victim. H6a predicted when offenders attributed their behavior more to controllable reasons, they expected more responsibility assigned to them. Findings were significant, ? = 0.26, z = 6.29, p < .01. The positive sign of the ? indicates the positive influence of controllability on responsibility expected to be assigned to offenders. H6b predicted a positive influence of controllability on expected anger. Results were significant, ? = 0.26, z = 6.85, p < .01. The positive sign of the ? coefficient suggests that when offenders make more controllable attribution, they expect more anger to be felt by the victim. H6 was supported. H7 stated that when offenders attributed their behaviors more to external reasons,
105 they expected less responsibility to be assigned to them and less anger felt by victims. Significant results were found for expected responsibility, ? = 0.04, z = 2.50, p < .01. However, the positive sign of the coefficient suggests that the direction of external attribution’s influence on expected responsibility is opposite to the prediction in H7a. The results showed that when offenders make more external attribution, they expect more responsibility to be assigned to them. No significant results were found for the influence of external attribution on expected anger, ? = -0.01, z = -0.48, p > .05. The negative sign of the coefficient showed that when more external attribution is made, offenders expect less anger to be felt by the victim. H7 was not supported. As for attributions to uncontrollable reasons, H8 predicted uncontrollability negatively influenced expected responsibility (H8a) and anger (H8b). H8a was supported, ? = -0.12, z = -2.88, p < .01. The negative sign of the coefficient indicates that when offenders make more uncontrollable attribution, they expect victims to assign less responsibility to them. Significant results were found for H8b, ? = 0.13, z = 3.66, p < .01. However, the positive sign of the coefficient indicates that when more uncontrollable attribution is made, offenders expect more anger felt by victims, which is opposite to the prediction in H8b. H8 was partially supported. Expected responsibility was predicted to positively influence expected anger in Hypothesis 9. The results supported this relationship, ? = 0.57, z = 10.54, p < .05. The positive sign of the coefficient indicates that more expected responsibility leads to more expected anger. H9 was supported. H10 predicted a negative influence of expected anger on offenders’ selection of accounts: When offenders expect more anger felt by victims, they will be less likely to
106 use a defensive account. The findings were statistically significant, ? = -0.14, z = -4.91, p < .01. The negative sign of the coefficient suggests that a defensive account is less likely to be selected when offenders expect more anger felt by the victims. H10 was supported. Overall the model did not have a good fit. Among the ten hypotheses and one research question, five hypotheses were supported and two other hypotheses (H5 and H8) were partially supported. The two coefficients that tested the manipulations were significant also, consistent with the manipulation checks that were conducted outside of the model testing, ? = 0.74, z = 10.53, p < .01 for outcome severity manipulation check and ? = 0.31, z = 6.03, p < .01 for relational closeness manipulation check (for the quadratic relational closeness). Post-Hoc Analyses Post-hoc analyses were conducted to further explore the influence of relational closeness and outcome severity on offenders’ choice of accounts. First, the nonsignificant hypotheses in the model testing were examined separately. Second, modification indices for the model tested in the above section were examined for additional paths to improve the model fit. Third, the model was revised based on the results in the first two steps and the revised model was tested using LISREL 8.70 (Jöreskog & Sörbom, 2004). Hypothesis Testing The non-significant results in the model testing may be due to multicolinearity among the five exogenous variables. The correlation matrix of the five exogenous variables in the proposed model had a determinant of .11. Among ten correlations formed by these five variables, seven of them were statistically significant. See Table 12 for all
107 Table 12 Intercorrelations between the Five Exogenous Variables Variables ?1 ?2 ?3 ?4 ?5 ** p < .01. Note. ?1 refers to outcome severity manipulation (low outcome severity was coded as -1 and high outcome severity was coded 1) and ?2 refers to relational closeness manipulation (quadratic coding: low relational closeness was coded as 1, moderate relational closeness was coded as -2, and high relational closeness was coded as 1). ?3, ?4, and ?5 refer to the interactions between outcome severity and relational closeness. ?3 was calculated by multiplying the outcome severity manipulation and relational closeness manipulation. ?4 refers to the coding of the high severity and close relationship condition (coded as 1) versus all other five conditions (coded as 0). ?5 refers to the coding of the high severity and not-close-at-all relationship (coded as 1) and other five conditions are coded as 0. ?1 1 ?2 .01 1 ?3 .03 .01 1 ?4 .46** .33** .57** 1 ?5 .43** .31** -.55** -.21** 1
108 the correlation coefficients among the five exogenous variables. According to Kline (2005), a small determinant of the correlation or covariance matrix can lead to nonsignificant structural coefficients, which may partly explain the non-significant results in the current research. The paths that were not significant in the model testing were examined again separately. Specifically, H2, H3, H4, H5b, and H7b were re-tested. H2 hypothesized a curvilinear effect of relational closeness on offenders’ choice of accounts. Specifically, when a conflict occurs between two parties who have a moderate level of relational closeness, offenders are less likely to use a more defensive account than when the conflict occurs between close friends or between strangers. A linear regression was conducted to test H2, where the quadratic relational closeness manipulation check was entered as the independent variable and LDA (the composite measure of accounts) was entered as the dependent variable. The results were significant, ? = -0.64, F(1, 233) = 9.68, p < .05, R² = .04. However, the negative sign of the ? coefficient suggests that when a conflict occurs between two parties that are moderately close, offenders are more likely to use defensive accounts than when a conflict occurs between friends or between strangers. The direction was opposite to the prediction in the hypothesis. Hypothesis 3 proposed an interaction between outcome severity and relational closeness on offenders’ choice of accounts. When a conflict occurs between close friends, a less defensive account is more likely to be used in a severe situation than in a notsevere situation. When the conflict occurs between two parties who have a moderate level of relationship, a less defensive account will be used regardless of the severity of the outcome. When the conflict happens between strangers, a more defensive account is more
109 likely to be used in a severe situation than in a not-severe situation. An ANOVA was used to examine H3. LDA, the likelihood of selecting a defensive account, was used as the dependent variable and the interaction between outcome severity and relational closeness was entered as the independent variable. The interaction variable (i.e., ?3 in the model tested previously) was created by multiplying outcome severity manipulation (high severity coded as 1 and low severity coded as -1) and relational closeness manipulation (linear coding: high relational closeness coded as 1, moderate relational closeness coded as 0, and low relational closeness coded as -1). The results were significant, F(2, 232) = 4.97, p < .05, ?² = .04. The direction of the influence was consistent with the hypothesis. When a conflict occurs between high relationally close parties, offenders are less likely to use a more defensive account in a severe-outcome situation (M = 60.83, SD = 4.53) than in a not severe-outcome situation (M = 61.21, SD = 4.51). When the conflict occurs between low relational closeness parties, offenders are more likely to use a more defensive account in a severe-outcome situation (M = 61.21, SD = 4.51) than in a not-severe-outcome situation (M = 60.83, SD = 4.53). When the conflict occurs between moderate relational closeness parties, offenders’ likelihood of using accounts did not differ by different levels of outcome severity (M = 62.93, SD = 4.22). H3 was supported. Hypothesis 4 predicted relational closeness and outcome severity interact to influence offenders’ attributions. H4 had four subsets of hypotheses. H4a and H4b predicted that when a severe conflict occurs between high relationally close parties, offenders are more likely to make internal and controllable attributions than in all other situations. H4c and H4d hypothesized that when a severe outcome conflict occurs
110 between low relationally close parties, offenders are more likely to make external and uncontrollable attributions than in all other conditions. To examine H4, the six experimental conditions created by two levels of outcome severity (high vs. low) and three levels of relational closeness (high vs. moderate vs. low) were coded into two variables. The first variable (i.e., ?4 in the model tested previously) was coded the following way: The high outcome severity and high relational closeness condition was coded as 1 and all other five conditions were coded as 0. The second variable (i.e., ?5 in the model tested previously) was coded that the high outcome severity and low relational closeness condition was 1 and all other five conditions were 0. ?4 was entered as an independent variable in a MANOVA, the first component scores of the degree of internal attribution and controllability being the dependent variables. No significant results were found, F(1, 234) = 3.43, p = .07, ?² = .01 for the degree of internal attribution and F(1, 234) = 0.68, p = .41, ?² = .003 for controllability. Another MANOVA was conducted with ?5 as independent variable and the first components of the degree of external attribution and uncontrollability as the dependent variables. The results were not significant, F(1, 234) = 1.13, p = .29, ?² = .01 for the degree of external attribution and F (1, 234) = 0.08, p = .78, ?² = .00 for uncontrollability. Hypothesis 5b predicted that when offenders make more internal attribution, more anger is expected to be felt by the victim. The first components of the degree of internal attribution and expected anger were obtained to be used in a regression analysis. A linear regression was conducted with the degree of internal attribution as the independent variable and expected anger as the dependent variable. The results were significant, ? = 0.56, F(1, 234) = 104.23, p < .01, R² = .31. The positive sign of the coefficient suggests
111 that when offenders made more internal attribution, more anger was expected to be elicited by the victim. Hypothesis 7b proposed a negative influence of external attribution on expected anger. The first components of the degree of external attribution and expected anger were used in a regression analysis to test H7b. The results were significant, ? = 0.28, F(1, 234) = 19.36, p < .01, R² = .08. Although the results were significant, the positive sign of the coefficient indicates that when more external attribution is made, more anger is expected, which is opposite to the prediction. The significant paths that were indicated by the exploratory analyses were kept in the model. That is, the influence from perceived quadratic relational closeness on people’s choice of accounts, the influence from the degree of internal attribution on expected anger, the influence of the degree of external attribution on expected anger, and the interaction of outcome severity and relational closeness on people’s choice of accounts remained in the model. All the paths that indicated the effect of outcome severity and relational closeness on the attributions people make were removed from the model due to non-significant results in both the model testing and the exploratory analyses. Modification Indices Modification indices for the proposed model (see Figure 2) suggested the maximum ?² change would occur when a path was established from expected responsibility to controllable attribution (??² = 168.34). However, it was expected that people’s attributions influence their perceptions of expected responsibility and expected anger, and then influence their choice of accounts, not vice versa. The study was also
112 conducted in a way that attribution was first made and then participants were asked to expect how much responsibility to be assigned. Several other paths that had large value of modification indices, such as a link that indicated the influence from expected responsibility on internal attribution (??² = 158.64), had the same problem. Those paths, although they have large modification indices, were not considered in the revised model. It was hypothesized that outcome severity and relational closeness interacted to influence different attributions. These hypotheses were not supported. However, the model’s modification indices suggested that outcome severity alone may influence different attributions via perceived outcome severity. If a path is added from perceived outcome severity to the degree of internal attribution, the expected change in the ?² value is 36.26. If a path is established from perceived outcome severity to controllability attribution, the expected ?² change is 64.30. If a path is added from perceived outcome severity to the degree of external attribution, the expected ??² equals 26.48. If a path is added from perceived outcome severity to uncontrollability attribution, the expected ?² change is 9.05. Those paths were considered in revising the model. Furthermore, modification indices suggested a path from perceived outcome severity to expected anger (??² = 57.71). It was hypothesized that when an offender perceives the outcome of a conflict to be severe, he or she expects more responsibility to be assigned to him or her; in turn, more expected responsibility is expected to lead to more expected anger. The influence of perceived outcome severity on expected anger is mediated by expected responsibility. However, a direct influence is also possible. Thus, the path from perceived outcome severity to expected anger is considered in the revised model.
113 Model Revision Based on the exploratory analyses and modification indices, the following changes are made to revise the model: (a) Two interaction variables (?4 and ?5, see Figure 2 for the detailed description of these two variables) did not influence attributions as predicted in Hypothesis 4. The relevant paths were removed from the model. (b) The influence of perceived outcome severity on expected responsibility was mediated by attributions (i.e., the degree of internal attribution, controllability, the degree of external attribution, and uncontrollability). Four paths were added to demonstrate the mediation by attributional processes. (c) An additional path was added from perceived outcome severity to expected anger. See Figure 5 for the revised model. The dotted lines represent the paths added to the original model based on exploratory analyses and the modification indices. Testing the Revised Model The revised model was tested with structural equation modeling using LISREL 8.70 (Jöreskog & Sörbom, 2004). The model consisted of structural and measurement components. Model specifications are listed in Table 13, and unstandardized structural coefficients are reported in Table 14 and Figure 6. The covariance matrix analyzed in LISREL is found in Appendix O. LISREL syntax for model testing can be found in Appendix P. Figure 5 only illustrates the structural model of the relationship between the exogenous and endogenous variables. The measurement models are not included in the figure but they were included in the model testing. For the revised model, the chi-square was significant, ?²(883, N = 236) = 2210.26, p < .01. Compared to the model before revision, chi-square value was improved by
114
OS
POS
IA
Eresp
CON
EA Eanger Uncon LDA RC PRC
Inter1
Figure 5. Structural relations based on post-hoc analysis. Note. All exogenous variables are allowed to covary and all error terms are allowed to covary. The dotted lines represent the paths added based on the exploratory analyses and modification indices. The labels and coding for each variable can be found in Figure 2.
115 Table 13 Structural Equations of the Revised Model Based on Modification Indices and Exploratory Analyses ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = ?31 ?41 ?51 ?61 ?71 ?81 ?91 ?92 ?73 ?83 ?74 ?84 ?75 ?85 ?76 ?86 ?87 ?98 ?93 ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 ?11 ?22 ?2 ?3 ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9
Note. Bold symbols represent the paths that were added based on exploratory analyses and modification indices. See Figure 2 for the labels and coding of each variable.
116 Table 14 Unstandardized Structural Coefficients for the Revised Model ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = 1.18** 0.74** 1.30** 0.24** 0.29** 0.48** 0.11** -0.02 0.08** 0.29** 0.04** -0.11** 0.01 0.15** -0.01 0.05 0.37** -0.14** ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 0.74** 0.31** ?2 ?3 ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 -0.02 ?9
* p < .05; ** p < .01. Note. Bold values represent the paths that were added based on exploratory analyses and modification indices. See Figure 2 for the labels and coding of each variable.
117 0.29** 0.08** 0.01 1.18** 0.74** OS POS 1.30** 0.24** Uncon 0.48** 0.11** LDA 0.31** -0.02 EA 0.04** -0.11** -0.01 0.05 -0.14** 0.37** Eanger 0.74** CON 0.15** 0.29** Eresp
IA
RC
PRC
-0.02 Inter1 Figure 6. The revised model with unstandardized structural coefficients. ** p < .01. See Figure 2 for the labels and coding of each variable.
118 265.63. Given that ?df equaled 83, the revised model was a significant improvement over the original model. For the revised model, the ratio of ?²/df equaled 2.50, which indicated a good model fit according to Kline’s (2005) standard. Before the revision, the original model had the following fit indices: SRMR = .21, CFI = .94 and RMSEA = .08 with a 90% CI of (.08, .09). After the revision, model fit indices were improved: SRMR = .12, CFI = .95, and RMSEA = .08 with a 90% CI of (.08, .08). Although based on Hu and Bentler’s (1999) criteria the revised model still could not be considered a good fit, it had been improved from the original model, and the CFI had reached the cutoff standard suggested by Hu and Bentler. After examining the ?² value, the ratio of ?²/df, and fit indices, the revised model is considered to have a moderate fit. Examining the paths in the revised model indicated almost the same pattern as shown in the original model: Paths that were significant in the original model were statistically significant in the revised model (except the link between uncontrollability and expected anger), and paths that were not significant in the original model testing but significant in the exploratory analyses still remained non-significant in the revised model. The five paths that were added based on modification indices were significant (see Table 14 for all unstandardized structural coefficients). When testing effects outside of the structural equation model, relational closeness (quadratic coding) had a significant main effect on offenders’ choice of accounts and also relational closeness interacted with outcome severity to influence offenders’ choice of accounts significantly. However, the effect of relational closeness on offenders’ choice of accounts was not significant once outcome severity came into the picture. Outcome severity influenced offenders’ choice of accounts directly and indirectly via perceived
119 outcome severity, attributions offenders made, and consequences offenders expected. Compared with relational closeness, outcome severity is a better predictor about the accounts that an offender will use in conflict situations.
120 CHAPTER V Discussion This chapter consists of four parts. A summary of the study is provided in the first part. The second part interprets the results and discusses the study’s implications. The third part consists of the study’s limitations and directions for the future research. The significance of the study and a conclusion form the last part of the chapter. Summary of the Study The current research addressed the following question: What attributional processes do offenders engage in to determine the account used in a conflict? This study examined the influence of outcome severity and relational closeness on offenders’ choice of accounts, mediated by attributional processes (i.e., internal, external, controllable, and uncontrollable attributions) and offenders’ anticipated consequences (i.e., expected responsibility and expected anger). Two levels of outcome severity and three levels of relational closeness were employed, creating six experimental conditions. In each condition, offenders read a hypothetical conflict scenario by imagining they were offenders in the conflict and they were interacting with another person. Prior to the formal study, six pilot studies were conducted to (a) develop the manipulations that would be used in the formal study, (b) develop a measure of accounts for the formal study, and (c) test and improve the items that measured all the constructs. The manipulations and instruments that were developed and tested in the pilot studies were then used in the formal study. The formal study recruited 238 participants enrolled at the University of Maryland. Two participants had incomplete data. Pairwise or listwise deletion was used,
121 depending on the type of statistical analyses used. Each participant was randomly assigned to one of the six experimental conditions. Questionnaires were employed in the study. The questionnaire started with a hypothetical scenario that described a conflict in a class project; this scenario was followed by the manipulation checks. The second part of the questionnaire asked participants to report their attributions. Participants also reported their expectations about the responsibility assigned to them and the anger felt by the victim in the scenario. Participants were then asked to report the likelihood of selecting each account developed from Pilot Study 4.6 Participants’ demographic information was collected in the last part of the questionnaire. After participants completed their questionnaire, they were fully debriefed. Magnitude scales were used in this study. Items that violated the assumption of normality were transformed by a power transformation. Confirmatory factor analyses (CFAs) were conducted to examine the dimensions of each construct. When ?² value and fit indices indicated that there was more than one factor for some variables, principal components analyses (PCAs) were conducted to explore the potential components. No rotation was used in the PCAs. The results from CFAs and PCAs determined the items that were included in the structural equation model. The hypotheses were tested using structural equation modeling. Outcome severity, relational closeness, and their interactions were dummy coded and entered as exogenous variables (see Figure 2 for the details of labels and coding of each variable). Manipulation checks were used as mediating variables; offenders’ choice of account was used as an outcome variable; and attributional processes and anticipated consequences were also expected to mediate the
The likelihood results of the formal study were combined with the defensiveness results collected from Pilot Study 5 to create the composite measure (LDA) used in model and hypothesis testing. The details about the creation of LDA can be found on pp. 86-88.
6
122 influence of the independent variables on offenders’ choice of accounts. The post-hoc analyses were conducted for model fit improvement and better understanding of accounts that offenders use in conflict situations. Summary of Results Hypothesis 1: Influence of Outcome Severity on Expected Responsibility Outcome severity was predicted to positively affect responsibility assigned to offenders (Walster, 1966). Previous research has suggested that victims or observers perceive offenders to be less careful when the outcome is severe; therefore, victims or observers tend to assign more responsibility to offenders (Walster, 1966). From the offenders’ perspective, when they perceive the outcome as severe, offenders would expect little chance for them not to be blamed for what has happened. Therefore, offenders expect victims to assign more responsibility to them when the outcome is severe than when the outcome is not severe. The results showed a significant positive effect of outcome severity on expected responsibility. Offenders, when the outcome is severe, tend to expect more responsibility to be assigned to them. Research Question 1: Effect of Outcome Severity on the Choice of Accounts Researchers have argued for opposing possibilities of outcome severity’s effect on people’s choice of accounts. McLaughlin, O’Hair, et al. (1983) stated that when an outcome is severe, people tend to use less defensive accounts, such as concessions, for the purpose of easing the tension aroused by their behaviors and to maintain a good image. However, Schönbach (1990) argued that although there is a possibility that offenders may use a less defensive account to present a good image in the situation, when an outcome is really severe, taking responsibility may be a great burden for offenders.
123 Instead, offenders may reject or deny their responsibility to save their face. They will choose more defensive accounts when the severity of an outcome is more severe. The results favored Schönbach’s prediction: When offenders perceived that the outcome was severe, offenders were more likely to use a more defensive account, such as a justification, than when the outcome is not severe. Hypotheses 2: Influence of Relational Closeness on the Choice of Accounts It is argued in H2 that when a conflict occurs between parties that are moderately close, offenders are less likely to use a defensive account than when the conflict occurs between parties that are close or distant. A curvilinear effect of relational closeness on offenders’ choice of accounts was proposed in H2. However, the path indicated by H2 in the structural equation model was not significant. When testing H2 outside of the model, the results suggested that relational closeness indeed influences offenders’ choice of accounts, in a way opposite to the prediction in H2: Offenders were more likely to use a defensive account when moderately close to the victim than when relational closeness between the offender and the victim was high or low. One possibility is when manipulating the moderate relational closeness in the hypothetical conflict scenario, participants were told that they have known the other person in the conflict because they have taken several courses together but they did not really hang out after school. Although participants were told that they had interactions with the other party in the past (mostly in class), it did not necessarily indicate that they would take courses together again; that is, participants may not perceive a high possibility of future interaction with the victim. Another possibility is the influence of relational closeness on the choice of accounts may depend on how severe the outcome is.
124 Hypotheses3: Influence of Relational Closeness and Outcome Severity on the Choice of Accounts Hypothesis 3 examined the interaction of relational closeness and outcome severity on offenders’ choice of accounts. It was argued that when a severe conflict occurs between two parties who are close, offenders are less likely to use a defensive account than when the conflict is not severe. When the conflict occurs in a moderate relational closeness situation, offenders are less likely to use a defensive account than when the conflict is between two parties that are distant or close, regardless of the severity of outcome. When a conflict occurs between strangers, offenders are more likely to use a defensive account than when the conflict is not severe. So offenders select their account depending on the interaction between outcome severity and relational closeness. Similar as Hypothesis 2, when H3 was tested in the structural equation model, the path was not significant. However, when testing the hypothesis outside of the model, significant results suggested that outcome severity and relational closeness interact to influence offenders’ choice of accounts. Hypotheses 4: Attributional Processes Hypothesis 4 specified the effect of outcome severity and relational closeness on offenders’ attributional processes. The findings did not support the predictions in H4. The interaction variables created by outcome severity and relational closeness did not influence the attributions offenders made. In the post-hoc analysis, the modification indices suggested that outcome severity alone may influence the attributions offenders make and then influence the expectation of consequences and the choice of accounts. Additional paths were added to the model to test the influence of outcome severity and
125 attributions offenders make. The findings of the revised model suggested that when the outcome is severe, offenders tend to make more attributions (more internal attribution, more controllability attribution, more external attribution, and more uncontrollability attribution) than when outcome is not severe. Maybe when offenders perceive that their behavior leads to a severe outcome, they search for any possible reason for the behavior in question so that they can decide which reason works best to explain their behavior to the victim. Hypotheses 5: Internal Attribution and Anticipated Consequences H5 proposed that the attributions offenders make influences their expectations about the responsibilities to be assigned to them and expected anger: When causes are perceived to be internal, offenders acknowledge that they are the major causes of the conflict; thus, they expect more responsibility assigned to them by victims (H5a). Furthermore, when offenders make more internal attributions, they expect victims to feel angrier toward their wrong-doings (H5b). The findings indicated that when offenders make more internal attribution, they expect more responsibility to be assigned to them. However, the degree of internal attribution made by offenders does not influence the expectation of anger felt by the victim. Given that the expected responsibility significantly influences the expected anger, as indicated by the results for Hypothesis 9, it’s possible that the direct influence of internal attribution on the expected anger is cancelled out by the mediating effect. Hypotheses 6: Controllability and Anticipated Consequences H6 proposed a positive influence of controllability attribution on anticipated consequences: When offenders perceive that they have control over what has happened,
126 they expect the victims to assign more responsibility to them because offenders could have done otherwise to prevent the conflict. Furthermore, offenders expect that the victims have more anger as well. The findings suggested that the controllability attribution positively affected the amount of responsibility that was expected to be assigned to offenders and the amount of anger expected to be felt by victims. When offenders realize that they could have prevented what has happened, they need to watch out for the anger from the victim. Hypotheses 7: External Attribution and Anticipated Consequences H7 specified a negative influence of external attribution on expected consequences. When an external attribution is made by offenders, they expect victims to assign less responsibility to them and feel less angry. The results did not tell such a story: External attributions did not significantly influence offenders’ expected responsibility and expected anger; furthermore, the positive sign of ? coefficient implied that when offenders make more external attribution, they expect the victims to assign more responsibility to them, which was opposite to the prediction. As discussed by Konovsky and Jaster (1989), external attribution may not work all the time. When offenders attribute their wrong-doings to external causes, they may be perceived as lacking in the ability to take care of their own behaviors or to be simply dishonest about what they have done. When victims believe that the offenders are not sincere, they may assign more responsibility to the offenders as a punishment. Hypotheses 8: Uncontrollability and Anticipated Consequences H8 specified that when offenders make more uncontrollability attribution, they believe there is nothing much they could have done to prevent what has happened.
127 Therefore, offenders do not see themselves as being responsible for the conflict and they expect less responsibility to be assigned to them and less anger to be felt by victims. It was found that when the causes are uncontrollable, offenders do not expect themselves to be held responsible for their wrong-doings. However, the results indicated that even when uncontrollable reasons are believed to lead to the conflict, offenders still expect victims to be angry, which is opposite to the prediction. An explanation of these results is that although uncontrollable reasons could shift some amount of responsibility from the offenders, what has happened does negatively influence the victims. Offenders may still expect certain negative emotions, such as anger, could arise naturally as the consequence of the offenders’ wrong-doing. Hypothesis 9: Expected Responsibility and Expected Anger Judgment of responsibility was believed to positively influence the feeling of anger (Allred et al., 1997; Liu & Yao, 2007). H9 stated that when offenders expect more responsibility to be assigned to them by victims, they also expect victims to be angrier. These findings were consistent with the hypothesis. When one is judged to be responsible for a negative incident, more feelings of anger are expected. Hypothesis 10: Expected Anger and the Choice of Accounts When offenders expected more anger to be felt by victims, to reduce the negative emotion and possibly decrease the responsibility, they are less likely to use a defensive account. The findings supported this hypothesis. Offenders apparently determine their accounts based on their expectations of the victims’ anger. If they sense the victims get furious because of their behaviors, they tend to be nice and be less defensive; however, if they believe victims are not so angry about their behavior, they are more likely to take a
128 defensive stand when explaining their behaviors. How do People Decide the Accounts to Use in Conflict Situations? When people know that their behaviors lead to a conflict, how will they explain their behaviors? According to the findings of the current research, people choose their accounts mainly based on how serious the outcome is. The relationship between the parties involved in a conflict may also play a role in how offenders explain their behaviors. For example, if a conflict happens between two acquaintances, e.g., classmates, offenders tend to choose a more defensive account to explain their wrongdoings, compared to when the conflict occurs between two close friends or between two strangers. But none of this matters if the conflict has severe consequences. When offenders perceive that a severe outcome has been caused by their behaviors, no matter with whom they have the conflict, offenders may think about defending themselves, perhaps as a method to protect their own images. But the determination of accounts is well calculated. When offenders perceive that their behaviors lead to a severe outcome, they seek out all possible reasons, internal, external, controllable, or uncontrollable, and they imagine how much responsibility the victim may assign to them and how angry the victim may be. After deliberately considering the causes of the conflict and the reaction from the victim, offenders will select the account that they think benefits them most. How do offenders know which account works best? They imagine how angry the victim may be and how much responsibility the victim is willing to assign to them. If offenders believe that the victims will assign more responsibility to them and the victims will become angry, the offenders will then be nice and use a less defensive account, such
129 as admitting their wrong-doings or offering some compensation. On the other hand, if offenders think that the victims are not so angry about the wrong-doing, they tend to take a more defensive stand, such as pointing to external reasons as the cause of the conflict or denying their connection to the behavior in question. Limitations of the Study and the Direction for Future Research This study has several limitations regarding the conceptualization of accounts, the measure for the accounts, and the hypothetical scenario used in the formal study. Limitation Regarding the Conceptualization of Accounts The current research examined four types of accounts: concession, justification, excuse, and refusal. Offenders’ actions were assumed to be self-serving: Offenders choose different types of accounts to defend themselves. Therefore, the four types of accounts represent the extent to which offenders defend themselves in a conflict situation. As indicated by the results from Pilot Study 5, concession was found to be the least defensive type of accounts and refusal was found to be the most defensive type of accounts. The defensiveness levels of justification and excuse were in the middle and justification was more defensive than excuse, which was opposite to what Schönbach (1990) proposed. However, one caveat to this assumption is that not all accounts are selfserving. An offender may use an account to honestly describe what has occurred in a conflict situation, so that the account is not intended to defend oneself at all. The defensiveness continuum of accounts cannot be applied to all types of accounts. Future research may explore other continua to describe different types of accounts and may consider other types of accounts other than the four types investigated in the current research.
130 Further, the current research argues that offenders use different types of accounts to maintain their self image or maintain the relationship with others. However, the goals of different types of accounts are assumed, not tested, in this dissertation. Offenders may consider different goals in different relationships, rather than simply trying to maintain the relationship. Future research should scrutinize the different goals offenders may try to fulfill and examine the account-making process through a goal perspective. Limitation Regarding the Dependent Measures Each construct proposed in the theoretical model was measured by 5 to 7 items, which were developed and tested in several pilot studies. Reliability coefficients for all the measures were acceptable and in some cases good. However, fit indices from the majority of the measurement models indicate non-unidimensional factor structures. For example, the measures for the four types of attributions all had more than one underlying factors, which may contribute to the non-significant results about attributional processes. Previous research on accounts often asked participants to self-report the actual accounts they used in an incident and use coders to code these responses. The current research took a different approach. A list of accounts was developed and provided to participants to indicate their likelihood of choosing each account. Although this approach provides a measure that can be adopted in other conflict situations, it has its limitation. The accounts people make in a conflict situation may be much more complicated than the list of accounts provided in the measure. Also people often use more than one type of accounts. The measure used in the current research constrained participants to only a limited amount of accounts. Future research may use both methods (i.e., self report of actual accounts and selection from a list of provided accounts) to better understand
131 people’s choice of accounts and may consider examining the sequences of accounts, instead of a single account. Further, this study only focused on the offenders’ perspective for selecting accounts. To determine the accounts to be used in a conflict situation, offenders seem to go through a series of calculations: They consider the reasons that lead to their wrongdoing and they imagine the victims’ reactions, such as how much responsibility the victims will assign to the offenders and how angry the victims will be. But all these variables were assessed only from the offenders’ perspective. Whether the accounts offenders select will work out (i.e., accepted by the victims) remains a question. The acceptance of accounts is beyond the scope of the study but is an important topic for the future research. Limitation Regarding the Hypothetical Conflict Scenario This study used one hypothetical conflict scenario and asked participants to imagine they were the offenders in the conflict. Although using hypothetical scenarios had its advantages, the problems with hypothetical scenarios bring cautions about interpreting and applying the study’s results. The current research is based on the assumption that people deliberately process information at the time of conflict and they are engaged in making attributions, expecting consequences for their communication, and selecting an account. However, Nisbett and Wilson (1977) pointed out that “in dissonance and attribution studies…people appear to be unable to report that a cognitive process has occurred” (p. 246), which may explain why the paths to the attributions in the original model were not significant. It is also possible that the processes described in an experiment may be different from the processes people experience in an actual
132 interaction. Future research may consider using actual interactions to examine the processes offender may experience in a conflict situation. Future Directions: Cultural Differences in Accounts The current research only studied the attributional processes involving accounts in the United States. Although some hypotheses were not supported, overall the model had a moderate fit. The question is whether the model can be extended to different cultures. Some features of account-giving and attributional processes provided foundations for possible cultural differences in the proposed model. First, McLaughlin, Cody et al. (1983), Gonzales, Manning, and Haugen (1992), and Takaku (2000) have argued that accounts have implications for face concerns: In general, different types of accounts pose different degrees of threats to offenders’ and victims’ face. Given the consistent findings on cross-cultural difference in face concerns (e.g., Cai & Fink, 2002; Ting-Toomey & Oetzel, 2002), it is expected that culture will influence people’s choice of accounts. McLaughlin, Cody et al. (1983) and Gonzales et al. (1992) argued that concessions, an explanation that admits responsibility and offers compensation, make offenders vulnerable to others’ blame. But using concessions indicates that offenders feel sorry about their behaviors and are willing to do something to compensate the victims, which saves the victims’ face. McLaughlin, Cody et al. and Gonzales et al. specifically argued that using concessions, among all four types of accounts, poses a maximum threat to offenders’ face but a minimum threat to the victims’ face. Excuses also threaten offenders’ self face because excuses, like concessions, admit offenders’ responsibility for the behavior that led to a conflict. However, when people use excuses, they can point to
133 external reasons as the causes of the behavior in question, the level of threat to offenders’ face is less than that posed by a concession. On the other hand, when offenders use external reasons to explain their behavior, although offenders admit that they need to take responsibility for their behavior, they are not so guilty because they don’t have too much control over what has happened. Excuses pose a greater threat to victims’ face than do concessions. Further, justifications suggest that offenders believe that their behaviors are consistent with a higher goal. Therefore, offenders’ self face is protected because offenders perceive that they do not have much freedom when it comes to the behavior in question; they have to behave in a certain way to fulfill a superordinate goal. Although justifications acknowledge offenders’ wrong-doing, offenders do not intend to accept that they violate a social norm, which poses a higher threat to victims’ face than either concessions or excuses. McLaughlin, O’Hair, et. al. (1983) and Gonzales et al. (1992) stated that when refusals are used, offenders could deny their connection with the behavior in question or deny that the behavior in question is inappropriate; as a result, refusals protect offenders’ face to the maximum degree but pose the greatest threat to victims’ face. Studies on conflict management styles have suggested that face concerns vary as a function of culture (e.g., Ting-Toomey & Oetzel, 2002). Most of the literature on face concerns and conflict management has focused on Chinese and Americans (e.g., Brew & Cairns, 2004) and has suggested that Chinese have a higher concern for others’ face than Americans do;on the other hand, the literature has argued that Americans’ concern is more with self face than others’ face (Ting-Toomey & Oetzel, 2002). Although there are
134 some arguments against this conclusion in research (e.g., Cai & Donohue, 1997), given this discussion on face concern and accounts, Chinese offenders are expected to be more likely to use a less defensive account in a conflict situation whereas American offenders may be more likely to use a more defensive account to handle conflict. A second reason for cultural differences in account use is that attributional patterns are different across cultures. For example, the fundamental attribution error (FAE, i.e., a tendency to overestimate the influence of personal traits and underestimate the influence of situational factors; Ross, 1977) is found in the United States but the generality of the FAE cross-culturally is arguable. Studies have found that Chinese, as compared to Americans, are more likely to rely on external factors to explain an actor’s behaviors (e.g., Choi, Nisbett, & Norenzayan, 1999). However, when a negative event occurs, Asians are more likely to attribute their own failure to internal reasons (e.g., lack of effort) whereas Americans attribute the failure to environmental factors (e.g., the difficulty of an exam) (Salili, Maehr, & Gillmore, 1976). Findings from those attribution studies suggest that when a conflict occurs, Chinese are more likely to make internal attributions about their own behavior whereas Americans are more likely to make external attributions. Consequently, to reduce the responsibility assigned by victims and alleviate the anger that accompanies a conflict, Chinese are less likely to use a defensive account than Americans. Future research may test the model in different cultures and examine cultural influences on attributional processes involved in the selection of accounts. Significance of the Study and Conclusion This study attempted to answer one question: What attributional processes do
135 offenders engage in to determine the account to use in a conflict situation? A model was established to examine the influence of outcome severity and relational closeness on offenders’ choice of accounts, mediated by the attributions offenders make and offenders’ anticipation of the victim’s reactions. The data moderately supported the model and showed that when the outcome of a conflict is severe, offenders tend to make more attributions, expect more responsibility assigned to them, and expect more anger felt by the victim, compared to when the outcome is not severe. Based on the expected reaction from the victim, offenders will choose an account to explain their wrong-doings (i.e., the more expected anger felt by victims, the less likely a defensive account would be chosen). The driving force for this study came from the discrepancy between the significant role of accounts in conflict management and the research that has been done to explore this area. When looking at a simple interpersonal conflict (e.g., the vignette at the beginning of the current research), explaining one’s behavior constitutes a major part of the conflict process. Offenders’ explanation of their behaviors actually lay the foundation for how a conflict will be resolved. Offenders’ accounts directly influence victims’ perceptions of the causes of the conflict, the offenders’ intentions, the amount of responsibility that can be assigned to offenders, and in general, the way the conflict can be handled (Sitkin & Bies, 1993). Although accounts play an important role in conflict management, relatively little research has explored this simple way of handling conflict (e.g., McLaughlin, O’Hair et al., 1983). The current research, one of few studies examining how offenders select their accounts, adds to the understanding of accountdevelopment processes.
136 The current research contributes to the literatures of conflict management and communication in general because it provides a theoretical model for understanding the account-giving processes via an attributional perspective. The current study focused on interpersonal conflict; however, the results can be cautiously applied to other contexts, such as organizational communication. In addition, the model can be revised and evaluated in diverse ethnicities and cultures, contributing to the cross-cultural communication literature. The current research has not explored the account-giving process in different cultural contexts; but the discussion of face concerns and attributions points to possible cross-cultural studies on accounts in the future. The current research contributes to the account literatures in three different aspects. First, it changes the focus of account studies from typology development to understanding the account-selecting processes. For a long time, McLaughlin et al. (1992) have promoted greater understanding of how people create accounts, in addition to developing different types of accounts. Researchers have been looking into this issue but not many studies have been generated. The current research answers the call to examine one account-making process and provides important information about how offenders determine the accounts used in a conflict situation. Second, the current research took an approach to examine accounts that was different from past research. Traditionally, participants are asked to self report a negative event and provide the actual accounts they used at that time. After that, trained coders code the accounts into different categories. Although this method may have more external validity, it limits researchers’ ability to examine specific antecedent conditions of interest. The current research employed an experimental design to study the variables
137 of interest and, instead of asking participants to report their accounts, provided a list of accounts to which they responded. This technique is easy to administer and can be adapted to different conflict situations. Third, researchers have been mapping types of accounts on different continua; for example, Schönbach (1990) used a defensiveness continuum to describe concessions, justifications, excuses, and refusals. Schönbach argued that concessions are least defensive, followed by justifications, and excuses are more defensive than justifications but less defensive than refusals. However, Schönbach did not provide any empirical evidence for this proposition. One pilot study for the current research actually tested the defensiveness levels of these four types of accounts and found that concessions, indeed, are the least defensive account whereas refusals are the most defensive. The defensiveness of justifications and excuses are significantly different from each other and justifications are more defensive than excuses, which is opposite to Schönbach’s original proposition. Whether Schönbach’s proposition is wrong cannot be determined by results from a single study. The current research provides certain empirical findings on the defensiveness of each type of accounts but more studies are needed for an affirmative answer in the future. Different types of accounts have their own features. When used in an interpersonal conflict, each type of accounts reflects different expectations offenders have about victims’ reactions and possibly implies different goals offenders try to achieve. The current research found that offenders will take a less defensive stand to explain their behavior when they expect the victim to be angry and to assign more responsibility to them, in another word, when offenders expect the victim to be tough. However, when
138 offenders expect the victim to be soft, that is to say, offenders do not expect the victim to be so angry and to assign much responsibility to them, offenders then will play hard. A Chinese idiom describes it well, qi ruan pa ying (bully the soft ones but be afraid of the tough ones). It’s in the human nature.
139 Appendix A Instruments for Manipulation Checks in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Perceived Outcome Severity 1. To what extent does your mistake influence Person B? 2. How severe is the outcome of this situation for Person B? 3. How severely does your mistake affect Person B? ______ ______ ______
4. How much influence does your mistake have on Person B’s scholarship for the next year? 5. How much difficulty does your mistake have on Person B’s obtaining his or her scholarship for the next year? 6. To what extent do you think the outcome is severe? Perceived Relational Closeness 1. To what extent do you think Person B is a good friend of yours? 2. To what extent do you think you and Person B are close to each other? 3. How much knowledge do you have of Person B? 4. How familiar are you with Person B? 5. How close are you with Person B? 6. How well do you know Person B? ______ ______ ______ ______ ______ ______ ______ ______ ______
Note. The labels (i.e., perceived outcome severity and perceived relational closeness) are for readers’ convenience. They were not shown in the actual questionnaire. The same idea was followed for all the items listed below in Appendices B, C, D, and E.
140 Appendix B Instruments for Attributions in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Degree of Internal Attributions 1. How likely is that what has happened was due to your own carelessness? 2. How likely is that what has happened was due to internal traits, such as your personality? ______ ______
3. How likely is that what has happened was due to you not paying enough attention when the readings were assigned? ______
4. How likely is that what has happened was because you did not care about what grade you and Person B could get for this assignment? ______
5. To what extent do you believe what has happened was because you did not spend enough time on the assignment? 6. To what extent do you think what has happened was because of you? ______ ______
7. To what extent do you think your personality determines what has happened? ______ Controllability 1. How likely is that what has happened was avoidable? 2. How much control do you have over what has happened? 3. How likely is it that you could have prevented this mistake? ______ ______ ______
4. To what extent do you believe you could have done otherwise to prevent what has happened? 5. To what extent do you think you could have done differently to avoid what has ______
141 happened? Degree of External Attributions 1. How likely is that what has happened was due to the professor not giving clear instructions for the assignment? ______ ______
2. How likely is that what has happened was caused by other people instead of you? ______ 3. How likely is that what has happened was because the wordings in the instruction were confusing? ______
4. How likely is that what has happened was due to some external reason (for example, someone mistakenly took your notes away so you had nothing to rely for the instructions for the project)? ______
5. To what extent do you think what has happened was due to Person B’s decision to meet the night before the due date? ______
6. To what extent do you think what has happened was because of other people? ______ 7. To what extent do you think environmental reasons determine what has happened? ______ Uncontrollability 1. How likely is that what happened was out of your control? ______
2. To what extent do you believe there is not much you could do to prevent what has happened? 3. How likely is that you have no control over what has happened? 4. To what extent do you think you cannot prevent what has happened? 5. To what extent do you think what has happened was inevitable? ______ ______ ______ ______
142 Note. Based on results from confirmatory factor analysis and principal components analysis, not all seven items that measured the degree of internal attribution were included in the model that was tested. The four items used in the model that was tested were item 1, item 3, item 5, and item 6.
143 Appendix C Instruments for Anticipated Consequences in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Expected Responsibility 1. How much responsibility would Person B assign to you for what has happened? ______ 2. To what extent would Person B think you are responsible for what has happened? ______ 3. To what extent would Person B think you are chargeable for what has happened? ______ 4. To what extent would Person B think you are accountable for what has happened? ______ 5. To what extent would Person B think you should take full responsibility for what has happened? ______ 6. To what extent would Person B think what has happened should be attributed to you? ______ Expected Anger 1. How angry do you expect Person B to be? 2. How irritated do you expect Person B to be? 3. How furious do you expect Person B to be? 4. How mad do you expect Person B to be? ______ ______ ______ ______
144 5. How fuming do you expect Person B to be? 6. How annoyed do you expect Person B to be? ______ ______
145 Appendix D Instruments for the Likelihood of Selecting an Account in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Concessions Concession 1. I am sorry that I misunderstood the instructions. Is there any way that I can make it up to you? Concession 2. I am sorry that I misunderstood the instruction. I will take the full responsibility and it’s all my fault. Concession 3. I am sorry that I misunderstood the instruction. I will talk with the professor to see whether we can come up with a possible solution. ______ ______ ______
Concession 4. I am sorry that I misunderstood the instruction. I hope you know that I am shocked about this grade and feel horrible about bringing your grade down. ______
Concession 5. I am sorry that I misunderstood the instruction. Could I possibly write a letter to complement your scholarship application to explain what happened? Justifications Justification 1. I am sorry that the grade influenced your scholarship. But I honestly thought I was doing the correct work. ______ ______
Justification 2. I am sorry that the grade influenced your scholarship. You know that I am a good student and I would never want this grade myself. It’s a complete misunderstanding. Justification 3. I am sorry that the grade influenced your scholarship. I knew how important the grade was to you but I did try my best to get a good grade. ______ ______
146 Justification 4. I am sorry that the grade influenced your scholarship. But I worked very hard on this assignment. We should have gotten together to edit the paper earlier. ______ Justification 5. I am sorry that the grade influenced your scholarship. But I thought I did on the correct readings and I did a thorough job on the readings that I reported on. ______ Excuses Excuse 1. I am sorry that I misunderstood the instruction. But the wording in the direction was really confusing and that may be the reason why I wrote the wrong portion of the paper. ______
Excuse 2. I am sorry that I misunderstood the instruction. I was being careless and confused with all the paper work. ______
Excuse 3. I am sorry that I misunderstood the instruction. It’s just ……there were other things that contributed to the mistake. ______
Excuse 4. I am sorry that I misunderstood the instruction. I was so nervous about this course and tried to do well on this assignment; but maybe my nervousness led me to write the wrong portion of the paper. ______
Excuse 5. I am sorry that I misunderstood the instruction. My sister accidently took my notes away and I did not have anything to rely on to write the paper. I thought I remembered the instruction correctly. Refusals Refusal 1. I know the readings I read were the right material for the paper. I did not find anything wrong to what I contributed to the paper. ______ ______
Refusal 2. I have to say I did more work than you and I don’t think the grading is fair. ______
147 Refusal 3. If this course means a lot to you, you should have put more focus on the collaboration. ______
Refusal 4. You made me do all the work and not call me on what the paper should be about. ______
Refusal 5. I told you we should not wait until the last minute. We deserve the grade we got. It’s all your fault not contributing to the paper and we got a poor grade. ______
148 Appendix E Pilot Study 1: Questionnaire for Collecting Conflict Scenarios CONFLICT STRATEGY STUDY Department of Communication University of Maryland, College Park Spring, 2007 PLEASE READ CAREFULLY In this study, we are interested in learning what you say when you have to account for yourselves: That is, what you say when you feel that, for one reason or another, you have to answer to someone else because of something you did or failed to do, which resulted into a conflict between you and another person. For example, you might have to account to your girlfriend or boyfriend for the fact that when you used his or her laptop to surf online, the web page you visited contained virus that wiped out all his or her important documents; or your classmate might expect you to account for your failure to return the lecture notes to her, which left her little time to prepare for the final and she got an overall low grade for that class. Sometimes we give accounts even if no one asks for one, because we think that our behavior needs explanations, or because we think that the other person thinks that our behavior was questionable. For example, you were responsible for making powerpoint slides for a group presentation. In the middle of the presentation, one of your slides accidentally made weird noises from the animation you forgot to delete. Although your group members did not say anything, you felt you ought to say something to explain that weird noise. So, we would like you to recall a recent conflict you had with another person resulted from something you did, and answer the following questions on the next page. Please read all instructions carefully before you answer each question. At the very end, you will have an opportunity to comment on our study. Please do not leave any questions out. This would cause that we have to exclude your answers from our study. Please note that there are no correct or incorrect answers. We are simply interested in what your account was in the situation you recall.
149 1. In the table, please recall a situation in which a conflict occurred recently between you and another person because of something YOU did. You should know that conflicts do not always have satisfactory endings. So you can describe any conflict you actually had due to something you did. Make sure to describe the occurrence in detail. When did the conflict happen (approximately)? Relationship between you and the other person in the conflict. What happened (please describe the incidence in detail, from the start of the conflict to its solutions)? To your knowledge, what was/were the cause(s) of the conflict (please list all reasons that you can think of)? To your knowledge, what did you expect the other person would think the cause(s) was/were (please all reasons that you expect the other person would think of)? What was your emotion during the conflict (e.g., did you feel angry)? Please list all emotions you experienced at the moment of the conflict. What was your expectation of the other person’s emotion at the time of conflict (e.g., did you expect the other person feel angry)? Please list all emotions you expected the other person might have experienced at the moment of the conflict.
150
2. When the conflict described in question 1 happened, did the other person say anything that made you feel you ought to explain your behaviors to the other person?
If there was a verbalized request for an explanation, Please recall as best as you can the actual words used by the other person
Yes. He/She said: “_________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ _________________________________________________.”
If no verbalized request for an explanation was presented, please recall as best as you can why you thought, at that moment, an explanation was necessary.
No. But I thought I should explain my behavior because __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ _________________________________________________.
151 3. Please recall, in as close to the exact words as possible, what you said to explain your behavior to the other person in the conflict described in question 1. To explain my behavior, I said: “____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ ___________________________________________________________________.”
4. Are you satisfied with the endings of the conflict mentioned in question 1 (circle one): YES NO
5. If given a second chance, what else would you say other than the explanation you provided above? To explain my behavior, I would have said: “____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ ___________________________________________________________________.” We would also like to know a little more about you. Please complete the following information. 1. What is your gender? (circle one) MALE FEMALE
2. What is your age? I am ________ years old.
152 3. Which race/ethnicity label describes you best? (circle one) African-American / Black Asian-American / Asian Caucasian / White Hispanic / Latino Native American Other (Please specify): _____________
4. What is your nationality (which country issues you a passport)? _____________ 5. Is English your native (first) language? (circle one) YES NO
If not, what is your native language? _____________________ 6. What is your major in college? ____________ 7. What year are you in college? (circle one) FRESHMAN JUNIOR GRADUATE SOPHOMORE SENIOR OTHER (Please specify): ___________________
Do you have any comments regarding our study? _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ Thank you for completing all the questions. We appreciate your participation!
153 Appendix F Pilot Study 2: Questionnaire for Sorting Task Sorting out Conflict Information Department of Communication University of Maryland, College Park August, 2007 Purpose: The study contains information from another study. You will read about different conflicts. Your task is to sort those conflicts according to the severity of the conflict (severe vs. not severe) and the closeness of the relationship between the people in that conflict (not close at all, somewhat close, very close). Remember, there is no right or wrong way to sort those scenarios. You are sorting the scenarios based on your own judgment.
Instructions: You will receive 93 numbered index cards. On each card there is one conflict. Some of them are described in detail and some are less detailed. Step 1: Please read over all 93 conflicts carefully, getting familiar with them and making sure you understand each of them. If you have any questions, please feel free to ask the researcher.
154 Step 2: After you get familiar with these 93 conflicts, please sort them into the sorting table on the next page (Table 1). You will sort these conflicts based on two criteria: the severity of the conflict and closeness of the relationship between the participants to the conflict. *Severity of the conflict refers to how serious the consequence might be because of the conflict mentioned on the card. *Closeness of the Relationship refers to the extent that the people in the conflict are close to each other. Please put the number of the card in the most appropriate cell. And remember: Each conflict can be put in ONE AND ONLY ONE CELL in the table. If you are not sure which cell to put a conflict into, please write down the card number in Table 2 (Other) and explain the reason why you cannot determine which cell the conflict should be in.
155 Sorting Table (Table 1)
Severity of the Conflict
Severe
Not severe
Closeness of the Relationship
Not close at all
Somewhat close
Very close
156
Other (Table 2) Card Number Please specify the reason why you cannot determine the cell from Table 1 you should put this scenario into.
157
Step 3: Now you have finished sorting all 93 conflicts. Before we move to the next step, do you want to take a final look at your sorting? Do you want to change any sorting?
158 Step 4: We would also like to know a little more about you. Please complete the following information. 1. What is your gender? (circle one) MALE FEMALE
2. What is your age? I am ________ years old. 3. Which race/ethnicity label describes you best? (circle one) African-American / Black Asian-American / Asian Caucasian / White Hispanic / Latino Native American Other (Please specify): _____________
4. What is your nationality or citizenship? _____________ 5. Is English your native (first) language? (circle one) YES NO
If not, what is your native language? _____________________ 6. What is your major in college? ____________ 7. What year are you in college? (circle one) FRESHMAN JUNIOR GRADUATE SOPHOMORE SENIOR OTHER (Please specify): ___________________
8. Please provide your SONA ID here: ____________________________
159 Appendix G Conflict Incidents Reported in Pilot Study 1 How sorters sorted the 93 conflicts reported in Pilot Study 1 was examined and the following five incidents were applicable in all six experimental conditions. The first incident was selected to develop the hypothetical conflict scenario used in the formal study because the conflict occurred in a context familiar to future participants. Incident 1: “There were five students including me that decided to form a virtual study group. I was not in class the day that everyone decided who would outline what chapters. I was given the wrong chapters so I did the same work as another student!” Incident 2: “Basically my roommate was angry at me because I didn't ‘completely’ wake him up to study for his finals exam. I believe it was him that caused the conflict because he made a huge deal about it. I came back at 12pm and I woke him up then but he was angry.” Incident 3: “Even though my boyfriend told me he was coming to pick me up, I wasn’t ready and we missed the movie that he had really wanted to see. He was mad because he had given me over an hour to get ready and I had been doing other things instead. I apologized many times.” Incident 4: “I copied my work schedule down incorrectly, so I showed up to work an hour late. I ended up showing up late, excused by my manager, but I felt like I'd disappointed him.” Incident 5: “It happened about two weeks ago. I was sitting directly opposite to the air conditioner in the office and I was freezing to death. So I reset the temperature. However, one of my officemates immediately stood up and turned the AC temperature
160 back, which upset me. I said that I could not handle the room that cold, and he pretended not to hear my complaints.”
161 Appendix H Hypothetical Conflict Scenarios Used in the Formal Study 1. Scenario for the condition of close relationship and high outcome severity: You and Person B grew up together and attended the same university. The two of you have been best friends for many years. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was very upset because such a low grade resulted in him or her not getting next year's scholarship.
2. Scenario for the condition of close relationship and low outcome severity: You and Person B grew up together and attended the same university. The two of you have been best friends for many years. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 10% of the overall grade. Both group members would receive the same grade for this assignment.
162 The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be influenced. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 70%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was not happy but he or she did great on other assignments and believed it would not influence his or her getting next year's scholarship.
3. Scenario for the condition of somewhat close relationship and high outcome severity: You and Person B were in couple of classes together before. Although the two of you know each other, you only see each other in class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was very upset because
163 such a low grade resulted in him or her not getting next year's scholarship.
4. Scenario for the condition of somewhat close relationship and low outcome severity: You and Person B were in couple of classes together before. Although the two of you know each other, you only see each other in class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 10% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be influenced. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 70%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was not happy but he or she did great on other assignments and believed it would not influence his or her getting next year's scholarship.
5. Scenario for the condition of not–close-at–all relationship and high outcome severity: You and Person B do not really know each other. You happened to sit next to each other on the first day of class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade.
164 Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was very upset because such a low grade resulted in him or her not getting next year's scholarship.
6. Scenario for the condition of not-close-at–all relationship and low outcome severity: You and Person B do not really know each other. You happened to sit next to each other on the first day of class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 10% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be influenced. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 70%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was not happy but he or she did great
165 on other assignments and believed it would not influence his or her getting next year's scholarship.
166 Appendix I Pilot Study 3: Questionnaire for Manipulation Checks Scenario Assessment Department of Communication University of Maryland, College Park Spring, 2008 Note. The same instructions were used in Pilot Studies 3, 5, 6, and the formal study. Researchers at the University of Maryland are interested in knowing how people explain themselves in interactions. What do you need to do in this study? 1. Read a hypothetical incident. There are two persons (Person A and Person B) involved in the incident. Please image you were Person A. 2. Provide possible explanations by imagining you were Person A. 3. Answer some questions regarding the incident. Ready? Start!
*Part I: Situation * You and Person B grew up together and attended the same university. The two of you have been best friends for many years. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong
167 Part II: Providing explanations Imagine that you were the one that worked with Person B on this group project. What would you say to Person B to explain what you did (i.e., you wrote your portion of the review paper on wrong readings, which contributed to receiving a 40% for the review paper)?
168 Part III: Questions about the Situation Still imagine yourself in the situation. Read the instructions and answer the following questions. Please use the following instructions to answer the questions: Use a number from 0 (zero) to
infinity to respond to each question. Zero means not at all and 100 means a moderate amount; you may use numbers from zero to infinity but please use a specific number. For example, suppose you were asked the question “how much do you like chocolate?” If you like chocolate a moderate amount, then you would rate your answer as 100. If you like chocolate only a little, then you may rate your answer as 20; but if you really like chocolate, you may rate your answer as 300. For each question, you would rate your answer using any numbers based on the following scale: 0 = not at all 100 = moderate amount Use any number from zero on up
To ensure you understand how to use the scale, first practice using two questions that are not relevant to the situation you just read. Question 1. How much do you like the food offered at the dining hall of your school? 2. How much do you like the service at the main library of your school? Scale
0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up
Your Rating
When you feel ready to use the scale you just practiced, turn to the next page. Please remember: Answer questions by IMAGINING the incident happened between you and Person B.
169 Question 1. How severe is the outcome of this situation? 2. How severely did your mistake affect Person B? 3. How much influence would your mistake have on Person B’s fellowship application? 4. How much difficulty would your mistake have on Person B’s ability to achieve his or her goals? 5. How close are you with Person B? 6. How much knowledge do you have of Person B? 7. How familiar are you with Person B? 8. How well do you know Person B? Scale
0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up
Your Rating
1. My age is ____________ years. 2. I am (circle one) FEMALE MALE
3. I am a citizen of _______________________ (country). 4. My ethnicity is or most closely to African-American / Black Asian-American / Asian Caucasian / White 5. My native language is (circle one) Hispanic / Latino Native American Other (Please specify): _____________ ENGLISH Other (please specify): ______________________
6. I am a
(circle one) FRESHMAN JUNIOR SOPHOMORE SENIOR
170 GRADUATE STUDENT OTHER (please specify): ___________________.
171 Appendix J Instruments for Defensiveness of Accounts in Pilot Study 5 Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Defensiveness 1. To what extent were you being self-protective? 2. To what extent were you being guarded? 3. To what extent were you sticking out for yourself? 4. To what extent were you being assertive? 5. To what extent were you trying to shift responsibility? 6. To what extent were you being defensive? ______ ______ ______ ______ ______ ______
172 Appendix K List of Open-Ended Questions Used in Pilot Study 6 Note. The purpose of using the following questions was to obtain comments from participants regarding the clarity of the questionnaire that measured all dependent variables. 1. Could you describe your general impression of the questionnaire? 2. What do you think the purpose of the study is? 3. What do you think the researchers want to measure? 4. Do you feel the questions are clear? 5. Are there any questions you found confusing or difficult to answer? Please explain. 6. Do you have any other comments on the study?
173 Appendix L Descriptives of the Indicators in the Formal Study after Transformation
Indicators SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 REL1 REL2 REL3 REL4 REL5 REL6 IA1 IA2 IA3 IA4 IA5 IA6 IA7 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RES1 RES2
Mean 6.59 6.63 6.56 6.31 6.04 6.00 4.88 4.67 4.99 5.06 4.73 4.91 10.87 5.85 10.35 2.53 7.23 10.42 4.85 6.84 6.09 6.77 6.13 5.93 11.81 6.04 12.48 6.77 14.13 6.88 5.95 2.79 3.35 2.93 3.10 2.47 4.83 4.91
SD 1.50 2.06 1.84 2.54 2.06 1.91 2.32 2.45 2.12 2.08 2.36 2.19 3.45 3.58 3.49 3.01 3.44 3.38 3.61 2.10 2.33 1.87 1.70 1.46 6.02 4.91 5.66 4.78 7.32 4.52 4.70 2.45 2.35 2.58 2.55 2.51 1.26 1.21
Range Trim Value ? 8.81 9.06 7.66 11.22 9.49 8.80 8.80 8.80 8.80 8.80 8.80 8.80 16.39 10.85 16.39 7.94 13.02 16.68 10.85 10.47 10.47 6.69 7.52 8.27 23.04 12.59 23.04 12.59 27.72 12.59 12.59 6.31 6.31 6.41 6.31 6.31 6.92 6.92 501 900 700 1000 620 500 500 500 500 500 500 500 500 200 500 100 300 520 200 1000 1000 1000 800 500 300 100 300 100 420 100 100 100 100 100 104 100 1000 1000 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.34 0.34 0.34 0.34 0.34 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.40 0.40 0.40 0.40 0.40 0.28 0.28
Skewness -0.23 0.52 0.32 0.31 -0.16 -0.31 -0.18 -0.09 0.22 0.22 0.07 0.26 -0.14 -0.40 0.15 0.61 -0.22 0.30 -0.05 0.37 -0.01 0.84 0.67 -0.26 0.11 -0.03 0.31 -0.27 0.26 -0.26 0.01 0.03 -0.37 0.01 -0.10 0.30 0.04 -0.03
Standard Error 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16
174
Indicators RES3 RES4 RES5 RES6 ANG1 ANG2 ANG3 ANG4 ANG5 ANG6 Concession1 Concession2 Concession3 Concession4 Concession5 Justification1 Justification2 Justification3 Justification4 Justification5 Excuse1 Excuse2 Excuse3 Excuse4 Excuse5 Refusal1 Refusal2 Refusal3 Refusal4 Refusal5
Mean 4.86 4.90 4.60 4.75 4.96 4.95 4.59 4.88 4.44 5.06 10.39 9.41 13.07 10.42 8.28 9.10 9.22 7.69 8.20 8.12 6.99 5.86 3.77 3.56 2.25 1.89 1.30 2.74 1.50 1.85
SD 1.20 1.20 1.33 1.30 1.21 1.17 1.40 1.23 1.43 1.16 3.96 4.07 4.72 3.87 4.61 4.26 4.34 4.06 4.65 4.51 3.70 3.63 3.31 3.19 2.88 2.75 2.20 3.16 2.35 2.73
Range Trim Value ? 6.92 6.92 6.92 6.92 5.46 7.19 7.19 7.41 6.97 7.19 16.39 16.39 18.54 16.39 16.39 16.39 16.39 14.21 16.39 16.39 13.02 10.85 8.64 7.98 7.75 7.10 5.81 7.94 5.85 7.18 1000 1000 1000 1000 1000 900 900 1000 810 900 500 500 1000 500 500 500 500 364 500 500 300 200 120.5 100 94.6 80.1 50 100 50.7 80 0.28 0.28 0.28 0.28 0.29 0.29 0.29 0.29 0.29 0.29 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45
Skewness 0.19 0.23 -0.16 -0.16 0.70 0.26 -0.08 0.43 -0.28 0.09 -0.16 0.09 0.76 -0.12 0.08 0.07 -0.08 -0.23 0.06 0.08 -0.20 -0.42 0.04 0.08 0.77 0.95 1.27 0.57 1.08 1.03
Standard Error 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16
175 [See the table below for item label used above]. Indicator Label SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 REL1 REL2 REL3 REL4 REL5 REL6 IA1 IA2 IA3 IA4 IA5 IA6 IA7 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RES1 RES2 RES3 RES4 RES5 RES6 Description Perceived Outcome Severity Item 1 Perceived Outcome Severity Item 2 Perceived Outcome Severity Item 3 Perceived Outcome Severity Item 4 Perceived Outcome Severity Item 5 Perceived Outcome Severity Item 6 Perceived Relational Closeness Item 1 Perceived Relational Closeness Item 2 Perceived Relational Closeness Item 3 Perceived Relational Closeness Item 4 Perceived Relational Closeness Item 5 Perceived Relational Closeness Item 6 Internal Attribution Item 1 Internal Attribution Item 2 Internal Attribution Item 3 Internal Attribution Item 4 Internal Attribution Item 5 Internal Attribution Item 6 Internal Attribution Item 7 Controllability Item 1 Controllability Item 2 Controllability Item 3 Controllability Item 4 Controllability Item 5 External Attribution Item 1 External Attribution Item 2 External Attribution Item 3 External Attribution Item 4 External Attribution Item 5 External Attribution Item 6 External Attribution Item 7 Uncontrollability Item 1 Uncontrollability Item 2 Uncontrollability Item 3 Uncontrollability Item 4 Uncontrollability Item 5 Expected Responsibility Item 1 Expected Responsibility Item 2 Expected Responsibility Item 3 Expected Responsibility Item 4 Expected Responsibility Item 5 Expected Responsibility Item 6
176 Indicator Label ANG1 ANG2 ANG3 ANG4 ANG5 ANG6 Concession1 Concession2 Concession3 Concession4 Concession5 Justification1 Justification2 Justification3 Justification4 Justification5 Excuse1 Excuse2 Excuse3 Excuse4 Excuse5 Refusal1 Refusal2 Refusal3 Refusal4 Refusal5 Description Expected Anger Item 1 Expected Anger Item 2 Expected Anger Item 3 Expected Anger Item 4 Expected Anger Item 5 Expected Anger Item 6 Concession 1 Concession 2 Concession 3 Concession 4 Concession 5 Justification 1 Justification 2 Justification 3 Justification 4 Justification 5 Excuse 1 Excuse 2 Excuse 3 Excuse 4 Excuse 5 Refusal 1 Refusal 2 Refusal 3 Refusal 4 Refusal 5
177 Appendix M Covariance Matrix in LISREL for the Proposed Model in Figure 2 Covariance Matrix Item SEV1 -------2.24 2.09 2.04 2.38 2.04 1.90 0.38 2.12 2.26 0.61 1.89 1.30 1.05 1.46 1.11 0.81 2.49 1.26 2.75 0.45 2.82 1.15 0.17 0.44 0.30 0.64 0.65 0.24 0.95 0.93 0.90 0.89 0.95 1.00 1.09 1.00 1.16 SEV2 -------SEV3 -------SEV4 -------SEV5 -------SEV6 --------
SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3
4.27 3.39 4.30 3.40 3.37 0.49 2.62 2.42 0.36 1.69 1.70 1.27 1.76 1.26 1.01 3.39 1.20 2.99 0.40 5.51 1.63 0.24 0.30 0.08 0.68 0.85 0.22 1.38 1.30 1.32 1.31 1.31 1.47 1.81 1.56 2.01
3.40 3.73 3.06 2.92 0.41 2.26 2.15 0.42 1.79 1.48 1.14 1.67 1.05 0.85 3.20 0.95 3.17 0.76 4.84 1.45 0.75 0.30 0.33 0.69 0.83 0.34 1.24 1.18 1.15 1.17 1.27 1.29 1.58 1.39 1.81
6.51 4.51 3.69 0.61 3.36 3.05 1.04 1.92 2.27 1.64 2.18 1.47 1.20 4.16 1.96 3.22 0.80 6.96 2.21 0.16 0.79 0.49 0.95 1.13 0.60 1.62 1.55 1.55 1.50 1.25 1.58 2.17 1.90 2.43
4.26 3.18 0.44 2.79 2.57 0.71 2.08 1.74 1.19 1.67 1.17 1.08 3.72 1.36 3.22 0.86 4.42 1.81 0.49 0.62 0.45 0.99 0.97 0.56 1.22 1.19 1.22 1.20 1.15 1.30 1.60 1.40 1.82
3.66 0.34 2.45 2.23 0.75 2.06 1.65 1.20 1.59 1.17 1.02 2.78 1.25 2.66 0.54 3.83 1.32 0.65 0.39 0.29 0.76 0.78 0.36 1.17 1.10 1.13 1.11 1.17 1.25 1.53 1.31 1.71
178 Item SEV1 -------1.18 1.09 1.06 -0.06 0.63 0.03 -0.03 0.13 0.09 SEV2 -------1.87 1.91 1.54 -0.02 1.36 0.02 -0.07 0.24 0.25 SEV3 -------1.74 1.72 1.34 -0.02 1.13 0.09 -0.01 0.21 0.19 SEV4 -------2.23 2.31 1.82 0.03 1.47 0.05 -0.13 0.25 0.32 SEV5 -------1.73 1.76 1.40 -0.03 1.22 0.12 -0.02 0.23 0.23 SEV6 -------1.58 1.66 1.33 -0.04 1.18 0.13 -0.10 0.20 0.24
ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
Covariance Matrix Item QREL -------1.42 0.70 0.38 0.14 0.64 0.52 0.00 0.37 0.18 0.04 0.44 -0.26 0.32 -0.52 -0.21 -0.57 -0.88 0.06 -0.29 -0.19 -0.19 -0.06 0.37 0.33 0.30 0.30 IA1 -------IA2 -------IA3 -------IA4 -------CON1 --------
QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4
12.04 6.69 2.77 6.84 3.94 3.19 3.77 3.22 2.70 7.06 0.42 2.88 -0.35 2.31 0.74 0.77 0.06 0.19 0.18 0.21 -0.30 2.11 2.02 1.96 1.99
12.26 4.05 6.25 3.93 2.53 3.96 3.29 2.72 7.87 1.06 6.44 0.94 3.38 0.99 0.81 0.32 0.50 0.21 0.23 -0.09 1.98 2.12 1.97 1.98
11.88 2.11 1.91 1.59 1.15 1.08 1.62 3.59 0.72 1.99 1.08 5.34 1.44 2.03 0.46 0.89 0.95 1.11 0.44 0.07 0.17 0.05 0.12
11.40 3.71 2.75 3.91 3.39 2.85 3.77 -0.41 3.33 -1.08 0.62 -0.59 0.93 -0.06 -0.19 -0.29 -0.07 0.33 1.96 2.05 1.95 2.00
4.41 2.34 2.71 1.90 1.67 2.68 0.59 1.70 -1.17 2.18 0.46 -0.16 -0.32 -0.26 -0.22 -0.24 -0.26 1.21 1.25 1.27 1.20
179 Item QREL -------0.23 0.25 0.28 0.23 0.17 0.20 0.15 0.14 -0.03 0.12 0.61 0.02 0.14 0.03 IA1 -------1.46 1.85 2.08 1.98 1.84 2.01 1.76 2.11 -0.45 0.09 0.59 -0.01 0.16 0.03 IA2 -------2.12 2.11 2.12 2.15 2.26 2.03 2.09 2.45 -0.42 0.26 0.38 0.02 0.12 0.04 IA3 -------0.09 0.15 0.33 0.60 0.29 0.25 0.29 0.76 -0.11 -0.06 0.40 0.29 0.09 -0.06 IA4 -------2.03 2.15 1.82 1.81 1.67 1.78 1.66 1.68 -0.66 -0.05 0.73 0.11 0.16 -0.01 CON1 -------0.89 1.20 1.33 1.25 1.12 1.28 0.97 1.17 -0.26 0.12 0.18 -0.10 0.07 0.08
RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
COVARIANCE MATRIX Item CON2 -------5.46 2.21 0.85 1.17 4.31 2.11 1.64 0.18 2.60 0.39 0.03 0.15 -0.36 -0.84 -0.70 -0.73 1.07 1.07 1.02 1.04 0.73 1.10 CON3 -------CON4 -------CON5 -------EA1 -------EA2 --------
CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6
3.51 2.34 1.72 2.19 0.20 2.47 -0.90 3.59 0.17 -0.01 -0.45 -0.61 -0.52 -0.34 -0.39 1.49 1.56 1.46 1.47 1.18 1.51
2.88 1.67 2.12 -0.31 2.46 -0.85 2.98 0.11 -0.39 -0.27 -0.68 -0.60 -0.54 -0.39 1.04 1.10 0.98 0.99 0.75 1.03
2.13 1.26 0.20 1.19 -0.78 2.53 0.34 0.17 -0.16 -0.34 -0.11 -0.20 -0.05 0.71 0.76 0.75 0.70 0.61 0.83
36.06 13.83 17.51 8.89 13.84 8.25 6.66 5.75 4.78 4.11 4.45 4.17 1.82 2.11 1.97 1.97 1.56 1.76
24.04 6.77 6.27 9.71 12.28 7.94 6.81 4.61 3.85 3.95 3.82 0.73 0.47 0.68 0.43 0.92 0.71
180 Item CON2 -------1.02 0.96 1.07 1.06 1.15 0.98 -0.20 0.02 0.08 -0.02 0.04 0.03 CON3 -------1.50 1.42 1.47 1.50 1.36 1.37 -0.17 0.12 -0.08 -0.06 0.04 0.03 CON4 -------1.12 1.07 1.12 1.12 1.02 1.20 -0.16 0.12 -0.09 0.01 0.06 0.01 CON5 -------0.86 0.85 0.80 0.81 0.80 0.87 -0.16 0.06 -0.03 -0.03 0.01 0.04 EA1 -------1.71 1.95 2.23 1.95 2.04 2.59 0.07 0.55 0.57 0.74 0.50 -0.25 EA2 -------0.56 0.74 0.67 0.52 0.81 1.03 0.28 0.24 -0.28 0.53 0.12 -0.15
ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
Covariance Matrix Item EA3 -------31.79 6.94 9.49 8.77 3.69 4.59 3.39 3.66 4.01 2.99 1.72 1.90 1.70 1.66 1.88 1.56 1.76 1.97 1.94 1.84 1.64 2.38 0.19 0.85 0.23 EA4 -------EA5 -------EA6 -------EA7 -------UNCON1 --------
EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC
22.90 6.28 7.47 5.58 2.65 4.36 3.57 3.74 3.16 -0.35 -0.06 -0.03 -0.05 0.11 -0.66 -0.45 -0.16 -0.23 -0.37 -0.15 -0.01 0.14 0.42 -0.16
53.65 11.51 5.61 1.81 2.32 5.47 5.16 3.75 1.71 1.85 1.66 1.56 1.09 1.28 2.56 2.72 3.57 3.02 3.32 3.27 1.18 1.64 -0.42
20.49 6.76 3.48 3.33 4.67 5.20 3.50 0.76 0.71 0.83 0.62 0.58 0.54 0.67 0.67 0.76 0.70 0.81 1.17 0.41 0.57 -0.49
22.16 3.20 4.32 3.47 4.64 2.99 0.11 -0.08 -0.05 -0.15 0.08 0.03 0.06 0.43 0.44 0.06 0.32 0.56 0.03 -0.16 -0.08
5.98 2.97 2.92 3.04 2.66 0.03 -0.09 -0.02 -0.07 0.05 0.00 0.15 0.28 0.15 0.14 0.23 0.24 0.07 0.18 -0.02
181 Item EA3 -------0.08 0.36 -0.13 EA4 -------0.28 -0.04 -0.05 EA5 -------0.29 0.24 0.35 EA6 -------0.22 0.15 -0.05 EA7 -------0.42 0.10 -0.17 UNCON1 -------0.22 0.09 -0.03
INTER1 INTER2 INTER3
Covariance Matrix Item UNCON2 UNCON3 UNCON4 UNCON5 RESP1 -----------------------------------5.53 2.99 3.22 2.89 -0.07 -0.17 -0.11 -0.15 0.02 -0.33 -0.04 0.19 0.19 -0.05 0.17 0.09 0.18 0.23 0.20 0.09 -0.03 RESP2 --------
UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER6 LDA OS RC INTER1 INTER2 INTER3
6.69 5.07 3.42 -0.11 -0.11 -0.08 -0.15 0.11 -0.18 0.14 0.20 0.28 0.20 0.23 0.12 0.31 -0.01 0.19 0.10 0.00
6.48 3.67 -0.02 -0.14 -0.05 -0.14 0.22 -0.06 0.20 0.23 0.39 0.26 0.26 0.14 0.40 0.09 0.20 0.13 0.03
6.30 -0.08 -0.18 -0.13 -0.16 0.17 -0.25 0.03 0.24 0.25 0.16 0.20 0.19 0.16 -0.15 0.28 0.10 -0.04
1.61 1.36 1.42 1.39 1.20 1.40 1.16 1.07 1.09 1.01 1.01 -0.08 0.21 0.03 -0.06 0.03 0.06
1.47 1.35 1.34 1.18 1.34 1.14 1.07 1.08 1.03 1.00 -0.09 0.18 -0.02 -0.10 0.01 0.06
Covariance Matrix Item RESP3 -------1.46 1.38 1.21 1.36 1.10 1.02 1.05 RESP4 -------RESP5 -------RESP6 -------ANGER1 ANGER2 ---------------
RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3
1.42 1.17 1.35 1.10 1.01 1.03
1.78 1.35 0.97 0.90 0.98
1.73 1.17 1.10 1.15
1.45 1.26 1.45
1.36 1.33
182 Item RESP3 -------1.07 1.00 0.95 -0.08 0.21 -0.04 -0.11 0.02 0.06 RESP4 -------1.07 0.99 0.95 -0.09 0.19 0.01 -0.09 0.02 0.06 RESP5 -------0.98 0.91 0.83 -0.12 0.24 0.04 -0.07 0.03 0.05 RESP6 -------1.13 1.09 1.01 -0.11 0.27 -0.05 -0.01 0.04 0.06 ANGER1 ANGER2 --------------1.41 1.41 1.16 -0.09 0.47 -0.07 -0.02 0.08 0.08 1.23 1.26 1.12 -0.08 0.37 -0.06 0.01 0.06 0.06
ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
Covariance Matrix Item ANGER3 ANGER4 ANGER5 ANGER6 LDA -----------------------------------1.97 1.56 1.89 1.26 -0.04 0.67 -0.09 0.02 0.12 0.11 OS --------
ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
1.61 1.49 1.20 -0.08 0.52 -0.07 0.00 0.09 0.07
2.05 1.22 -0.04 0.64 -0.09 -0.01 0.11 0.11
1.34 -0.07 0.35 -0.10 0.04 0.08 0.04
0.20 0.05 -0.13 -0.01 -0.01 0.01
1.00 0.00 0.02 0.18 0.16
Covariance Matrix Item RC -------1.98 0.01 0.18 0.16 INTER1 -------INTER2 -------INTER3 --------
RC INTER1 INTER2 INTER3
0.68 0.18 -0.17
0.15 -0.03
0.14
183 [See the table below for item label used in the study and indicator label used in the LISREL program. Note all the indicators except LDA are found in Appendices A to D. LDA is a composite measure of the likelihood of each account and the defensiveness of each account. The detailed discussion regarding the development of LDA can be found on pp. 85-87] Indicator Label used in LISREL SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RES1 RES2 RES3 RES4 RES5 RES6 ANG1 ANG2 ANG3 ANG4 Item Label used in the study Perceived Outcome Severity1 Perceived Outcome Severity2 Perceived Outcome Severity3 Perceived Outcome Severity4 Perceived Outcome Severity5 Perceived Outcome Severity6 Perceived Quadratic Relational Closeness Internal Attribution1 Internal Attribution2 Internal Attribution3 Internal Attribution4 Controllability1 Controllability2 Controllability3 Controllability4 Controllability5 External Attribution1 External Attribution2 External Attribution3 External Attribution4 External Attribution5 External Attribution6 External Attribution7 Uncontrollability1 Uncontrollability2 Uncontrollability3 Uncontrollability4 Uncontrollability5 Expected Responsibility1 Expected Responsibility2 Expected Responsibility3 Expected Responsibility4 Expected Responsibility5 Expected Responsibility6 Expected Anger1 Expected Anger2 Expected Anger3 Expected Anger4
184 Indicator Label used in LISREL ANG5 ANG6 LDA OS RC INTER1 INTER2 INTER3 Item Label used in the study Expected Anger5 Expected Anger6 Composite Measure of Accounts Outcome Severity Manipulation (?1) Quadratic Relational Closeness manipulation (?2) Interaction1 (?3) Interaction2 (?4) Interaction3 (?5)
185 Appendix N LISREL Syntax for Testing the Proposed Model in Figure 2
Testing the proposed model DA NI=46 NO=236 MA=CM CM=C:\origin1116.psf LA sev1 sev2 sev3 sev4 sev5 sev6 quadrel ia1 ia2 ia3 ia4 con1 con2 con3 con4 con5 ea1 ea2 ea3 ea4 ea5 ea6 ea7 uncon1 uncon2 uncon3 uncon4 uncon5 resp1 resp2 resp3 resp4 resp5 resp6 anger1 anger2 anger3 anger4 anger5 anger6 LDA ex1 ex2 ex3 ex4 ex5 MO NX=5 NK=5 NY=41 NE=9 FI LY=FU,FI LX=ID TE=DI,FR TD=ZE TH=ZE,FI BE=FU,FI GA=FU,FI PS=DI,FR TH=ZE,FI FR GA 1 1 GA 2 2 GA 9 3 GA 3 4 GA 4 4 GA 5 5 GA 6 5 FR BE 9 1 BE 7 1 BE 9 2 BE 9 8 BE 8 7 BE 7 3 BE 8 3 BE 7 4 BE 8 4 BE 7 5 BE 8 5 BE 7 6 FR BE 8 6 FR LY 2 1 LY 3 1 LY 4 1 LY 5 1 LY 6 1 LY 9 3 LY 10 3 LY 11 3 LY 13 4 LY 14 4 LY 15 4 FR LY 16 4 LY 18 5 LY 19 5 LY 20 5 LY 21 5 LY 22 5 LY 23 5 LY 25 6 LY 26 6 LY 27 6 FR LY 28 6 FR LY 30 7 LY 31 7 LY 32 7 LY 33 7 LY 34 7 LY 36 8 LY 37 8 LY 38 8 LY 39 8 LY 40 8 FI TE 41 41 TE 7 7 VA 1 LY 1 1 LY 7 2 LY 8 3 LY 12 4 LY 17 5 LY 24 6 LY 29 7 LY 35 8 LY 41 9 PD OU AL MI AD=OFF
186 Appendix O Covariance Matrix in LISREL for the Revised Model in Figure 6 Covariance Matrix Item SEV1 -------2.24 2.09 2.04 2.38 2.04 1.90 0.38 2.12 2.26 0.61 1.89 1.30 1.05 1.46 1.11 0.81 2.49 1.26 2.75 0.45 2.82 1.15 0.17 0.44 0.30 0.64 0.65 0.24 0.95 0.93 0.90 0.89 0.95 1.00 1.09 1.00 1.16 SEV2 -------SEV3 -------SEV4 -------SEV5 -------SEV6 --------
SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3
4.27 3.39 4.30 3.40 3.37 0.49 2.62 2.42 0.36 1.69 1.70 1.27 1.76 1.26 1.01 3.39 1.20 2.99 0.40 5.51 1.63 0.24 0.30 0.08 0.68 0.85 0.22 1.38 1.30 1.32 1.31 1.31 1.47 1.81 1.56 2.01
3.40 3.73 3.06 2.92 0.41 2.26 2.15 0.42 1.79 1.48 1.14 1.67 1.05 0.85 3.20 0.95 3.17 0.76 4.84 1.45 0.75 0.30 0.33 0.69 0.83 0.34 1.24 1.18 1.15 1.17 1.27 1.29 1.58 1.39 1.81
6.51 4.51 3.69 0.61 3.36 3.05 1.04 1.92 2.27 1.64 2.18 1.47 1.20 4.16 1.96 3.22 0.80 6.96 2.21 0.16 0.79 0.49 0.95 1.13 0.60 1.62 1.55 1.55 1.50 1.25 1.58 2.17 1.90 2.43
4.26 3.18 0.44 2.79 2.57 0.71 2.08 1.74 1.19 1.67 1.17 1.08 3.72 1.36 3.22 0.86 4.42 1.81 0.49 0.62 0.45 0.99 0.97 0.56 1.22 1.19 1.22 1.20 1.15 1.30 1.60 1.40 1.82
3.66 0.34 2.45 2.23 0.75 2.06 1.65 1.20 1.59 1.17 1.02 2.78 1.25 2.66 0.54 3.83 1.32 0.65 0.39 0.29 0.76 0.78 0.36 1.17 1.10 1.13 1.11 1.17 1.25 1.53 1.31 1.71
187 Item SEV1 -------1.18 1.09 1.06 -0.06 0.63 0.03 -0.03 SEV2 -------1.87 1.91 1.54 -0.02 1.36 0.02 -0.07 SEV3 -------1.74 1.72 1.34 -0.02 1.13 0.09 -0.01 SEV4 -------2.23 2.31 1.82 0.03 1.47 0.05 -0.13 SEV5 -------1.73 1.76 1.40 -0.03 1.22 0.12 -0.02 SEV6 -------1.58 1.66 1.33 -0.04 1.18 0.13 -0.10
ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
Covariance Matrix Item QREL -------1.42 0.70 0.38 0.14 0.64 0.52 0.00 0.37 0.18 0.04 0.44 -0.26 0.32 -0.52 -0.21 -0.57 -0.88 0.06 -0.29 -0.19 -0.19 -0.06 0.37 0.33 0.30 0.30 0.23 0.25 0.28 0.23 IA1 -------IA2 -------IA3 -------IA4 -------CON1 --------
QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2
12.04 6.69 2.77 6.84 3.94 3.19 3.77 3.22 2.70 7.06 0.42 2.88 -0.35 2.31 0.74 0.77 0.06 0.19 0.18 0.21 -0.30 2.11 2.02 1.96 1.99 1.46 1.85 2.08 1.98
12.26 4.05 6.25 3.93 2.53 3.96 3.29 2.72 7.87 1.06 6.44 0.94 3.38 0.99 0.81 0.32 0.50 0.21 0.23 -0.09 1.98 2.12 1.97 1.98 2.12 2.11 2.12 2.15
11.88 2.11 1.91 1.59 1.15 1.08 1.62 3.59 0.72 1.99 1.08 5.34 1.44 2.03 0.46 0.89 0.95 1.11 0.44 0.07 0.17 0.05 0.12 0.09 0.15 0.33 0.60
11.40 3.71 2.75 3.91 3.39 2.85 3.77 -0.41 3.33 -1.08 0.62 -0.59 0.93 -0.06 -0.19 -0.29 -0.07 0.33 1.96 2.05 1.95 2.00 2.03 2.15 1.82 1.81
4.41 2.34 2.71 1.90 1.67 2.68 0.59 1.70 -1.17 2.18 0.46 -0.16 -0.32 -0.26 -0.22 -0.24 -0.26 1.21 1.25 1.27 1.20 0.89 1.20 1.33 1.25
188 Item QREL -------0.17 0.20 0.15 0.14 -0.03 0.12 0.61 0.02 IA1 -------1.84 2.01 1.76 2.11 -0.45 0.09 0.59 -0.01 IA2 -------2.26 2.03 2.09 2.45 -0.42 0.26 0.38 0.02 IA3 -------0.29 0.25 0.29 0.76 -0.11 -0.06 0.40 0.29 IA4 -------1.67 1.78 1.66 1.68 -0.66 -0.05 0.73 0.11 CON1 -------1.12 1.28 0.97 1.17 -0.26 0.12 0.18 -0.10
ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
COVARIANCE MATRIX Item CON2 -------5.46 2.21 .85 1.17 4.31 2.11 1.64 0.18 2.60 0.39 0.03 0.15 -0.36 -0.84 -0.70 -0.73 1.07 1.07 1.02 1.04 0.73 1.10 1.02 0.96 1.07 1.06 1.15 0.98 -0.20 CON3 -------CON4 -------CON5 -------EA1 -------EA2 --------
CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA
3.51 2.34 1.72 2.19 0.20 2.47 -0.90 3.59 0.17 -0.01 -0.45 -0.61 -0.52 -0.34 -0.39 1.49 1.56 1.46 1.47 1.18 1.51 1.50 1.42 1.47 1.50 1.36 1.37 -0.17
2.88 1.67 2.12 -0.31 2.46 -0.85 2.98 0.11 -0.39 -0.27 -0.68 -0.60 -0.54 -0.39 1.04 1.10 0.98 0.99 0.75 1.03 1.12 1.07 1.12 1.12 1.02 1.20 -0.16
2.13 1.26 0.20 1.19 -0.78 2.53 0.34 0.17 -0.16 -0.34 -0.11 -0.20 -0.05 0.71 0.76 0.75 0.70 0.61 0.83 0.86 0.85 0.80 0.81 0.80 0.87 -0.16
36.06 13.83 17.51 8.89 13.84 8.25 6.66 5.75 4.78 4.11 4.45 4.17 1.82 2.11 1.97 1.97 1.56 1.76 1.71 1.95 2.23 1.95 2.04 2.59 0.07
24.04 6.77 6.27 9.71 12.28 7.94 6.81 4.61 3.85 3.95 3.82 0.73 0.47 0.68 0.43 0.92 0.71 0.56 0.74 0.67 0.52 0.81 1.03 0.28
189 Item CON2 -------0.02 0.08 -0.02 CON3 -------0.12 -0.08 -0.06 CON4 -------0.12 -0.09 0.01 CON5 -------0.06 -0.03 -0.03 EA1 -------0.55 0.57 0.74 EA2 -------0.24 -0.28 0.53
OS RC INTER1
Covariance Matrix Item EA3 -------31.79 6.94 9.49 8.77 3.69 4.59 3.39 3.66 4.01 2.99 1.72 1.90 1.70 1.66 1.88 1.56 1.76 1.97 1.94 1.84 1.64 2.38 0.19 0.85 0.23 0.08 EA4 -------EA5 -------EA6 -------UNCON1 UNCON2 ---------------
EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
22.90 6.28 7.47 5.58 2.65 4.36 3.57 3.74 3.16 -0.35 -0.06 -0.03 -0.05 0.11 -0.66 -0.45 -0.16 -0.23 -0.37 -0.15 -0.01 0.14 0.42 -0.16 0.28
53.65 11.51 5.61 1.81 2.32 5.47 5.16 3.75 1.71 1.85 1.66 1.56 1.09 1.28 2.56 2.72 3.57 3.02 3.32 3.27 1.18 1.64 -0.42 0.29
20.49 6.76 3.48 3.33 4.67 5.20 3.50 0.76 0.71 0.83 0.62 0.58 0.54 0.67 0.67 0.76 0.70 0.81 1.17 0.41 0.57 -0.49 0.22
22.16 3.20 4.32 3.47 4.64 2.99 0.11 -0.08 -0.05 -0.15 0.08 0.03 0.06 0.43 0.44 0.06 0.32 0.56 0.03 -0.16 -0.08 0.42
5.98 2.97 2.92 3.04 2.66 0.03 -0.09 -0.02 -0.07 0.05 0.00 0.15 0.28 0.15 0.14 0.23 0.24 0.07 0.18 -0.02 0.22
Covariance Matrix Item UNCON2 UNCON3 UNCON4 UNCON5 ----------------------------5.53 2.99 RESP1 -------RESP2 --------
UNCON2 UNCON3
6.69
190 Item UNCON2 UNCON3 UNCON4 UNCON5 ----------------------------3.22 2.89 -0.07 -0.17 -0.11 -0.15 0.02 -0.33 -0.04 0.19 0.19 -0.05 0.17 0.09 0.18 0.23 0.20 5.07 3.42 -0.11 -0.11 -0.08 -0.15 0.11 -0.18 0.14 0.20 0.28 0.20 0.23 0.12 0.31 -0.01 0.19 6.48 3.67 -0.02 -0.14 -0.05 -0.14 0.22 -0.06 0.20 0.23 0.39 0.26 0.26 0.14 0.40 0.09 0.20 RESP1 -------RESP2 --------
UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER6 LDA OS RC INTER1
6.30 -0.08 -0.18 -0.13 -0.16 0.17 -0.25 0.03 0.24 0.25 0.16 0.20 0.19 0.16 -0.15 0.28
1.61 1.36 1.42 1.39 1.20 1.40 1.16 1.07 1.09 1.01 1.01 -0.08 0.21 0.03 -0.06
1.47 1.35 1.34 1.18 1.34 1.14 1.07 1.08 1.03 1.00 -0.09 0.18 -0.02 -0.10
Covariance Matrix Item RESP3 -------1.46 1.38 1.21 1.36 1.10 1.02 1.05 1.07 1.00 0.95 -0.08 0.21 -0.04 -0.11 RESP4 -------RESP5 -------RESP6 -------ANGER1 ANGER2 ---------------
RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
1.42 1.17 1.35 1.10 1.01 1.03 1.07 0.99 0.95 -0.09 0.19 0.01 -0.09
1.78 1.35 0.97 0.90 0.98 0.98 0.91 0.83 -0.12 0.24 0.04 -0.07
1.73 1.17 1.10 1.15 1.13 1.09 1.01 -0.11 0.27 -0.05 -0.01
1.45 1.26 1.45 1.41 1.41 1.16 -0.09 0.47 -0.07 -0.02
1.36 1.33 1.23 1.26 1.12 -0.08 0.37 -0.06 0.01
191 Covariance Matrix Item ANGER3 ANGER4 ANGER5 ANGER6 ----------------------------1.97 1.56 1.89 1.26 -0.04 0.67 -0.09 0.02 LDA -------OS --------
ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
1.61 1.49 1.20 -0.08 0.52 -0.07 0.00
2.05 1.22 -0.04 0.64 -0.09 -0.01
1.34 -0.07 0.35 -0.10 0.04
0.20 0.05 -0.13 -0.01
1.00 0.00 0.02
Covariance Matrix Item RC -------1.98 0.01 INTER1 --------
RC INTER1
0.68
Note. See the labels of variables in Appendix M.
192 Appendix P LISREL Syntax for Testing the Revised Model in Figure 6 Testing the revised model DA NI=44 NO=236 MA=CM CM=C:\revise1117.psf LA sev1 sev2 sev3 sev4 sev5 sev6 quadrel ia1 ia2 ia3 ia4 con1 con2 con3 con4 con5 ea1 ea2 ea3 ea4 ea5 ea6 ea7 uncon1 uncon2 uncon3 uncon4 uncon5 resp1 resp2 resp3 resp4 resp5 resp6 anger1 anger2 anger3 anger4 anger5 anger6 LDA ex1 ex2 ex3 MO NX=3 NK=3 NY=41 NE=9 FI LY=FU,FI LX=ID TE=DI,FR TD=ZE TH=ZE,FI BE=FU,FI GA=FU,FI PS=DI,FR TH=ZE,FI FR GA 1 1 GA 2 2 GA 9 3 FR BE 9 1 BE 7 1 BE 9 2 BE 9 8 BE 8 7 BE 7 3 BE 8 3 BE 7 4 BE 8 4 BE 7 5 BE 8 5 BE 7 6 FR BE 8 6 BE 3 1 BE 4 1 BE 5 1 BE 6 1 BE 8 1 FR LY 2 1 LY 3 1 LY 4 1 LY 5 1 LY 6 1 LY 9 3 LY 10 3 LY 11 3 LY 13 4 LY 14 4 LY 15 4 FR LY 16 4 LY 18 5 LY 19 5 LY 20 5 LY 21 5 LY 22 5 LY 23 5 LY 25 6 LY 26 6 LY 27 6 FR LY 28 6 FR LY 30 7 LY 31 7 LY 32 7 LY 33 7 LY 34 7 LY 36 8 LY 37 8 LY 38 8 LY 39 8 LY 40 8 FI TE 41 41 TE 7 7 VA 1 LY 1 1 LY 7 2 LY 8 3 LY 12 4 LY 17 5 LY 24 6 LY 29 7 LY 35 8 LY 41 9 PD OU AL MI AD=OFF
193 References Allred, K. G., Mallozzi, J. S., Matsui, F., & Raia, C. P. (1997). The influence of anger and compassion on negotiation performance. Organizational Behavior and Human Decision Processes, 70, 175-187. Averill, J. R. (1983). Studies on anger and aggression. American Psychologist, 38, 11451160. Bauer, C. L., & Fink, E. L. (1983). Fitting equations with power transformations: Examining variables with error. In R. N. Bostrom (Ed.), Communication yearbook 7 (pp. 146-199). Beverly Hills, CA: Sage. Bentler, P. M., & Chou, C. P. (1987). Practical issues in structural modeling. Sociological Methods & Research, 16, 78-117. Bertrand, M., & Mullainathan, S. (2003). Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. National Bureau of Economic Research. Retrieved November 26, 2007 fromhttp://papers.nber.org/papers/w9873 Bies, R. J. (1987). The predicament of injustice: The management of moral outrage. In L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior (Vol. 9, pp. 289-319). Greenwich, CT: JAI Press. Blake, R. R., & Mouton, J. S. (1964). The managerial grid. Houston, TX: Gulf. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Brew, F. P., & Cairns, D. R. (2004). Styles of managing interpersonal workplace conflict in relation to status and face concern: A study with Anglos and Chinese. International Journal of Conflict Management, 15, 27-56.
194 Cai, D. A., & Donohue, W. A. (1997). Determinants of facework in intercultural negotiation. Asian Journal of Communication, 7, 85-110. Cai, D. A., & Fink, E. L. (2002). Conflict style differences between individualists and collectivists. Communication Monographs, 69, 67-87. Carnevale, P. J., & Pruitt, D. G. (1992). Negotiation and mediation. Annual Review of Psychology, 43, 531-582. Choi, I., Nisbett, R. E., & Norenzayan, A. (1999). Causal attribution across cultures: Variation and universality. Psychological Bulletin, 125, 47-63. Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37-46. Cohen, J. (1988). Statistical power analysis for the behavior science (2nd ed.). New York: Academic Press. Converse, J. M., & Presser, S. (1986). Survey questions: Handcrafting the standardized questionnaire. Newbury Park, CA: Sage. El-Sheikh, M., Buckhalt, J. A., & Reiter, S. L. (2000). Gender-related effects in emotional responding to resolved and unresolved interpersonal conflict. Sex Roles: A Journal of Research. Retrieved November 26, 2007 fromhttp://www.findarticles.com/p/articles/mi_m2294/is_2000_Nov/ai_75959823 Fincham, F. D. (1992). The account episodes in close relationship. In M. L. McLaughlin, M. J. Cody, & S. J. Read (Eds.), Explaining one’s self to others: Reason-giving in a social context (pp. 167-182). Hillsdale, NJ: Lawrence Erlbaum Associates. Fink, C. F. (1968). Some conceptual difficulties in the theory of social conflict. Journal of Conflict Resolution, 12, 412-460.
195 Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York: McGraw-Hill. Fitzpatrick, M. A., & Winke, J. (1979). You always hurt the one you love: Strategies and tactics in interpersonal conflict. Communication Quarterly, 27, 3-11. Frey, L. R., Botan, C. H., & Kreps, G. L. (2000). Investigating communication: An introduction to research methods. Boston: Allyn & Bacon. Fritsche, I. (2002). Accounts for the violation of social norms: Integration and extension of sociological and social psychological typologies. Journal for the Theory of Social Behaviour, 32, 371-394. Gagné, P., & Hancock, G. R. (2006). Measurement model quality, sample size, and solution propriety in confirmatory factor models. Multivariate Behavioral Research, 41, 65-83. Gerber, E. (1994). The language of residence: Respondent understandings and census rules. Washington, DC: Center for Survey Methods Research, U. S. Bureau of the Census. Gleason, J. M., & Harris, V. A. (1976). Perceived freedom, accident severity and empathetic value as determinants of the attribution of responsibility. Social Behavior and Personality, 4, 171-176. Goffman, E. (1971). Relations in public. New York: Harper & Row. Gonzales, M. H., Manning, D. J., & Haugen, J. A. (1992). Explaining our sins: Factors influencing offenders accounts and anticipated victim responses. Journal of Personality and Social Psychology, 62, 958-971. Graham, S., Weiner, B., & Benesh-Weiner, M. (1995). An attributional analysis of the development of excuse giving in aggressive and nonaggressive African American
196 boys. Developmental Psychology, 31, 274-284. Hamilton, V. L., & Hagiwara, S. (1992). Roles, responsibility, and accounts across cultures. International Journal of Psychology, 27, 157-179. Hancock, G. R. (2006). Power analysis in covariance structure modeling. In R. C. Serlin (Series Ed.) & G. R. Hancock & R. O. Mueller (Vol. Eds.), Structural equation modeling: A second course (pp. 69-115). Greenwich, CO: Information Age. Harvey, J. H., & Weary, G. (1984). Current issues in attribution theory and research. Annual Review of Psychology, 35, 427-459. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Henderson, M., & Hewstone, M. (1984). Prison inmate’s explanations for interpersonal violence: Accounts and attributions. Journal of Consulting and Clinical Psychology, 52, 789-794. Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods, 3, 424– 453. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. Jackson, D. L. (2003). Revisiting sample size and number of parameter estimates: Some support for the N:q hypothesis. Structural Equation Modeling: A Multidisciplinary Journal, 10, 128-141. Jellison, J. M. (1990). Accounting: Societal implications. In M. J. Cody & M. L. McLaughlin (Eds.), The psychology of tactical communication (pp. 283-298).
197 Philadelphia: Multilingual Matters. Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in person perception. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 220-266). New York: Academic Press. Jöreskog, K., & Sörbom, D (2004). LISREL for Windows (Version 8.70) [Computer software]. Chicago: Scientific Software International. Kelley, H. H. (1971). Attribution in social interaction. Morristown, NJ: General Learning Press. Kelley, H. H. (1986). Attribution theory in social psychology. Nebraska Symposium on Motivation, 15, 192-238. Kelley, H. H. (1986). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska Symposium on Motivation, 15, 192-238. Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). New York: Guilford Press. Konovsky, M. A., & Jaster, F. (1989). “Blaming the victim” and other ways business men and women account for questionable behavior. Journal of Business Ethics, 8, 391-398. Liu, M., & Yao, S. (2007, November). Why do Chinese and Americans differ in attributions of responsibility? Investigating the universality vs. culture-specificity of cognitive appraisals. Paper presented at the annual conference of National Communication Association, Chicago, IL. Mack, R. W., & Snyder, R. C. (1957). The analysis of social conflict—Toward an overview and synthesis. Journal of Conflict Resolution, 1, 212-248. Martin, E. (2006). Vignettes and respondent debriefing for questionnaire
198 design and evaluation. U. S. Bureau of Census Research Report Series. Retrieved December 1, 2007 fromhttp://www.census.gov/srd/papers/pdf/rsm2006-08.pdf McLaughlin, M. L., Cody, M. J., & French, K. (1990). Account-giving and the attribution of responsibility: Impressions of traffic offenders. In M. J. Cody & M. L. McLaughlin (Eds.), The psychology of tactical communication (pp. 244-267). Philadelphia: Multilingual Matters. McLaughlin, M. L., Cody, M. J., & Read, S. J. (1992). Explaining one’s self to others: Reason-giving in a social context. Hillsdale, NJ: Lawrence Erlbaum Associates. McLaughlin, M. L., Cody, M. J., & Rosenstein, N. E. (1983). Account sequences in conversations between strangers. Communication Monographs, 50, 102-125. McLaughlin, M. L., O’Hair, H. D., & Cody, M. J. (1983). The management of failure events: Some contextual determinants of accounting behavior. Human Communication Research, 9, 208-224. Mills, C. W. (1940). Situated actions and vocabularies of motive. American Sociological Review, 5, 904-913. Mitchell, T. R., & Kalb, L. S. (1981). Effects of outcome knowledge and outcome valence on supervisors’ evaluations. Journal of Applied Psychology, 66, 604-612. Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231-259. Ohbuchi, K., Fukushima, O., & Tedeschi, J. T. (1999). Cultural values in conflict management: Goal orientation, goal attainment, and tactical decision. Journal of Cross-Cultural Psychology, 30, 51-71. Orbuch, T. L. (1997). People’s accounts count: The sociology of accounts. Annual
199 Review of Sociology, 23, 455-478. Pruitt, D. G., & Rubin, J. Z. (1986). Social conflict: Escalation, stalemate, and settlement. New York: Random House. Putnam, L. L., & Poole, M. S. (1987). Conflict and negotiation. In F. M. Jablin, L. L. Putnam, K. H. Roberts, & L. W. Porter (Eds.), Handbook of organizational communication (pp. 549-599). Beverly Hills, CA: Sage. Rahim, M. A. (1983). A measure of styles of handling interpersonal conflict. Academy of Management Journal, 26, 368-376. Rahim, M. A. (1992). Managing conflict in organizations. New York: Praeger. Rahim, M. A., & Bonoma, T. V. (1979). Managing organizational conflict: A model for diagnosis and intervention. Psychological Reports, 44, 1323-1344. Read, S. J. (1992). Constructing accounts: The role of explanatory coherence. In M. L. McLaughlin, M. J. Cody, & S. J. Read (Eds.), Explaining one’s self to others: Reason-giving in a social context (pp. 3-20). Hillsdale, NJ: Lawrence Erlbaum. Rest, J. R. (1984). The major components of morality. In W. M. Kurtines & J. L. Gewirtz (Eds.), Morality, moral behavior and moral development (pp. 24-38). New York: Wiley. Richardson, D. S., & Green, L. R. (2006). Direct and indirect aggression: Relationships as social context. Journal of Applied Social Psychology, 36, 2492-2508. Robbennolt, J. (2000). Outcome severity and judgments of “responsibility”: A metaanalytic review. Journal of Applied Social Psychology, 30, 2575-2609. Roloff, M. E., & Soule, K. P. (2002). Interpersonal conflict: A review. In M. L. Knapp & J. A. Daly (Eds.), Handbook of interpersonal communication (3rd ed., pp. 475-
200 528). Thousand Oaks, CA: Sage. Ross, D. (1982). The causal dimension scale: A measure of how individuals perceive causes. Journal of Personality and Social Psychology, 42, 1137-1145. Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 10, pp. 174-214). New York: Academic Press. Salili, F., Maehr, M. L., & Gillmore, G. (1976). Achievement and morality: A crosscultural analysis of causal attribution and evaluation. Journal of Personality and Social Psychology, 33, 327-337. Schlenker, B. R. (1980). Impression management: The self-concept, social identity, and interpersonal relations. Monterey, CA: Brooks/Cole. Schönbach, P. (1980). A category system of account phases. European Journal of Social Psychology, 10, 195-200. Schönbach, P. (1990). Account episodes: The management or escalation of conflict. Cambridge, UK: Cambridge University Press. Scott, M. B., & Lyman, S. M. (1968). Accounts. American Sociological Review, 3, 46-62. Shaver, K. G. (1970). Defensive attribution: Effects of severity and relevance on the responsibility assigned for an accident. Journal of Personality and Social Psychology, 14, 101-113. Sillar, A. L. (1980). The sequential and distributional structure of conflict interactions as a function of attributions concerning the locus of responsibility and stability of conflicts. In D. Nimmo (Ed.), Communication yearbook 4 (pp. 217-235). New Brunswick, NJ: Transaction Books.
201 Sitkin, S. B., & Bies, R. J. (1993). Accounts in conflict situations: Using explanation to manage conflict. Human Relations, 46, 349-370. Sykes, G. M., & Matza, D. (1957). Techniques of neutralization: A theory of delinquency. American Sociological Review, 22, 664-670. Takaku, S. (2000). Culture and status as influences on account giving: A comparison between the United States and Japan. Journal of Applied Social Psychology, 30, 371-388. Tedeschi, J. T., & Reiss, M. (1981). Verbal strategies in impression management. In C. Antaki (Ed.), The psychology of ordinary explanations of social behaviour (pp. 271-309). London: Academic Press. Tennen, H., & Affleck, G. (1990). Blaming others for threatening events. Psychological Bulletin, 108, 209-232. Thomas, E. A. C., & Parpal, M. (1987). Liability as a function of plaintiff and defendant fault. Journal of Personality and Social Psychology, 53, 843-857. Thomas, K. W. (1976). Conflict and conflict management. In M. Dunnette (Ed.), Handbook of industrial and organizational psychology (pp. 889-935). Chicago: Rand McNally. Thomas, K. W. (1992). Conflict and conflict management: Reflections and update. Journal of Organizational Behavior, 13, 265-274. Ting-Toomey, S., & Oetzel, J. G. (2002). Cross-cultural face concerns and conflict styles. In W. B. Gudykunst & B. Mody (Eds.), Handbook of international and intercultural communication (2nd ed., pp. 143-164). Thousand Oaks, CA: Sage. Walster, E. (1960). Assignment of responsibility for an accident. Journal of Personality
202 and Social Psychology, 3, 73-79. Wang, Q. (2006). Linking goals to avoidance in interpersonal conflict situations: A cognitive approach. Unpublished dissertation, University of Maryland, College Park. Weiner, B. (1986). An attribution theory of motivation and emotion. New York: Springer Verlag. Weiner, B. (1992). Excuses in everyday interaction. In M. L. McLaughlin, M. J. Cody, & S. J. Read (Eds.), Explaining one’s self to others: Reason-giving in a social context (pp. 131-146). Hillsdale, NJ: Lawrence Erlbaum Associates. Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New York: Springer-Verlag. Weiner, B. (2000). Intrapersonal and interpersonal theories of motivation from an attributional perspective. Educational Psychology Review, 12, 1-14. Weiner, B. (2006). Social motivation, justice, and the moral emotions: An attributional approach. Mahwah, NJ: Lawrence Erlbaum. Weiner, B., Amirkhan, J., Folkes, V. S., & Verette, J. A. (1987). An attributional analysis of excuse giving: Studies of a naïve theory of emotion. Journal of Personality and Social Psychology, 52, 316-324. Weiner, B., Figueroa-Muñoz, A., & Kakihara, C. (1991). The goals of excuses and communication strategies related to causal perceptions. Personality and Social Psychology Bulletin, 17, 4-13. Weiner, B., & Handel, S. (1985). Anticipated emotional consequences of causal communications and reported communication strategy. Developmental Psychology, 21, 102-107.
203 Wilmot, W. W., & Hocker, J. L. (2005). Interpersonal conflict (7th ed.). New York: McGraw Hill. Wilson, R. J., & Jonah, B. A. (1988). Assignment of responsibility and penalties for an impaired driving incident. Journal of Applied Social Psychology, 18, 564-583. Wilson, S. R. (2005). Seeking and resisting compliance: Why people say what they do when trying to influence others. Thousand Oaks, CA: Sage. Wrench, J. S., Thomas-Maddox, C., Richmond, V. P., & McCroskey, J. C. (2008). Quantitative research methods for communication: A hands-on approach. New York: Oxford University Press. Yirmiya, N., & Weiner, B. (1986). Perceptions of controllability and anticipated anger. Cognitive Development, 1, 273-280.
doc_270369076.pdf
Conflict management is the process of limiting the negative aspects of conflict while increasing the positive aspects of conflict.
ABSTRACT Title of Dissertation: ATTRIBUTIONAL PROCESSES IN ACCOUNTING FOR CONFLICT BEHAVIORS Shuo Yao, Doctor of Philosophy, 2009 Dissertation Directed By: Professors Deborah A. Cai & Edward L. Fink Department of Communication
One simple way to handle interpersonal conflict is to use accounts to explain one’s behaviors. Although accounts play a significant role in managing conflict, relatively little research has explored the processes offenders use to determine the accounts selected in conflict situations. The goal of this dissertation is to investigate the attributional processes offenders use that determine their accounts in conflict. Ten hypotheses were proposed about how the severity of the conflict outcome and the closeness between the parties involved in the conflict influence offenders’ choice of accounts. A structural equation model was developed and tested based on the proposed hypotheses. An experiment was conducted, with two levels of outcome severity and three levels of relational closeness. Offenders’ attributions (i.e., the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability), anticipated consequences (i.e., expected responsibility and expected anger), and offenders’ expected choice of accounts (i.e., the likelihood of selecting concessions, justifications, excuses, and refusals) were measured. Two hundred thirty-eight participants were recruited and randomly assigned to one of the six experimental conditions. Participants read a hypothetical conflict scenario,
imagined that they were the offender in the scenario, and completed a questionnaire that had the dependent measures. Results indicated that outcome severity influenced offenders’ choice of accounts directly and indirectly. Offenders tended to choose a more defensive account when they perceived the outcome to be severe than when the outcome was not severe. The influence of outcome severity on offenders’ choice of accounts was also mediated by the attributions offenders made, the responsibility expected to be assigned to offenders, and anger expected to be felt by victims. When offenders perceived the outcome to be severe, offenders made more attributions, expected more responsibility to be assigned to them, and expected that victims felt angrier about offenders’ behavior than when the outcome was not severe. Consequently, when offenders expected more anger from victims, they tended to be less defensive. Interpretations and implications of results, the limitations of the study, and future directions were discussed.
ATTRIBUTIONAL PROCESSES IN ACCOUNTING FOR CONFLICT BEHAVIORS
by Shuo Yao
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2009
Advisory Committee: Professor Deborah A. Cai, Co-Chair Professor Edward L. Fink, Co-Chair Professor Torsten O. Reimer Professor Harold Sigall Professor Monique M. Turner
© Copyright by Shuo Yao 2009
ii
Dedication
This dissertation is dedicated to my parents and my sister, for their love and support.
iii
Acknowledgement
I would like to express my gratitude to all those who gave me the possibility to complete this dissertation. I am deeply indebted to my two advisors, Dr. Deborah A. Cai and Dr. Edward L. Fink. Dr. Cai always makes herself available when I need her. I cannot recall how many times we sat in her office and talked about research ideas or problems in my life. She always inspires me and guides me on my way. Two years ago I was in the most difficult time of my life. If it was not Dr. Cai who listened to me patiently and encouraged me to concentrate on my work, I probably would not be able to finish the program. My other advisor, Dr. Fink, is the most intelligent and diligent scholars I have ever encountered. I remember my fellow students and I used to say that Dr. Fink is a “walking encyclopedia.” No matter what questions we have, we can always get some answers from him. Dr. Fink met with me extensively to work on this dissertation. Even when I am away to teach at another university, he told me to call him anytime when I have a question. Without Dr. Fink’s help, this dissertation would be impossible to complete. Being able to be one of his students is the luckiest thing that has ever happened to me. My appreciation to my two advisors cannot be expressed fully by words. I also want to thank my other committee members: Drs. Monique M. Turner, Torsten O. Reimer, and Harold Sigall. I am so lucky that I could be in Dr. Turner’s research team during my first year at Maryland. The experience taught me way much than what I could have learned from books. Dr. Turner also sets a role model for me not only in doing research but also in teaching. Dr. Reimer is always polite and helpful. When I worked with him on one of his research ideas, he always surprised me with new
iv ideas or perspectives to conduct the study. I am also grateful to have Dr. Sigall as one of my committee members. I would like to thank Dr. Timothy J Ng. When I was about to defend my dissertation prospectus, Dr. Sigall was sick and could not attend the defense. On a short notice, Dr. Ng still generously agreed to replace Dr. Sigall and attended the defense. My former colleagues from the Department of Communication, University of Maryland, College Park, supported me in my research work. I want to thank them all for their help, support, interest, and valuable ideas. Especially, I am obliged to Drs. Qi Wang, Xiaoying Xie, and Bing Han. Qi was my roommate for four years and we share so many happy moments. Qi is also the person that I will first contact if I have any question about my research and teaching. I arrived at Maryland the same year as Bing and Xiaoying. We took the same classes together and I am so lucky to have these two smart people as my cohort. I also want to thank my friend, Dr. Jennifer Vardeman, who helped me in various aspects of life, especially when I was struggling two years ago. My friend, Stephanie Materese George, can always find a way to put a smile on my face. Dr. Lan Ni talks with me over the phone and helps me go through the transition stage from a student to a professor. Hua Jiang, Hongmei Shen, Ai Zhang, Sanja Sipek, Sabine Chai, Ioana Cionea, Sejal R. Patel, John Banas, Elena Bessarabova Banas, Eirca Lamm, Timothy Barney, Art Herbig, Ahnlee Jang, Yi Luo, I thank you for your help and support. And, especially, I would like to thank Jiangming Teng, who accompanied me during the darkest time of my life. His support and understanding enabled me to complete this work.
v
Table of Contents
Dedication ...................................................................................................................... i Acknowledgements ....................................................................................................... ii Table of Contents .......................................................................................................... v List of Tables ............................................................................................................... vi List of Figures ............................................................................................................. vii CHAPTER I: Introduction ............................................................................................ 1 CHAPTER II: Attributional Processes in Accounting for Conflict Behaviors ............ 6 Interpreonal Conflict and Conflict management....................................................... 6 Accounts ................................................................................................................... 8 Outcome Severity.................................................................................................... 16 Relational Closeness ............................................................................................... 20 Attribution Theory .................................................................................................. 22 Weiner’s Attribution Theory of Motivation and Emotion .................................. 23 CHAPTER III: Method ............................................................................................... 31 Overall Design, Stimulus Message, and Instruments ............................................. 31 Overall Design .................................................................................................... 31 Hypothetical Interpersonal Conflict Scenario ..................................................... 32 Role Playing in a Hypothetical Conflict Scenario .............................................. 33 Naming the Hypothetical Other Party................................................................. 34 Instruments .......................................................................................................... 34 Summary ............................................................................................................. 37 Pilot Studies ............................................................................................................ 37 Pilot Study 1: Collecting Conflict Scenarios ...................................................... 38 Pilot Study 2: Developing a Hypothetical Conflict Scenario ............................. 40 Pilot Study 3: Manipulation Checks ................................................................... 42 Pilot Study 4: Creating the Measure of Accounts ............................................... 49 Pilot Study 5: Testing the Defensiveness Level of Each Account...................... 50 Pilot Study 6: Pilot Testing Dependent Measures .............................................. 65 The Formal Study ................................................................................................... 72 Analytical Strategies ........................................................................................... 72 Design ................................................................................................................. 74 Sample................................................................................................................. 74 Procedure ............................................................................................................ 75 CHAPTER IV: Results ............................................................................................... 78 Preliminary Analyses .............................................................................................. 78 Data Entry Checks .............................................................................................. 78 Missing Data ....................................................................................................... 78 Data Transformation ........................................................................................... 79 Instruments .......................................................................................................... 80 Manipulation Checks .......................................................................................... 96 Model Assessment and Hypothesis Testing ........................................................... 97 Model Assessment .............................................................................................. 97 Hypothesis Testing............................................................................................ 101
vi Post-Hoc Analyses ................................................................................................ 106 Hypothesis Testing............................................................................................ 106 Modification Indices ......................................................................................... 111 Model Revising ................................................................................................. 113 Testing the Revised Model ............................................................................... 113 CHAPTER V: Discussion ........................................................................................ 120 Summary of the Study .......................................................................................... 120 Summary of the Results ........................................................................................ 122 Hypothesis 1: Influence of Outcome Severity on Expected Responsibility ..... 122 Research Question 1: Effect of Outcome Severity on the Choice of Accounts 122 Hypothesis 2: Influence of Relational Closeness on the Choice of Accounts .. 123 Hypothesis 3: Interaction and the Choice of Accounts ..................................... 124 Hypothesis 4: Attributional Processes .............................................................. 124 Hypothesis 5: Internal Attribution and Anticipated Consequences .................. 125 Hypothesis 6: Controllability and Anticipated Consequences.......................... 125 Hypothesis 7: External Attribution and Anticipated Consequences ................. 126 Hypothesis 8: Uncontrollability and Anticipated Consequences...................... 126 Hypothesis 9: Expected Responsibility and Expected Anger ........................... 127 Hypothesis 10: Expected Anger and the Choice of Accounts .......................... 127 How do People Decide the Accounts in Conflict Situations ............................ 128 Limitations of the Study and the Direction for Future Research .......................... 129 Limitation Regarding the Conceptulization of Accounts ................................. 129 Limitation Regarding the Dependent Measures ............................................... 130 Limitation Regarding the Hypothetical Conflict Scenario ............................... 131 Future Direction: Cultural Differences in Accounts ......................................... 132 Significance of the Study and Conclusion ............................................................ 134 Appendix A Instruments for Manipulation Checks in the Formal Study ................. 139 Appendix B Instruments for Attributions in the Formal Study ................................ 140 Appendix C Instruments for Anticipated Consequences in the Formal Study ......... 143 Appendix D Instruments for the Likelihood of Selecting an Account in the Formal Study ......................................................................................................................... 145 Appendix E Pilot Study 1: Questionnaire for Collecting Conflict Scenarios ........... 148 Appendix F Pilot Study 2: Questionnaire for Sorting Task ...................................... 153 Appendix G Conflict Incidents Reported in Pilot Study 1 ....................................... 159 Appendix H Hypothetical Conflict Scenarios Used in the Formal Study ................ 161 Appendix I Pilot Study 3: Questionnaire for Manipulation Checks ......................... 166 Appendix J Instruments for Defensiveness of Accounts in Pilot Study 5 ................ 171 Appendix K List of Open-Ended Questions Used in Pilot Study 6 .......................... 172 Appendix L Descriptives of the Indicatiors in the Formal Study after Transformation ................................................................................................................................... 173 Appendix L Covariance Matrix in LISREL for the Proposed Model in Figure 2 .... 177 Appendix M LISREL Syntax for Testing the Proposed Model in Figure 2 ............. 185 Appendix N Covariance Matrix in LISREL for the Revised Model in Figure 6 ...... 186 Appendix O LISREL Syntax for Testing the Revised Model in Figure 6 ................ 192 References ................................................................................................................. 193
vii
List of Tables
Table 1. Table 2. Major Types of Accounts, Sources, and Definitions. Coding Scheme for Four Types of Accounts, Adapted from Fritsche’s (2002) Review (pp. 389-394). Table 3. Table 4. 51 12
Intercorrelations between Items for Defensiveness of Each Account. 55 The Cronbach’s Alpha for the Sum of Items that Measured Defensiveness Level of Each Account (N = 22). 62
Table 5.
Eigenvalues of the First Principal Components of Endogenous Variables with the Proportions of the Variance Explained. 88
Table 6.
The Cronbach’s Alpha for the Sum of Measurement Items for All Endogenous variables in the Formal Study. 89
Table 7.
Manipulation Checks, Indicator Loadings, and Measurement Model Indices. 91
Table 8.
Attributional Processes Variables, Indicator Loadings, and Measurement Model Indices. 93
Table 9.
Expected Consequences Variables, Indicator Loadings, and Measurement Model Indices. 95 98 99 107
Table 10. Table 11. Table 12. Table 13.
Structural Equations of the Model Proposed by the Hypotheses. The Unstandardized Structural Coefficients for the Model. Intercorrelations between the Five Exogenous Variables.
Structural Equations of the Revised Model Based on Modification Indices and Exploratory Analyses. 115 116
Table 14.
Unstandardized Structural Coefficients for the Revised Model.
viii
List of Figures
Figure 1. Types of accounts on the defensiveness continuum, as proposed by Schönbach (1990). Figure 2. Figure 3. Structural relations based on the hypotheses. Defensiveness level of concessions, justifications, excuses, and refusals. Figure 4. Structural relations proposed in the hypotheses with unstandardized structural coefficients. Figure 5. Figure 6. Structural relations based on post-hoc analysis. The revised model with unstandardized structural coefficients. 100 114 117 64 16 29
1 CHAPTER I Introduction Person A: You are supposed to cover this training session with me together but you never showed up! How many times are you going to leave me to work this group by myself? Person B: My other job has had me working sixty hours a week for the past few weeks to close out an important project! (Pause) Person B: But it will end soon. By that time I can be back to work with you regularly. I understand where you are coming from. (Extract from a self-report conflict scenario in Pilot Study 1 of the current research). Despite growing interest in behavioral strategies for conflict management, little attention has been given to a simple way to handle conflicts: the use of accounts. Accounts are statements made by offenders to explain their unanticipated or normcontradictive behaviors (Scott & Lyman, 1968). As shown in the above example, when an interpersonal conflict occurs, communication plays a significant role in the conflict process (Putnam & Poole, 1987). People often use accounts to try to influence the affected party’s perception of the conflict, to avoid further escalation of the conflict or to maintain their relationship with the affected party. Verbal explanation is especially important for offenders whose behavior has a negative influence on victims in conflict situations. What offenders say to explain their behaviors directly influences a victim’s perceptions of (1) the responsibility assigned to the offender, (2) the offenders’ intentions, and (3) the way that the conflict can be handled (Sitkin & Bies, 1993).
2 Studying communication that occurs in conflict situations, especially accounts offenders use to explain their behaviors, provides a foundation for conflict management research because accounts often set the tone for how a conflict can be resolved. Further, examining offenders’ choice of accounts in a conflict situation can provide insight to the account literature. Past research on accounts has focused on developing different types of accounts and examining their functions in interpersonal situations (e.g., Jellison, 1990; Tedeschi & Reiss, 1981). Relatively little work has explored how people construct their accounts (for exceptions see Read, 1992; Rest, 1984). Read argued that offenders construct accounts based on a typical form, such as information about the goals of offenders, factors that stimulate those goals, behaviors or strategies to achieve those goals, outcomes (e.g., whether those goals are satisfied), and “the physical and social situations in which the action occurred” (Read, 1992, p. 6). Rest looked into how offenders reconstruct the definition of situations to mitigate the consequences of their actions. Rest pointed out that the defensive reinterpretation, a process to reframe offenders’ understanding about the causes of their actions and their judgments of responsibility, was used by offenders to protect them in the course of a negative event. Read’s (1992) and Rest’s (1984) studies are examples of research examining the cognitive processes offenders use to determine their accounts. McLaughlin, Cody, and Read (1992) suggested that more attention should be given to understanding how people create accounts. The current research answers the call from McLaughlin et al. (1992) by focusing on one specific process, the attributional process, to understand the accounts that offenders use in conflict situations. The question addressed by this research is: What attributional processes do offenders use to determine
3 their accounts in a conflict? Attribution theories provide a framework for understanding how people make sense of their world (e.g., Weiner, 1986, 1992). Causal inferences organize people’s perceptions of reality and make the experiential world more understandable and predicable (Kelley, 1971). By using attribution theories, this study can test the cognitive and emotional processes that offenders engage in after their behavior leads to an interpersonal conflict. Weiner’s (1986, 1995, 2006) attribution theory of motivation and emotion provides the theoretical background for this research. Weiner argued that to explain a norm-contradictive behavior, offenders often start by examining antecedent conditions that lead to the behavior. In a conflict situation, the antecedent conditions could be the relationship between the offender and the victim or the severity of the situation. According to Weiner, once offenders have analyzed the antecedent conditions, they will make causal inferences about why the behavior has occurred. The behavior could have happened because of environmental reasons or because of the offenders’ personal traits. Weiner further argued that based on the causal inferences made, offenders will have expectations about consequences if such inferences were presented to the victim. Expected consequences refer to the level of responsibility to be assigned to offenders and the level of the victim’s anger. Offenders then will decide the accounts to be conveyed to victims based on the expected consequences. The current research examined offenders’ account-giving processes by focusing on two antecedent conditions—outcome severity and relational closeness—and causal inferences associated with the offending behavior. Outcome severity refers to the level of
4 severity in consequences caused by offenders’ behavior. Attribution studies have suggested that people are “motivate[d] to be highly engaged in an attributional search in response to more threatening or unusual events because there is a greater need to explain such events” than events that have a less severe outcome (Robbenolt, 2000, p. 2576). Research also indicated that different levels of outcome severity lead to a different amount of responsibility to be assigned to offenders (Walster, 1960). When the outcome of a conflict is severe, offenders expect victims to assign more responsibility to offenders. To reduce offenders’ negative feelings about themselves, offenders may blame themselves and use an account to comfort victims, such as by offering compensation or showing regret. When the outcome of a conflict is not severe, offenders expect victims to assign less responsibility to offenders; thus, offenders will be more likely to defend themselves, compared to the situation in which the outcome is severe. The current research also posits that people’s perception about conflict and managing conflict varies, depending on with whom they are in conflict. Relational closeness, the level of closeness between the offender and the victim, influences the choice of accounts. Fitzpatrick and Winke (1979) pointed out that when a conflict occurs between close friends, people perceive that they have a lower risk of relationship termination than if the conflict is between people who are not close. Offenders may then take a defensive stand to handle the conflict. If the conflict occurs between acquaintances, given the desire to maintain the relationship with the victim, offenders often will be mitigating to cope with the conflict. However, when a conflict occurs between strangers, with a low possibility of future interaction, offenders will try to defend themselves and, as a result, will be aggressive.
5 The current research focuses on these two contextual variables (i.e., outcome severity and relational closeness between offenders and victims) of interpersonal conflicts and examines the cognitive and emotional processes involved in the provision of accounts. An experiment that manipulated three levels of relational closeness (high vs. moderate vs. low) and two levels of outcome severity (high vs. low) in a hypothetical conflict scenario was conducted to further our understanding of the offenders’ choice of accounts in a conflict situation.
6 CHAPTER II Attributional Processes in Accounting for Conflict Behaviors Interpersonal Conflict and Conflict Management Although no definition of conflict is universally agreed upon, the central characteristic of conflict is generally viewed as incompatibility of goals (Roloff & Soule, 2002). For example, Mack and Snyder (1957) defined conflict as a “particular kind of social interaction process or ‘interaction relationship’ between parties who have mutually exclusive or incompatible values” (cited in Fink, 1968, p. 432). Similarly, Wilmot and Hocker (2005) defined conflict as “an expressed struggle between at least two interdependent parties who perceive incompatible goals, scarce resources, and interference from others in achieving their goals” (p. 9). Conflicts can occur under different circumstances, such as when two parties’ behavioral preferences are inconsistent, when there are not enough resources to fulfill each party’s desire, or when there is a conflict of values or goals (Rahim, 1992; Rahim & Bonoma, 1979). The primary focus of conflict research has been the examination of people’s preference for conflict management styles (Thomas, 1992). Rahim’s (1983) typology of conflict management styles has been widely studied. These styles are dominating, integrating, compromising, obliging, and avoiding. Although different versions of Rahim’s five conflict management styles have been used in conflict research (e.g., Ohbuchi, Fukushima & Tedeschi, 1999), a majority of conflict management styles can be categorized based on the dual-concern model (Carnevale & Pruitt, 1992; Pruitt & Rubin, 1986). Based on Blake and Mouton’s (1964) and Thomas’ (1976) research on conflict, Pruitt and Rubin (1986) developed the dual-concern model to represent different conflict
7 styles. The dual concerns refer to concerns for one’s own outcomes and for the outcome of others involved in the conflict. The dual-concern model suggests that dominating involves a high concern for one’s own outcome and a low concern for the other’s outcome; integrating reflects a high concern for both own and other’s outcome; compromising represents a moderate concern for one’s own and the other’s outcome; obliging involves a low concern for one’s own outcome but a high concern for the other’s outcome; and avoiding reflects a low concern for both own and other’s outcome. Although conflict management style preferences have been widely studied (e.g., Ohbuchi et al., 1999), relatively little work has explicated the communication aspects of conflict management—accounts offenders use to explain the behaviors that have led to a conflict. Conflict management styles have been studied as a general way to handle conflict; for example, Cai and Fink (2002) found that compromising and integrating are preferred more by collectivists than by individualists. This result suggested that someone high on collectivism is more likely to use compromising or integrating to handle a conflict than those high on individualism. Accounts are different from conflict management styles in that accounts depend on specific contexts. For example, McLaughlin, O’Hair, and Cody (1983) studied contextual determinants of accounts and found that when there was a reproach, when the offense was severe, and when the offender felt guilty, offenders were more likely to use concessions to explain their behaviors. Contexts play an important role in determining the accounts that will be used in conflict situations. The use of a particular account is not necessarily connected to individuals’ dispositional characteristics and not necessarily correlated with individuals’ conflict management styles.
8 Accounts Accounts have been recognized as one impression management strategy. For example, Goffman (1971) discussed accounts as a remedial strategy to protect an offender from attacks to his or her social identity (Goffman, 1971). This theoretical orientation is also found in Scott and Lyman’s (1968) typology of accounts. According to Scott and Lyman, an account is “a statement made by a social actor to explain unanticipated or untoward behavior” (p. 46). Offenders use accounts to influence victims’ perception of themselves or of the causes of the conduct in a specific incident (Mills, 1940). As argued by Mills, “When an agent vocalizes or imputes motives, he is not trying to describe his experienced social action. He is not merely stating ‘reasons.’ He is influencing others—and himself” (p. 907). Most early studies of accounts were descriptive and focused on the development of account taxonomies (e.g., Scott & Lyman, 1968; Tedeschi & Reiss, 1981). For example, in a classic study on accounts (Sykes & Matza, 1957), the authors examined the techniques (i.e., techniques of neutralization) that juvenile delinquents used to justify their deviant behaviors and to protect themselves from self-blame or blame from others. Sykes and Matza (1957) classified the techniques of neutralization that juvenile delinquents used to provide explanations for their behaviors into five categories: denial of responsibility (i.e., a claim that something happened due to uncontrollable forces), denial of injury (i.e., a claim that no harm has been caused), denial of victim (i.e., a claim that the victim deserved the treatment given), condemnation of the condemners (i.e., a claim that condemners are reacting out of personal spite), or appeal to higher loyalties (i.e., a claim that loyalty to important others requires such behaviors). By using the above
9 techniques, juvenile delinquents could adjust their understanding of social norms to be consistent with their behaviors, thus avoiding self-blame and disregarding blame from others. Inspired by Sykes and Matza’s (1957) research, Scott and Lyman (1968) distinguished two types of accounts: excuses and justifications. An excuse refers to an explanation in which people admit the inappropriateness of the behavior in question but cite some external reasons to deny their full responsibility. In the example cited at the beginning of the current research, Person B used having another job as an excuse for why he or she did not fulfill the commitment to Person A. By doing so, Person B acknowledged that the behavior (i.e., not showing up to the training session) was not appropriate; however, the other job for an important project was seemingly more timeconsuming and demanding, which excused Person B’s misconduct (i.e., not showing up to the training session). Thus, Person B would not take full responsibility for missing the training session. The other type of account, justification, is defined as an explanation that denies the behavior in question is wrong but admits responsibility for it. Often people who use this account claim that their conduct fulfills a superordinate goal. Scott and Lyman (1968) offered the following example of a justification: A soldier could assert that killing in a war is for a superordinate goal—fighting for the cause of freedom. Therefore, although the solider should take the responsibility for killing, he or she does not think the behavior is inappropriate because killing in a war is necessary and perhaps even desirable. Scholars have extended Scott and Lyman’s (1968) account taxonomy (e.g., Bies,
10 1987; Schlenker, 1980; Schönbach, 1990; Sitkin & Bies, 1993; Tedeschi & Reiss, 1981). Much work has focused on clarifying the distinction between excuses and justifications (e.g., Schlenker, 1980). Schönbach expanded the account taxonomy by developing two additional categories: concession and refusal. Concession is defined as an explanation that admits the inappropriateness of the behavior in question and also admits partial or total responsibility for the behavior. Concessions are often accompanied with a confession, expression of regret or guilt, or an offer of restitution (Fincham, 1992). In the example cited at the beginning of the current research, if a concession were to be used, Person B would probably say something like this: “I am sorry for being absent from all those training sessions. It’s my fault. Maybe I can cover the next two sessions by myself so you can have some time to do your work.” In this concession, Person B admits that his or her behavior (i.e., not showing up to the training session) was not appropriate and offered compensation. On the other hand, when people choose refusals, they deny guilt or responsibility, claim that the behavior in question has not occurred, deny personal involvement with the behavior in question, or attribute guilt to others, even to the accuser (Konovsky & Jaster, 1989; Schönbach, 1980). If a refusal would be used in the above example, Person B probably would argue like this: “You did not tell me I had to be there for the training session. If you wanted me to be there, you should have let me know ahead of time, rather than complaining afterwards.” By attributing the guilt to Person A (i.e., not informing Person B about his or her duty), Person B used a refusal to deny his or her responsibility. Although a few more types of accounts have been discussed in the account literature, the majority of the new types of accounts are similar to Scott and Lyman’s
11 (1968) and Schönbach’s (1980) conceptualizations (e.g., Bies, 1987; Hamilton & Hagiwara, 1992; McLaughlin, O’Hair et al., 1983). See Table 1 for definitions of major types of accounts as proposed by various authors. For example, Bies (1987) described four different types of accounts: causal accounts, ideological accounts, referential accounts, and penitential accounts. Bie’s framework is relatively similar to Schönbach’s taxonomy. Causal accounts are in essence excuses; ideological accounts are similar to justifications; and penitential accounts consist of apology and remorse, which often are used in concessions. The only difference between Bies’ framework and Schönbach’s taxonomy is that Bies de-emphasized refusals but highlighted referential accounts (i.e., when social comparison and consensus information are used to explain the behavior in question). Scholars perceived apology as another type of account (e.g., Hamilton & Hagiwara, 1992). Apology admits that harm was done but does not necessarily remove or reduce responsibility. Using an apology assuages negative emotions aroused by the behavior in question. Apologies can be identified as a part of concessions (Fincham, 1992). McLaughlin, O’Hair et al. (1983) argued that under certain circumstances (for instance, when the embarrassment associated with the behavior in question is severe) the best choice to deal with the conflict may be silence. Silence is defined as avoiding any reference to or verbal explanation about the untoward behavior. Although silence or avoidance sometimes may be the best way to prevent interpersonal conflicts from escalating, it is not a type of accounts because it does not provide any verbal explanation about the behaviors in question. Because concessions, justifications, excuses, and refusals overlap with most types of accounts discussed in the account literature, these four types
12 Table 1 Major Types of Accounts, Sources, and Definitions Types of accounts Excuse Sources Definition Explanations in which people admit the inappropriateness of the behavior in question but cite external reasons to deny their full responsibility. Justification Scott & Lyman (1968) Explanations in which people admit full responsibility for what has happened but indicate that the behavior in question is to fulfill a superordinate goal; thus, the behavior is not inappropriate. Concession Schönbach (1980) Explanations in which people admit partial or total responsibility or guilt, express regret concerning the behavior in question, and offer compensation. Refusal Schönbach (1980) Explanations in which people deny guilt or responsibility, claim that the behavior in question has not occurred, and attribute guilt to others. Causal Bies (1987) Explanations in which people use a reason to mitigate offenders’ responsibility, or explanations that are more commonly
Scott & Lyman (1968)
13 referred to as excuses. Ideological Bies (1987) Explanations in which the actor attempts to invoke a superordinate goal or to place a positive label on the deed. Referential Bies (1987) Explanations in which social comparison and consensus information is presented. Penitential Bies (1987) Explanations that consist of apology and remorse. Consensus Hamilton & Hagiwara (1992) External excuse Hamilton & Hagiwara (1992) Explanations that use social comparison and declare that “everybody’s doing it.” Explanations that include a broad range of circumstances and interventions by human or natural causes. Internal excuse Hamilton & Hagiwara (1992) Explanations that refers to something inside the person that caused the outcome, whether or not it was controllable by that person. Apology Hamilton & Hagiwara (1992) Explanations that admit the wrong-doing and the responsibility; an apology is expected to assuage feelings.
14 of accounts will be examined in the current research. More than developing account taxonomies, McLaughlin, Cody, and Rosenstein (1983) and Schönbach (1990) advanced the study of accounts by using a continuum to lay out the underlying dimension for the four types of accounts (i.e., concessions, justifications, excuses, and refusals). McLaughlin, Cody et al. used an aggravatingmitigating continuum whereas Schönbach investigated each type of accounts in terms of its defensiveness, that is, to the extent to which using each type of accounts helps offenders to defend themselves. Regardless of the terminology, the continua proposed by these two groups of researchers are similar. According to Schönbach (1990), refusals, which deny the connection between offenders and the behavior in question, are most defensive. Further, excuses are less defensive than refusals because when excuses are used, offenders admit that their behavior is wrong and point to external reasons to reduce some responsibility. Schönbach argued that justifications are less defensive than excuses because offenders admit partial or full responsibility for their behavior, although they claim their behaviors are not inappropriate. According to Schönbach, concessions, which admit offenders’ responsibility or guilt for the behavior in question, are the least defensive type of accounts. On McLaughlin, Cody et al.’s aggravating-mitigating continuum, concessions and excuses are mitigating types of accounts whereas refusals and justifications are aggravating types of accounts. In particular, McLaughlin, Cody et al. argued that concessions are most mitigating and excuses are next to most mitigating type of accounts. Refusals are most aggravating and justifications are less aggravating, compared to refusals.
15 McLaughlin, Cody et al.’s (1983) assumed that aggravating accounts and mitigating accounts are different and exclusive: When one type of accounts is mitigating, it will not be considered as an aggravating type of accounts, and vice versa. For example, according to McLaughlin, Cody et al., excuses are a type of mitigating accounts, not an aggravating account. And justifications are a type of aggravating accounts, not a mitigating account. However, the definitions of these two types of accounts do not necessarily suggest this interpretation. By using a justification, an offender could mitigate the conflict to some extent because a justification admits the offender’s responsibility for what has happened. On the other hand, an excuse could aggregate the conflict because a victim may regard an offender’s using excuses as being dishonest and therefore think that the offenders is not willing to take his or her responsibility. A type of account may be both mitigating and aggravating at the same time. There was no empirical support to both the mitigating-aggravating continuum proposed by McLaughlin, Cody et al. and the defensiveness continuum proposed by Schönbach. Given the unclear distinction between mitigating and aggravating accounts, the current research decided to follow Schönbach’s defensiveness continuum to study accounts. Figure 1 indicates the possible ranking of four types of accounts on a defensiveness continuum as proposed by Schönbach. The use of accounts is associated with the personal responsibility assigned to offenders (McLaughlin, Cody, & French, 1990). For example, justifications and excuses are more likely to be connected with personal responsibility than concessions because justifications and excuses often “call attention to the character and motives of the defendant” (McLaughlin et al., 1990, p. 252). One factor that influences people’s judgment of responsibility is outcome severity.
16 Low Defensiveness High Defensiveness
Concession
Justification
Excuse
Refusal
Figure 1. Types of accounts on the defensiveness continuum, as proposed by Schönbach (1990). Outcome Severity Outcome severity determines the amount of responsibility assigned to a potentially responsible actor (Tennen & Affleck, 1990). For example, in Walster’s (1966) study, participants listened to a story about a car accident and rated the responsibility of offenders. Walster found a positive relationship between outcome severity and responsibility assigned to the offender; that is, when the outcome was severe, participants attributed more responsibility to the offender than when the outcome was not severe. However, later research provided mixed support for this proposition (Robbennolt, 2000). A number of studies have replicated Walters’s findings (e.g., Gleason & Harris, 1976; Tennen & Affleck, 1990; Wilson & Jonah, 1988), but other studies have either failed to find a relationship between outcome severity and expected responsibility (e.g., Thomas & Parpal, 1987) or have found their relationship to be negative (e.g., Shaver, 1970). Robbennolt (2000) conducted a meta-analysis to further understand the relationship between outcome severity and judgment of responsibility. Seventy-five studies were collected for the meta-analysis. Although the strength of correlation varied in terms of types of judgment that participants were asked to make (e.g., the correlation was smallest for liability judgments and largest for assessments of punishment), Robbenolt’s meta-analysis supported Walster's hypothesis that, overall, people assign
17 more responsibility to offenders when an outcome of a negative event is more severe than when the outcome is less severe. Robbenolt reported that the effect sizes of all studies that she examined ranged from -.26 (Shaver, 1970) to .75 (Mitchell & Kalb, 1981). The overall effect size was positive and significantly different from zero (r = .08, p < .01).1 The majority of studies on outcome severity and responsibility focus on the observer’s perspective: having a third person assign responsibility to the offender. Relatively little research has been conducted from the offender’s perspective. The question is whether offenders understand the violation the same way that observers do. When the outcome of a conflict is severe, does the offender expect more responsibility to be assigned to themselves than when the outcome is not severe? McLaughlin et al. (1990) pointed out that severe outcomes often include distinct damage to victims, such as bodily harm or loss of a large sum of money. With evidence of damage, those offenses are less deniable and offenders perceive a greater chance that they would be held responsible by others. Thus, it is hypothesized that offenders will expect more responsibility assigned to them in a severe outcome situation than a not severe outcome situation. H1: Offenders expect more responsibility to be assigned to them when the outcome is severe than when the outcome is not severe. McLaughlin et al. (1990) argued that in a severe outcome situation, an offender holds out little hope that he or she will not be found guilty for his or her behaviors. An offender will then “direct all of his or her efforts toward mitigation” (McLaughlin et al., 1990, p. 266). A less defensive account is more likely to be used by the offender to
According to Robbennolt (2000), the results for each study she examined “were converted into correlation coefficients (r) describing the relationship between outcome severity and the responsibility-related judgment” (p. 2586). A weighted average based on sample size was calculated if more than one effect between outcome severity and responsibility judgment was reported.
1
18 achieve his or her goal of mitigation when the outcome is severe than when the outcome is not severe. McLaughlin, O’Hair et al.’s (1983) study provided support for this hypothesis. McLaughlin, O’Hair et al. had participants report an incident in which their behavior led to a negative event. Participants provided the actual accounts they used in the situation and rated the severity of the negative event. The results suggested that offenders were likely to use concessions when the negative event was severe than when it was not severe. However, McLaughlin, O’Hair et al.’s (1983) results may only provide one side of the story. Schönbach (1990) argued that the self-reported incidents in McLaughlin, O’Hair et al.’s study were incidents that only had minor consequences, such as being late for a meeting. Schönbach argued that if the offense were extremely severe, such as causing bodily harm, the pressure on offenders for taking full responsibility becomes unbearable. To direct some responsibility to factors other than themselves, offenders are expected to use defensive accounts, such as justifications. Schönbach cited studies from criminal justice to support this idea. For example, Henderson and Hewstone (1984) coded explanations provided by inmates about their current convictions. They found offenders overwhelmingly pointed to external reasons, such as characteristics of victims or something about the environment, to explain their criminal behaviors. Justifications and excuses were identified as the most frequently used types of accounts in those inmates’ explanations than concessions. In the case of a severe outcome, there is a possibility that offenders will use a more defensive strategy to take some responsibility off their shoulders than when the outcome is not severe. Schönbach (1990) argued that there are two possibilities about the influence of
19 outcome severity on the defensiveness of accounts. On one hand, defensiveness of the chosen account may be negatively associated with outcome severity. McLaughlin, O’Hair et al.’s (1983) study provided support for this idea. When the outcome of a conflict is severe, offenders expect negative responses from victims: Victims will be angry and will assign more responsibility to offenders, compared to when the outcome of the conflict is not severe. To reduce those negative responses, offenders are more likely to use a less defensive account to amend the possible damages offenders’ behavior has brought to victims. On the other hand, the increase in outcome severity leads to a higher level of defensiveness, which results in a more defensive account. When the outcome is severe, taking full responsibility becomes a greater burden for offenders. For example, admitting the wrong-doing fully to one self should damage offenders’ self image and make offenders look immoral. Offenders will try to defend themselves against the damages to their face by attributing their behavior to situational rather than dispositional causes, or justify their behavior by linking it to a superordinate goal. Results from Henderson and Hewstone’s (1984) study provide support to the idea that offenders use defensive accounts to reduce the responsibility assigned to offenders when the outcome is severe. Schönbach called for more work to examine the influence of outcome severity on the level of defensiveness for a chosen account. Thus, a research question is proposed in the current research: RQ1: How does outcome severity influence the defensiveness of an account chosen by offenders? The direction of the effect of outcome severity on accounts is unclear and may be clarified by understanding the relational closeness between offenders and victims
20 (McLaughlin, O’Hair et al. 1983). Relational Closeness Research on aggression and the relationship between friends provides indirect support for the influence of relational closeness on choice of accounts. Richardson and Green (2006) found that in a conflict, people tend to use direct aggression in response to anger from their romantic partners but use indirect aggression toward their friends, who are less close to them than their romantic partners. According to Richardson and Green, direct aggression “confronts the target face to face, delivering harm through direct verbal or physical means” (p. 2493) whereas indirect aggression refers to when people involved in a conflict “[deliver] harm circuitously, through another person or object” (p. 2493). Richardson and Green’s finding is consistent with Fitzpatrick and Winke’s (1979) argument that in a close relationship offenders are less concerned with the threat of relationship termination than when the relationship is not close. As a result, offenders are likely to express their opinions directly to their close friends and defend their own behaviors, believing that their close friends will not be offended or they can compensate to their close friends at a later time. Following this logic, if a relationship is not so close, offenders need to be cautious when they choose an account to explain their behavior because a defensive account may escalate the conflict and ultimately lead to the dissolution of the relationship. For example, when a conflict occurs between acquaintances, offenders realize that their relationship with victims is much less stable and more likely to be dissolved than if the conflict were with close friends. When a conflict occurs between parties that have a moderately close relationship, compared to the conflict that occurs between close friends, offenders will use a less defensive account,
21 hoping it will reduce the possibility for relationship termination. On the other hand, if there is no established relationship, such as when a conflict occurs between strangers, offenders will be likely to defend or justify their behaviors by using a defensive account because they do not perceive a high possibility of future interaction with the victim. Based on the above arguments, it is hypothesized that: H2: Relational closeness has a curvilinear effect on the defensiveness of accounts used in a conflict situation. In particular, less defensive accounts are used when a conflict occurs between two parties that have a moderately close relationship and more defensive accounts are used when a conflict occurs between parties that are more distant or are closer. When discussing the influence of relational closeness on conflict strategies, neither Richard and Green (2006) nor Fitzpatrick and Winke (1979) took outcome severity into consideration. The findings from Richard and Green’s study may be valid only when the outcome of a conflict is not very severe. When the outcome of a conflict between close friends is not severe, offenders may feel that they can compensate their victims later. Thus, at the moment of the conflict, offenders could justify their behaviors or defend themselves. However, if the outcome is severe enough to possibly damage the relationship, offenders will be likely to admit responsibility and use a less defensive account than when the outcome is not severe. When offenders and victims are acquaintances, the risk of relationship termination and the desire to preserve a good image causes offenders to be willing to use a less defensive account regardless of severity of the outcome. Furthermore, when facing a stranger, with whom one has a small chance of future interaction, offenders will be more likely to use a defensive account if the
22 outcome is severe than if the outcome is not severe. Thus, the current research predicts the following relationship: H3: The relational closeness and the severity of outcome interact to influence the defensiveness of the account used in a conflict situation. (H3a) When a severe conflict occurs between close friends, offenders are more likely to use a less defensive account than when the conflict is not severe. (H3b) When the conflict is between parties that are moderately close, offenders are more likely to use a less defensive account than when the conflict is between parties that are more distant or closer, regardless of the severity of outcome. (H3c) When a severe conflict occurs between strangers, offenders are more likely to use a more defensive account than when the conflict is not severe. The above hypothesis predicts that the defensiveness of the account used in a conflict situation depends on outcome severity and relational closeness. But the hypothesis does not explain the process by which offenders choose certain accounts to explain their behaviors. Weiner’s (1986, 1992, 2006) attribution theory of motivation and emotions provides the framework to examine the cognitive and emotional processes that offenders engage in to determine the accounts used in conflict situations. Attribution Theory Research on accounts is closely related to attribution theory (Fiske & Taylor, 1991; Harvey & Weary, 1984; Heider, 1958; Jones & Davis, 1965). Attribution theory has focused on the process by which people create explanations for their own or others’ actions (Heider, 1958; Jones & Davis, 1965; Kelley, 1986). These explanations allow offenders in a conflict situation to interpret their behaviors and decide an account to use
23 (Orbuch, 1997). Offenders can attribute their behaviors in terms of locus (internal vs. external) and controllability (controllable vs. uncontrollable). Causal inferences influence what tactics people choose to handle conflict. For example, Sillar (1980) found that when people attributed the causes of conflict to others (e.g., their roommates), they used more distributive acts (e.g., explicitly expressing negative evaluations of the other party) to handle conflict than integrative acts (e.g., using neutral or positive statements to describe the other party involved in conflict). How offenders make attributions can also influence their choice of accounts. Weiner’s Attribution Theory of Motivation and Emotion Weiner (1986, 2000) identified three dimensions of causality—locus, stability, and controllability—to explain attribution processes. Locus refers to whether a cause is internal or external to the person. Stability refers to whether a cause is relatively enduring across situations or is situation-specific. Controllability refers to whether a cause can be subject to volitional influence. For example, if a meeting is cancelled because of a hurricane, the cause of the scheduling conflict is considered external, unstable, and uncontrollable. Weiner’s (1986, 1992, 2006) work has focused on using locus and controllability to explain people’s account-giving processes. According to Weiner (1995), when a social contract has been broken, offenders often first consider the antecedent conditions about what happened and why the social contract was broken. Furthermore, offenders make causal inferences, such as whether the reasons for the violation of the social contract are related to offenders or related to environmental factors. Weiner argued that based on the causal inferences, offenders imagine the consequences if the reasons were communicated to victims. Perceptions of
24 the causes of the behavior, such as offenders’ ability, effort, and luck, lead to inferences about responsibility, which in turn generate emotional responses, such as anger, from victims.2 With the anticipated consequences in their mind, offenders will decide on the explanations that they want to communicate to victims or to other people. There could be many antecedent conditions that lead to a conflict. Here the focus is on two contextual factors: outcome severity and relational closeness. People tend to attribute their own success or positive behaviors to internal reasons and failure or negative behaviors to external reasons (i.e., the ego-protective bias; Kelley, 1986). When offenders’ behavior leads to a conflict, offenders are likely to make external attributions about their behavior in question. The more severe the outcome of the conflict is, the more likely offenders will tend to explain their behaviors by pointing to factors outside of them or factors that are not in their control. This attribution may change depending on how close offenders are with the victim. Specifically, when offenders and victims are relationally close, offenders understand the importance of keeping the relationship with victims and also offenders understand that victims may know them very well to figure out why the behavior in question has occurred. When the outcome of the conflict between close friends is severe, offenders tend to attribute their behavior to internal and controllable reasons, as a sign to show sincerity and honesty to their friends and as a sign to show their desire to maintain the relationship. When a conflict occurs between strangers, the concern for the relationship is minimal. Offenders are likely to experience ego-protective bias, especially when the outcome of conflict is severe. Thus, the following prediction is made:
Weiner (1986, 2000) discussed two emotions: anger and sympathy. The current research will only examine anger because the study is placed in a conflict context, within which the arousal of negative emotions is more likely.
2
25 H4: The relational closeness and outcome severity interact to affect offenders’ attributions. When a severe conflict occurs between close friends, offenders will make more internal attribution (H4a) and more controllable attribution (H4b) than when the conflict is not severe. When a severe conflict occurs between strangers, offenders will make more external attribution (H4c) and more uncontrollable attribution (H4d) than when the conflict is not severe. Perceptions of causes, such as internal or external causes and controllable or uncontrollable causes, lead to judgment of responsibility, which in turn generates emotional responses such as anger from victims. When causes are perceived as internal or controllable, offenders often are judged personally responsible for their behavior because they could have done otherwise to prevent the negative event. On the contrary, when causes are perceived to be external or uncontrollable, offenders often are judged as having little or no personal responsibility for their behavior because there was not much they could have done to prevent what had happened. The judgment of responsibility is associated with expected anger. When offenders expect more responsibility to be assigned to them, they will expect victims to feel angrier about the behavior in question. Weiner and his associates studied the influence of attributions on expected responsibility assigned to offenders and expected anger felt by victims (Weiner, Amirkhan, Folkes, & Verette, 1987; Weiner, Figueroa-Muñoz, & Kakihara, 1991; Weiner & Handel, 1985; Yirmiya & Weiner, 1986). For example, Graham, Weiner, and Benesh-Weiner (1995) found that when victims perceived causes of a negative event as controllable, victims held offenders more personally responsible for their behavior and more anger was elicited, as compared to when victims perceived the causes as
26 uncontrollable. When victims assign more responsibility to offenders and feel angrier about offenders’ behavior, they tend to evaluate offenders negatively and show little interest in future interaction with the offenders (Weiner et al., 1987). Weiner’s research suggested that attributing causes of actions to external and uncontrollable reasons benefits offenders because victims are likely to forgive offenders’ behavior and assign less responsibility to them and victims also will feel less angry about the offenders’ behavior, compared to the situation when offenders attribute causes of actions to internal and controllable reasons. Although Weiner’s (1986, 2000) attribution theory of motivation and emotions was mainly applied to the excuse-giving process, his framework has provided insights into understanding offenders’ choice of other types of accounts, such as concession, refusal, and justification. When offenders perceive the causes of their actions as internal or controllable, they expect that victims assign more responsibility to them. A less defensive account acknowledges offenders’ wrong-doing and their desire to take full or partial responsibility, which helps to make victims believe that the offenders’ behaviors were not intentional, and consequently, decreases the possible punishment directed towards offenders. Thus, a less defensive type of accounts, such as concessions, is a better choice than a more defensive type of accounts, such as refusals. On the other hand, when causes are perceived as external or uncontrollable, there is not much that offenders can do to prevent what has already occurred; offenders will be judged as having little or no personal responsibility for their behaviors (Liu & Yao, 2007). Thus, offenders are more likely to use a more defensive account when they perceive the causes of conflict as external or uncontrollable than when they perceive the causes of conflict to be internal or
27 controllable. According to Weiner (1995), assignment of responsibility is associated with emotional reactions toward the behavior in question. Specifically, the assignment of personal responsibility is associated with anger and a tendency to attack or punish. Inspired by Weiner’s theory, Allred, Mallozzi, Matsui, and Raia (1997) investigated the influence of anger and compassion on negotiation performance. Their findings supported Weiner’s (1995) argument: Participants who attributed a higher level of responsibility to their partners reported feeling more angry than participants who attributed a lower level of responsibility to their partners. Liu and Yao’s (2007) study of cultural differences in the association between emotional arousal and attribution of personal responsibility had similar findings. Averill’s (1983) study on anger provided further support to the proposition regarding the relationship between attributions and emotional response. Averill asked participants to report recent events that made them angry. Nearly 80% of the reported incidents that elicited anger involved controllable behaviors. On the contrary, when victims perceived that the causes of a conflict were not controllable, victims were more likely to forgive offenders’ behaviors because victims perceived that the behaviors in question were not intentional. In this case, responsibility was less likely to be assigned to offenders and less anger was aroused than when the causes were believed to be controllable. Based on the above arguments, the following hypotheses are proposed: H5: When offenders make more internal attributions about their behavior in a conflict situation, offenders expect more responsibility to be assigned to them (H5a) and expect more anger to be felt by victims (H5b).
28 H6: When offenders perceive the causes of a conflict as more controllable, offenders expect more responsibility to be assigned to them (H6a) and expect more anger to be felt by victims (H6b). H7: When offenders make more external attributions in a conflict situation, offenders expect less responsibility to be assigned to them (H7a) and expect less anger to be felt by victims (H7b). H8: When offenders perceive the causes of a conflict as more uncontrollable, offenders expect less responsibility to be assigned to them (H8a) and expect less anger to be felt by victims (H8b). H9: When offenders expect more responsibility to be assigned to them, they expect more anger to be felt by victims. H10: When more anger is expected, a less defensive account is used. A causal model is established to explain the attributional processes offenders use to determine the accounts selected in conflict situations (see Figure 2). The rationale of all links has been presented in the research question and the ten hypotheses discussed in this chapter.
29
OS (?1)
POS (?1)
RC (?2)
PRC (?2)
Inter1 (?3) LDA (?9)
IA (?3) Inter2 (?4) CON (?4) Eresp (?7)
EA (?5) Inter3 (?5) Uncon (?6)
Eanger (?8)
Figure 2. Structural relations based on the hypotheses. Note. All exogenous variables are allowed to covary and ?s are also allowed to covary (errors in equations). OS (i.e., ?1) refers to outcome severity manipulation (low outcome severity was coded as -1 and high outcome severity was coded 1) and RC (i.e., ?2) is relational closeness manipulation (quadratic coding: low relational closeness was coded as 1, moderate relational closeness was coded as -2, and high relational closeness was coded as 1). Inter1 (i.e., ?3), Inter2 (i.e., ?4), and Inter 3 (i.e., ?5) are the interactions between outcome severity and relational closeness. In the model testing, ?3 was calculated by multiplying the outcome severity manipulation (coded as 1, -1) and relational closeness manipulation (linear coding: low relational closeness was coded as -
30 1, moderate relational closeness was coded as 0, and high relational closeness was coded as 1). ?4 refers to the coding of the high severity and close relationship condition (coded as 1) versus all other five conditions (coded as 0). ?5 refers to the coding of the interaction between outcome severity and relational closeness: The high severity and low relational closeness condition is coded as 1 and other five conditions are coded as 0. POS (i.e., ?1) is perceived outcome severity, with six indicators. PRC (i.e., ?2) refers to perceived quadratic relation closeness, the manipulation check for the quadratic coding of relational closeness. ?2 was created by squaring the first component score of the six items that measured relational closeness. ?2 only has one indicator in the model. IA (i.e., ?3) is the degree of internal attribution, measured by four of seven items (three items were deleted based on confirmatory factor analysis and principal components analysis results; see details in the result section for the formal study). CON (i.e., ? 4) is controllability, with five indicators. EA (i.e., ?5) is the degree of external attribution, measured by seven items. Uncon (i.e., ?6) is uncontrollability, with five indicators. Eresp (i.e., ?7) refers to expected responsibility assigned to the offender, measured by six items. Eanger (i.e., ?8) is expected anger, measured by six items. And LDA (i.e., ?9) refers to the measure of the likelihood of selecting a defensive account. ? 9 only has one indicator and the detail for obtaining the composite measure for ?9 was discussed in Pilot Study 5 and the formal study.
31 CHAPTER III Method This chapter describes the proposed stimulus message and instruments, the pilot studies that tested these proposed materials, and the methods of the formal study. Scales used in the current research were created by the researcher, and pilot studies were conducted to evaluate the effectiveness of the inductions and the measurements to be used in the final study. The formal study tested the causal model (see Figure 2) and the hypotheses. The pilot studies and the formal study were approved by the University of Maryland Institutional Review Board on May 28, 2007 (for the first two pilot studies) and on February 12, 2008 (Pilot Study 3 to Pilot Study 6 and the formal study). Data for Pilot Study 5 were collected at Villanova University and all other studies were conducted at University of Maryland. Approvals for conducting Pilot Study 5 at Villanova University were obtained from the University of Maryland and the Villanova University Institutional Review Boards on April 10 and April 15, 2008, respectively. Data were collected during the summer of 2007 and between March and June, 2008. Overall Design, Stimulus Message, and Instruments Overall Design An experiment was conducted to test the causal model, research question, and the hypotheses. Outcome severity and relational closeness were manipulated in a hypothetical scenario that described an interpersonal conflict in a group project at school. Two levels of outcome severity and three levels of relational closeness (high vs. moderate vs. low) were created in the hypothetical scenario. Six experimental conditions were developed. Attributions (the degree of internal attribution, the degree of external
32 attribution, controllability, and uncontrollability), anticipated consequences (expected responsibility assigned to offenders and expected anger from victims), the defensiveness level of each type of accounts (concession, justification, excuse, and refusal), and the likelihood of selecting each account were measured.3 Participants were randomly assigned to one of six conditions. They read a hypothetical conflict scenario, were told to imagine they were the offender in the scenario, and answered questions afterwards from the offender’s perspective. Hypothetical Interpersonal Conflict Scenario Two criteria need to be fulfilled to create a hypothetical conflict scenario. First, the conflict incident used in the scenario should be familiar to participants so that participants can easily relate to the situation and answer questions without difficulty. Second, the current research focuses on the accounts offenders use to explain their behaviors. However, not every incident invites an explanation. Participants may use avoidance to handle conflicts, but avoidance is not the focus of the current research. To ensure an account is offered, the need to explain a wrong-doing should be clearly demonstrated in the scenario. Pilots Studies 1 and 2 were used to create a hypothetical scenario that meet the above two criteria. Details are provided in the section on pilot studies. Based on the results of Pilot Studies 1 and 2, a scenario about a conflict between group members for a class project was chosen to be developed as the stimulus message in the proposed study. Group work is a familiar practice for college students. Because college students were the
Pilot Study 5 measured the defensiveness level of each account and the formal study measured the likelihood of selecting each account. Results from Pilot Study 5 and the formal study were combined to create the composite measure of choosing accounts as the dependent measure in model testing. Details can be found in the discussion of Pilot Study 5.
3
33 major source of participants in the studies, using a hypothetical scenario that involves a conflict between group members is easily understood by participants. To ensure that an account would be necessary, the hypothetical conflict scenario clearly specified an offense conducted by the offender. In particular, the hypothetical scenario described a group project between two classmates: writing a review paper together. One wrote his or her portion of the review paper based on the wrong readings, which reduced the grade for both group members. Participants were told to take the role of the offender and that an explanation is necessary when the scenario stated that the offender’s behavior brought negative consequences to the victim. The details of the hypothetical conflict scenario can be found in the section on Pilot Studies 1 and 2. Role Playing in a Hypothetical Conflict Scenario A hypothetical conflict scenario was used for several reasons. First, Martin (2006) argued that a hypothetical scenario allows the researcher to systematically manipulate variables of interest, minimizing the influence of possible extraneous variables on dependent variables. Using a hypothetical conflict scenario, different levels of outcome severity and relational closeness may be manipulated and their influence on offenders’ choice of accounts can be examined. Second, a hypothetical scenario can present respondents with situations that they may actually encounter, allowing researchers to "tap into the expectations and reactions which they (respondents) would have in similar social circumstances" (Gerber, 1994, p. 4). Third, as argued by Converse and Presser (1986), a hypothetical scenario can keep respondents’ focus on the variables intended by researchers. Given the above reasons, a hypothetical scenario was used in the current research and participants were asked to take the role of the offender in the scenario and
34 answer questions from the offender’s perspective. Naming the Hypothetical Other Party Participants were told to imagine that they were the offender in the group project described in the hypothetical scenario; the scenario used “you” to refer to the participants. The teammate (i.e., the victim) in the scenario needs a name to ensure the flow of the story. Although participants may find it easier to relate to scenarios that use actual names (Wang, 2006), actual names often are associated with specific gender and racial information, which may influence how participants process the stimulus message. For example, a labor market study by Bertrand and Mullainathan (2003) found that a very White-sounding name on a resume received significantly more callbacks for interviews than a very African-American-sounding name. Moreover, El-Sheikh, Buckhalt, and Reiter (2000) discovered that different degrees of emotional arousal are associated with gender in hypothetical scenarios. Men perceived more emotional arousal with a male hypothetical character than women perceived with a female counterpart in a hypothetical situation (cited in Wang, 2006). To ensure that participants will not be influenced by the gender and racial information attached with actual names, the current research decided to use “Person B” instead of an actual name for the victim in the scenario. Instruments Perceived outcome severity. The hypothetical scenario varied two levels of outcome severity (high severity vs. low severity). A list of measures was developed to assess whether the manipulation of outcome severity was successful (see Appendix A). An example item for outcome severity is, “how severely does your mistake affect Person B?” Magnitude scales were used in the study where 0 indicates not severe at all and 100
35 indicates a moderate level of severity. The items were tested and revised based on Pilot Study 3. Perceived relational closeness. Three levels of relational closeness were manipulated in the hypothetical scenario. A list of measures was developed to examine whether the manipulation of relational closeness was successful (see Appendix A). An example question is, “how familiar are you with Person B?” Magnitude scales were used in the study where 0 indicates not familiar at all and 100 indicates a moderate level of familiarity. The items were tested and revised based on Pilot Study 3. Attributions. Although studies (e.g., Ross, 1982) have examined attributions in terms of causal (i.e., the degree of internal attribution is at one end of the continuum and the degree of external attribution is at the other end) and controllability (i.e., controllability and uncontrollability are at the two ends of the continuum) dimensions, whether internal attribution is opposite to external attribution and controllability is opposite to uncontrollability remains questionable (Harvey & Weary, 1984). The current research does not want to assume the relationships among the four types of attribution. Thus, these four types of attributions were measured separately: the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability. Different attributions regarding the offender’s behavior were listed and participants indicated their belief about how likely each reason influenced his or her behavior (see Appendix B). Again, magnitude scales were used. Example questions include “how likely is that what has happened was due to your own carelessness?” for the degree of internal attribution; “how likely is that what has happened was due to the professor not giving
36 clear instructions for the assignment?” for the degree of external attribution; “how likely is that you could have prevented this mistake?” for controllability; and “how likely is that what has happened was out of your control?” for uncontrollability. In all magnitude scales, 0 = not likely at all, and 100 indicates a moderate level of likelihood. Anticipated consequences. Anticipated consequences referred to two variables: expected responsibility assigned to the offender and expected anger felt by victims. For expected responsibility, participants were asked to indicate their expectation about how much responsibility would be assigned to them. A magnitude scale was used with 0 meaning not expected to be responsible at all and 100 indicating a moderate level of expected responsibility. A sample item is, “how much responsibility would Person B assign to you for what has happened?” For expected anger, participants indicated the level of expected anger felt by victims by a magnitude scale, where 0 means not angry at all and 100 indicates a moderate level of expected anger. An example item is, “how annoyed do you expect Person B to be?” The complete list of items that measure both expected responsibility and expected anger is included in Appendix C. Expected choice of accounts. Unlike traditional accounts research in which coders code actual accounts that participants provide to explain their behaviors into different types of accounts, a list of twenty accounts (5 concessions, 5 justifications, 5 excuses, and 5 refusals) was developed in Pilot Studies 3 and 4. In the formal study participants were asked to indicate the likelihood that they would choose each account in the hypothetical conflict situation by using a magnitude scale (see Appendix D for the complete list of measures on choice of accounts). On that scale, zero means not likely at all and 100
37 indicates a moderate level of likelihood. The hypotheses proposed in the current research predicted that the influence of outcome severity and relational closeness was not only on the likelihood of selecting accounts but also on the defensiveness level of accounts. For example, Hypothesis 3a states that, “when a severe conflict occurs between close friends, offenders are more likely to use a less defensive account than when the conflict is not severe.” Only using the likelihood of selecting accounts is not enough to test the hypotheses. To create a measure that incorporates both the likelihood and the defensiveness level of each account, Pilot Study 5 was conducted to determine the defensiveness of each type of accounts and the results of Pilot Study 5 were used with the results in the formal study to create the composite measure needed for hypothesis testing (i.e., LDA: the likelihood of selecting a defensive account). The details of creating the composite measure were presented in the discussion of Pilot Study 5 and the formal study. Summary. This section has introduced the overall experimental design and the operationalizations of the relevant constructs proposed in the study: the hypothetical scenario, the name choices in the hypothetical scenario, and the instruments—for manipulation checks, the attributional processes, anticipated consequences, and the choice of accounts. The realism of the experimental inductions and the effectiveness of the instruments proposed for the current research were tested in six pilot studies, which are described below. Pilot Studies Six pilot studies were conducted in the summer term (June to July, 2007) and
38 between March and June, 2008. The majority of participants in each pilot study were undergraduate students enrolled at the University of Maryland (except that Pilot Study 5 was conducted at Villanova University). Participants received a small amount of extra credit from their course instructors in exchange for their participation. Students who participated in the summer term were informed not to participate in any other portions of this research that would take place later. Because all the participants had to sign up for the studies through an online participation pool program, students’ online identification numbers were used to ensure that they could only participate in one of the studies. All participants were debriefed in detail after they completed their participation. Pilot Study 1: Collecting Conflict Scenarios Purpose. The purpose of this pilot study was to collect self-reported conflict incidents to provide information for developing a hypothetical scenario that could be used in the formal study. Participants. Self-reported conflict scenarios were collected in the United States and China.4 Forty-nine participants were recruited from undergraduate communication courses at the University of Maryland. The age range for American respondents was 18
National cultures (i.e., China vs. the U. S.) were considered as one independent variable in the prospectus for the current research. Data were collected in both the United States and China in Pilot Studies 1, 2, and 3. Before the researcher could continue collecting data for the rest of the pilot studies and the formal study, two issues arose that interrupted the data collection process. First, it became apparent that the incentives were different for participants in each culture. Although obtaining extra credit in exchange of the participation is a common practice at the American university where the study was conducted, Chinese instructors would not give permission for their students to receive extra credit. Chinese students participated in the study voluntarily. Differences in incentives may be a factor that contributes to the differences in the results, making the conclusions problematic. Second, massive protests against media bias occurred in China after collecting data for Pilot Study 3, with college students being the major protesters. Given that this study measures perceptions in a hypothetical conflict situation, there is a concern that Chinese perceptions about conflict might be influenced by these protests. Consequently, the validity of the results may be questionable. Therefore, the researcher decided not to include national culture as a variable in the formal study. Results for Pilot Studies 1 and 2 still included Chinese data to create a hypothetical conflict scenario that could be used in data collection in China in the future. Pilot Study 3 only analyzed data collected in the United States because the formal study would only recruited American participants; results based on Chinese data were not relevant.
4
39 to 39 (M = 21.73, SD = 3.38, Mdn = 21.00). Seventy-six percent of participants were female (n = 37). Forty-four Chinese participants were recruited from undergraduate and graduate courses in two Chinese northern cities (i.e., Beijing, the capital of China and Changchun, the capital of Ji Lin Province). The age range for Chinese participants was 19 to 44 (M = 24.10, SD = 5.08, Mdn = 23.00). Fifty-seven percent of Chinese participants were female (n = 25). Self-reported conflicts were collected in both China and the United States to develop a hypothetical conflict scenario that could be used in different cultures for the future studies. Procedure. This pilot study was conducted as an online survey. American participants completed the survey through an online participation pool program and Chinese participants completed the survey through an online survey Web site. Information about the purpose of the study was provided, followed by a consent form. Participants were asked to read the consent form and provide their electronic signature if they agreed to participate in the study. Only after they agreed, participants were allowed to continue with the questionnaire. The questionnaire had two sections. The first section asked participants to recall a situation in which a conflict occurred recently between them and another person. Specifically, participants were asked to recall a conflict that happened due to their behavior. Participants described the conflict in detail; for example, participants were asked to provide information about the relationship between them and the other person involved in the conflict, the cause of the conflict from participants’ perspective and from the other person’s perspective, the consequences of the conflict, the emotions participants experienced and they expected the other party to have experienced at the time of the
40 conflict, etc.. In the second section, participants were instructed to recall in detail their explanation about the behavior that led to the conflict they described. All the questions were open-ended, allowing participants to provide details of a recent conflict they experienced. See Appendix E for the complete questionnaire. Results. A total of 93 different conflict scenarios were generated. A sorting task (Pilot Study 2) was conducted to sort the reported incidents based on the outcome severity and the relational closeness between the two parties involved in the conflict. Pilot Study 2: Developing a Hypothetical Conflict Scenario Purpose. The purpose of Pilot Study 2 was to develop a hypothetical conflict scenario based on the qualitative responses collected in Pilot Study 1. Participants. Seven Americans and six Chinese were recruited to participate in the second pilot study: sorting ninety-three conflict scenarios that were collected in Pilot Study 1. The ninety-three conflict scenarios included some incidents collected in China. Therefore, Chinese participants were recruited to ensure the sorting would be not biased by having only American participants. The age range for American sorters was 17 to 21 (M = 19.14, SD = 1.35, Mdn = 19.00), whereas Chinese sorters’ age range was 26 to 33 (M = 29.33, SD = 2.50, Mdn = 29.00). Six of the American sorters were female and five of the Chinese sorters were female. American sorters were undergraduate students enrolled in communication courses at the University of Maryland. Due to the difficulty of locating Chinese natives in undergraduate courses, Chinese graduate students at the same university were asked to participate in the sorting task. Procedure. Sorters arrived at a pre-arranged classroom alone. The researcher explained the purpose of the study and obtained the consent from participants that agreed
41 to be part of the study. Each sorter was provided with ninety-three numbered index cards. On each index card, there was one conflict scenario collected in Pilot Study 1. Sorters were instructed to first read over all ninety-three conflicts carefully and ask questions if they had difficulty understanding any scenario. After the researcher made sure that sorters fully understood the 93 scenarios, they were asked to sort the scenarios based on its outcome severity and relational closeness. Outcome severity was defined in the instructions as how serious the consequence of the conflict might be and relational closeness was the extent to which the people involved in the conflict were close to each other. Based on the hypotheses, a sorting table with two levels of outcome severity (high vs. low) and three levels of relationship closeness (high vs. moderate vs. low) was presented to the sorters. Sorters were instructed to put the number of the card in the most appropriate cell in the sorting table. They were told that each conflict could be put in one and only one cell in the table. See Appendix F for the complete instructions and the sorting table used in the Pilot Study 2. Results. Looking at frequency of participants’ sorting of ninety-three scenarios indicated that five incidents were evenly sorted into six cells (see Appendix G for these five incidents reported by participants in Pilot Study 1), which allowed the researcher to revise the incidents easily based on the independent variables. Among all five incidents, one scenario about a conflict in school projects was chosen for the hypothetical conflict scenario to be used in the formal study because participants in the formal study were to be mainly undergraduate students. Working on a group project for a class is a familiar practice to the participants. The selected incident was revised to reflect the variations of two independent
42 variables (outcome severity and relational closeness). Six versions of the hypothetical conflict scenario were generated, with two levels of outcome severity and three levels of relational closeness. All six versions had 12 sentences, with an average of 17 words per sentence and a Fresch reading ease rating of 74.5 (Microsoft Office, 2003). All scenarios started by defining relational closeness between an offender and a victim, followed by the description of a conflict between the offender and the victim and the consequence of the conflict. For example, in the high relational closeness conditions, the relationship between the offender and the victim reads: “You and Person B grew up together and attended the same university. The two of you have been best friends for many years.” The moderate relational closeness condition says “You and Person B were in couple of classes together before. Although the two of you know each other, you only see each other in class.” When the relational closeness between the offender and the victim is low, the scenario reads “You and Person B do not really know each other. You happened to sit next to each other on the first day of class.” Similarly, the two levels of outcome severity were manipulated in the scenarios. The details of the manipulations are discussed in the method section of the formal study; see Appendix H for all six versions of the hypothetical conflict scenario. Pilot Study 3: Manipulation Checks Purpose. The purpose of this pilot study was to test whether the manipulations of the two independent variables were effective. Although the hypothetical conflict scenario was created in a way to be suitable in both China and the United States, only data collected from Americans participants were analyzed because the formal study would not be conducted outside of the United States.
43 Participants. Participants (N = 80) were recruited from undergraduate communication courses at the University of Maryland. The age range was 18 to 25 (M = 20.38, SD = 1.58, Mdn = 20.00). Eighty-one percent of participants were female (n = 65). Sixty-five percent of participants identified themselves as Caucasian. Study design and procedure. A 3 (relational closeness: high vs. moderate vs. low) X 2 (the severity of an outcome: high severity vs. low severity) independent groups experiment was conducted. Participants were recruited through a departmental participation pool program, which connected participants to an external online survey Web site to participate in the study. Participants were randomly assigned to read one of six hypothetical conflict scenarios and they were asked to imagine they were the offender in the scenario. After reading the stimulus message, each participant was first asked to provide an explanation about the behavior described in the scenario (i.e., wrote their portion of paper based on the wrong readings) from the perspective of the offender. After that, two questions that were not relevant to the hypothetical scenario were provided to participants for them to practice using magnitude scales before they answered the questions. Those two questions were: “How much do you like the food offered at the dining hall of your school?” and “How much do you like the service at the main library of your school?” Participants were told to continue with the questionnaire only when they felt ready to use the scale. The questionnaire had measures for perceived outcome severity and perceived relational closeness (see Appendix I for the instructions and measures used in Pilot Study 3). After participants completed the questionnaire, they were debriefed. The same two practice questions were used in Pilot Studies 5 and 6 and the formal study.
44 Data coding and data transformation. Outcome severity has two levels (high and low) and relational closeness has three levels (high, moderate, and low). When coding the data, high severity is coded as 1 and low severity is coded as 0. Three levels of relationship closeness (high, moderate, and low) were coded as 1, 0, and -1, respectively. Participants used any nonnegative number to indicate their views about the hypothetical conflict. Data collected in Pilot Study 3 were submitted to a descriptive analysis; the skewness and kurtosis were examined to see whether a transformation was necessary. The ratio of the skewness over the standard error for each item was calculated and if the ratio was greater than 1.96, the data were significantly positively skewed and if the ratio was smaller than -1.96, the data were significantly negatively skewed (Frey, Botan, & Kreps, 2000). Using the (-1.96, 1.96) cutoff rule, all eight indicators (four for perceived relational closeness and four for perceived outcome severity) resulted into positively skewed data. A data transformation was needed. A power transformation is considered one strategy to transform positively skewed data resulting from magnitude scales (Bauer & Fink, 1983). A transformation can also helps with the assumption of homescedasticity that was used for testing the manipulation checks. When the variance of the residuals is not uniform across all levels of the independent variables, the homoscedasticity assumption is violated. Kline (2005) argued that a transformation “may remedy heteroscedasticity due to nonnormality” (p. 5). Before transformation, the data set was first trimmed at the 90th percentile to eliminate the extreme values (e.g., the 90th percentile for one item to measure outcome severity was 500; thus, any score larger than 500 was converted to 500). The data after trimming were submitted to a power transformation using the following formula:
45 New indicator = (original indicator) ?, different values of ? were tried in the power transformation and when ? = 0.40, the skewness of each indicator was not significant (p > .05). Therefore, 0.40 was used as the value of ? in this pilot study. For the power transformations conducted in the following pilot studies and the formal study, the same process was used: Different values of ? were tried and the one that ensured the non-significant skewness of all or the majority of indicators was selected. Instruments. A 4-item magnitude scale was used to assess the effectiveness of outcome severity manipulation (see Appendix I). An example item is “How much influence does your mistake have on Person B’s scholarship for the next year?” Magnitude scales were used in this pilot study where 0 indicated not severe at all and 100 indicated a moderate level of outcome severity. Confirmative factor analyses (CFAs) were conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results indicated that the scale for perceived outcome severity was not unidimensional, ?²(2, N = 80) = 7.01, p < .05; NFI = 0.97; CFI = 0.98; SRMR = 0.03. However, when a principal components analysis with no rotation was conducted on the four items for perceived outcome severity, only one eigenvalue was larger than 1 (eigenvalue = 3.10), which explained 76% of the variance in the items. The first component score was saved as the manipulation check. The Cronbach’s alpha for the sum of all four items testing perceived outcome severity was .71. Another 4-item magnitude scale measured the perceived relational closeness. Participants were asked questions such as “How familiar are you with Person B?” Magnitude scales were used in this pilot study where 0 indicated not close at all and 100
46 indicated a moderate level of perceived relational closeness. Confirmative factor analyses (CFAs) were conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results supported a one-factor structure for the perceived relational closeness, ?²(2, N = 90) = 1.73, p > .05; NFI = 1.00; CFI = 1.00; SRMR = 0.00. A principal components analysis without rotation was conducted to obtain the first component score for perceived relational closeness, which was used as the manipulation check for relational closeness. The Cronbach’s alpha for the sum of all four items testing perceived relational closeness was .96.5 Results. Two-way ANOVAs were performed to assess whether the manipulations of outcome severity and relational closeness were successful. In the first ANOVA, the dependent variable was perceived outcome severity and the independent variables were the manipulations of outcome severity and relational closeness. Results indicated that (a) the manipulation of outcome severity influenced perceived outcome severity significantly, F(1, 74) = 12.09, p < .01, ?² = .14, (b) relational closeness did not have a significant influence on perceived outcome severity, F(2, 74) = 0.60, p = .55, ?² = .02, and (c) no significant interaction between the manipulations of outcome severity and relational closeness was found, F(2, 74) = 0.11, p = .90, ?² = .00. Participants in the high outcome-severity condition (M = .38, SD = 1.12) regarded the outcome to be more severe than those in the low outcome-severity condition (M = -.36, SD = .71). The manipulation
In Pilot Study 3, both the scales for perceived outcome severity and perceived relational closeness had four items. The reliability for the scale of the perceived outcome severity was acceptable but not very satisfactory. Additionally, CFA results did not yield a unidimensional factor structure for perceived outcome severity. Because only four items were used, dropping items from the scale was not appropriate (a saturated measurement model would result if any item were dropped from the perceived outcome severity scale). To enhance the reliability of the scales for the formal study, additional items need to be added (Wrench, Thomas-Maddox, Richmond, & McCroskey, 2008), which also allows the researcher to drop one or more items if the factor structure were not unidimensional. Two items were added to the scale for perceived outcome severity, which were phrased similar to the other items. Two additional items were added to the scale for perceived relational closeness as well.
5
47 of outcome severity was successful. The second ANOVA had perceived relational closeness as the dependent variable and the manipulations of outcome severity and relational closeness as the independent variables. The results suggested that (a) relational closeness was manipulated effectively, F(2, 74) = 72.58, p < .01, ?² = .60, (b) the manipulation of outcome severity did not have a significant effect on perceived relationship, F(1, 74) = 0.47, p = .50, ?² = .01, and (c) no significant interaction was found between the manipulations of outcome severity and relational closeness, F(2, 74) = 0.33, p = .72, ?² = .01. Participants in the high relational closeness condition (M = 1.19, SD = .86) reported being closest to Person B than those in the moderate closeness condition (M = -.43, SD = .47) and in the low relational closeness condition (M = -.64, SD = .34). A linear relationship between the three levels of relational closeness was established, F(1, 79) = 104.13, p < .01, ?² = .56. Limitation and revision. The effect size for perceived outcome severity was only .14. According to Cohen (1988), the threshold for a small effect size is .20. Therefore, the hypothetical scenario needed to be revised to get a larger effect size for the outcome severity manipulation. To revise the scenario, five undergraduate students were interviewed on campus (4 were female). Two versions of the hypothetical scenarios (high vs. low outcome severity with relational closeness controlled) were read by these five interviewees. The researcher then interviewed the participants with questions such as whether the story was realistic, what the researcher intended to vary in two scenarios, and how to enlarge the difference between low and high level of outcome severity. Based on interviewees’ suggestions, appropriate revisions were made to the hypothetical conflict scenario and a
48 second test was conducted to specifically assess the manipulation of outcome severity. Two versions of the hypothetical conflict scenarios were used in the second test with relational closeness controlled at the high closeness level. In the second test, sixteen participants were recruited from an introductory communication course at the University of Maryland. Their age range was 18 to 19 (M = 18.35, SD = 0.49, Mdn = 18.00). Seventy-one percent of participants were male (n = 12), and fifty-nine percent of participants identified themselves as Caucasian. Participants were randomly assigned to read one of two versions of the hypothetical conflict scenario (high vs. low outcome severity). Participants were not asked to provide explanations but simply completed the manipulation checks, imagining they were the offender in the scenario; the same four-item scale in the first test of Pilot Study 3 was used. The data were positively skewed and a power transformation with ? = 0.50 was conducted to meet the assumption of normality for the analyses of the variance. A CFA was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The Chi-square value suggested a unidimensional factor structure although the fit indices were not satisfactory, ?²(2, N = 16) = 3.74, p > .05; NFI = 0.90; CFI = 0.93; SRMR = 0.05. The Cronbach’s alpha for the sum of all four items testing perceived outcome severity was .83. A principal components analysis without rotation was conducted and the first principal component score was saved for the analysis of the variance. In the ANOVA, the dependent variable was perceived outcome severity and the independent variable was the manipulation of outcome severity. Results indicated a significant main effect of outcome severity manipulation, F(1, 14) = 6.66, p < .05, ?² = .32. Participants in the high outcome-severity condition (M = -.12, SD = .20) perceived the consequence of the
49 conflict to be more severe than participants in the low outcome-severity condition (M = -.36, SD = .17). Based on these results, the induction of outcome severity was successful. The revised scenarios were used in the later studies. Pilot Study 4: Creating the Measure of Accounts Purpose. Pilot Study 4 was conducted to develop a measure for accounts. Coders and procedure. In Pilot Study 3, participants were asked to imagine they were the offender and to provide an explanation about the behavior described in the hypothetical scenario presented to them. Eighty explanations were collected from participants. Two trained coders coded the eighty explanations into four types of accounts examined in the current research (i.e., concession, justification, excuse, and refusal), and coders were also asked to select the five most frequently used explanations for each type of accounts. Coders used a coding scheme adapted from Fritsche's (2002) review of accounts (see Table 2). Results. Two undergraduate student coders (both females) were trained on the list of the twenty actual explanations collected in Pilot Study 3 until they obtained acceptable inter-coder reliability. They were instructed to code each explanation in terms of four types of accounts (i.e., concession, justification, excuse, and refusal). The overall initial agreement among two coders was 76% and intercoder reliability was calculated using Cohen’s (1960) kappa. The kappa equaled .68. Differences in coding were then discussed and resolved by the two coders. The coders also listed the five most frequently used explanations in each type of accounts, which created a list of twenty statements—five statements for each account type. Those statements were developed into a measure of accounts used in Pilot Studies 5 and 6 and in the formal study (Appendix D presents the
50 measure of accounts). Pilot Study 5: Testing the Defensiveness of Each Account Purpose. Schönbach (1990) argued that each type of accounts is associated with a certain defensiveness level. He arbitrarily placed the four types of accounts on a continuum of defensiveness (see Figure 1 in the section of literature review). However, little empirical research examined the location of each type of accounts on the defensiveness continuum. Pilot Study 5 served two purposes. First, the results would provide empirical evidence to support or disconfirm Schönbach’s (1990) ranking of the four types of accounts on the defensiveness continuum. Second, in the formal study, participants indicated their likelihood of using each type of accounts in a conflict situation. However, in the hypotheses, the likelihood of using a less or more defensive account was predicted. To be consistent with the hypotheses, a composite score (LDA: the likelihood of selecting a defensive account) was developed by using the results on defensiveness of each type of accounts from Pilot Study 5 and the results on likelihood of selecting each type of accounts from the formal study. LDA, was used for the hypothesis testing. The formula to calculate LDA is as follows: LDA = ? (Di • Li) / ? Li where Di refers to the defensiveness level of each type of accounts and Li refers to the likelihood of selecting each type of accounts, where i = 4 to represent the four types of accounts (i.e., concession, justification, excuse, and refusal). Participants. Twenty-two participants were recruited from undergraduate communication students at Villanova University. Their age range was 20 to 21 (M = 20.59, SD = 0.50, Mdn = 21.00). Eighty-two percent of participants were female (n = 18),
51 Table 2 Coding Scheme for Four Types of Accounts, Adapted from Fritsche’s (2002) Review (pp. 389-394) Types of accounts Refusal Descriptions An offender responds to a reproach that the behavior has been norm contradictive by questioning the legitimacy of the reproach. This is made possible by explicit refusal to deal with the reproach, by rejecting the validity of the salient social norm, by denying the behavior or relevance of the salience norm, or by attacking the reproaching instance. Excuse An offender responds to a reproach that the behavior has been norm contradictive by denying the connection between self and the critical behavior. The offender points to a restricted freedom of decision, which can emerge from a lack of control, an absence of planning or missing an opportunity to plan the behavior, as well as from external constraints and incentives that are not under the offender’s control. Justification An offender responds to a reproach that the behavior has been norm contradictive by denying that there is a connection between the behavior and the salient norm. The offender disputes that he or she has violated a salient norm. Concession An offender neither denies the connection to the deed, nor
52 its norm-contradictive character, nor bypassing it. Instead the guilt is admitted and the behavior is regretted.
and eighty-two percent of participants identified themselves as Caucasian (n = 18). Procedure. Participants completed the study through an online survey Web site. The link to the study was distributed to participants by an e-mail. Each participant was asked to provide an electronic signature, indicating their agreement to participate in the study, before completing the questionnaire. The questionnaire had two parts. The first part introduced a conflict scenario to participants. The purpose of this pilot study was to examine the defensiveness level of each type of accounts, not to test the influence of outcome severity and relational closeness on how people account for their behaviors in conflict situations. Therefore, outcome severity and relational closeness were not varied in the hypothetical conflict scenario presented in this pilot study. The second part of the questionnaire had four types of accounts generated in Pilot Study 4. Each type (i.e., concession, justification, excuse, and refusal) had five accounts. Twenty accounts were listed in total. For example, one of the refusals was, “I know the readings I read were the right material for the paper. I did not find anything wrong to what I contributed to the paper.” Participants indicated the defensiveness level of each account by using a six-item magnitude scale, which generated a total of one hundred and twenty questions (6 X 20) in the questionnaire. The six items for each account were the same. An example question reads, “To what extent were you being self-protective?” (Appendix J lists the scale measuring the defensiveness level of the accounts). To familiarize participants with using magnitude scales, the two practice questions in Pilot
53 Study 3 were used. Data transformation. Participants used nonnegative numbers to indicate their views about the defensiveness level of each account provided by the researcher. The data were submitted to the descriptive analysis for the skewness and kurtosis to see whether a transformation was needed. A (-1.96, 1.96) cutoff rule was used to determine whether the data were approximately normal (Kline, 2005). Based on the cutoff rule, fourteen of the one hundred and twenty items were approximately normal and the others were positively skewed. A data transformation was necessary to meet the assumption of normality for the later analyses. The data set was trimmed at the 90th percentile to eliminate the extreme values (e.g., the 90th percentile for one item to measure defensiveness was 400; thus, any score larger than 400 was converted to 400). A power transformation was used on the data set, with 0.40 selected as the value of ? that ensured the skewness of the majority of items after transformation was not significant (p > .05). Instruments. A list of twenty accounts was provided to participants. For each account, a 6-item magnitude scale was used to measure the defensiveness level of that account. For example, after reading one concession (e.g., “I am sorry that I misunderstood the instructions. Is there any way that I can make it up to you?”), participants were asked: “To what extent were you being self-protective?”, “To what extent were you guarded?”, “To what extent were you sticking out for yourself?”, “To what extent were you being assertive?”, “To what extent were you shifting responsibility?”, and “To what extent were you being defensive?” Correlations between the six items for each account were calculated (see Table 3). Four of the six items correlated with each other significantly across all twenty accounts (i.e., being self-
54 protective, being guarded, sticking out for one self, and being defensive). Therefore, the other two items (i.e., being assertive and shifting responsibility) were dropped from the scales. Twenty reliabilities were calculated with the remaining four items of the defensiveness scales. The Cronbach’s alphas for the sum of the four items measuring the defensiveness of all twenty accounts were acceptable, ranging from .68 to .92 (see the complete list of reliability coefficients in Table 4). The averages of the four-item scales after transformation were calculated for the following analysis. Results. An ANOVA with a repeated measure was conducted with four factors (i.e., the defensiveness level for concessions, justifications, excuses, and refusals). The
55 Table 3 Intercorrelations between Items for Defensiveness of Each Account Items 1 2 3 4 5 6
Concession 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Concession 2 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Concession 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 1 .47* 1 .33** .26 1 .39 .12 .82** 1 .28 .45* .19 .17 1 .37 .49* .28 .09 .56** 1 .67** 1 .64** .73** 1 .20 .45* .70** 1 .34 .08 .05 -.02 1 .46* .54** .67** .34 .25 1 1 .84** 1 .40 .54** 1 .35 .52** .62** 1 .19 .37 .56** .35 1 .36 .55** .59** .42 .59** 1
56 6. Defensive Concession 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Concession 5 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 2 (n = 22) 1 .33 1 .85** .24 1 .04 -.39 .21 1 -.22 .24 -.13 .22 1 .41 .16 .42* .30 .23 1 1 .21 1 .62** .61** 1 .61** .14 .38 1 .37 .38 .61** .12 1 .19 .81** .60** .11 .47* 1 1 .79** 1 .67** .57** 1 .11 .46* .44* 1 .38 .49 .52 .20 1 .30 .37 .56** .56** .23 1 1
57 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Justification 5 (n = 22) 1. Self-protective 2. Guarded 1 .32 1 .88** .13 .23 -.18 .17 .82** .85** .23 1 .44* 1 .95** .45* 1 .48* -.04 .27 1 .56** .20 .55** .42 1 .86** .48* .82** .44* .57** 1 1 .34 1 .91** .17 1 .35 .21 .35 1 -.04 .55** -.18 .28 1 .84** .28 .86** .30 -.04 1 1 .35 1 .93** .33 1 .00 -.00 .02 1 -.08 .28 -.09 -.02 1 .60** .14 .59** -.01 -.01 1
58 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 2 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 1 .86** 1 .71** .50* 1 .46* .41 .65** 1 .26 .39 .29 .28 .46* .75** .37 .38 1 .86** 1 .74** .69* 1 .15 .28 .24 1 .74** .76** .76** .05 1 .55* .54* .74** .20 .71** 1 1 .55** 1 .66** .22 1 .23 .03 .27 1 .45* .13 .43* .12 1 .75** .44* .59** .35 .60** 1 1 .32 1 -.03 -.11 1 .75** .25 .08 1
59 5. Shifting responsibility 6. Defensive Excuse 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Excuse 5 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 1 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive 1 .53* 1 .53* .23 1 .05 -.32 .19 1 .17 .30 .16 .07 1 .52* .63** .70** .05 .22 1 1 .77** 1 .41 .59** 1 .49* .53* .34 1 .51* .34 .43* .14 1 .57** .47* -.10 .60** .11 1 1 .59** 1 .71** .50* 1 .46* .41 .65** 1 .26 .39 .29 .28 1 .46* .75** .37 .38 .14 1 1 .14 1
60 Refusal 2 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 3 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 4 (n = 22) 1. Self-protective 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive Refusal 5 (n = 22) 1. Self-protective 1 .58** .56** -.02 .80** .84** 1 .42 1 .50* .48* 1 .42* .21 .60** 1 .68** .41 .78** .86** 1 .56** .44* .90** .53* .70** 1 1 .83** 1 .76** .76** 1 .62** .49* .74** 1 .69** .55** .79** .87** 1 .71** .56** .86** .90** .94** 1 1 .39 1 .82** .23 1 .78** .15 .94** 1 .88** .37 .77** .78** 1 .51* .65** .53* .50* .42 1
61 2. Guarded 3. Sticking out for one self 4. Assertive 5. Shifting responsibility 6. Defensive * p < .05; ** p < .01. Note. Each number refers to a single item that measures the defensiveness level of the accounts. See Appendix D for the complete list of accounts and labels of accounts presented to participants. 1 .50* 1 -.20 .21 1 .61** .79** .08 1 .53* .59** -.01 .89** 1
62 Table 4 The Cronbach’s Alpha for the Sum of Items that Measured Defensiveness Level of Each Account (N = 22) Defensiveness Concession 1 Concession 2 Concession 3 Concession 4 Concession 5 Justification 1 Justification 2 Justification 3 Justification 4 Justification 5 Excuse 1 Excuse 2 Excuse 3 Excuse 4 Excuse 5 Refusal 1 Refusal 2 Refusal 3 Refusal 4 Cronbach’s ? .82 .87 .70 .81 .80 .68 .76 .82 .86 .77 .81 .84 .88 .83 .78 .77 .79 .92 .81
63 Refusal 5 .85
Note. See Appendix D for the complete list of the accounts and labels of the accounts presented to participants.
type of accounts was the between-subject variable and the within-subject variable was the defensiveness level of the accounts. A significant within-subject effect was found, F(1, 21) = 30.14, p < .01, ?² = .59. In addition, the results indicated a significant linear and a quadratic effect of the defensiveness level of the accounts, F(1, 21) = 33.42, p < .01, ?² = .61 for the linear effect and F(1, 21) = 6.99, p < .05, ?² = .25 for the quadratic effect (Figure 3 pictures the defensiveness level of four types of accounts). Consistent with Schönbach’s (1990) argument, concessions had the lowest defensiveness level (M = 4.58, SD = 1.77) and refusals’ defensiveness level was highest (M = 8.29, SD = 2.32). On the defensiveness continuum proposed by Schönbach, justifications and excuses had moderate levels of defensiveness and excuses had a higher defensiveness level than justifications. The results indicated otherwise: Justifications’ defensiveness level (M = 7.63, SD = 1.95) was higher than the defensiveness level of excuses (M = 6.22, SD = 1.47). The difference was significant, t(21) = 4.53, p < .01. As discussed in Pilot Study 4, the defensiveness level of the accounts would be used with the likelihood of selecting each account in the formal study to constitute the composite measure (LDA) in the formal study. The average defensiveness level of each type of accounts was used as the Di in the formula to calculate LDA (LDA = ? [Di • Li] / ? Li). Specifically, Di for concession, justification, excuse, and refusals was 4.58, 7.63, 6.22, and 8.29, respectively.
64
9
8
Defensiveness Level
7
6
5
4 Concession Justification Excuse Refusal
Accounts
Figure 3. Defensiveness level of concessions, justifications, excuses, and refusals.
65 Pilot Study 6: Pilot Testing Dependent Measures Purpose. Pilot Study 6 was conducted to test the measures of all the dependent variables. Participants. Sixty-three undergraduate students at the University of Maryland participated in this pilot study. The age range of participants was 18 to 22 (M = 19.40, SD = 1.06, Mdn = 19.00). Sixty percent of participants were female (n = 38). And 60% of participants identified themselves as Caucasian (n = 38). Procedure. When participants arrived, the researcher explained the purpose of the study and asked them to sign the study’s consent form before they completed the questionnaire. The questionnaire had three parts: the stimulus message, all the dependent measures, and open-ended questions about the clarity of the dependent measures. Only two versions of the hypothetical conflict scenario were used in this pilot study: the scenario that induces a high outcome severity and a high level of relational closeness and the scenario that induces a low outcome severity and a low level of relational closeness. After reading the scenario, two practice questions were provided to participants to learn how to use magnitude scales. These two questions were the same ones used in Pilot Studies 3 and 5. After participants felt ready to use magnitude scales, they were asked to continue with the second part of the questionnaire: questions for the degree of internal attribution (five indicators), the degree of external attribution (five indicators), controllability (five indicators), uncontrollability (five indicators), expected responsibility (five indicators), expected anger (six indicators), likelihood of using concessions (five indicators), likelihood of using justifications (five indicators), likelihood of using excuses (five indicators), and likelihood of using refusals (five indicators). A total of fifty-one
66 indicators were used. All the measures used magnitude scales. The third part of the questionnaire listed open-ended questions regarding participants’ opinions about the dependent measures. For example, participants were asked to explain any confusion in the questionnaire. Appendix K listed all questions regarding the clarity of the questions used in the Pilot Study 6. Data transformation. Participants used nonnegative numbers to answer the questions regarding the hypothetical conflict. The data were submitted to descriptive analysis to see whether a transformation was needed. Following the (-1.96, 1.96) cutoff rule, all fifty-one indicators were significantly positively skewed. A transformation was necessary. The data were trimmed at the 90th percentile to eliminate the extreme values (e.g., the 90th percentile for one item to measure defensiveness was 400; thus, any score larger than 400 was converted to 400). Power transformations were used with ? ranges from 0.30 to 0.50 for different variables. Instruments and results. The pilot study measured the following variables: the degree of internal attribution, the degree of external attribution, controllability, uncontrollability, expected responsibility, expected anger, and likelihood of selecting each account (four types of accounts were used and each type had five accounts; in total twenty accounts were listed in the questionnaire). Degree of internal attribution. To measure the degree of internal attribution, a 5item magnitude scale was used (e.g., “How likely is it that your mistake was due to your own carelessness?”). A CFA was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results supported a one-factor structure although the fit indices were not satisfactory, ?²(5, N = 63) = 10.24, p > .05; NFI = 0.87; CFI = 0.93; SRMR = 0.08. A
67 principal components analysis without rotation was conducted. The results indicated two components with eigenvalues larger than 1. The first component had an eigenvalue of 2.27 and explained 45% of the variance. The second component’s eigenvalue equaled 1.05 and 21% of the variance was explained by the second component. The Cronbach’s alpha for the sum of the five items was .68. The degree of external attributions was measured by a 5-item magnitude scale (e.g., one item was, “How likely is it that your mistake was due to the professor not giving clear instructions for the assignment?”). CFA results did not support a one-factor structure, ?²(5, N = 63) = 14.22, p < .05; NFI = 0.89; CFI = 0.92; SRMR = 0.07. However, a principal components analysis yielded only one component with an eigenvalue greater than 1 (eigenvalue = 2.79). Fifty-five percent of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .79. A 5-item magnitude scale was used to measure controllability (e.g., one item was, “how likely is it that you could have prevented this mistake?”). CFA results supported a one-factor structure, ?²(5, N = 63) = 9.89, p > .05; NFI = 0.96; CFI = 0.98; SRMR = 0.03. A principal components analysis without rotation yielded two components with eigenvalues larger than 1. The eigenvalue of the first component was 2.96, explaining 59% of the variance and the second component’s eigenvalue equaled 1.00, explaining 20% of the variance. The Cronbach’s alpha for the sum of the five items was .89. Uncontrollability was measured by a 5-item magnitude scale (e.g., one item was, “How likely is that what happened was out of your control?”). CFA results indicated a one-factor structure, ?²(5, N = 63) = 7.93, p > .05; NFI = 0.95; CFI = 0.98; SRMR = 0.05. A principal components analysis with no rotation indicated that only one component had
68 an eigenvalue larger than 1 (eigenvalue = 2.92) and 58% of the variance explained by this component. The Cronbach’s alpha for the sum of the five items was .83. Anticipated consequences. To measure participants’ expected responsibility, a 5item magnitude scale was used (e.g., one item was, “how much responsibility do you have if Person B did not achieve his or her goal of being awarded the fellowship?”). CFA results did not support a one-factor structure, ?²(5, N = 63) = 38.67, p < .05; NFI = 0.66; CFI = 0.67; SRMR = 0.13. A principal components analysis without rotation found two components with eigenvalue larger than 1. The first component had an eigenvalue of 2.91 and explained 58% of the variance, and the second component had an eigenvalue of 1.32. Twenty-six percent of the variance was explained by the second component. The Cronbach’s alpha for the sum of the five items was .82. A 6-item magnitude scale was used to measure participants’ expected anger (e.g., “How mad do you expect Person B to be?”). CFA results did not support a one-factor structure, ?²(9, N = 63) = 45.64, p < .05; NFI = 0.90; CFI = 0.91; SRMR = 0.04. However, a principal components analysis without rotation indicated that only one component had an eigenvalue larger than 1. The first component had an eigenvalue of 5.12, explaining 85% of the variance. The Cronbach’s alpha for the sum of the five items was .97. Likelihood of selecting each account. Five accounts of concessions (developed in Pilot Study 4) were provided to participants. Participants were asked to rate the likelihood of using each concession when explaining their behavior in the conflict scenario. A magnitude scale was used. CFA results did not support a one-factor structure, ?²(5, N = 63) = 14.59, p < .05; NFI = 0.91; CFI = 0.94; SRMR = 0.07; however, the principal components analysis without rotation found only one component with an eigenvalue
69 greater than 1 (eigenvalue = 3.00), and 60% of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .81. Participants were asked to rate the likelihood of using each of the five justifications when explaining their behavior in the conflict scenario by using a magnitude scale. A one-factor structure was not supported by CFA results, ?²(5, N = 63) = 27.57, p < .05; NFI = 0.88; CFI = 0.89; SRMR = 0.08. However, the principal components analysis without rotation found only one component with an eigenvalue larger than 1 (eigenvalue = 3.35) and 67% of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .87. Participants also rated the likelihood of using each of the five excuses by using a magnitude scale. CFA results supported a one-factor structure marginally, ?²(5, N = 63) = 11.04, p = .05; NFI = 1.00; CFI = 1.00; SRMR = 0.06. A principal components analysis with no rotation found only one component with an eigenvalue larger than 1 (eigenvalue = 2.80), explaining 56% of the variance. The Cronbach’s alpha for the sum of the five items was .78. Five refusals were presented to participants to indicate their likelihood of using each refusal in the conflict situation. CFA results did not support a one-factor structure, ?²(5, N = 63) = 27.21, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.12. However, a principal components analysis without rotation found only one component with an eigenvalue larger than 1. The eigenvalue of the first component equaled 2.89, and 58% of the variance was explained by this component. The Cronbach’s alpha for the sum of the five items was .79. Revisions of the measures. Participants answered six open-ended questions
70 regarding the clarity of the questionnaire (see Appendix K). For example, participants were asked "What do you think the researchers want to measure?" Other questions examined the clarity and difficult levels of the questions. The researcher also interviewed sixteen participants about their perceptions of the items on the questionnaire. The majority of participants were able to identify three or four variables the researchers intended to measure, and most questions were perceived to be clearly stated and easy to answer. Revisions were made based on participants’ comments. For example, to measure expected responsibility of the offender, one question was revised from "How much responsibility do you have for what has happened" to "How much responsibility would Person B assign to you for what has happened?" The Cronbach’s alphas for all dependent measures were acceptable, ranging from .68 to .97. The degree of internal attribution had the lowest reliability coefficient (? = .68). A principal components analysis with no rotation indicated two components with eigenvalue larger than 1 for the degree of internal attribution. Two items (i.e., item 4: "How likely is it that your mistake was because you did not care about what grade you and Person B could get for this assignment?" and item 5: "To what extent do you believe what has happened was because you did not spend enough time on the assignment?") loaded highly on both the first and second components (.54 and .63 for item 4 and .50 and .47 for item 5). When eliminating both items, the Cronbach’s alpha for the sum of the remaining three items increased to .73; however, if these items were eliminated, the measurement model would be saturated and therefore fit perfectly, which is problematic. To ensure the measurement model for the degree of internal attribution was overidentified, three new items, phrased similarly as the remaining three items, were added to
71 the scale, and all six items were used in the formal study. For the scale of expected responsibility, two items loaded high on both first and second components (.63 and .71 for item 4, and .70 and .64 for item 5). These two items were "How much responsibility can be assigned to others (e.g., the professor or Person B) if Person B did not achieve his or her goal?" (item 4), and "How accountable were others (e.g., the professor or Person B) if Person B did not achieve his or her goal?" (item 5). After eliminating the two items from the scale, the Cronbach’s alpha for the sum of the remaining three items increased to .87. The measurement model for the expected responsibility with the remaining three items was saturated. Three new items were added to the scale and all six items were used in the formal study. CFA results indicated that the scales for the degree of external attribution, likelihood of using concessions, likelihood of using justifications, and likelihood of using refusals were not unidimensional. However, the principal components analyses found only one component with an eigenvalue greater than 1 for all four scales. The scales measuring the likelihood of using concessions, justifications, and refusals remained the same. However, based on participants’ comments, two more items were added to the scale for degree of external attribution. Based on the results of Pilot Study 3 and Pilot Study 6, in the formal study outcome severity and relational closeness manipulation checks were each measured by 6 items. The degree of internal attribution was measured by a 7-item magnitude scale, as was the degree of external attribution. A 5-item scale was used to measure controllability. Uncontrollability also was measured by a 5-item scale. Expected responsibility and expected anger were each measured by a 6-item scale. And five accounts were listed
72 under each type of accounts (i.e., concession, justification, excuse, and refusal). In total, 20 accounts were provided to participants to investigate their likelihood of using those accounts. The Formal Study Data collection for the formal study was conducted at the University of Maryland between May 6th and June 18th, 2008. This section presents the analytical strategy, the determination of the sample size and the sample description, the variables of interest, the experimental design, and the data collection procedure. Analytical Strategy The hypotheses and the proposed model were tested with analyses of variance and with structural equation modeling. Analyses of variance (ANOVAs). ANOVAs were used mainly to examine the success of the manipulations. In particular, the statistical difference between two levels of outcome severity (high vs. low) and the statistical difference between three levels of relational closeness (high vs. moderate vs. low) were tested. Structural equation modeling (SEM). SEM was used to test the causal relationships proposed by the hypotheses (see Figure 2). Severity of outcome (?1; high outcome severity was coded as 1 and low outcome severity was coded as -1), relational closeness (?2; quadratic coding: high relational closeness was coded as -1, moderate relational closeness as 2, and low relational closeness as -1), and the interactions (?3, ?4, and ?5) between outcome severity and relational closeness were the independent variables; manipulation checks for outcome severity (?1) and relational closeness (?2) were mediating variables; and the degree of internal attribution (?3), controllability (?4),
73 the degree of external attributions (?5), uncontrollability (?6), expected responsibility (?7), expected anger (?8), and LDA (?9) were outcome variables. See Figure 2 for labels and coding of each variable used in model testing. Two stages were used to test the structural equation model. First, confirmation factor analyses were conducted on each variable to determine the underlying dimensions. Second, the theoretical and the measurement model were tested together using LISREL 8.70 (Jöreskog & Sörbom, 2004). The influence of the outcome severity manipulation on expected responsibility and LDA was mediated by the outcome severity manipulation check (i.e., perceived outcome severity). The relational closeness manipulation (quadratic coding) and relational closeness manipulation check (i.e., perceived quadratic relational closeness) were used to test the curvilinear relationship between relational closeness and offenders’ choice of accounts. A direct influence of the interaction between outcome severity (linear coding) and relational closeness (linear coding) on LDA was proposed by Hypothesis 3. According to Hypothesis 4, the interaction between outcome severity and relational closeness also indirectly influences the defensiveness of the accounts through the attributional process variables (e.g., degree of internal and external attribution). Both direct and indirect paths were tested using LISREL8.70 (Jöreskog & Sörbom, 2004). Confirmatory factor analyses (CFAs) were performed using LISREL 8.70 (Jöreskog & Sörbom, 2004) to test measurement model fit. The goodness-of-fit criteria was based mainly on Hu and Bentler’s (1998, 1999) recommendations. Hu and Bentler (1999) found variation of performance among the fit indices. They suggested that to distinguish good models from poor ones, a two-index presentation strategy needs to be used that the maximum likelihood (ML) based SRMR is supplemented with another ML-
74 based fit index, such as TLI, RNI, CFI, Gamma Hat, Mc, or RMSEA. Hu and Bentler (1998) specified that a cutoff value close to .95 for CFI, a cutoff value close to .90 for Mc, a cutoff value close to .08 for SRMR, and a cutoff value close to .06 for RMSEA could “result in lower Type II error rates (with acceptable costs of Type-I error rates)” (p. 27). Hu and Bentler’s criteria have been largely accepted (e.g., Hancock, 2006). The above goodness-of-fit criteria also were used to show how well the data fit the proposed model. Testing total effects of the independent variables on the outcome variables indicated whether the hypotheses were supported. Significant coefficients from the independent variables to the outcome variables indicated the extent to which the independent variables influenced the dependent variables. Design A 2 (Outcome severity: severe vs. not severe) X 3 (Relational closeness: high vs. moderate vs. low) independent groups experimental design was employed. A scenario describing an interpersonal conflict in a course project at school was used to induce different levels of outcome severity and relational closeness. All participants were randomly assigned to one of six conditions formed by severity of an outcome and closeness of relationship. The cognitive (the degree of internal attribution, the degree of external attribution, controllability, uncontrollability, and expected responsibility) and emotional (expected anger) processes participants engaged in and the likelihood of selection of different accounts (concessions, justifications, excuses, and refusals) were measured. Sample Proposed sample size. Structural equation modeling (SEM) was used in data
75 analysis. Different researchers recommend different ways to determine sample size (e.g., Bentler & Chou, 1987; Jackson, 2003). The current research adopted the recommendation from Gagné and Hancock (2006). Gagné and Hancock took into consideration the ratio of the number of indicators per factor (p/f) and their loading magnitudes when deciding whether a sample size is acceptable. The ratio of the number of indicators per factor for the current research was 3.86. Based on Gagné and Hancock’s guideline, an acceptable estimation of sample size for the current research was between 200 and 400. Sample. Participants were 238 students; they were undergraduates (32.8% freshmen, 25.6% sophomores, 28.2% juniors, 13.0% seniors, and 0.4% others) recruited from the University of Maryland. Their age range was 18 to 37 (M = 19.76, SD = 1.78, Mdn = 20.00). One hundred and twenty-nine of the participants were female (54.2%). The majority were Caucasian (58.8%), 17.6% were Asian Americans, 13.4% were African American, and 4.6% were Hispanic. Approximately 5.5% of the participants checked the category “other.” Participants received extra credit in exchange of their participation in the study. Procedure Participants arrived at a pre-assigned classroom, in a group of five or six. The researcher informed the participants the purpose and the procedure of the study and asked them to sign a consent form before they participated in the study. All people coming to the study agreed to participate. After the signed consent forms were collected, a package including a single conflict scenario and a questionnaire was distributed. Participants were randomly assigned to receive one of the six versions of the hypothetical conflict scenario (as mentioned before, Appendix H lists all six versions of the hypothetical conflict
76 scenario). Participants were told to imagine that the hypothetical situation happened between them and another person (i.e., “Person B”). The scenario described a class project between participants (“you”) and Person B. The participants’ behavior in the scenario (i.e., writing his or her portion of a review paper based on the wrong readings) resulted in a poor grade for both the offender (i.e., the role the participant took) and the victim (i.e., Person B in the scenario), which, they were told, influenced the awarding of Person B’s scholarship for the next year. To induce relational closeness, participants and Person B’s relationship was described as follow: (a) They grew up together and have been best friends for many years (high level of closeness), (b) they were in a couple of classes together before, and they know each other but only see each other in class (moderate level of closeness), or (3) they don’t really know each other and they happened to sit together on the first day of class (low level of closeness). Outcome severity was manipulated in several ways. First, in the high outcomeseverity condition, the class project (i.e., writing a review paper) constituted a large portion of the class grade (i.e., 40%); the project was described as only counting for 10% of the grade in the low outcome-severity condition. Second, the influence of the class grade on Person B’s scholarship was different. In high outcome-severity condition, participants were told that “if the grade for this course was not good, his or her (i.e., Person B) scholarship for the next year would definitely be in jeopardy.” In the low outcome-severity condition, Person B told participants that “if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be
77 influenced.” Third, the consequence of the conflict was different. In the high outcomeseverity condition, after obtaining a poor grade for the class project, “Person B was very upset because such a low grade resulted in him or her not getting next year's scholarship,” compared to “Person B was not happy but he or she did great on other assignments and believed it would not influence his or her getting next year's scholarship” (low outcomeseverity). Each participant read one version of the hypothetical conflict scenario. Participants were asked to complete a questionnaire by taking the role of the offender. The questionnaire had the manipulation checks (i.e., for perceived outcome severity and perceived relational closeness), measures of attributions (i.e., the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability), measures of anticipated consequences (i.e., expected responsibility and expected anger), and measures of accounts (i.e., likelihood of selecting each account). Participants’ demographic information was collected at the end of the questionnaire. After completing the questionnaire, participants were debriefed in detail.
78 CHAPTER IV Results This chapter consists of two parts. The first part presents data preparation for the primary analyses, reliability and confirmatory factor analyses for measurement models, and manipulation checks. The second part presents the primary data analyses and tests of hypotheses. Preliminary Analyses Data Entry Checks Frequencies of all the variables were examined to minimize data entry error. All frequencies fell in the preset range defined by the lower boundary and the upper boundary for each item. Furthermore, all the responses in the questionnaires were checked against the data entries. Out of 238 participants’ data (each participant answered 72 question including demographic information), 6 errors were found, yielding an error rate of 6 ÷ (238 x 72) = .0004. The six incorrect inputs were corrected. Missing Data The completed questionnaires were numbered in the order of when they were turned in. Two participants had incomplete data. The missing data in both questionnaires did not show a systematic pattern. Kline (2005) suggested that non-systematic missing data can be ignored. He further argued that in statistical analyses such as principal components analyses and analyses of variance, pairwise deletion should be used to maintain the statistical power (Kline, 2005). However, pairwise deletion may pose a problem for confirmatory factor analyses and SEM; Kline suggested that when the number of cases with missing data is small, listwise deletion is preferable to pairwise
79 deletion for CFA and SEM. Thus, pairwise deletion was used when ANOVA was conducted for the manipulation checks and listwise deletion was used when CFA and SEM were conducted for testing the proposed model. Data Transformation Magnitude scales were used on the questionnaire for all dependent measures and manipulation checks. Participants used nonnegative numbers to answer the questions. Based on the results of the six pilot studies, perceived outcome severity, perceived relational closeness, expected responsibility, and expected anger each had six indicators; the degree of internal attribution and the degree of external attribution each had seven indicators; controllability and uncontrollability each had five indicators; and likelihood of selecting an account had five indicators for each of four types of accounts (20 items in total). Therefore, the total number of indicators for these measures was sixty-eight. The sixty-eight indicators were submitted to a descriptive analysis to see whether any data transformation would be needed. The skewness and kurtosis were examined. The ratio of skewness to its standard error was calculated, and an indicator’s distribution was considered approaching normality when the ratio ranged from -1.96 to 1.96 (see Frey et al., 2000). Following the |1.96| cutoff rule, all sixty-eight indicators were positively skewed. A data transformation was necessary. Before transformation, to eliminate extreme values, all data were trimmed at the 90th percentile (e.g., if 500 was the 90th percentile for an item, any score larger than 500 was converted to 500). Power transformation was used to treat the skewed data resulting from magnitude scales (Bauer & Fink, 1983): New indicator = (original indicator) ?,
80 the researcher tried different values of ? and used the ones that ensured the majority of indicators had non-significant statistics in skewness. The values of ? ranged from 0.28 to 0.55 for the formal study data. Transformed data were used in confirmatory factor analyses, principal components analyses, ANOVAs for the manipulation checks, and model testing. Appendix L provides descriptive statistics for the indicators used in the formal study after transformation. Instruments As mentioned above, Appendices A to D listed all the questions that measured the following variables. Manipulation checks. Manipulation checks involved two variables: perceived outcome severity and perceived relational closeness. Perceived outcome severity was measured by a 6-item magnitude scale, where 0 indicated not severe at all and 100 indicated a moderate level of severity. An example question is, “How severely does your mistake affect Person B?” A CFA was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). The results did not support a one-factor structure, ?²(9, N = 236) = 71.16, p < .05; NFI = 0.96; CFI = 0.97; SRMR = 0.03. When a principal components analysis without rotation was conducted, the results indicated only one component had an eigenvalue larger than 1 (eigenvalue = 4.86), explaining 81% of the variance. The Cronbach’s alpha for the sum of the six items was .95. The first component was used as the manipulation check. For model testing, all six items after transformation were used in testing the proposed model. Perceived relational closeness was also measured by a 6-item magnitude scale, where 0 indicated not close at all and 100 indicated a moderate level of relational
81 closeness. An example item was, “How familiar are you with Person B?” CFA results did not support a one-factor structure, ?²(9, N = 236) = 88.63, p < .05; NFI = 0.95; CFI = 0.95; SRMR = 0.03. A principal components analysis without rotation was conducted to examine the underlying components. The results indicated that only one component had an eigenvalue larger than 1 (eigenvalue = 5.49), explaining 92% of the variance. The Cronbach’s alpha for the sum of the six items was .98. The first component was obtained for two purposes. First, the first component score of the perceived relational closeness was used in ANOVA for assessing whether the manipulation of relational closeness (linear coding) was successful. Second, the first component score was used to create a manipulation check for quadratic relational closeness. A curvilinear relationship between relational closeness and selection of accounts was predicted in Hypothesis 2. To test this curvilinear relationship in the model, a manipulation check for the quadratic relational closeness needed to be included to mediate the relationship between relational closeness (quadratic coding) and selection of accounts. The following formula was used to create the perceived quadratic relational closeness: Perceived quadratic relational closeness = (first component score – mean)², where the mean for the first component score equals zero. Thus, the perceived quadratic relational closeness simply was created by squaring the first component score of the six items that measured the perceived relational closeness. In the model testing, only one indicator (i.e., perceived quadratic relational closeness) was included to represent this variable. Attributions. The current research did not assume the relationships between the degree of internal attribution and the degree of external attribution and between
82 controllability and uncontrollability. Four types of attributions were measured separately; they are the degree of internal attribution, the degree of external attribution, controllability, and uncontrollability. Different reasons regarding the offender’s behavior were listed and participants indicated their perception of how likely each reason influenced his or her behavior, using magnitude scales where 0 = not likely at all and 100 indicated a moderate level of likelihood. The degree of internal attribution was measured by a 7-item magnitude scale. For example, participants were asked, “How likely is that what has happened was due to your own carelessness?” CFA results did not support a one-factor structure, ?²(14, N = 236) = 123.79, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.11. A principal components analysis without rotation was conducted to examine the underlying components. The results indicated two components had eigenvalues larger than 1. The first component’s eigenvalue was 2.74, explaining 39% of the variance. The second component’s eigenvalue equaled 1.37, explaining 20% of the variance. Item 2 (i.e., “How likely is that what has happened was due to internal traits, such as your personality?”), item 4 (i.e., “How likely is that what has happened was because you did not care about what grade you and Person B could get for this assignment?”), and item 7 (i.e., “To what extent do you think your personality determines what has happened?”) loaded high on both first and second components. After deleting these three items, a CFA with the remaining four items had good fit indices but the chi-square value was significant, ?²(2, N = 236) = 7.80, p < .05; NFI = 0.97; CFI = 0.98; SRMR = 0.04. A principal components analysis with no rotation was conducted on the remaining four items, indicating only one component had an eigenvalue larger than 1 (eigenvalue = 2.26) and explaining 56% of the variance. The
83 Cronbach’s alpha for the sum of these four items was .73. These remaining four items were used for the model testing. The first component score for the remaining four items was saved to examine the relationship between the degree of internal attribution and the degree of external attribution. The degree of external attribution was measured by a 7-item magnitude scale. An example question was, “How likely is that what has happened was due to the professor not giving clear instructions for the assignment?” CFA results suggested a multidimensional factor structure, ?²(14, N = 236) = 58.17, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.06. A principal components analysis without rotation found that only one component had an eigenvalue larger than 1 (eigenvalue = 2.90), explaining 41% of the variance. The Cronbach’s alpha for the sum of the seven items was .75. All seven items were used in the model testing. The first component score for all seven items was saved to investigate the relationship between the degree of internal attribution and the degree of external attribution. The correlation between the degree of internal attribution and the degree of external attribution was significant, r = .33, p < .01. Controllability was measured by a 5-item magnitude scale. For example, participants were asked to indicate “How much control do you have over what has happened?” Confirmatory factor analysis was conducted using LISREL 8.70 (Jöreskog & Sörbom, 2004). Results did not indicate a unidimensional factor structure, ?²(5, N = 236) = 27.41, p < .05; NFI = 0.96; CFI = 0.97; SRMR = 0.04. However, a principal components analysis without rotation found only one component with an eigenvalue larger than 1. The first component’s eigenvalue equaled 3.27, explaining 65% of the variance. All five items were used in the model testing. The Cronbach’s alpha for the sum
84 of the five items was .85. The first component score for the five items was saved to test the relationship between controllability and uncontrollability. A 5-item magnitude scale was used to measure uncontrollability. An example question was, “To what extent do you believe there is not much you could do to prevent what has happened?” The CFA did not support a one-factor structure, ?²(5, N = 236) = 28.72, p < .05; NFI = 1.00; CFI = 1.00; SRMR = 0.05. The principal components analysis with no rotation found only one component that had an eigenvalue larger than 1. The first component’s eigenvalue equaled 3.12, explaining 62% of the variance. All five items were kept for the model testing. The Cronbach’s alpha for the sum of the five items was .85. The first component score was used to test the relationship between controllability and uncontrollability. The correlation was significant, r = -.13, p < .05. Anticipated consequences. Anticipated consequences were represented by two variables: expected responsibility assigned to the offender and expected anger felt by victims. For expected responsibility, participants were asked to indicate their expectation about how much responsibility would be assigned to them. A 6-item magnitude scale was used with 0 indicated not responsible at all and 100 indicated a moderate level of expected responsibility. For example, one item is, “How much responsibility would Person B assign to you for what has happened?” CFA results did not support a one-factor structure, ?²(9, N = 236) = 34.60, p < .05; NFI = 0.99; CFI = 0.99; SRMR = 0.02. A principal components analysis with no rotation found only one component with an eigenvalue larger than 1 (eigenvalue = 5.24), explaining 87% of the variance. All six items were used in the model testing. The Cronbach’s alpha for the sum of the six items was .97.
85 For expected anger, participants indicated the expected anger felt by victims by a 6-item magnitude scale. For example, an item is, “How annoyed do you expect Person B to be?” One-factor structure was not supported by CFA results, ?²(9, N = 236) = 196.93, p < .05; NFI = 0.91; CFI = 0.91; SRMR = 0.04. However, only one component with an eigenvalue larger than 1 was found in a principal components analysis with no rotation. The eigenvalue of the first component was 5.17, explaining 86% of the variance. All six items were used for the model testing. The Cronbach’s alpha for the sum of the six items was .97. Likelihood of selecting each account. A list of 20 accounts (5 statements for each type of accounts) was provided to participants to indicate the likelihood of selecting each account by using a magnitude scale, where 0 indicated not likely at all and 100 indicated a moderate level of likelihood. For example, a concession stated, “I am sorry that I misunderstood the instruction. Is there any way that I can make it up to you?” An example of a justification was like the following, “I am sorry that the grade influenced your scholarship. But I worked very hard on this assignment. We should have gotten together to edit the paper earlier.” An example excuse was, “I am sorry that I misunderstood the instruction. But the wording in the direction was really confusing and that may be the reason why I wrote the wrong portion of the paper.” An example refusal was, “I know the readings I read were the right material for the paper. I did not find anything wrong to what I contributed to the paper.” Participants were asked to rate their likelihood of using each account. To convert the results into a data set for likelihood (i.e., with a range from 0% to 100%), the original data were first trimmed at the 90th percentile and a power transformation was conducted,
86 with ? = 0.45. After transformation, the sum of all transformed likelihood data were calculated (named the Ltotal in the following formula) and used as the base for the likelihood of each account. The likelihood of each account was calculated by dividing the transformed data by Ltotal. For example, the likelihood of selecting a concession (named the Lc in the following formula) was calculated as: Lci = Ci / Ltotal, where C refers to the likelihood of selecting a concession and i ranges from 1 to 5. When the likelihood of selecting the five concessions was obtained (i.e., Lc1, Lc2, Lc3, Lc4, and Lc5), the average of Lc1 to Lc5 was calculated, resulting the likelihood of selecting concessions (Lc). The likelihood of selecting justifications (Lj), excuses (Le) and refusals (Lr) were calculated the same way. As discussed above, participants indicated how likely it was that they would select each account. However, the dissertation’s hypotheses refer to both likelihood of selecting each account and the defensiveness level of each account. Only using the data collected in the formal study would not be sufficient and not appropriate to test the hypotheses. To create an appropriate measure for accounts used in the hypotheses testing, the defensiveness level of each type of accounts resulted from Pilot Study 5 was combined with the likelihood of selecting a defensive account (i.e., LDA) resulted from the formal study. The formula to calculate LDA is as follows: LDA = ? (Di • Li) / ? Li, where Di refers to the defensiveness level of each type of accounts and Li refers to the likelihood of using each type of accounts (i.e., Lc, Lj, Le, and Lr). Based on the results from Pilot Study 5, Di for concessions, justifications, excuses, and refusals were 4.58,
87 7.63, 6.22, and 8.29, respectively. Therefore, LDA = (4.58 • Lc + 7.63 • Lj + 6.22 • Le + 8.29 • Lr) / (Lc + Lj + Le + Lr), the LDA value indicates the likelihood of selecting a defensive account. The higher the value of LDA, the more likely a more defensive account will be selected. Overall, for manipulation checks, the first components of perceived outcome severity and perceived relational closeness were submitted to ANOVA. For hypothesis and model testing, perceived outcome severity had six indicators. Perceived quadratic relational closeness had one indicator. Four indicators were used for the degree of internal attribution where the degree of external attribution had seven indicators. Controllability and uncontrollability each had five indicators. Expected responsibility assigned to the offender had six indicators and expected anger had six indicators. The composite measure of accounts (LDA) had one indicator. The model had five exogenous variables and nine endogenous variables, with forty-six indicators in all. Table 5 summarizes the results of principal components analyses of all instruments except LDA. Table 6 summarizes the Cronbach’s alpha for all instruments except LDA. Table 7 through Table 9 present each indicator’s factor loading from CFA and the measurement model fit indices. Loadings and fit indices for perceived outcome severity and perceived relational closeness are included in Table 7, loadings and fit indices for attributions are included in Table 8, and loadings and fit indices for expected responsibility and expected anger are presented in Table 9.
88 Table 5 Eigenvalues of the First Principal Components of Endogenous Variables with the Proportions of the Variance Explaineda Variable Perceived Outcome Severity Perceived Relational Closenessb Degree of Internal Attributionc Degree of External Attribution Controllability Uncontrollability Expected Responsibility Expected Anger Eigenvalue 4.86 5.49 2.26 2.90 3.27 3.12 5.24 5.17 % Variance Explained 81.04% 91.54% 56.49% 41.38% 65.43% 62.49% 87.25% 86.23%
Note. a. The table listed all endogenous variables except LDA (i.e., the likelihood of selecting a defensive account) because LDA only had one indicator and no principal components analysis was conducted on a single-indicator variable. b. The principal components analysis results reported here are based on the six items that measured the perceived relational closeness for the linear relation manipulation check. c. Degree of internal attribution was measured originally by a 7-item magnitude scale. The principal components analyses indicated three items loaded highly on both the first and second components (the eigenvalues for both components were larger than 1). These three items were deleted from the scale. The statistics reported in this table represent the results after deletion.
89 Table 6 The Cronbach’s Alpha for the Sum of Measurement Items for All Endogenous Variables in the Formal Studya Variables Perceived outcome Severity Perceived Relational Closenessb Degree of Internal Attributionc Degree of External Attribution Controllability Uncontrollability Expected Responsibility Expected Anger Likelihood of Concessions Likelihood of Justifications Likelihood of Excuses Likelihood of Refusals Valid N 237 237 237 237 237 237 238 238 237 238 238 238 Cronbach’s ? .95 .98 .73 .75 .85 .85 .97 .97 .80 .90 .78 .87 Number of items 6 6 4 7 5 5 6 6 5 5 5 5
Note. The indicators for each variable are found in Appendices A to D. a. Although model testing used LDA for the choice of accounts, the formal study used four scales to measure the likelihood of selecting each type of accounts. Therefore, the table also listed the reliability coefficients for these four measurement scales. b. The reliability results reported here was based on the six items that measured the perceived relational closeness for the linear manipulation check. c. Degree of internal attribution was measured originally by a 7-item magnitude
90 scale. The principal components analyses indicated three items loaded highly on both the first and second components (the eigenvalues for both components were larger than 1). The three items were deleted from the scale. The statistics reported in this table represent the results after deletion.
91 Table 7 Manipulation Checks, Indicator Loadings, and Measurement Model Indices Manipulation Checks and Indicators Unstandardized Loadings (Standardized) Outcome Severity To what extent does your mistake influence Person B? How severe is the outcome of this situation for Person B? How severely does your mistake affect Person B? How much influence does your mistake have on Person B’s scholarship for the next year? How much difficulty does your mistake have on Person B’s obtaining his or her scholarship for the next year? To what extent do you think the outcome is severe? 1.56 (.60)** 1.67 (.61)** 1.00 (.73)** 1.78 (.55)** 1.57 (.56)** 2.04 (.61)**
?² (9, N = 236) = 71.16, p < .05, RMSEA = .17, SRMR = .03, CFI = .97. Relational Closenessa To what extent do you think Person B is a good friend of yours? 1.00 (.58)**
To what extent do you think you and Person B are close to each other? 1.08 (.56)** How much knowledge do you have of Person B? How familiar are you with Person B? How close are you with Person B? How well do you know Person B? 0.95 (.55)** 0.94 (.55)** 1.07 (.54)** 1.01 (.52)**
?² (9, N = 236) = 88.63, p < .05, RMSEA = .19, SRMR = .03, CFI = .95. * p < .05; ** p < .01.
92 Note. Bold values represent fixed unstandardized loadings for reference indicators. a. The confirmatory factor analysis results reported here are based on the six items that measured the perceived relational closeness for the linear manipulation check.
93 Table 8 Attributional Processes Variables, Indicator Loadings, and Measurement Model Indices Attributional Processes and Indicators Unstandardized Loadings (Standardized) Degree of Internal Attributionsa How likely…due to your own carelessness? How likely...due to you not paying enough attention? How likely…because you did not spend enough time? To what extent…because of you? 1.00 (.74)** 1.06 (.96)** 0.24 (.43)** 0.96 (.93)**
?² (2, N = 236) = 7.80, p < .05, RMSEA = .11, SRMR = .04, CFI = .98. Degree of External Attributions How likely…due to the professor? How likely…caused by other people? How likely....because the wording was confusing? How likely…due to some external reasons? How likely…because of Person B? To what extent…because of other people? To what extent…because of environmental reasons? 1.00 (.78)** 0.86 (.76)** 0.75 (.86)** 0.55 (.90)** 0.78 (.91)** 0.78 (.77)** 0.50 (.91)**
?² (14, N = 236) = 58.17, p < .05, RMSEA = .00, SRMR = .06, CFI = 1.00. Controllability How likely...was avoidable? How much control do you have…? How likely…you could have prevented the mistake? 1.00 (.73)** 0.86 (.84)** 1.14 (.57)**
94 To what extent…you could have done otherwise? To what extent…you could have done differently? 0.91 (.66)** 0.69 (.74)**
?² (5, N = 236) = 27.41, p < .05, RMSEA = .14, SRMR = .04, CFI = .97. Uncontrollability How likely...was out of your control? To what extent…there is not much you could do? How likely…you have no control? To what extent…you cannot prevent…? To what extent…was inevitable? 1.00 (.81)** 1.04 (.78)** 1.39 (.67)** 1.46 (.63)** 1.13 (.77)**
?² (5, N = 236) = 28.72, p < .05, RMSEA = .00, SRMR = .05, CFI = 1.00. * p < .05; ** p < .01. Note. Bold values represent fixed unstandardized loadings for reference indicators. See the complete list of items for attributions in Appendix B. a. Degree of internal attribution was measured originally by a 7-item magnitude scale. The principal components analyses indicated three items loaded highly on both the first and second components (the eigenvalues for both components were larger than 1). The three items were deleted from the scale. The statistics reported in this table represent the results after deletion.
95 Table 9 Expected Consequences Variables, Indicator Loadings, and Measurement Model Indices Expected Consequences and Indicators Unstandardized Loadings (Standardized) Expected Anger How angry do you expect Person B to be? How irritated do you expect Person B to be? How furious do you expect Person B to be? How mad do you expect Person B to be? How fuming do you expect Person B to be? How annoyed do you expect Person B to be? 1.00 (.58)** 0.92 (.62)** 1.07 (.65)** 1.01 (.61)** 1.03 (.69)** 0.88 (.65)**
?² (9, N = 236) = 196.93, p < .05, RMSEA = .30, SRMR = .04, CFI = .91. Expected Responsibility How much responsibility…assigned to you? To what extent…Person B think you are responsible? To what extent…Person B think you are chargeable? To what extent…Person B think you are accountable? To what extent…should be attributed to you? 1.00 (.63)** 0.98 (.61)** 0.98 (.61)** 0.97 (.61)** 0.87 (.75)**
?² (9, N = 236) = 34.60, p < .05, RMSEA = .11, SRMR = .02, CFI = .99. * p < .05; ** p < .01. Note. Bold values represent fixed unstandardized loadings for reference indicators. See the complete list of items for anticipated consequences in Appendix C.
96 Manipulation Checks Two-way ANOVAs were conducted to see whether manipulations of outcome severity and relational closeness were successful. In the first ANOVA, the dependent variable was perceived outcome severity and the independent variables were the manipulations of outcome severity (high vs. low, coded as 1, -1) and relational closeness (high vs. moderate vs. low, coded as 1, 0, -1). Results indicated that (a) outcome severity had a significant influence on perceived outcome severity, F(1, 231) = 160.35, p < .01, ?² = .41, (b) relational closeness did not have a significant influence on perceived outcome severity, F(2, 231) = 0.13, p = .88, ?² = .00, and (c) no significant interaction between outcome severity and relational closeness was found, F(2, 231) = 1.28, p = .28, ?² = .01. Participants in the high severity conditions (M = .62, SD = .84) regarded the outcome to be more severe than those in the low severity conditions (M = -.66, SD = .70). The manipulation of outcome severity was successful. The second ANOVA had perceived relational closeness as the dependent variables and the manipulations of outcome severity (high vs. low, coded as 1, -1) and relational closeness (high vs. moderate vs. low, coded as 1, 0, -1) as the independent variables. The results suggested that (a) the manipulation of relational closeness significantly influenced perceived relational closeness, F(2, 231) = 150.54, p < .01, ?² = .57, (b) the manipulation of outcome severity did not have a significant effect on perceived relational closeness, F(1, 231) = 3.58, p = .06, ?² = .02, and (c) no significant interaction was found between the manipulations of outcome severity and relational closeness, F(2, 231) = 0.87, p = .42, ?² = .01. Participants in high relational closeness conditions (M = 1.00, SD = .77) reported being closest to Person B than those in the
97 moderate closeness conditions (M = -.35, SD = .49) and in the low relational closeness condition (M = -.72, SD = .69). A linear relationship between three levels of relational closeness was established, F(1, 226) = 238.877, p < .01, ?² = .50. The manipulation of relational closeness was successful. Model Assessment and Hypothesis Testing The hypotheses were tested with structural equation modeling using LISREL 8.70 (Jöreskog & Sörbom, 2004). The model consists of structural and measurement components. The model was tested to see the effect of outcome severity and relational closeness on individuals’ account selection processes, mediated by manipulation checks (i.e., perceived outcome severity and perceived quadratic relational closeness). Model Assessment The model (Figure 2) shows all links that were proposed by the hypotheses. Model specifications are listed in Table 10 and the unstandardized structural coefficients are reported in Table 11 and Figure 4. The covariance matrix analyzed in LISREL is presented in Appendix M. LISREL syntax for model testing can be found in Appendix N. Figure 2 only illustrates the structural model. The measurement models are not presented in the figure but they were included in the model testing. Overall model fit and model comparison. Chi-square results did not indicate a good model fit, ?²(966, N = 236) = 2475.89, p < .05, RMSEA = .08 with a 90% CI of (.08, .09), SRMR = .21, NFI = .91, GFI = .94. A comparison with the null model was conducted. In the null model, no covariance was assumed between any pair of constructs. For the null model, ?²(1,035, N = 236) = 30724.93, p < .05. Comparing both models resulted in ??²(69, N = 236) = 28249.04, p < .01. The tested model was a significant
98 Table 10 Structural Equations of the Model Proposed by the Hypotheses ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = ?91 ?92 ?71 ?73 ?83 ?74 ?84 ?75 ?85 ?76 ?86 ?87 ?98 ?93 ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 ?11 ?22 ?34 ?44 ?55 ?65 ?2 ?3 ?4 ?5 ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9
Note. ?1 and ?2 represent the manipulation checks (i.e., perceived outcome severity and perceived quadratic relational closeness). ?3 through ?6 are the degree of internal attribution, controllability, the degree of external attribution, and uncontrollability. ?7 is expected responsibility assigned to the offender and ?8 is expected anger felt by victims. ?9 refers to LDA (the composite measure for the likelihood of selecting a defensive account). ?1 through ?5 are outcome severity and relational closeness manipulations and the interactions variables (see Figure 2 for the explanation of the interaction variables).
99 Table 11 The Unstandardized Structural Coefficients for the Model ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = 0.10** -0.02 0.36** 0.09** 0.26** 0.04** -0.12** -0.01 0.26** -0.01 0.13** 0.57** -0.12** -0.02 ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 0.74** 0.31** 0.94 0.30 -0.89 0.03 ?2 ?3 ?4 ?5
** p < .01. Note. See Figure 2 for the explanation of each label.
100 0.74** OS POS
0.10** RC 0.31** PRC 0.36** -0.02 -0.02 Inter1 0.09** IA 0.94 Inter2 0.30 CON -0.01 0.26** 0.26** EA Inter3 -0.89 0.03 Uncon Figure 4. Structural relations proposed in the hypotheses with unstandardized structural coefficients. ** p < .01. See Figure 2 for the explanation of each label. -0.12** 0.04** -0.01 Eanger 0.13** 0.57** Eresp -0.12** LDA
101 improvement over the null model. Although the ?² value was used as an indication of model fit, some scholars argue that the ?² is not an appropriate measure of model fit with a sample larger than 200 (Kline, 2005). Kline proposed that the ratio of ?²/df as a more effective measure of fit when N > 200; Kline suggested a ratio less than 3.00 was acceptable. For the current model, ?²/df = 2.56. The proposed model still could be considered as a good fit based on Kline’s suggestion. SEM scholars (e.g., Bollen, 1989; Hu & Bentler, 1998, 1999) also have recommended that multiple fit indices should be considered to evaluate the fit of a model, given the advantages and disadvantages of each fit index. The goodness-of-fit criteria used here are based on Hu and Bentler’s (1998, 1999) recommendation of a twoindex presentation strategy, using SRMR supplemented with other indices such as CFI, RNI, and RMSEA. Hu and Bentler (1999) specified that a cutoff value close to .95 for CFI, a cutoff value close to .08 for SRMR, a cutoff value close to .95 for RNI, and a cutoff value close to .06 for RMSEA. The proposed model had following fit indices: SRMR = .21, CFI = .94 and RMSEA = .08 with a 90% CI of (.08, .09). Based on Hu and Bentler’s criteria, the proposed model did not have a good fit. Hypotheses Testing Ten hypotheses were proposed regarding the attributional processes individuals engage in to determine their accounts. The effects of the exogenous variables on their relevant endogenous variables were examined. H1 predicts that outcome severity influences the expected responsibility assigned to offenders. In particular, when the outcome is severe, offenders expect more responsibility assigned to them than when the outcome is not severe. The effect of
102 outcome severity on the expected responsibility was significant, ? = 0.36, z = 6.50, p < .01. H1 was supported. The positive sign of the coefficient indicated that more responsibility was expected to be assigned to offenders when the outcome was severe than when the outcome was not severe. One research question was proposed regarding the influence of outcome severity on the selection of an account. The effect of outcome severity on the selection of an account was significant, ? = 0.10, z = 3.45, p < .01. As discussed in the construction of LDA (the composite measure of the likelihood of selecting a defensive account), the higher the value of LDA, a more defensive account will be more likely to be selected. The positive sign of the ? coefficient indicates that when the outcome was severe, offenders were more likely to use a more defensive account than when the outcome was not severe. The results are consistent with Schönbach’s (1990) idea. Schönbach argued that when facing severe outcome, offenders often fear the pressure to take full responsibility for their offenses. To shift some responsibility to reasons other than themselves, the offenders tend to take a more defensive stand than when the outcome of their behavior is not severe. H2 proposed that relational closeness influences people’s choice of accounts curvilinearly. It is predicted that a defensive account is less likely to be used when two parties involved in a conflict have a moderate level of relational closeness than when two parties are strangers or when two parties are close friends. The results were not significant, ? = -0.02, z = -30.77, p > .05. The negative sign of the coefficient implies that when conflict occurs between two parties with a moderate level of relational closeness, offenders are more likely to use a defensive account than when the two parties are close
103 friends or strangers. The direction was opposite to the hypothesis. H2 was not supported. H3 predicts that outcome severity and relational closeness interact to influence the offenders’ selection of accounts in a conflict situation. It is hypothesized that when a severe conflict occurs between two close friends, a less defensive account is more likely to be used than when the conflict is not severe. However, when a severe conflict happens between two strangers, offenders are more likely to use a more defensive account than when the conflict is not severe. When two parties involved in a conflict have a moderate level of relational closeness, a less defensive account is more likely to be used than when the conflict occurs between parties that are distant or close, regardless of outcome severity. The independent variable was the interaction variable (?3). H3 was not supported, r = -0.02, z = -0.44, p > .05. There was no interaction of outcome severity and relational closeness on offenders’ selection of accounts. H4 had four subsets of hypotheses regarding the attributional processes that offenders engage in after a conflict has occurred. It was predicted that when a severe conflict occurs between close friends, internal (H4a) and controllable attributions (H4b) are more likely than in all other conditions; however, when a severe conflict occurs between strangers, offenders tend to attribute their own negative behaviors to external (H4c) and uncontrollable reasons (H4d) more than in all other conditions. No significant results were found, r = 0.94, z = 1.87, p > .05 for internal attribution; ? = 0.30, z = 1.09, p > .05 for controllability; ? = -0.89, z = -1.17, p > .05 for external attribution; and ? = 0.03, z = 0.10, p > .05 for uncontrollability. H4 was not supported. Although the results were not significant, the signs of the coefficients suggest that when a severe conflict occurs between close friends, internal and controllable attributions are more likely to be made by
104 offenders than other conditions. However, when a severe conflict happens between strangers, offenders is less likely to make external attribution and more likely to make uncontrollable attribution than other conditions. H5 to H8 examined the relationship between attributions and expected responsibility and expected anger. Specifically, H5a specified a positive influence of internal attribution on expected responsibility, and H5b predicted a positive influence of internal attribution on expected anger. H5a was supported, ? = 0.09, z = 3.72, p < .01. The positive sign of the ? coefficient indicates that when offenders attribute their behaviors more to their internal traits, they expect more responsibility to be assigned to them. H5b was not supported, ? = -0.01, z = -0.32, p > .05. The negative sign of the coefficient suggests an opposite direction to what was predicted in H5b: When offenders attribute their behaviors more to their internal characteristics, they expect less anger would be felt by the victim. H5 was partially supported. H6 predicted that controllability positively influences expected responsibility to be assigned to offenders and expected anger felt by the victim. H6a predicted when offenders attributed their behavior more to controllable reasons, they expected more responsibility assigned to them. Findings were significant, ? = 0.26, z = 6.29, p < .01. The positive sign of the ? indicates the positive influence of controllability on responsibility expected to be assigned to offenders. H6b predicted a positive influence of controllability on expected anger. Results were significant, ? = 0.26, z = 6.85, p < .01. The positive sign of the ? coefficient suggests that when offenders make more controllable attribution, they expect more anger to be felt by the victim. H6 was supported. H7 stated that when offenders attributed their behaviors more to external reasons,
105 they expected less responsibility to be assigned to them and less anger felt by victims. Significant results were found for expected responsibility, ? = 0.04, z = 2.50, p < .01. However, the positive sign of the coefficient suggests that the direction of external attribution’s influence on expected responsibility is opposite to the prediction in H7a. The results showed that when offenders make more external attribution, they expect more responsibility to be assigned to them. No significant results were found for the influence of external attribution on expected anger, ? = -0.01, z = -0.48, p > .05. The negative sign of the coefficient showed that when more external attribution is made, offenders expect less anger to be felt by the victim. H7 was not supported. As for attributions to uncontrollable reasons, H8 predicted uncontrollability negatively influenced expected responsibility (H8a) and anger (H8b). H8a was supported, ? = -0.12, z = -2.88, p < .01. The negative sign of the coefficient indicates that when offenders make more uncontrollable attribution, they expect victims to assign less responsibility to them. Significant results were found for H8b, ? = 0.13, z = 3.66, p < .01. However, the positive sign of the coefficient indicates that when more uncontrollable attribution is made, offenders expect more anger felt by victims, which is opposite to the prediction in H8b. H8 was partially supported. Expected responsibility was predicted to positively influence expected anger in Hypothesis 9. The results supported this relationship, ? = 0.57, z = 10.54, p < .05. The positive sign of the coefficient indicates that more expected responsibility leads to more expected anger. H9 was supported. H10 predicted a negative influence of expected anger on offenders’ selection of accounts: When offenders expect more anger felt by victims, they will be less likely to
106 use a defensive account. The findings were statistically significant, ? = -0.14, z = -4.91, p < .01. The negative sign of the coefficient suggests that a defensive account is less likely to be selected when offenders expect more anger felt by the victims. H10 was supported. Overall the model did not have a good fit. Among the ten hypotheses and one research question, five hypotheses were supported and two other hypotheses (H5 and H8) were partially supported. The two coefficients that tested the manipulations were significant also, consistent with the manipulation checks that were conducted outside of the model testing, ? = 0.74, z = 10.53, p < .01 for outcome severity manipulation check and ? = 0.31, z = 6.03, p < .01 for relational closeness manipulation check (for the quadratic relational closeness). Post-Hoc Analyses Post-hoc analyses were conducted to further explore the influence of relational closeness and outcome severity on offenders’ choice of accounts. First, the nonsignificant hypotheses in the model testing were examined separately. Second, modification indices for the model tested in the above section were examined for additional paths to improve the model fit. Third, the model was revised based on the results in the first two steps and the revised model was tested using LISREL 8.70 (Jöreskog & Sörbom, 2004). Hypothesis Testing The non-significant results in the model testing may be due to multicolinearity among the five exogenous variables. The correlation matrix of the five exogenous variables in the proposed model had a determinant of .11. Among ten correlations formed by these five variables, seven of them were statistically significant. See Table 12 for all
107 Table 12 Intercorrelations between the Five Exogenous Variables Variables ?1 ?2 ?3 ?4 ?5 ** p < .01. Note. ?1 refers to outcome severity manipulation (low outcome severity was coded as -1 and high outcome severity was coded 1) and ?2 refers to relational closeness manipulation (quadratic coding: low relational closeness was coded as 1, moderate relational closeness was coded as -2, and high relational closeness was coded as 1). ?3, ?4, and ?5 refer to the interactions between outcome severity and relational closeness. ?3 was calculated by multiplying the outcome severity manipulation and relational closeness manipulation. ?4 refers to the coding of the high severity and close relationship condition (coded as 1) versus all other five conditions (coded as 0). ?5 refers to the coding of the high severity and not-close-at-all relationship (coded as 1) and other five conditions are coded as 0. ?1 1 ?2 .01 1 ?3 .03 .01 1 ?4 .46** .33** .57** 1 ?5 .43** .31** -.55** -.21** 1
108 the correlation coefficients among the five exogenous variables. According to Kline (2005), a small determinant of the correlation or covariance matrix can lead to nonsignificant structural coefficients, which may partly explain the non-significant results in the current research. The paths that were not significant in the model testing were examined again separately. Specifically, H2, H3, H4, H5b, and H7b were re-tested. H2 hypothesized a curvilinear effect of relational closeness on offenders’ choice of accounts. Specifically, when a conflict occurs between two parties who have a moderate level of relational closeness, offenders are less likely to use a more defensive account than when the conflict occurs between close friends or between strangers. A linear regression was conducted to test H2, where the quadratic relational closeness manipulation check was entered as the independent variable and LDA (the composite measure of accounts) was entered as the dependent variable. The results were significant, ? = -0.64, F(1, 233) = 9.68, p < .05, R² = .04. However, the negative sign of the ? coefficient suggests that when a conflict occurs between two parties that are moderately close, offenders are more likely to use defensive accounts than when a conflict occurs between friends or between strangers. The direction was opposite to the prediction in the hypothesis. Hypothesis 3 proposed an interaction between outcome severity and relational closeness on offenders’ choice of accounts. When a conflict occurs between close friends, a less defensive account is more likely to be used in a severe situation than in a notsevere situation. When the conflict occurs between two parties who have a moderate level of relationship, a less defensive account will be used regardless of the severity of the outcome. When the conflict happens between strangers, a more defensive account is more
109 likely to be used in a severe situation than in a not-severe situation. An ANOVA was used to examine H3. LDA, the likelihood of selecting a defensive account, was used as the dependent variable and the interaction between outcome severity and relational closeness was entered as the independent variable. The interaction variable (i.e., ?3 in the model tested previously) was created by multiplying outcome severity manipulation (high severity coded as 1 and low severity coded as -1) and relational closeness manipulation (linear coding: high relational closeness coded as 1, moderate relational closeness coded as 0, and low relational closeness coded as -1). The results were significant, F(2, 232) = 4.97, p < .05, ?² = .04. The direction of the influence was consistent with the hypothesis. When a conflict occurs between high relationally close parties, offenders are less likely to use a more defensive account in a severe-outcome situation (M = 60.83, SD = 4.53) than in a not severe-outcome situation (M = 61.21, SD = 4.51). When the conflict occurs between low relational closeness parties, offenders are more likely to use a more defensive account in a severe-outcome situation (M = 61.21, SD = 4.51) than in a not-severe-outcome situation (M = 60.83, SD = 4.53). When the conflict occurs between moderate relational closeness parties, offenders’ likelihood of using accounts did not differ by different levels of outcome severity (M = 62.93, SD = 4.22). H3 was supported. Hypothesis 4 predicted relational closeness and outcome severity interact to influence offenders’ attributions. H4 had four subsets of hypotheses. H4a and H4b predicted that when a severe conflict occurs between high relationally close parties, offenders are more likely to make internal and controllable attributions than in all other situations. H4c and H4d hypothesized that when a severe outcome conflict occurs
110 between low relationally close parties, offenders are more likely to make external and uncontrollable attributions than in all other conditions. To examine H4, the six experimental conditions created by two levels of outcome severity (high vs. low) and three levels of relational closeness (high vs. moderate vs. low) were coded into two variables. The first variable (i.e., ?4 in the model tested previously) was coded the following way: The high outcome severity and high relational closeness condition was coded as 1 and all other five conditions were coded as 0. The second variable (i.e., ?5 in the model tested previously) was coded that the high outcome severity and low relational closeness condition was 1 and all other five conditions were 0. ?4 was entered as an independent variable in a MANOVA, the first component scores of the degree of internal attribution and controllability being the dependent variables. No significant results were found, F(1, 234) = 3.43, p = .07, ?² = .01 for the degree of internal attribution and F(1, 234) = 0.68, p = .41, ?² = .003 for controllability. Another MANOVA was conducted with ?5 as independent variable and the first components of the degree of external attribution and uncontrollability as the dependent variables. The results were not significant, F(1, 234) = 1.13, p = .29, ?² = .01 for the degree of external attribution and F (1, 234) = 0.08, p = .78, ?² = .00 for uncontrollability. Hypothesis 5b predicted that when offenders make more internal attribution, more anger is expected to be felt by the victim. The first components of the degree of internal attribution and expected anger were obtained to be used in a regression analysis. A linear regression was conducted with the degree of internal attribution as the independent variable and expected anger as the dependent variable. The results were significant, ? = 0.56, F(1, 234) = 104.23, p < .01, R² = .31. The positive sign of the coefficient suggests
111 that when offenders made more internal attribution, more anger was expected to be elicited by the victim. Hypothesis 7b proposed a negative influence of external attribution on expected anger. The first components of the degree of external attribution and expected anger were used in a regression analysis to test H7b. The results were significant, ? = 0.28, F(1, 234) = 19.36, p < .01, R² = .08. Although the results were significant, the positive sign of the coefficient indicates that when more external attribution is made, more anger is expected, which is opposite to the prediction. The significant paths that were indicated by the exploratory analyses were kept in the model. That is, the influence from perceived quadratic relational closeness on people’s choice of accounts, the influence from the degree of internal attribution on expected anger, the influence of the degree of external attribution on expected anger, and the interaction of outcome severity and relational closeness on people’s choice of accounts remained in the model. All the paths that indicated the effect of outcome severity and relational closeness on the attributions people make were removed from the model due to non-significant results in both the model testing and the exploratory analyses. Modification Indices Modification indices for the proposed model (see Figure 2) suggested the maximum ?² change would occur when a path was established from expected responsibility to controllable attribution (??² = 168.34). However, it was expected that people’s attributions influence their perceptions of expected responsibility and expected anger, and then influence their choice of accounts, not vice versa. The study was also
112 conducted in a way that attribution was first made and then participants were asked to expect how much responsibility to be assigned. Several other paths that had large value of modification indices, such as a link that indicated the influence from expected responsibility on internal attribution (??² = 158.64), had the same problem. Those paths, although they have large modification indices, were not considered in the revised model. It was hypothesized that outcome severity and relational closeness interacted to influence different attributions. These hypotheses were not supported. However, the model’s modification indices suggested that outcome severity alone may influence different attributions via perceived outcome severity. If a path is added from perceived outcome severity to the degree of internal attribution, the expected change in the ?² value is 36.26. If a path is established from perceived outcome severity to controllability attribution, the expected ?² change is 64.30. If a path is added from perceived outcome severity to the degree of external attribution, the expected ??² equals 26.48. If a path is added from perceived outcome severity to uncontrollability attribution, the expected ?² change is 9.05. Those paths were considered in revising the model. Furthermore, modification indices suggested a path from perceived outcome severity to expected anger (??² = 57.71). It was hypothesized that when an offender perceives the outcome of a conflict to be severe, he or she expects more responsibility to be assigned to him or her; in turn, more expected responsibility is expected to lead to more expected anger. The influence of perceived outcome severity on expected anger is mediated by expected responsibility. However, a direct influence is also possible. Thus, the path from perceived outcome severity to expected anger is considered in the revised model.
113 Model Revision Based on the exploratory analyses and modification indices, the following changes are made to revise the model: (a) Two interaction variables (?4 and ?5, see Figure 2 for the detailed description of these two variables) did not influence attributions as predicted in Hypothesis 4. The relevant paths were removed from the model. (b) The influence of perceived outcome severity on expected responsibility was mediated by attributions (i.e., the degree of internal attribution, controllability, the degree of external attribution, and uncontrollability). Four paths were added to demonstrate the mediation by attributional processes. (c) An additional path was added from perceived outcome severity to expected anger. See Figure 5 for the revised model. The dotted lines represent the paths added to the original model based on exploratory analyses and the modification indices. Testing the Revised Model The revised model was tested with structural equation modeling using LISREL 8.70 (Jöreskog & Sörbom, 2004). The model consisted of structural and measurement components. Model specifications are listed in Table 13, and unstandardized structural coefficients are reported in Table 14 and Figure 6. The covariance matrix analyzed in LISREL is found in Appendix O. LISREL syntax for model testing can be found in Appendix P. Figure 5 only illustrates the structural model of the relationship between the exogenous and endogenous variables. The measurement models are not included in the figure but they were included in the model testing. For the revised model, the chi-square was significant, ?²(883, N = 236) = 2210.26, p < .01. Compared to the model before revision, chi-square value was improved by
114
OS
POS
IA
Eresp
CON
EA Eanger Uncon LDA RC PRC
Inter1
Figure 5. Structural relations based on post-hoc analysis. Note. All exogenous variables are allowed to covary and all error terms are allowed to covary. The dotted lines represent the paths added based on the exploratory analyses and modification indices. The labels and coding for each variable can be found in Figure 2.
115 Table 13 Structural Equations of the Revised Model Based on Modification Indices and Exploratory Analyses ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = ?31 ?41 ?51 ?61 ?71 ?81 ?91 ?92 ?73 ?83 ?74 ?84 ?75 ?85 ?76 ?86 ?87 ?98 ?93 ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 ?11 ?22 ?2 ?3 ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9
Note. Bold symbols represent the paths that were added based on exploratory analyses and modification indices. See Figure 2 for the labels and coding of each variable.
116 Table 14 Unstandardized Structural Coefficients for the Revised Model ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 = = = = = = = = = 1.18** 0.74** 1.30** 0.24** 0.29** 0.48** 0.11** -0.02 0.08** 0.29** 0.04** -0.11** 0.01 0.15** -0.01 0.05 0.37** -0.14** ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 ?9 ?1 0.74** 0.31** ?2 ?3 ? ?1 ?2 ?3 ?4 ?5 ?6 ?7 ?8 -0.02 ?9
* p < .05; ** p < .01. Note. Bold values represent the paths that were added based on exploratory analyses and modification indices. See Figure 2 for the labels and coding of each variable.
117 0.29** 0.08** 0.01 1.18** 0.74** OS POS 1.30** 0.24** Uncon 0.48** 0.11** LDA 0.31** -0.02 EA 0.04** -0.11** -0.01 0.05 -0.14** 0.37** Eanger 0.74** CON 0.15** 0.29** Eresp
IA
RC
PRC
-0.02 Inter1 Figure 6. The revised model with unstandardized structural coefficients. ** p < .01. See Figure 2 for the labels and coding of each variable.
118 265.63. Given that ?df equaled 83, the revised model was a significant improvement over the original model. For the revised model, the ratio of ?²/df equaled 2.50, which indicated a good model fit according to Kline’s (2005) standard. Before the revision, the original model had the following fit indices: SRMR = .21, CFI = .94 and RMSEA = .08 with a 90% CI of (.08, .09). After the revision, model fit indices were improved: SRMR = .12, CFI = .95, and RMSEA = .08 with a 90% CI of (.08, .08). Although based on Hu and Bentler’s (1999) criteria the revised model still could not be considered a good fit, it had been improved from the original model, and the CFI had reached the cutoff standard suggested by Hu and Bentler. After examining the ?² value, the ratio of ?²/df, and fit indices, the revised model is considered to have a moderate fit. Examining the paths in the revised model indicated almost the same pattern as shown in the original model: Paths that were significant in the original model were statistically significant in the revised model (except the link between uncontrollability and expected anger), and paths that were not significant in the original model testing but significant in the exploratory analyses still remained non-significant in the revised model. The five paths that were added based on modification indices were significant (see Table 14 for all unstandardized structural coefficients). When testing effects outside of the structural equation model, relational closeness (quadratic coding) had a significant main effect on offenders’ choice of accounts and also relational closeness interacted with outcome severity to influence offenders’ choice of accounts significantly. However, the effect of relational closeness on offenders’ choice of accounts was not significant once outcome severity came into the picture. Outcome severity influenced offenders’ choice of accounts directly and indirectly via perceived
119 outcome severity, attributions offenders made, and consequences offenders expected. Compared with relational closeness, outcome severity is a better predictor about the accounts that an offender will use in conflict situations.
120 CHAPTER V Discussion This chapter consists of four parts. A summary of the study is provided in the first part. The second part interprets the results and discusses the study’s implications. The third part consists of the study’s limitations and directions for the future research. The significance of the study and a conclusion form the last part of the chapter. Summary of the Study The current research addressed the following question: What attributional processes do offenders engage in to determine the account used in a conflict? This study examined the influence of outcome severity and relational closeness on offenders’ choice of accounts, mediated by attributional processes (i.e., internal, external, controllable, and uncontrollable attributions) and offenders’ anticipated consequences (i.e., expected responsibility and expected anger). Two levels of outcome severity and three levels of relational closeness were employed, creating six experimental conditions. In each condition, offenders read a hypothetical conflict scenario by imagining they were offenders in the conflict and they were interacting with another person. Prior to the formal study, six pilot studies were conducted to (a) develop the manipulations that would be used in the formal study, (b) develop a measure of accounts for the formal study, and (c) test and improve the items that measured all the constructs. The manipulations and instruments that were developed and tested in the pilot studies were then used in the formal study. The formal study recruited 238 participants enrolled at the University of Maryland. Two participants had incomplete data. Pairwise or listwise deletion was used,
121 depending on the type of statistical analyses used. Each participant was randomly assigned to one of the six experimental conditions. Questionnaires were employed in the study. The questionnaire started with a hypothetical scenario that described a conflict in a class project; this scenario was followed by the manipulation checks. The second part of the questionnaire asked participants to report their attributions. Participants also reported their expectations about the responsibility assigned to them and the anger felt by the victim in the scenario. Participants were then asked to report the likelihood of selecting each account developed from Pilot Study 4.6 Participants’ demographic information was collected in the last part of the questionnaire. After participants completed their questionnaire, they were fully debriefed. Magnitude scales were used in this study. Items that violated the assumption of normality were transformed by a power transformation. Confirmatory factor analyses (CFAs) were conducted to examine the dimensions of each construct. When ?² value and fit indices indicated that there was more than one factor for some variables, principal components analyses (PCAs) were conducted to explore the potential components. No rotation was used in the PCAs. The results from CFAs and PCAs determined the items that were included in the structural equation model. The hypotheses were tested using structural equation modeling. Outcome severity, relational closeness, and their interactions were dummy coded and entered as exogenous variables (see Figure 2 for the details of labels and coding of each variable). Manipulation checks were used as mediating variables; offenders’ choice of account was used as an outcome variable; and attributional processes and anticipated consequences were also expected to mediate the
The likelihood results of the formal study were combined with the defensiveness results collected from Pilot Study 5 to create the composite measure (LDA) used in model and hypothesis testing. The details about the creation of LDA can be found on pp. 86-88.
6
122 influence of the independent variables on offenders’ choice of accounts. The post-hoc analyses were conducted for model fit improvement and better understanding of accounts that offenders use in conflict situations. Summary of Results Hypothesis 1: Influence of Outcome Severity on Expected Responsibility Outcome severity was predicted to positively affect responsibility assigned to offenders (Walster, 1966). Previous research has suggested that victims or observers perceive offenders to be less careful when the outcome is severe; therefore, victims or observers tend to assign more responsibility to offenders (Walster, 1966). From the offenders’ perspective, when they perceive the outcome as severe, offenders would expect little chance for them not to be blamed for what has happened. Therefore, offenders expect victims to assign more responsibility to them when the outcome is severe than when the outcome is not severe. The results showed a significant positive effect of outcome severity on expected responsibility. Offenders, when the outcome is severe, tend to expect more responsibility to be assigned to them. Research Question 1: Effect of Outcome Severity on the Choice of Accounts Researchers have argued for opposing possibilities of outcome severity’s effect on people’s choice of accounts. McLaughlin, O’Hair, et al. (1983) stated that when an outcome is severe, people tend to use less defensive accounts, such as concessions, for the purpose of easing the tension aroused by their behaviors and to maintain a good image. However, Schönbach (1990) argued that although there is a possibility that offenders may use a less defensive account to present a good image in the situation, when an outcome is really severe, taking responsibility may be a great burden for offenders.
123 Instead, offenders may reject or deny their responsibility to save their face. They will choose more defensive accounts when the severity of an outcome is more severe. The results favored Schönbach’s prediction: When offenders perceived that the outcome was severe, offenders were more likely to use a more defensive account, such as a justification, than when the outcome is not severe. Hypotheses 2: Influence of Relational Closeness on the Choice of Accounts It is argued in H2 that when a conflict occurs between parties that are moderately close, offenders are less likely to use a defensive account than when the conflict occurs between parties that are close or distant. A curvilinear effect of relational closeness on offenders’ choice of accounts was proposed in H2. However, the path indicated by H2 in the structural equation model was not significant. When testing H2 outside of the model, the results suggested that relational closeness indeed influences offenders’ choice of accounts, in a way opposite to the prediction in H2: Offenders were more likely to use a defensive account when moderately close to the victim than when relational closeness between the offender and the victim was high or low. One possibility is when manipulating the moderate relational closeness in the hypothetical conflict scenario, participants were told that they have known the other person in the conflict because they have taken several courses together but they did not really hang out after school. Although participants were told that they had interactions with the other party in the past (mostly in class), it did not necessarily indicate that they would take courses together again; that is, participants may not perceive a high possibility of future interaction with the victim. Another possibility is the influence of relational closeness on the choice of accounts may depend on how severe the outcome is.
124 Hypotheses3: Influence of Relational Closeness and Outcome Severity on the Choice of Accounts Hypothesis 3 examined the interaction of relational closeness and outcome severity on offenders’ choice of accounts. It was argued that when a severe conflict occurs between two parties who are close, offenders are less likely to use a defensive account than when the conflict is not severe. When the conflict occurs in a moderate relational closeness situation, offenders are less likely to use a defensive account than when the conflict is between two parties that are distant or close, regardless of the severity of outcome. When a conflict occurs between strangers, offenders are more likely to use a defensive account than when the conflict is not severe. So offenders select their account depending on the interaction between outcome severity and relational closeness. Similar as Hypothesis 2, when H3 was tested in the structural equation model, the path was not significant. However, when testing the hypothesis outside of the model, significant results suggested that outcome severity and relational closeness interact to influence offenders’ choice of accounts. Hypotheses 4: Attributional Processes Hypothesis 4 specified the effect of outcome severity and relational closeness on offenders’ attributional processes. The findings did not support the predictions in H4. The interaction variables created by outcome severity and relational closeness did not influence the attributions offenders made. In the post-hoc analysis, the modification indices suggested that outcome severity alone may influence the attributions offenders make and then influence the expectation of consequences and the choice of accounts. Additional paths were added to the model to test the influence of outcome severity and
125 attributions offenders make. The findings of the revised model suggested that when the outcome is severe, offenders tend to make more attributions (more internal attribution, more controllability attribution, more external attribution, and more uncontrollability attribution) than when outcome is not severe. Maybe when offenders perceive that their behavior leads to a severe outcome, they search for any possible reason for the behavior in question so that they can decide which reason works best to explain their behavior to the victim. Hypotheses 5: Internal Attribution and Anticipated Consequences H5 proposed that the attributions offenders make influences their expectations about the responsibilities to be assigned to them and expected anger: When causes are perceived to be internal, offenders acknowledge that they are the major causes of the conflict; thus, they expect more responsibility assigned to them by victims (H5a). Furthermore, when offenders make more internal attributions, they expect victims to feel angrier toward their wrong-doings (H5b). The findings indicated that when offenders make more internal attribution, they expect more responsibility to be assigned to them. However, the degree of internal attribution made by offenders does not influence the expectation of anger felt by the victim. Given that the expected responsibility significantly influences the expected anger, as indicated by the results for Hypothesis 9, it’s possible that the direct influence of internal attribution on the expected anger is cancelled out by the mediating effect. Hypotheses 6: Controllability and Anticipated Consequences H6 proposed a positive influence of controllability attribution on anticipated consequences: When offenders perceive that they have control over what has happened,
126 they expect the victims to assign more responsibility to them because offenders could have done otherwise to prevent the conflict. Furthermore, offenders expect that the victims have more anger as well. The findings suggested that the controllability attribution positively affected the amount of responsibility that was expected to be assigned to offenders and the amount of anger expected to be felt by victims. When offenders realize that they could have prevented what has happened, they need to watch out for the anger from the victim. Hypotheses 7: External Attribution and Anticipated Consequences H7 specified a negative influence of external attribution on expected consequences. When an external attribution is made by offenders, they expect victims to assign less responsibility to them and feel less angry. The results did not tell such a story: External attributions did not significantly influence offenders’ expected responsibility and expected anger; furthermore, the positive sign of ? coefficient implied that when offenders make more external attribution, they expect the victims to assign more responsibility to them, which was opposite to the prediction. As discussed by Konovsky and Jaster (1989), external attribution may not work all the time. When offenders attribute their wrong-doings to external causes, they may be perceived as lacking in the ability to take care of their own behaviors or to be simply dishonest about what they have done. When victims believe that the offenders are not sincere, they may assign more responsibility to the offenders as a punishment. Hypotheses 8: Uncontrollability and Anticipated Consequences H8 specified that when offenders make more uncontrollability attribution, they believe there is nothing much they could have done to prevent what has happened.
127 Therefore, offenders do not see themselves as being responsible for the conflict and they expect less responsibility to be assigned to them and less anger to be felt by victims. It was found that when the causes are uncontrollable, offenders do not expect themselves to be held responsible for their wrong-doings. However, the results indicated that even when uncontrollable reasons are believed to lead to the conflict, offenders still expect victims to be angry, which is opposite to the prediction. An explanation of these results is that although uncontrollable reasons could shift some amount of responsibility from the offenders, what has happened does negatively influence the victims. Offenders may still expect certain negative emotions, such as anger, could arise naturally as the consequence of the offenders’ wrong-doing. Hypothesis 9: Expected Responsibility and Expected Anger Judgment of responsibility was believed to positively influence the feeling of anger (Allred et al., 1997; Liu & Yao, 2007). H9 stated that when offenders expect more responsibility to be assigned to them by victims, they also expect victims to be angrier. These findings were consistent with the hypothesis. When one is judged to be responsible for a negative incident, more feelings of anger are expected. Hypothesis 10: Expected Anger and the Choice of Accounts When offenders expected more anger to be felt by victims, to reduce the negative emotion and possibly decrease the responsibility, they are less likely to use a defensive account. The findings supported this hypothesis. Offenders apparently determine their accounts based on their expectations of the victims’ anger. If they sense the victims get furious because of their behaviors, they tend to be nice and be less defensive; however, if they believe victims are not so angry about their behavior, they are more likely to take a
128 defensive stand when explaining their behaviors. How do People Decide the Accounts to Use in Conflict Situations? When people know that their behaviors lead to a conflict, how will they explain their behaviors? According to the findings of the current research, people choose their accounts mainly based on how serious the outcome is. The relationship between the parties involved in a conflict may also play a role in how offenders explain their behaviors. For example, if a conflict happens between two acquaintances, e.g., classmates, offenders tend to choose a more defensive account to explain their wrongdoings, compared to when the conflict occurs between two close friends or between two strangers. But none of this matters if the conflict has severe consequences. When offenders perceive that a severe outcome has been caused by their behaviors, no matter with whom they have the conflict, offenders may think about defending themselves, perhaps as a method to protect their own images. But the determination of accounts is well calculated. When offenders perceive that their behaviors lead to a severe outcome, they seek out all possible reasons, internal, external, controllable, or uncontrollable, and they imagine how much responsibility the victim may assign to them and how angry the victim may be. After deliberately considering the causes of the conflict and the reaction from the victim, offenders will select the account that they think benefits them most. How do offenders know which account works best? They imagine how angry the victim may be and how much responsibility the victim is willing to assign to them. If offenders believe that the victims will assign more responsibility to them and the victims will become angry, the offenders will then be nice and use a less defensive account, such
129 as admitting their wrong-doings or offering some compensation. On the other hand, if offenders think that the victims are not so angry about the wrong-doing, they tend to take a more defensive stand, such as pointing to external reasons as the cause of the conflict or denying their connection to the behavior in question. Limitations of the Study and the Direction for Future Research This study has several limitations regarding the conceptualization of accounts, the measure for the accounts, and the hypothetical scenario used in the formal study. Limitation Regarding the Conceptualization of Accounts The current research examined four types of accounts: concession, justification, excuse, and refusal. Offenders’ actions were assumed to be self-serving: Offenders choose different types of accounts to defend themselves. Therefore, the four types of accounts represent the extent to which offenders defend themselves in a conflict situation. As indicated by the results from Pilot Study 5, concession was found to be the least defensive type of accounts and refusal was found to be the most defensive type of accounts. The defensiveness levels of justification and excuse were in the middle and justification was more defensive than excuse, which was opposite to what Schönbach (1990) proposed. However, one caveat to this assumption is that not all accounts are selfserving. An offender may use an account to honestly describe what has occurred in a conflict situation, so that the account is not intended to defend oneself at all. The defensiveness continuum of accounts cannot be applied to all types of accounts. Future research may explore other continua to describe different types of accounts and may consider other types of accounts other than the four types investigated in the current research.
130 Further, the current research argues that offenders use different types of accounts to maintain their self image or maintain the relationship with others. However, the goals of different types of accounts are assumed, not tested, in this dissertation. Offenders may consider different goals in different relationships, rather than simply trying to maintain the relationship. Future research should scrutinize the different goals offenders may try to fulfill and examine the account-making process through a goal perspective. Limitation Regarding the Dependent Measures Each construct proposed in the theoretical model was measured by 5 to 7 items, which were developed and tested in several pilot studies. Reliability coefficients for all the measures were acceptable and in some cases good. However, fit indices from the majority of the measurement models indicate non-unidimensional factor structures. For example, the measures for the four types of attributions all had more than one underlying factors, which may contribute to the non-significant results about attributional processes. Previous research on accounts often asked participants to self-report the actual accounts they used in an incident and use coders to code these responses. The current research took a different approach. A list of accounts was developed and provided to participants to indicate their likelihood of choosing each account. Although this approach provides a measure that can be adopted in other conflict situations, it has its limitation. The accounts people make in a conflict situation may be much more complicated than the list of accounts provided in the measure. Also people often use more than one type of accounts. The measure used in the current research constrained participants to only a limited amount of accounts. Future research may use both methods (i.e., self report of actual accounts and selection from a list of provided accounts) to better understand
131 people’s choice of accounts and may consider examining the sequences of accounts, instead of a single account. Further, this study only focused on the offenders’ perspective for selecting accounts. To determine the accounts to be used in a conflict situation, offenders seem to go through a series of calculations: They consider the reasons that lead to their wrongdoing and they imagine the victims’ reactions, such as how much responsibility the victims will assign to the offenders and how angry the victims will be. But all these variables were assessed only from the offenders’ perspective. Whether the accounts offenders select will work out (i.e., accepted by the victims) remains a question. The acceptance of accounts is beyond the scope of the study but is an important topic for the future research. Limitation Regarding the Hypothetical Conflict Scenario This study used one hypothetical conflict scenario and asked participants to imagine they were the offenders in the conflict. Although using hypothetical scenarios had its advantages, the problems with hypothetical scenarios bring cautions about interpreting and applying the study’s results. The current research is based on the assumption that people deliberately process information at the time of conflict and they are engaged in making attributions, expecting consequences for their communication, and selecting an account. However, Nisbett and Wilson (1977) pointed out that “in dissonance and attribution studies…people appear to be unable to report that a cognitive process has occurred” (p. 246), which may explain why the paths to the attributions in the original model were not significant. It is also possible that the processes described in an experiment may be different from the processes people experience in an actual
132 interaction. Future research may consider using actual interactions to examine the processes offender may experience in a conflict situation. Future Directions: Cultural Differences in Accounts The current research only studied the attributional processes involving accounts in the United States. Although some hypotheses were not supported, overall the model had a moderate fit. The question is whether the model can be extended to different cultures. Some features of account-giving and attributional processes provided foundations for possible cultural differences in the proposed model. First, McLaughlin, Cody et al. (1983), Gonzales, Manning, and Haugen (1992), and Takaku (2000) have argued that accounts have implications for face concerns: In general, different types of accounts pose different degrees of threats to offenders’ and victims’ face. Given the consistent findings on cross-cultural difference in face concerns (e.g., Cai & Fink, 2002; Ting-Toomey & Oetzel, 2002), it is expected that culture will influence people’s choice of accounts. McLaughlin, Cody et al. (1983) and Gonzales et al. (1992) argued that concessions, an explanation that admits responsibility and offers compensation, make offenders vulnerable to others’ blame. But using concessions indicates that offenders feel sorry about their behaviors and are willing to do something to compensate the victims, which saves the victims’ face. McLaughlin, Cody et al. and Gonzales et al. specifically argued that using concessions, among all four types of accounts, poses a maximum threat to offenders’ face but a minimum threat to the victims’ face. Excuses also threaten offenders’ self face because excuses, like concessions, admit offenders’ responsibility for the behavior that led to a conflict. However, when people use excuses, they can point to
133 external reasons as the causes of the behavior in question, the level of threat to offenders’ face is less than that posed by a concession. On the other hand, when offenders use external reasons to explain their behavior, although offenders admit that they need to take responsibility for their behavior, they are not so guilty because they don’t have too much control over what has happened. Excuses pose a greater threat to victims’ face than do concessions. Further, justifications suggest that offenders believe that their behaviors are consistent with a higher goal. Therefore, offenders’ self face is protected because offenders perceive that they do not have much freedom when it comes to the behavior in question; they have to behave in a certain way to fulfill a superordinate goal. Although justifications acknowledge offenders’ wrong-doing, offenders do not intend to accept that they violate a social norm, which poses a higher threat to victims’ face than either concessions or excuses. McLaughlin, O’Hair, et. al. (1983) and Gonzales et al. (1992) stated that when refusals are used, offenders could deny their connection with the behavior in question or deny that the behavior in question is inappropriate; as a result, refusals protect offenders’ face to the maximum degree but pose the greatest threat to victims’ face. Studies on conflict management styles have suggested that face concerns vary as a function of culture (e.g., Ting-Toomey & Oetzel, 2002). Most of the literature on face concerns and conflict management has focused on Chinese and Americans (e.g., Brew & Cairns, 2004) and has suggested that Chinese have a higher concern for others’ face than Americans do;on the other hand, the literature has argued that Americans’ concern is more with self face than others’ face (Ting-Toomey & Oetzel, 2002). Although there are
134 some arguments against this conclusion in research (e.g., Cai & Donohue, 1997), given this discussion on face concern and accounts, Chinese offenders are expected to be more likely to use a less defensive account in a conflict situation whereas American offenders may be more likely to use a more defensive account to handle conflict. A second reason for cultural differences in account use is that attributional patterns are different across cultures. For example, the fundamental attribution error (FAE, i.e., a tendency to overestimate the influence of personal traits and underestimate the influence of situational factors; Ross, 1977) is found in the United States but the generality of the FAE cross-culturally is arguable. Studies have found that Chinese, as compared to Americans, are more likely to rely on external factors to explain an actor’s behaviors (e.g., Choi, Nisbett, & Norenzayan, 1999). However, when a negative event occurs, Asians are more likely to attribute their own failure to internal reasons (e.g., lack of effort) whereas Americans attribute the failure to environmental factors (e.g., the difficulty of an exam) (Salili, Maehr, & Gillmore, 1976). Findings from those attribution studies suggest that when a conflict occurs, Chinese are more likely to make internal attributions about their own behavior whereas Americans are more likely to make external attributions. Consequently, to reduce the responsibility assigned by victims and alleviate the anger that accompanies a conflict, Chinese are less likely to use a defensive account than Americans. Future research may test the model in different cultures and examine cultural influences on attributional processes involved in the selection of accounts. Significance of the Study and Conclusion This study attempted to answer one question: What attributional processes do
135 offenders engage in to determine the account to use in a conflict situation? A model was established to examine the influence of outcome severity and relational closeness on offenders’ choice of accounts, mediated by the attributions offenders make and offenders’ anticipation of the victim’s reactions. The data moderately supported the model and showed that when the outcome of a conflict is severe, offenders tend to make more attributions, expect more responsibility assigned to them, and expect more anger felt by the victim, compared to when the outcome is not severe. Based on the expected reaction from the victim, offenders will choose an account to explain their wrong-doings (i.e., the more expected anger felt by victims, the less likely a defensive account would be chosen). The driving force for this study came from the discrepancy between the significant role of accounts in conflict management and the research that has been done to explore this area. When looking at a simple interpersonal conflict (e.g., the vignette at the beginning of the current research), explaining one’s behavior constitutes a major part of the conflict process. Offenders’ explanation of their behaviors actually lay the foundation for how a conflict will be resolved. Offenders’ accounts directly influence victims’ perceptions of the causes of the conflict, the offenders’ intentions, the amount of responsibility that can be assigned to offenders, and in general, the way the conflict can be handled (Sitkin & Bies, 1993). Although accounts play an important role in conflict management, relatively little research has explored this simple way of handling conflict (e.g., McLaughlin, O’Hair et al., 1983). The current research, one of few studies examining how offenders select their accounts, adds to the understanding of accountdevelopment processes.
136 The current research contributes to the literatures of conflict management and communication in general because it provides a theoretical model for understanding the account-giving processes via an attributional perspective. The current study focused on interpersonal conflict; however, the results can be cautiously applied to other contexts, such as organizational communication. In addition, the model can be revised and evaluated in diverse ethnicities and cultures, contributing to the cross-cultural communication literature. The current research has not explored the account-giving process in different cultural contexts; but the discussion of face concerns and attributions points to possible cross-cultural studies on accounts in the future. The current research contributes to the account literatures in three different aspects. First, it changes the focus of account studies from typology development to understanding the account-selecting processes. For a long time, McLaughlin et al. (1992) have promoted greater understanding of how people create accounts, in addition to developing different types of accounts. Researchers have been looking into this issue but not many studies have been generated. The current research answers the call to examine one account-making process and provides important information about how offenders determine the accounts used in a conflict situation. Second, the current research took an approach to examine accounts that was different from past research. Traditionally, participants are asked to self report a negative event and provide the actual accounts they used at that time. After that, trained coders code the accounts into different categories. Although this method may have more external validity, it limits researchers’ ability to examine specific antecedent conditions of interest. The current research employed an experimental design to study the variables
137 of interest and, instead of asking participants to report their accounts, provided a list of accounts to which they responded. This technique is easy to administer and can be adapted to different conflict situations. Third, researchers have been mapping types of accounts on different continua; for example, Schönbach (1990) used a defensiveness continuum to describe concessions, justifications, excuses, and refusals. Schönbach argued that concessions are least defensive, followed by justifications, and excuses are more defensive than justifications but less defensive than refusals. However, Schönbach did not provide any empirical evidence for this proposition. One pilot study for the current research actually tested the defensiveness levels of these four types of accounts and found that concessions, indeed, are the least defensive account whereas refusals are the most defensive. The defensiveness of justifications and excuses are significantly different from each other and justifications are more defensive than excuses, which is opposite to Schönbach’s original proposition. Whether Schönbach’s proposition is wrong cannot be determined by results from a single study. The current research provides certain empirical findings on the defensiveness of each type of accounts but more studies are needed for an affirmative answer in the future. Different types of accounts have their own features. When used in an interpersonal conflict, each type of accounts reflects different expectations offenders have about victims’ reactions and possibly implies different goals offenders try to achieve. The current research found that offenders will take a less defensive stand to explain their behavior when they expect the victim to be angry and to assign more responsibility to them, in another word, when offenders expect the victim to be tough. However, when
138 offenders expect the victim to be soft, that is to say, offenders do not expect the victim to be so angry and to assign much responsibility to them, offenders then will play hard. A Chinese idiom describes it well, qi ruan pa ying (bully the soft ones but be afraid of the tough ones). It’s in the human nature.
139 Appendix A Instruments for Manipulation Checks in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Perceived Outcome Severity 1. To what extent does your mistake influence Person B? 2. How severe is the outcome of this situation for Person B? 3. How severely does your mistake affect Person B? ______ ______ ______
4. How much influence does your mistake have on Person B’s scholarship for the next year? 5. How much difficulty does your mistake have on Person B’s obtaining his or her scholarship for the next year? 6. To what extent do you think the outcome is severe? Perceived Relational Closeness 1. To what extent do you think Person B is a good friend of yours? 2. To what extent do you think you and Person B are close to each other? 3. How much knowledge do you have of Person B? 4. How familiar are you with Person B? 5. How close are you with Person B? 6. How well do you know Person B? ______ ______ ______ ______ ______ ______ ______ ______ ______
Note. The labels (i.e., perceived outcome severity and perceived relational closeness) are for readers’ convenience. They were not shown in the actual questionnaire. The same idea was followed for all the items listed below in Appendices B, C, D, and E.
140 Appendix B Instruments for Attributions in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Degree of Internal Attributions 1. How likely is that what has happened was due to your own carelessness? 2. How likely is that what has happened was due to internal traits, such as your personality? ______ ______
3. How likely is that what has happened was due to you not paying enough attention when the readings were assigned? ______
4. How likely is that what has happened was because you did not care about what grade you and Person B could get for this assignment? ______
5. To what extent do you believe what has happened was because you did not spend enough time on the assignment? 6. To what extent do you think what has happened was because of you? ______ ______
7. To what extent do you think your personality determines what has happened? ______ Controllability 1. How likely is that what has happened was avoidable? 2. How much control do you have over what has happened? 3. How likely is it that you could have prevented this mistake? ______ ______ ______
4. To what extent do you believe you could have done otherwise to prevent what has happened? 5. To what extent do you think you could have done differently to avoid what has ______
141 happened? Degree of External Attributions 1. How likely is that what has happened was due to the professor not giving clear instructions for the assignment? ______ ______
2. How likely is that what has happened was caused by other people instead of you? ______ 3. How likely is that what has happened was because the wordings in the instruction were confusing? ______
4. How likely is that what has happened was due to some external reason (for example, someone mistakenly took your notes away so you had nothing to rely for the instructions for the project)? ______
5. To what extent do you think what has happened was due to Person B’s decision to meet the night before the due date? ______
6. To what extent do you think what has happened was because of other people? ______ 7. To what extent do you think environmental reasons determine what has happened? ______ Uncontrollability 1. How likely is that what happened was out of your control? ______
2. To what extent do you believe there is not much you could do to prevent what has happened? 3. How likely is that you have no control over what has happened? 4. To what extent do you think you cannot prevent what has happened? 5. To what extent do you think what has happened was inevitable? ______ ______ ______ ______
142 Note. Based on results from confirmatory factor analysis and principal components analysis, not all seven items that measured the degree of internal attribution were included in the model that was tested. The four items used in the model that was tested were item 1, item 3, item 5, and item 6.
143 Appendix C Instruments for Anticipated Consequences in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Expected Responsibility 1. How much responsibility would Person B assign to you for what has happened? ______ 2. To what extent would Person B think you are responsible for what has happened? ______ 3. To what extent would Person B think you are chargeable for what has happened? ______ 4. To what extent would Person B think you are accountable for what has happened? ______ 5. To what extent would Person B think you should take full responsibility for what has happened? ______ 6. To what extent would Person B think what has happened should be attributed to you? ______ Expected Anger 1. How angry do you expect Person B to be? 2. How irritated do you expect Person B to be? 3. How furious do you expect Person B to be? 4. How mad do you expect Person B to be? ______ ______ ______ ______
144 5. How fuming do you expect Person B to be? 6. How annoyed do you expect Person B to be? ______ ______
145 Appendix D Instruments for the Likelihood of Selecting an Account in the Formal Study Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Concessions Concession 1. I am sorry that I misunderstood the instructions. Is there any way that I can make it up to you? Concession 2. I am sorry that I misunderstood the instruction. I will take the full responsibility and it’s all my fault. Concession 3. I am sorry that I misunderstood the instruction. I will talk with the professor to see whether we can come up with a possible solution. ______ ______ ______
Concession 4. I am sorry that I misunderstood the instruction. I hope you know that I am shocked about this grade and feel horrible about bringing your grade down. ______
Concession 5. I am sorry that I misunderstood the instruction. Could I possibly write a letter to complement your scholarship application to explain what happened? Justifications Justification 1. I am sorry that the grade influenced your scholarship. But I honestly thought I was doing the correct work. ______ ______
Justification 2. I am sorry that the grade influenced your scholarship. You know that I am a good student and I would never want this grade myself. It’s a complete misunderstanding. Justification 3. I am sorry that the grade influenced your scholarship. I knew how important the grade was to you but I did try my best to get a good grade. ______ ______
146 Justification 4. I am sorry that the grade influenced your scholarship. But I worked very hard on this assignment. We should have gotten together to edit the paper earlier. ______ Justification 5. I am sorry that the grade influenced your scholarship. But I thought I did on the correct readings and I did a thorough job on the readings that I reported on. ______ Excuses Excuse 1. I am sorry that I misunderstood the instruction. But the wording in the direction was really confusing and that may be the reason why I wrote the wrong portion of the paper. ______
Excuse 2. I am sorry that I misunderstood the instruction. I was being careless and confused with all the paper work. ______
Excuse 3. I am sorry that I misunderstood the instruction. It’s just ……there were other things that contributed to the mistake. ______
Excuse 4. I am sorry that I misunderstood the instruction. I was so nervous about this course and tried to do well on this assignment; but maybe my nervousness led me to write the wrong portion of the paper. ______
Excuse 5. I am sorry that I misunderstood the instruction. My sister accidently took my notes away and I did not have anything to rely on to write the paper. I thought I remembered the instruction correctly. Refusals Refusal 1. I know the readings I read were the right material for the paper. I did not find anything wrong to what I contributed to the paper. ______ ______
Refusal 2. I have to say I did more work than you and I don’t think the grading is fair. ______
147 Refusal 3. If this course means a lot to you, you should have put more focus on the collaboration. ______
Refusal 4. You made me do all the work and not call me on what the paper should be about. ______
Refusal 5. I told you we should not wait until the last minute. We deserve the grade we got. It’s all your fault not contributing to the paper and we got a poor grade. ______
148 Appendix E Pilot Study 1: Questionnaire for Collecting Conflict Scenarios CONFLICT STRATEGY STUDY Department of Communication University of Maryland, College Park Spring, 2007 PLEASE READ CAREFULLY In this study, we are interested in learning what you say when you have to account for yourselves: That is, what you say when you feel that, for one reason or another, you have to answer to someone else because of something you did or failed to do, which resulted into a conflict between you and another person. For example, you might have to account to your girlfriend or boyfriend for the fact that when you used his or her laptop to surf online, the web page you visited contained virus that wiped out all his or her important documents; or your classmate might expect you to account for your failure to return the lecture notes to her, which left her little time to prepare for the final and she got an overall low grade for that class. Sometimes we give accounts even if no one asks for one, because we think that our behavior needs explanations, or because we think that the other person thinks that our behavior was questionable. For example, you were responsible for making powerpoint slides for a group presentation. In the middle of the presentation, one of your slides accidentally made weird noises from the animation you forgot to delete. Although your group members did not say anything, you felt you ought to say something to explain that weird noise. So, we would like you to recall a recent conflict you had with another person resulted from something you did, and answer the following questions on the next page. Please read all instructions carefully before you answer each question. At the very end, you will have an opportunity to comment on our study. Please do not leave any questions out. This would cause that we have to exclude your answers from our study. Please note that there are no correct or incorrect answers. We are simply interested in what your account was in the situation you recall.
149 1. In the table, please recall a situation in which a conflict occurred recently between you and another person because of something YOU did. You should know that conflicts do not always have satisfactory endings. So you can describe any conflict you actually had due to something you did. Make sure to describe the occurrence in detail. When did the conflict happen (approximately)? Relationship between you and the other person in the conflict. What happened (please describe the incidence in detail, from the start of the conflict to its solutions)? To your knowledge, what was/were the cause(s) of the conflict (please list all reasons that you can think of)? To your knowledge, what did you expect the other person would think the cause(s) was/were (please all reasons that you expect the other person would think of)? What was your emotion during the conflict (e.g., did you feel angry)? Please list all emotions you experienced at the moment of the conflict. What was your expectation of the other person’s emotion at the time of conflict (e.g., did you expect the other person feel angry)? Please list all emotions you expected the other person might have experienced at the moment of the conflict.
150
2. When the conflict described in question 1 happened, did the other person say anything that made you feel you ought to explain your behaviors to the other person?
If there was a verbalized request for an explanation, Please recall as best as you can the actual words used by the other person
Yes. He/She said: “_________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ _________________________________________________.”
If no verbalized request for an explanation was presented, please recall as best as you can why you thought, at that moment, an explanation was necessary.
No. But I thought I should explain my behavior because __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ __________________________________________________ _________________________________________________.
151 3. Please recall, in as close to the exact words as possible, what you said to explain your behavior to the other person in the conflict described in question 1. To explain my behavior, I said: “____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ ___________________________________________________________________.”
4. Are you satisfied with the endings of the conflict mentioned in question 1 (circle one): YES NO
5. If given a second chance, what else would you say other than the explanation you provided above? To explain my behavior, I would have said: “____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ ___________________________________________________________________.” We would also like to know a little more about you. Please complete the following information. 1. What is your gender? (circle one) MALE FEMALE
2. What is your age? I am ________ years old.
152 3. Which race/ethnicity label describes you best? (circle one) African-American / Black Asian-American / Asian Caucasian / White Hispanic / Latino Native American Other (Please specify): _____________
4. What is your nationality (which country issues you a passport)? _____________ 5. Is English your native (first) language? (circle one) YES NO
If not, what is your native language? _____________________ 6. What is your major in college? ____________ 7. What year are you in college? (circle one) FRESHMAN JUNIOR GRADUATE SOPHOMORE SENIOR OTHER (Please specify): ___________________
Do you have any comments regarding our study? _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ _____________________________________________________________________ Thank you for completing all the questions. We appreciate your participation!
153 Appendix F Pilot Study 2: Questionnaire for Sorting Task Sorting out Conflict Information Department of Communication University of Maryland, College Park August, 2007 Purpose: The study contains information from another study. You will read about different conflicts. Your task is to sort those conflicts according to the severity of the conflict (severe vs. not severe) and the closeness of the relationship between the people in that conflict (not close at all, somewhat close, very close). Remember, there is no right or wrong way to sort those scenarios. You are sorting the scenarios based on your own judgment.
Instructions: You will receive 93 numbered index cards. On each card there is one conflict. Some of them are described in detail and some are less detailed. Step 1: Please read over all 93 conflicts carefully, getting familiar with them and making sure you understand each of them. If you have any questions, please feel free to ask the researcher.
154 Step 2: After you get familiar with these 93 conflicts, please sort them into the sorting table on the next page (Table 1). You will sort these conflicts based on two criteria: the severity of the conflict and closeness of the relationship between the participants to the conflict. *Severity of the conflict refers to how serious the consequence might be because of the conflict mentioned on the card. *Closeness of the Relationship refers to the extent that the people in the conflict are close to each other. Please put the number of the card in the most appropriate cell. And remember: Each conflict can be put in ONE AND ONLY ONE CELL in the table. If you are not sure which cell to put a conflict into, please write down the card number in Table 2 (Other) and explain the reason why you cannot determine which cell the conflict should be in.
155 Sorting Table (Table 1)
Severity of the Conflict
Severe
Not severe
Closeness of the Relationship
Not close at all
Somewhat close
Very close
156
Other (Table 2) Card Number Please specify the reason why you cannot determine the cell from Table 1 you should put this scenario into.
157
Step 3: Now you have finished sorting all 93 conflicts. Before we move to the next step, do you want to take a final look at your sorting? Do you want to change any sorting?
158 Step 4: We would also like to know a little more about you. Please complete the following information. 1. What is your gender? (circle one) MALE FEMALE
2. What is your age? I am ________ years old. 3. Which race/ethnicity label describes you best? (circle one) African-American / Black Asian-American / Asian Caucasian / White Hispanic / Latino Native American Other (Please specify): _____________
4. What is your nationality or citizenship? _____________ 5. Is English your native (first) language? (circle one) YES NO
If not, what is your native language? _____________________ 6. What is your major in college? ____________ 7. What year are you in college? (circle one) FRESHMAN JUNIOR GRADUATE SOPHOMORE SENIOR OTHER (Please specify): ___________________
8. Please provide your SONA ID here: ____________________________
159 Appendix G Conflict Incidents Reported in Pilot Study 1 How sorters sorted the 93 conflicts reported in Pilot Study 1 was examined and the following five incidents were applicable in all six experimental conditions. The first incident was selected to develop the hypothetical conflict scenario used in the formal study because the conflict occurred in a context familiar to future participants. Incident 1: “There were five students including me that decided to form a virtual study group. I was not in class the day that everyone decided who would outline what chapters. I was given the wrong chapters so I did the same work as another student!” Incident 2: “Basically my roommate was angry at me because I didn't ‘completely’ wake him up to study for his finals exam. I believe it was him that caused the conflict because he made a huge deal about it. I came back at 12pm and I woke him up then but he was angry.” Incident 3: “Even though my boyfriend told me he was coming to pick me up, I wasn’t ready and we missed the movie that he had really wanted to see. He was mad because he had given me over an hour to get ready and I had been doing other things instead. I apologized many times.” Incident 4: “I copied my work schedule down incorrectly, so I showed up to work an hour late. I ended up showing up late, excused by my manager, but I felt like I'd disappointed him.” Incident 5: “It happened about two weeks ago. I was sitting directly opposite to the air conditioner in the office and I was freezing to death. So I reset the temperature. However, one of my officemates immediately stood up and turned the AC temperature
160 back, which upset me. I said that I could not handle the room that cold, and he pretended not to hear my complaints.”
161 Appendix H Hypothetical Conflict Scenarios Used in the Formal Study 1. Scenario for the condition of close relationship and high outcome severity: You and Person B grew up together and attended the same university. The two of you have been best friends for many years. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was very upset because such a low grade resulted in him or her not getting next year's scholarship.
2. Scenario for the condition of close relationship and low outcome severity: You and Person B grew up together and attended the same university. The two of you have been best friends for many years. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 10% of the overall grade. Both group members would receive the same grade for this assignment.
162 The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be influenced. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 70%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was not happy but he or she did great on other assignments and believed it would not influence his or her getting next year's scholarship.
3. Scenario for the condition of somewhat close relationship and high outcome severity: You and Person B were in couple of classes together before. Although the two of you know each other, you only see each other in class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was very upset because
163 such a low grade resulted in him or her not getting next year's scholarship.
4. Scenario for the condition of somewhat close relationship and low outcome severity: You and Person B were in couple of classes together before. Although the two of you know each other, you only see each other in class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 10% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be influenced. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 70%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was not happy but he or she did great on other assignments and believed it would not influence his or her getting next year's scholarship.
5. Scenario for the condition of not–close-at–all relationship and high outcome severity: You and Person B do not really know each other. You happened to sit next to each other on the first day of class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade.
164 Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was very upset because such a low grade resulted in him or her not getting next year's scholarship.
6. Scenario for the condition of not-close-at–all relationship and low outcome severity: You and Person B do not really know each other. You happened to sit next to each other on the first day of class. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 10% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, there is a possibility that his or her scholarship for the next year will be influenced. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 70%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong readings. Person B was not happy but he or she did great
165 on other assignments and believed it would not influence his or her getting next year's scholarship.
166 Appendix I Pilot Study 3: Questionnaire for Manipulation Checks Scenario Assessment Department of Communication University of Maryland, College Park Spring, 2008 Note. The same instructions were used in Pilot Studies 3, 5, 6, and the formal study. Researchers at the University of Maryland are interested in knowing how people explain themselves in interactions. What do you need to do in this study? 1. Read a hypothetical incident. There are two persons (Person A and Person B) involved in the incident. Please image you were Person A. 2. Provide possible explanations by imagining you were Person A. 3. Answer some questions regarding the incident. Ready? Start!
*Part I: Situation * You and Person B grew up together and attended the same university. The two of you have been best friends for many years. You and Person B took a tough course together last semester. You were teamed together and assigned to write a paper to summarize readings for the class. According to the syllabus, the paper counts for 40% of the overall grade. Both group members would receive the same grade for this assignment. The professor assigned the work evenly for each member. Person B told you that if the grade for this course was not good, his or her scholarship for the next year would definitely be in jeopardy. You and Person B worked independently and combined your work the night before the due date and turned it in as one group assignment. When you and Person B received the graded paper, you were shocked to learn that you had earned a score of only 60%. It turned out that you had misunderstood the directions and had written your part of the paper on the wrong
167 Part II: Providing explanations Imagine that you were the one that worked with Person B on this group project. What would you say to Person B to explain what you did (i.e., you wrote your portion of the review paper on wrong readings, which contributed to receiving a 40% for the review paper)?
168 Part III: Questions about the Situation Still imagine yourself in the situation. Read the instructions and answer the following questions. Please use the following instructions to answer the questions: Use a number from 0 (zero) to
infinity to respond to each question. Zero means not at all and 100 means a moderate amount; you may use numbers from zero to infinity but please use a specific number. For example, suppose you were asked the question “how much do you like chocolate?” If you like chocolate a moderate amount, then you would rate your answer as 100. If you like chocolate only a little, then you may rate your answer as 20; but if you really like chocolate, you may rate your answer as 300. For each question, you would rate your answer using any numbers based on the following scale: 0 = not at all 100 = moderate amount Use any number from zero on up
To ensure you understand how to use the scale, first practice using two questions that are not relevant to the situation you just read. Question 1. How much do you like the food offered at the dining hall of your school? 2. How much do you like the service at the main library of your school? Scale
0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up
Your Rating
When you feel ready to use the scale you just practiced, turn to the next page. Please remember: Answer questions by IMAGINING the incident happened between you and Person B.
169 Question 1. How severe is the outcome of this situation? 2. How severely did your mistake affect Person B? 3. How much influence would your mistake have on Person B’s fellowship application? 4. How much difficulty would your mistake have on Person B’s ability to achieve his or her goals? 5. How close are you with Person B? 6. How much knowledge do you have of Person B? 7. How familiar are you with Person B? 8. How well do you know Person B? Scale
0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up 0 = not at all 100 = moderate amount Use any number from zero on up
Your Rating
1. My age is ____________ years. 2. I am (circle one) FEMALE MALE
3. I am a citizen of _______________________ (country). 4. My ethnicity is or most closely to African-American / Black Asian-American / Asian Caucasian / White 5. My native language is (circle one) Hispanic / Latino Native American Other (Please specify): _____________ ENGLISH Other (please specify): ______________________
6. I am a
(circle one) FRESHMAN JUNIOR SOPHOMORE SENIOR170 GRADUATE STUDENT OTHER (please specify): ___________________.
171 Appendix J Instruments for Defensiveness of Accounts in Pilot Study 5 Scale: 0 = not at all 100 = moderate amount Use any number from zero on up
Defensiveness 1. To what extent were you being self-protective? 2. To what extent were you being guarded? 3. To what extent were you sticking out for yourself? 4. To what extent were you being assertive? 5. To what extent were you trying to shift responsibility? 6. To what extent were you being defensive? ______ ______ ______ ______ ______ ______
172 Appendix K List of Open-Ended Questions Used in Pilot Study 6 Note. The purpose of using the following questions was to obtain comments from participants regarding the clarity of the questionnaire that measured all dependent variables. 1. Could you describe your general impression of the questionnaire? 2. What do you think the purpose of the study is? 3. What do you think the researchers want to measure? 4. Do you feel the questions are clear? 5. Are there any questions you found confusing or difficult to answer? Please explain. 6. Do you have any other comments on the study?
173 Appendix L Descriptives of the Indicators in the Formal Study after Transformation
Indicators SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 REL1 REL2 REL3 REL4 REL5 REL6 IA1 IA2 IA3 IA4 IA5 IA6 IA7 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RES1 RES2
Mean 6.59 6.63 6.56 6.31 6.04 6.00 4.88 4.67 4.99 5.06 4.73 4.91 10.87 5.85 10.35 2.53 7.23 10.42 4.85 6.84 6.09 6.77 6.13 5.93 11.81 6.04 12.48 6.77 14.13 6.88 5.95 2.79 3.35 2.93 3.10 2.47 4.83 4.91
SD 1.50 2.06 1.84 2.54 2.06 1.91 2.32 2.45 2.12 2.08 2.36 2.19 3.45 3.58 3.49 3.01 3.44 3.38 3.61 2.10 2.33 1.87 1.70 1.46 6.02 4.91 5.66 4.78 7.32 4.52 4.70 2.45 2.35 2.58 2.55 2.51 1.26 1.21
Range Trim Value ? 8.81 9.06 7.66 11.22 9.49 8.80 8.80 8.80 8.80 8.80 8.80 8.80 16.39 10.85 16.39 7.94 13.02 16.68 10.85 10.47 10.47 6.69 7.52 8.27 23.04 12.59 23.04 12.59 27.72 12.59 12.59 6.31 6.31 6.41 6.31 6.31 6.92 6.92 501 900 700 1000 620 500 500 500 500 500 500 500 500 200 500 100 300 520 200 1000 1000 1000 800 500 300 100 300 100 420 100 100 100 100 100 104 100 1000 1000 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.35 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.34 0.34 0.34 0.34 0.34 0.55 0.55 0.55 0.55 0.55 0.55 0.55 0.40 0.40 0.40 0.40 0.40 0.28 0.28
Skewness -0.23 0.52 0.32 0.31 -0.16 -0.31 -0.18 -0.09 0.22 0.22 0.07 0.26 -0.14 -0.40 0.15 0.61 -0.22 0.30 -0.05 0.37 -0.01 0.84 0.67 -0.26 0.11 -0.03 0.31 -0.27 0.26 -0.26 0.01 0.03 -0.37 0.01 -0.10 0.30 0.04 -0.03
Standard Error 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16
174
Indicators RES3 RES4 RES5 RES6 ANG1 ANG2 ANG3 ANG4 ANG5 ANG6 Concession1 Concession2 Concession3 Concession4 Concession5 Justification1 Justification2 Justification3 Justification4 Justification5 Excuse1 Excuse2 Excuse3 Excuse4 Excuse5 Refusal1 Refusal2 Refusal3 Refusal4 Refusal5
Mean 4.86 4.90 4.60 4.75 4.96 4.95 4.59 4.88 4.44 5.06 10.39 9.41 13.07 10.42 8.28 9.10 9.22 7.69 8.20 8.12 6.99 5.86 3.77 3.56 2.25 1.89 1.30 2.74 1.50 1.85
SD 1.20 1.20 1.33 1.30 1.21 1.17 1.40 1.23 1.43 1.16 3.96 4.07 4.72 3.87 4.61 4.26 4.34 4.06 4.65 4.51 3.70 3.63 3.31 3.19 2.88 2.75 2.20 3.16 2.35 2.73
Range Trim Value ? 6.92 6.92 6.92 6.92 5.46 7.19 7.19 7.41 6.97 7.19 16.39 16.39 18.54 16.39 16.39 16.39 16.39 14.21 16.39 16.39 13.02 10.85 8.64 7.98 7.75 7.10 5.81 7.94 5.85 7.18 1000 1000 1000 1000 1000 900 900 1000 810 900 500 500 1000 500 500 500 500 364 500 500 300 200 120.5 100 94.6 80.1 50 100 50.7 80 0.28 0.28 0.28 0.28 0.29 0.29 0.29 0.29 0.29 0.29 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45 0.45
Skewness 0.19 0.23 -0.16 -0.16 0.70 0.26 -0.08 0.43 -0.28 0.09 -0.16 0.09 0.76 -0.12 0.08 0.07 -0.08 -0.23 0.06 0.08 -0.20 -0.42 0.04 0.08 0.77 0.95 1.27 0.57 1.08 1.03
Standard Error 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16
175 [See the table below for item label used above]. Indicator Label SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 REL1 REL2 REL3 REL4 REL5 REL6 IA1 IA2 IA3 IA4 IA5 IA6 IA7 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RES1 RES2 RES3 RES4 RES5 RES6 Description Perceived Outcome Severity Item 1 Perceived Outcome Severity Item 2 Perceived Outcome Severity Item 3 Perceived Outcome Severity Item 4 Perceived Outcome Severity Item 5 Perceived Outcome Severity Item 6 Perceived Relational Closeness Item 1 Perceived Relational Closeness Item 2 Perceived Relational Closeness Item 3 Perceived Relational Closeness Item 4 Perceived Relational Closeness Item 5 Perceived Relational Closeness Item 6 Internal Attribution Item 1 Internal Attribution Item 2 Internal Attribution Item 3 Internal Attribution Item 4 Internal Attribution Item 5 Internal Attribution Item 6 Internal Attribution Item 7 Controllability Item 1 Controllability Item 2 Controllability Item 3 Controllability Item 4 Controllability Item 5 External Attribution Item 1 External Attribution Item 2 External Attribution Item 3 External Attribution Item 4 External Attribution Item 5 External Attribution Item 6 External Attribution Item 7 Uncontrollability Item 1 Uncontrollability Item 2 Uncontrollability Item 3 Uncontrollability Item 4 Uncontrollability Item 5 Expected Responsibility Item 1 Expected Responsibility Item 2 Expected Responsibility Item 3 Expected Responsibility Item 4 Expected Responsibility Item 5 Expected Responsibility Item 6
176 Indicator Label ANG1 ANG2 ANG3 ANG4 ANG5 ANG6 Concession1 Concession2 Concession3 Concession4 Concession5 Justification1 Justification2 Justification3 Justification4 Justification5 Excuse1 Excuse2 Excuse3 Excuse4 Excuse5 Refusal1 Refusal2 Refusal3 Refusal4 Refusal5 Description Expected Anger Item 1 Expected Anger Item 2 Expected Anger Item 3 Expected Anger Item 4 Expected Anger Item 5 Expected Anger Item 6 Concession 1 Concession 2 Concession 3 Concession 4 Concession 5 Justification 1 Justification 2 Justification 3 Justification 4 Justification 5 Excuse 1 Excuse 2 Excuse 3 Excuse 4 Excuse 5 Refusal 1 Refusal 2 Refusal 3 Refusal 4 Refusal 5
177 Appendix M Covariance Matrix in LISREL for the Proposed Model in Figure 2 Covariance Matrix Item SEV1 -------2.24 2.09 2.04 2.38 2.04 1.90 0.38 2.12 2.26 0.61 1.89 1.30 1.05 1.46 1.11 0.81 2.49 1.26 2.75 0.45 2.82 1.15 0.17 0.44 0.30 0.64 0.65 0.24 0.95 0.93 0.90 0.89 0.95 1.00 1.09 1.00 1.16 SEV2 -------SEV3 -------SEV4 -------SEV5 -------SEV6 --------
SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3
4.27 3.39 4.30 3.40 3.37 0.49 2.62 2.42 0.36 1.69 1.70 1.27 1.76 1.26 1.01 3.39 1.20 2.99 0.40 5.51 1.63 0.24 0.30 0.08 0.68 0.85 0.22 1.38 1.30 1.32 1.31 1.31 1.47 1.81 1.56 2.01
3.40 3.73 3.06 2.92 0.41 2.26 2.15 0.42 1.79 1.48 1.14 1.67 1.05 0.85 3.20 0.95 3.17 0.76 4.84 1.45 0.75 0.30 0.33 0.69 0.83 0.34 1.24 1.18 1.15 1.17 1.27 1.29 1.58 1.39 1.81
6.51 4.51 3.69 0.61 3.36 3.05 1.04 1.92 2.27 1.64 2.18 1.47 1.20 4.16 1.96 3.22 0.80 6.96 2.21 0.16 0.79 0.49 0.95 1.13 0.60 1.62 1.55 1.55 1.50 1.25 1.58 2.17 1.90 2.43
4.26 3.18 0.44 2.79 2.57 0.71 2.08 1.74 1.19 1.67 1.17 1.08 3.72 1.36 3.22 0.86 4.42 1.81 0.49 0.62 0.45 0.99 0.97 0.56 1.22 1.19 1.22 1.20 1.15 1.30 1.60 1.40 1.82
3.66 0.34 2.45 2.23 0.75 2.06 1.65 1.20 1.59 1.17 1.02 2.78 1.25 2.66 0.54 3.83 1.32 0.65 0.39 0.29 0.76 0.78 0.36 1.17 1.10 1.13 1.11 1.17 1.25 1.53 1.31 1.71
178 Item SEV1 -------1.18 1.09 1.06 -0.06 0.63 0.03 -0.03 0.13 0.09 SEV2 -------1.87 1.91 1.54 -0.02 1.36 0.02 -0.07 0.24 0.25 SEV3 -------1.74 1.72 1.34 -0.02 1.13 0.09 -0.01 0.21 0.19 SEV4 -------2.23 2.31 1.82 0.03 1.47 0.05 -0.13 0.25 0.32 SEV5 -------1.73 1.76 1.40 -0.03 1.22 0.12 -0.02 0.23 0.23 SEV6 -------1.58 1.66 1.33 -0.04 1.18 0.13 -0.10 0.20 0.24
ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
Covariance Matrix Item QREL -------1.42 0.70 0.38 0.14 0.64 0.52 0.00 0.37 0.18 0.04 0.44 -0.26 0.32 -0.52 -0.21 -0.57 -0.88 0.06 -0.29 -0.19 -0.19 -0.06 0.37 0.33 0.30 0.30 IA1 -------IA2 -------IA3 -------IA4 -------CON1 --------
QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4
12.04 6.69 2.77 6.84 3.94 3.19 3.77 3.22 2.70 7.06 0.42 2.88 -0.35 2.31 0.74 0.77 0.06 0.19 0.18 0.21 -0.30 2.11 2.02 1.96 1.99
12.26 4.05 6.25 3.93 2.53 3.96 3.29 2.72 7.87 1.06 6.44 0.94 3.38 0.99 0.81 0.32 0.50 0.21 0.23 -0.09 1.98 2.12 1.97 1.98
11.88 2.11 1.91 1.59 1.15 1.08 1.62 3.59 0.72 1.99 1.08 5.34 1.44 2.03 0.46 0.89 0.95 1.11 0.44 0.07 0.17 0.05 0.12
11.40 3.71 2.75 3.91 3.39 2.85 3.77 -0.41 3.33 -1.08 0.62 -0.59 0.93 -0.06 -0.19 -0.29 -0.07 0.33 1.96 2.05 1.95 2.00
4.41 2.34 2.71 1.90 1.67 2.68 0.59 1.70 -1.17 2.18 0.46 -0.16 -0.32 -0.26 -0.22 -0.24 -0.26 1.21 1.25 1.27 1.20
179 Item QREL -------0.23 0.25 0.28 0.23 0.17 0.20 0.15 0.14 -0.03 0.12 0.61 0.02 0.14 0.03 IA1 -------1.46 1.85 2.08 1.98 1.84 2.01 1.76 2.11 -0.45 0.09 0.59 -0.01 0.16 0.03 IA2 -------2.12 2.11 2.12 2.15 2.26 2.03 2.09 2.45 -0.42 0.26 0.38 0.02 0.12 0.04 IA3 -------0.09 0.15 0.33 0.60 0.29 0.25 0.29 0.76 -0.11 -0.06 0.40 0.29 0.09 -0.06 IA4 -------2.03 2.15 1.82 1.81 1.67 1.78 1.66 1.68 -0.66 -0.05 0.73 0.11 0.16 -0.01 CON1 -------0.89 1.20 1.33 1.25 1.12 1.28 0.97 1.17 -0.26 0.12 0.18 -0.10 0.07 0.08
RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
COVARIANCE MATRIX Item CON2 -------5.46 2.21 0.85 1.17 4.31 2.11 1.64 0.18 2.60 0.39 0.03 0.15 -0.36 -0.84 -0.70 -0.73 1.07 1.07 1.02 1.04 0.73 1.10 CON3 -------CON4 -------CON5 -------EA1 -------EA2 --------
CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6
3.51 2.34 1.72 2.19 0.20 2.47 -0.90 3.59 0.17 -0.01 -0.45 -0.61 -0.52 -0.34 -0.39 1.49 1.56 1.46 1.47 1.18 1.51
2.88 1.67 2.12 -0.31 2.46 -0.85 2.98 0.11 -0.39 -0.27 -0.68 -0.60 -0.54 -0.39 1.04 1.10 0.98 0.99 0.75 1.03
2.13 1.26 0.20 1.19 -0.78 2.53 0.34 0.17 -0.16 -0.34 -0.11 -0.20 -0.05 0.71 0.76 0.75 0.70 0.61 0.83
36.06 13.83 17.51 8.89 13.84 8.25 6.66 5.75 4.78 4.11 4.45 4.17 1.82 2.11 1.97 1.97 1.56 1.76
24.04 6.77 6.27 9.71 12.28 7.94 6.81 4.61 3.85 3.95 3.82 0.73 0.47 0.68 0.43 0.92 0.71
180 Item CON2 -------1.02 0.96 1.07 1.06 1.15 0.98 -0.20 0.02 0.08 -0.02 0.04 0.03 CON3 -------1.50 1.42 1.47 1.50 1.36 1.37 -0.17 0.12 -0.08 -0.06 0.04 0.03 CON4 -------1.12 1.07 1.12 1.12 1.02 1.20 -0.16 0.12 -0.09 0.01 0.06 0.01 CON5 -------0.86 0.85 0.80 0.81 0.80 0.87 -0.16 0.06 -0.03 -0.03 0.01 0.04 EA1 -------1.71 1.95 2.23 1.95 2.04 2.59 0.07 0.55 0.57 0.74 0.50 -0.25 EA2 -------0.56 0.74 0.67 0.52 0.81 1.03 0.28 0.24 -0.28 0.53 0.12 -0.15
ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
Covariance Matrix Item EA3 -------31.79 6.94 9.49 8.77 3.69 4.59 3.39 3.66 4.01 2.99 1.72 1.90 1.70 1.66 1.88 1.56 1.76 1.97 1.94 1.84 1.64 2.38 0.19 0.85 0.23 EA4 -------EA5 -------EA6 -------EA7 -------UNCON1 --------
EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC
22.90 6.28 7.47 5.58 2.65 4.36 3.57 3.74 3.16 -0.35 -0.06 -0.03 -0.05 0.11 -0.66 -0.45 -0.16 -0.23 -0.37 -0.15 -0.01 0.14 0.42 -0.16
53.65 11.51 5.61 1.81 2.32 5.47 5.16 3.75 1.71 1.85 1.66 1.56 1.09 1.28 2.56 2.72 3.57 3.02 3.32 3.27 1.18 1.64 -0.42
20.49 6.76 3.48 3.33 4.67 5.20 3.50 0.76 0.71 0.83 0.62 0.58 0.54 0.67 0.67 0.76 0.70 0.81 1.17 0.41 0.57 -0.49
22.16 3.20 4.32 3.47 4.64 2.99 0.11 -0.08 -0.05 -0.15 0.08 0.03 0.06 0.43 0.44 0.06 0.32 0.56 0.03 -0.16 -0.08
5.98 2.97 2.92 3.04 2.66 0.03 -0.09 -0.02 -0.07 0.05 0.00 0.15 0.28 0.15 0.14 0.23 0.24 0.07 0.18 -0.02
181 Item EA3 -------0.08 0.36 -0.13 EA4 -------0.28 -0.04 -0.05 EA5 -------0.29 0.24 0.35 EA6 -------0.22 0.15 -0.05 EA7 -------0.42 0.10 -0.17 UNCON1 -------0.22 0.09 -0.03
INTER1 INTER2 INTER3
Covariance Matrix Item UNCON2 UNCON3 UNCON4 UNCON5 RESP1 -----------------------------------5.53 2.99 3.22 2.89 -0.07 -0.17 -0.11 -0.15 0.02 -0.33 -0.04 0.19 0.19 -0.05 0.17 0.09 0.18 0.23 0.20 0.09 -0.03 RESP2 --------
UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER6 LDA OS RC INTER1 INTER2 INTER3
6.69 5.07 3.42 -0.11 -0.11 -0.08 -0.15 0.11 -0.18 0.14 0.20 0.28 0.20 0.23 0.12 0.31 -0.01 0.19 0.10 0.00
6.48 3.67 -0.02 -0.14 -0.05 -0.14 0.22 -0.06 0.20 0.23 0.39 0.26 0.26 0.14 0.40 0.09 0.20 0.13 0.03
6.30 -0.08 -0.18 -0.13 -0.16 0.17 -0.25 0.03 0.24 0.25 0.16 0.20 0.19 0.16 -0.15 0.28 0.10 -0.04
1.61 1.36 1.42 1.39 1.20 1.40 1.16 1.07 1.09 1.01 1.01 -0.08 0.21 0.03 -0.06 0.03 0.06
1.47 1.35 1.34 1.18 1.34 1.14 1.07 1.08 1.03 1.00 -0.09 0.18 -0.02 -0.10 0.01 0.06
Covariance Matrix Item RESP3 -------1.46 1.38 1.21 1.36 1.10 1.02 1.05 RESP4 -------RESP5 -------RESP6 -------ANGER1 ANGER2 ---------------
RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3
1.42 1.17 1.35 1.10 1.01 1.03
1.78 1.35 0.97 0.90 0.98
1.73 1.17 1.10 1.15
1.45 1.26 1.45
1.36 1.33
182 Item RESP3 -------1.07 1.00 0.95 -0.08 0.21 -0.04 -0.11 0.02 0.06 RESP4 -------1.07 0.99 0.95 -0.09 0.19 0.01 -0.09 0.02 0.06 RESP5 -------0.98 0.91 0.83 -0.12 0.24 0.04 -0.07 0.03 0.05 RESP6 -------1.13 1.09 1.01 -0.11 0.27 -0.05 -0.01 0.04 0.06 ANGER1 ANGER2 --------------1.41 1.41 1.16 -0.09 0.47 -0.07 -0.02 0.08 0.08 1.23 1.26 1.12 -0.08 0.37 -0.06 0.01 0.06 0.06
ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
Covariance Matrix Item ANGER3 ANGER4 ANGER5 ANGER6 LDA -----------------------------------1.97 1.56 1.89 1.26 -0.04 0.67 -0.09 0.02 0.12 0.11 OS --------
ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1 INTER2 INTER3
1.61 1.49 1.20 -0.08 0.52 -0.07 0.00 0.09 0.07
2.05 1.22 -0.04 0.64 -0.09 -0.01 0.11 0.11
1.34 -0.07 0.35 -0.10 0.04 0.08 0.04
0.20 0.05 -0.13 -0.01 -0.01 0.01
1.00 0.00 0.02 0.18 0.16
Covariance Matrix Item RC -------1.98 0.01 0.18 0.16 INTER1 -------INTER2 -------INTER3 --------
RC INTER1 INTER2 INTER3
0.68 0.18 -0.17
0.15 -0.03
0.14
183 [See the table below for item label used in the study and indicator label used in the LISREL program. Note all the indicators except LDA are found in Appendices A to D. LDA is a composite measure of the likelihood of each account and the defensiveness of each account. The detailed discussion regarding the development of LDA can be found on pp. 85-87] Indicator Label used in LISREL SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RES1 RES2 RES3 RES4 RES5 RES6 ANG1 ANG2 ANG3 ANG4 Item Label used in the study Perceived Outcome Severity1 Perceived Outcome Severity2 Perceived Outcome Severity3 Perceived Outcome Severity4 Perceived Outcome Severity5 Perceived Outcome Severity6 Perceived Quadratic Relational Closeness Internal Attribution1 Internal Attribution2 Internal Attribution3 Internal Attribution4 Controllability1 Controllability2 Controllability3 Controllability4 Controllability5 External Attribution1 External Attribution2 External Attribution3 External Attribution4 External Attribution5 External Attribution6 External Attribution7 Uncontrollability1 Uncontrollability2 Uncontrollability3 Uncontrollability4 Uncontrollability5 Expected Responsibility1 Expected Responsibility2 Expected Responsibility3 Expected Responsibility4 Expected Responsibility5 Expected Responsibility6 Expected Anger1 Expected Anger2 Expected Anger3 Expected Anger4
184 Indicator Label used in LISREL ANG5 ANG6 LDA OS RC INTER1 INTER2 INTER3 Item Label used in the study Expected Anger5 Expected Anger6 Composite Measure of Accounts Outcome Severity Manipulation (?1) Quadratic Relational Closeness manipulation (?2) Interaction1 (?3) Interaction2 (?4) Interaction3 (?5)
185 Appendix N LISREL Syntax for Testing the Proposed Model in Figure 2
Testing the proposed model DA NI=46 NO=236 MA=CM CM=C:\origin1116.psf LA sev1 sev2 sev3 sev4 sev5 sev6 quadrel ia1 ia2 ia3 ia4 con1 con2 con3 con4 con5 ea1 ea2 ea3 ea4 ea5 ea6 ea7 uncon1 uncon2 uncon3 uncon4 uncon5 resp1 resp2 resp3 resp4 resp5 resp6 anger1 anger2 anger3 anger4 anger5 anger6 LDA ex1 ex2 ex3 ex4 ex5 MO NX=5 NK=5 NY=41 NE=9 FI LY=FU,FI LX=ID TE=DI,FR TD=ZE TH=ZE,FI BE=FU,FI GA=FU,FI PS=DI,FR TH=ZE,FI FR GA 1 1 GA 2 2 GA 9 3 GA 3 4 GA 4 4 GA 5 5 GA 6 5 FR BE 9 1 BE 7 1 BE 9 2 BE 9 8 BE 8 7 BE 7 3 BE 8 3 BE 7 4 BE 8 4 BE 7 5 BE 8 5 BE 7 6 FR BE 8 6 FR LY 2 1 LY 3 1 LY 4 1 LY 5 1 LY 6 1 LY 9 3 LY 10 3 LY 11 3 LY 13 4 LY 14 4 LY 15 4 FR LY 16 4 LY 18 5 LY 19 5 LY 20 5 LY 21 5 LY 22 5 LY 23 5 LY 25 6 LY 26 6 LY 27 6 FR LY 28 6 FR LY 30 7 LY 31 7 LY 32 7 LY 33 7 LY 34 7 LY 36 8 LY 37 8 LY 38 8 LY 39 8 LY 40 8 FI TE 41 41 TE 7 7 VA 1 LY 1 1 LY 7 2 LY 8 3 LY 12 4 LY 17 5 LY 24 6 LY 29 7 LY 35 8 LY 41 9 PD OU AL MI AD=OFF
186 Appendix O Covariance Matrix in LISREL for the Revised Model in Figure 6 Covariance Matrix Item SEV1 -------2.24 2.09 2.04 2.38 2.04 1.90 0.38 2.12 2.26 0.61 1.89 1.30 1.05 1.46 1.11 0.81 2.49 1.26 2.75 0.45 2.82 1.15 0.17 0.44 0.30 0.64 0.65 0.24 0.95 0.93 0.90 0.89 0.95 1.00 1.09 1.00 1.16 SEV2 -------SEV3 -------SEV4 -------SEV5 -------SEV6 --------
SEV1 SEV2 SEV3 SEV4 SEV5 SEV6 QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3
4.27 3.39 4.30 3.40 3.37 0.49 2.62 2.42 0.36 1.69 1.70 1.27 1.76 1.26 1.01 3.39 1.20 2.99 0.40 5.51 1.63 0.24 0.30 0.08 0.68 0.85 0.22 1.38 1.30 1.32 1.31 1.31 1.47 1.81 1.56 2.01
3.40 3.73 3.06 2.92 0.41 2.26 2.15 0.42 1.79 1.48 1.14 1.67 1.05 0.85 3.20 0.95 3.17 0.76 4.84 1.45 0.75 0.30 0.33 0.69 0.83 0.34 1.24 1.18 1.15 1.17 1.27 1.29 1.58 1.39 1.81
6.51 4.51 3.69 0.61 3.36 3.05 1.04 1.92 2.27 1.64 2.18 1.47 1.20 4.16 1.96 3.22 0.80 6.96 2.21 0.16 0.79 0.49 0.95 1.13 0.60 1.62 1.55 1.55 1.50 1.25 1.58 2.17 1.90 2.43
4.26 3.18 0.44 2.79 2.57 0.71 2.08 1.74 1.19 1.67 1.17 1.08 3.72 1.36 3.22 0.86 4.42 1.81 0.49 0.62 0.45 0.99 0.97 0.56 1.22 1.19 1.22 1.20 1.15 1.30 1.60 1.40 1.82
3.66 0.34 2.45 2.23 0.75 2.06 1.65 1.20 1.59 1.17 1.02 2.78 1.25 2.66 0.54 3.83 1.32 0.65 0.39 0.29 0.76 0.78 0.36 1.17 1.10 1.13 1.11 1.17 1.25 1.53 1.31 1.71
187 Item SEV1 -------1.18 1.09 1.06 -0.06 0.63 0.03 -0.03 SEV2 -------1.87 1.91 1.54 -0.02 1.36 0.02 -0.07 SEV3 -------1.74 1.72 1.34 -0.02 1.13 0.09 -0.01 SEV4 -------2.23 2.31 1.82 0.03 1.47 0.05 -0.13 SEV5 -------1.73 1.76 1.40 -0.03 1.22 0.12 -0.02 SEV6 -------1.58 1.66 1.33 -0.04 1.18 0.13 -0.10
ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
Covariance Matrix Item QREL -------1.42 0.70 0.38 0.14 0.64 0.52 0.00 0.37 0.18 0.04 0.44 -0.26 0.32 -0.52 -0.21 -0.57 -0.88 0.06 -0.29 -0.19 -0.19 -0.06 0.37 0.33 0.30 0.30 0.23 0.25 0.28 0.23 IA1 -------IA2 -------IA3 -------IA4 -------CON1 --------
QREL IA1 IA2 IA3 IA4 CON1 CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2
12.04 6.69 2.77 6.84 3.94 3.19 3.77 3.22 2.70 7.06 0.42 2.88 -0.35 2.31 0.74 0.77 0.06 0.19 0.18 0.21 -0.30 2.11 2.02 1.96 1.99 1.46 1.85 2.08 1.98
12.26 4.05 6.25 3.93 2.53 3.96 3.29 2.72 7.87 1.06 6.44 0.94 3.38 0.99 0.81 0.32 0.50 0.21 0.23 -0.09 1.98 2.12 1.97 1.98 2.12 2.11 2.12 2.15
11.88 2.11 1.91 1.59 1.15 1.08 1.62 3.59 0.72 1.99 1.08 5.34 1.44 2.03 0.46 0.89 0.95 1.11 0.44 0.07 0.17 0.05 0.12 0.09 0.15 0.33 0.60
11.40 3.71 2.75 3.91 3.39 2.85 3.77 -0.41 3.33 -1.08 0.62 -0.59 0.93 -0.06 -0.19 -0.29 -0.07 0.33 1.96 2.05 1.95 2.00 2.03 2.15 1.82 1.81
4.41 2.34 2.71 1.90 1.67 2.68 0.59 1.70 -1.17 2.18 0.46 -0.16 -0.32 -0.26 -0.22 -0.24 -0.26 1.21 1.25 1.27 1.20 0.89 1.20 1.33 1.25
188 Item QREL -------0.17 0.20 0.15 0.14 -0.03 0.12 0.61 0.02 IA1 -------1.84 2.01 1.76 2.11 -0.45 0.09 0.59 -0.01 IA2 -------2.26 2.03 2.09 2.45 -0.42 0.26 0.38 0.02 IA3 -------0.29 0.25 0.29 0.76 -0.11 -0.06 0.40 0.29 IA4 -------1.67 1.78 1.66 1.68 -0.66 -0.05 0.73 0.11 CON1 -------1.12 1.28 0.97 1.17 -0.26 0.12 0.18 -0.10
ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
COVARIANCE MATRIX Item CON2 -------5.46 2.21 .85 1.17 4.31 2.11 1.64 0.18 2.60 0.39 0.03 0.15 -0.36 -0.84 -0.70 -0.73 1.07 1.07 1.02 1.04 0.73 1.10 1.02 0.96 1.07 1.06 1.15 0.98 -0.20 CON3 -------CON4 -------CON5 -------EA1 -------EA2 --------
CON2 CON3 CON4 CON5 EA1 EA2 EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA
3.51 2.34 1.72 2.19 0.20 2.47 -0.90 3.59 0.17 -0.01 -0.45 -0.61 -0.52 -0.34 -0.39 1.49 1.56 1.46 1.47 1.18 1.51 1.50 1.42 1.47 1.50 1.36 1.37 -0.17
2.88 1.67 2.12 -0.31 2.46 -0.85 2.98 0.11 -0.39 -0.27 -0.68 -0.60 -0.54 -0.39 1.04 1.10 0.98 0.99 0.75 1.03 1.12 1.07 1.12 1.12 1.02 1.20 -0.16
2.13 1.26 0.20 1.19 -0.78 2.53 0.34 0.17 -0.16 -0.34 -0.11 -0.20 -0.05 0.71 0.76 0.75 0.70 0.61 0.83 0.86 0.85 0.80 0.81 0.80 0.87 -0.16
36.06 13.83 17.51 8.89 13.84 8.25 6.66 5.75 4.78 4.11 4.45 4.17 1.82 2.11 1.97 1.97 1.56 1.76 1.71 1.95 2.23 1.95 2.04 2.59 0.07
24.04 6.77 6.27 9.71 12.28 7.94 6.81 4.61 3.85 3.95 3.82 0.73 0.47 0.68 0.43 0.92 0.71 0.56 0.74 0.67 0.52 0.81 1.03 0.28
189 Item CON2 -------0.02 0.08 -0.02 CON3 -------0.12 -0.08 -0.06 CON4 -------0.12 -0.09 0.01 CON5 -------0.06 -0.03 -0.03 EA1 -------0.55 0.57 0.74 EA2 -------0.24 -0.28 0.53
OS RC INTER1
Covariance Matrix Item EA3 -------31.79 6.94 9.49 8.77 3.69 4.59 3.39 3.66 4.01 2.99 1.72 1.90 1.70 1.66 1.88 1.56 1.76 1.97 1.94 1.84 1.64 2.38 0.19 0.85 0.23 0.08 EA4 -------EA5 -------EA6 -------UNCON1 UNCON2 ---------------
EA3 EA4 EA5 EA6 EA7 UNCON1 UNCON2 UNCON3 UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
22.90 6.28 7.47 5.58 2.65 4.36 3.57 3.74 3.16 -0.35 -0.06 -0.03 -0.05 0.11 -0.66 -0.45 -0.16 -0.23 -0.37 -0.15 -0.01 0.14 0.42 -0.16 0.28
53.65 11.51 5.61 1.81 2.32 5.47 5.16 3.75 1.71 1.85 1.66 1.56 1.09 1.28 2.56 2.72 3.57 3.02 3.32 3.27 1.18 1.64 -0.42 0.29
20.49 6.76 3.48 3.33 4.67 5.20 3.50 0.76 0.71 0.83 0.62 0.58 0.54 0.67 0.67 0.76 0.70 0.81 1.17 0.41 0.57 -0.49 0.22
22.16 3.20 4.32 3.47 4.64 2.99 0.11 -0.08 -0.05 -0.15 0.08 0.03 0.06 0.43 0.44 0.06 0.32 0.56 0.03 -0.16 -0.08 0.42
5.98 2.97 2.92 3.04 2.66 0.03 -0.09 -0.02 -0.07 0.05 0.00 0.15 0.28 0.15 0.14 0.23 0.24 0.07 0.18 -0.02 0.22
Covariance Matrix Item UNCON2 UNCON3 UNCON4 UNCON5 ----------------------------5.53 2.99 RESP1 -------RESP2 --------
UNCON2 UNCON3
6.69
190 Item UNCON2 UNCON3 UNCON4 UNCON5 ----------------------------3.22 2.89 -0.07 -0.17 -0.11 -0.15 0.02 -0.33 -0.04 0.19 0.19 -0.05 0.17 0.09 0.18 0.23 0.20 5.07 3.42 -0.11 -0.11 -0.08 -0.15 0.11 -0.18 0.14 0.20 0.28 0.20 0.23 0.12 0.31 -0.01 0.19 6.48 3.67 -0.02 -0.14 -0.05 -0.14 0.22 -0.06 0.20 0.23 0.39 0.26 0.26 0.14 0.40 0.09 0.20 RESP1 -------RESP2 --------
UNCON4 UNCON5 RESP1 RESP2 RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER6 LDA OS RC INTER1
6.30 -0.08 -0.18 -0.13 -0.16 0.17 -0.25 0.03 0.24 0.25 0.16 0.20 0.19 0.16 -0.15 0.28
1.61 1.36 1.42 1.39 1.20 1.40 1.16 1.07 1.09 1.01 1.01 -0.08 0.21 0.03 -0.06
1.47 1.35 1.34 1.18 1.34 1.14 1.07 1.08 1.03 1.00 -0.09 0.18 -0.02 -0.10
Covariance Matrix Item RESP3 -------1.46 1.38 1.21 1.36 1.10 1.02 1.05 1.07 1.00 0.95 -0.08 0.21 -0.04 -0.11 RESP4 -------RESP5 -------RESP6 -------ANGER1 ANGER2 ---------------
RESP3 RESP4 RESP5 RESP6 ANGER1 ANGER2 ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
1.42 1.17 1.35 1.10 1.01 1.03 1.07 0.99 0.95 -0.09 0.19 0.01 -0.09
1.78 1.35 0.97 0.90 0.98 0.98 0.91 0.83 -0.12 0.24 0.04 -0.07
1.73 1.17 1.10 1.15 1.13 1.09 1.01 -0.11 0.27 -0.05 -0.01
1.45 1.26 1.45 1.41 1.41 1.16 -0.09 0.47 -0.07 -0.02
1.36 1.33 1.23 1.26 1.12 -0.08 0.37 -0.06 0.01
191 Covariance Matrix Item ANGER3 ANGER4 ANGER5 ANGER6 ----------------------------1.97 1.56 1.89 1.26 -0.04 0.67 -0.09 0.02 LDA -------OS --------
ANGER3 ANGER4 ANGER5 ANGER6 LDA OS RC INTER1
1.61 1.49 1.20 -0.08 0.52 -0.07 0.00
2.05 1.22 -0.04 0.64 -0.09 -0.01
1.34 -0.07 0.35 -0.10 0.04
0.20 0.05 -0.13 -0.01
1.00 0.00 0.02
Covariance Matrix Item RC -------1.98 0.01 INTER1 --------
RC INTER1
0.68
Note. See the labels of variables in Appendix M.
192 Appendix P LISREL Syntax for Testing the Revised Model in Figure 6 Testing the revised model DA NI=44 NO=236 MA=CM CM=C:\revise1117.psf LA sev1 sev2 sev3 sev4 sev5 sev6 quadrel ia1 ia2 ia3 ia4 con1 con2 con3 con4 con5 ea1 ea2 ea3 ea4 ea5 ea6 ea7 uncon1 uncon2 uncon3 uncon4 uncon5 resp1 resp2 resp3 resp4 resp5 resp6 anger1 anger2 anger3 anger4 anger5 anger6 LDA ex1 ex2 ex3 MO NX=3 NK=3 NY=41 NE=9 FI LY=FU,FI LX=ID TE=DI,FR TD=ZE TH=ZE,FI BE=FU,FI GA=FU,FI PS=DI,FR TH=ZE,FI FR GA 1 1 GA 2 2 GA 9 3 FR BE 9 1 BE 7 1 BE 9 2 BE 9 8 BE 8 7 BE 7 3 BE 8 3 BE 7 4 BE 8 4 BE 7 5 BE 8 5 BE 7 6 FR BE 8 6 BE 3 1 BE 4 1 BE 5 1 BE 6 1 BE 8 1 FR LY 2 1 LY 3 1 LY 4 1 LY 5 1 LY 6 1 LY 9 3 LY 10 3 LY 11 3 LY 13 4 LY 14 4 LY 15 4 FR LY 16 4 LY 18 5 LY 19 5 LY 20 5 LY 21 5 LY 22 5 LY 23 5 LY 25 6 LY 26 6 LY 27 6 FR LY 28 6 FR LY 30 7 LY 31 7 LY 32 7 LY 33 7 LY 34 7 LY 36 8 LY 37 8 LY 38 8 LY 39 8 LY 40 8 FI TE 41 41 TE 7 7 VA 1 LY 1 1 LY 7 2 LY 8 3 LY 12 4 LY 17 5 LY 24 6 LY 29 7 LY 35 8 LY 41 9 PD OU AL MI AD=OFF
193 References Allred, K. G., Mallozzi, J. S., Matsui, F., & Raia, C. P. (1997). The influence of anger and compassion on negotiation performance. Organizational Behavior and Human Decision Processes, 70, 175-187. Averill, J. R. (1983). Studies on anger and aggression. American Psychologist, 38, 11451160. Bauer, C. L., & Fink, E. L. (1983). Fitting equations with power transformations: Examining variables with error. In R. N. Bostrom (Ed.), Communication yearbook 7 (pp. 146-199). Beverly Hills, CA: Sage. Bentler, P. M., & Chou, C. P. (1987). Practical issues in structural modeling. Sociological Methods & Research, 16, 78-117. Bertrand, M., & Mullainathan, S. (2003). Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. National Bureau of Economic Research. Retrieved November 26, 2007 fromhttp://papers.nber.org/papers/w9873 Bies, R. J. (1987). The predicament of injustice: The management of moral outrage. In L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior (Vol. 9, pp. 289-319). Greenwich, CT: JAI Press. Blake, R. R., & Mouton, J. S. (1964). The managerial grid. Houston, TX: Gulf. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Brew, F. P., & Cairns, D. R. (2004). Styles of managing interpersonal workplace conflict in relation to status and face concern: A study with Anglos and Chinese. International Journal of Conflict Management, 15, 27-56.
194 Cai, D. A., & Donohue, W. A. (1997). Determinants of facework in intercultural negotiation. Asian Journal of Communication, 7, 85-110. Cai, D. A., & Fink, E. L. (2002). Conflict style differences between individualists and collectivists. Communication Monographs, 69, 67-87. Carnevale, P. J., & Pruitt, D. G. (1992). Negotiation and mediation. Annual Review of Psychology, 43, 531-582. Choi, I., Nisbett, R. E., & Norenzayan, A. (1999). Causal attribution across cultures: Variation and universality. Psychological Bulletin, 125, 47-63. Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37-46. Cohen, J. (1988). Statistical power analysis for the behavior science (2nd ed.). New York: Academic Press. Converse, J. M., & Presser, S. (1986). Survey questions: Handcrafting the standardized questionnaire. Newbury Park, CA: Sage. El-Sheikh, M., Buckhalt, J. A., & Reiter, S. L. (2000). Gender-related effects in emotional responding to resolved and unresolved interpersonal conflict. Sex Roles: A Journal of Research. Retrieved November 26, 2007 fromhttp://www.findarticles.com/p/articles/mi_m2294/is_2000_Nov/ai_75959823 Fincham, F. D. (1992). The account episodes in close relationship. In M. L. McLaughlin, M. J. Cody, & S. J. Read (Eds.), Explaining one’s self to others: Reason-giving in a social context (pp. 167-182). Hillsdale, NJ: Lawrence Erlbaum Associates. Fink, C. F. (1968). Some conceptual difficulties in the theory of social conflict. Journal of Conflict Resolution, 12, 412-460.
195 Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York: McGraw-Hill. Fitzpatrick, M. A., & Winke, J. (1979). You always hurt the one you love: Strategies and tactics in interpersonal conflict. Communication Quarterly, 27, 3-11. Frey, L. R., Botan, C. H., & Kreps, G. L. (2000). Investigating communication: An introduction to research methods. Boston: Allyn & Bacon. Fritsche, I. (2002). Accounts for the violation of social norms: Integration and extension of sociological and social psychological typologies. Journal for the Theory of Social Behaviour, 32, 371-394. Gagné, P., & Hancock, G. R. (2006). Measurement model quality, sample size, and solution propriety in confirmatory factor models. Multivariate Behavioral Research, 41, 65-83. Gerber, E. (1994). The language of residence: Respondent understandings and census rules. Washington, DC: Center for Survey Methods Research, U. S. Bureau of the Census. Gleason, J. M., & Harris, V. A. (1976). Perceived freedom, accident severity and empathetic value as determinants of the attribution of responsibility. Social Behavior and Personality, 4, 171-176. Goffman, E. (1971). Relations in public. New York: Harper & Row. Gonzales, M. H., Manning, D. J., & Haugen, J. A. (1992). Explaining our sins: Factors influencing offenders accounts and anticipated victim responses. Journal of Personality and Social Psychology, 62, 958-971. Graham, S., Weiner, B., & Benesh-Weiner, M. (1995). An attributional analysis of the development of excuse giving in aggressive and nonaggressive African American
196 boys. Developmental Psychology, 31, 274-284. Hamilton, V. L., & Hagiwara, S. (1992). Roles, responsibility, and accounts across cultures. International Journal of Psychology, 27, 157-179. Hancock, G. R. (2006). Power analysis in covariance structure modeling. In R. C. Serlin (Series Ed.) & G. R. Hancock & R. O. Mueller (Vol. Eds.), Structural equation modeling: A second course (pp. 69-115). Greenwich, CO: Information Age. Harvey, J. H., & Weary, G. (1984). Current issues in attribution theory and research. Annual Review of Psychology, 35, 427-459. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Henderson, M., & Hewstone, M. (1984). Prison inmate’s explanations for interpersonal violence: Accounts and attributions. Journal of Consulting and Clinical Psychology, 52, 789-794. Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods, 3, 424– 453. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. Jackson, D. L. (2003). Revisiting sample size and number of parameter estimates: Some support for the N:q hypothesis. Structural Equation Modeling: A Multidisciplinary Journal, 10, 128-141. Jellison, J. M. (1990). Accounting: Societal implications. In M. J. Cody & M. L. McLaughlin (Eds.), The psychology of tactical communication (pp. 283-298).
197 Philadelphia: Multilingual Matters. Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in person perception. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 220-266). New York: Academic Press. Jöreskog, K., & Sörbom, D (2004). LISREL for Windows (Version 8.70) [Computer software]. Chicago: Scientific Software International. Kelley, H. H. (1971). Attribution in social interaction. Morristown, NJ: General Learning Press. Kelley, H. H. (1986). Attribution theory in social psychology. Nebraska Symposium on Motivation, 15, 192-238. Kelley, H. H. (1986). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska Symposium on Motivation, 15, 192-238. Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). New York: Guilford Press. Konovsky, M. A., & Jaster, F. (1989). “Blaming the victim” and other ways business men and women account for questionable behavior. Journal of Business Ethics, 8, 391-398. Liu, M., & Yao, S. (2007, November). Why do Chinese and Americans differ in attributions of responsibility? Investigating the universality vs. culture-specificity of cognitive appraisals. Paper presented at the annual conference of National Communication Association, Chicago, IL. Mack, R. W., & Snyder, R. C. (1957). The analysis of social conflict—Toward an overview and synthesis. Journal of Conflict Resolution, 1, 212-248. Martin, E. (2006). Vignettes and respondent debriefing for questionnaire
198 design and evaluation. U. S. Bureau of Census Research Report Series. Retrieved December 1, 2007 fromhttp://www.census.gov/srd/papers/pdf/rsm2006-08.pdf McLaughlin, M. L., Cody, M. J., & French, K. (1990). Account-giving and the attribution of responsibility: Impressions of traffic offenders. In M. J. Cody & M. L. McLaughlin (Eds.), The psychology of tactical communication (pp. 244-267). Philadelphia: Multilingual Matters. McLaughlin, M. L., Cody, M. J., & Read, S. J. (1992). Explaining one’s self to others: Reason-giving in a social context. Hillsdale, NJ: Lawrence Erlbaum Associates. McLaughlin, M. L., Cody, M. J., & Rosenstein, N. E. (1983). Account sequences in conversations between strangers. Communication Monographs, 50, 102-125. McLaughlin, M. L., O’Hair, H. D., & Cody, M. J. (1983). The management of failure events: Some contextual determinants of accounting behavior. Human Communication Research, 9, 208-224. Mills, C. W. (1940). Situated actions and vocabularies of motive. American Sociological Review, 5, 904-913. Mitchell, T. R., & Kalb, L. S. (1981). Effects of outcome knowledge and outcome valence on supervisors’ evaluations. Journal of Applied Psychology, 66, 604-612. Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231-259. Ohbuchi, K., Fukushima, O., & Tedeschi, J. T. (1999). Cultural values in conflict management: Goal orientation, goal attainment, and tactical decision. Journal of Cross-Cultural Psychology, 30, 51-71. Orbuch, T. L. (1997). People’s accounts count: The sociology of accounts. Annual
199 Review of Sociology, 23, 455-478. Pruitt, D. G., & Rubin, J. Z. (1986). Social conflict: Escalation, stalemate, and settlement. New York: Random House. Putnam, L. L., & Poole, M. S. (1987). Conflict and negotiation. In F. M. Jablin, L. L. Putnam, K. H. Roberts, & L. W. Porter (Eds.), Handbook of organizational communication (pp. 549-599). Beverly Hills, CA: Sage. Rahim, M. A. (1983). A measure of styles of handling interpersonal conflict. Academy of Management Journal, 26, 368-376. Rahim, M. A. (1992). Managing conflict in organizations. New York: Praeger. Rahim, M. A., & Bonoma, T. V. (1979). Managing organizational conflict: A model for diagnosis and intervention. Psychological Reports, 44, 1323-1344. Read, S. J. (1992). Constructing accounts: The role of explanatory coherence. In M. L. McLaughlin, M. J. Cody, & S. J. Read (Eds.), Explaining one’s self to others: Reason-giving in a social context (pp. 3-20). Hillsdale, NJ: Lawrence Erlbaum. Rest, J. R. (1984). The major components of morality. In W. M. Kurtines & J. L. Gewirtz (Eds.), Morality, moral behavior and moral development (pp. 24-38). New York: Wiley. Richardson, D. S., & Green, L. R. (2006). Direct and indirect aggression: Relationships as social context. Journal of Applied Social Psychology, 36, 2492-2508. Robbennolt, J. (2000). Outcome severity and judgments of “responsibility”: A metaanalytic review. Journal of Applied Social Psychology, 30, 2575-2609. Roloff, M. E., & Soule, K. P. (2002). Interpersonal conflict: A review. In M. L. Knapp & J. A. Daly (Eds.), Handbook of interpersonal communication (3rd ed., pp. 475-
200 528). Thousand Oaks, CA: Sage. Ross, D. (1982). The causal dimension scale: A measure of how individuals perceive causes. Journal of Personality and Social Psychology, 42, 1137-1145. Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 10, pp. 174-214). New York: Academic Press. Salili, F., Maehr, M. L., & Gillmore, G. (1976). Achievement and morality: A crosscultural analysis of causal attribution and evaluation. Journal of Personality and Social Psychology, 33, 327-337. Schlenker, B. R. (1980). Impression management: The self-concept, social identity, and interpersonal relations. Monterey, CA: Brooks/Cole. Schönbach, P. (1980). A category system of account phases. European Journal of Social Psychology, 10, 195-200. Schönbach, P. (1990). Account episodes: The management or escalation of conflict. Cambridge, UK: Cambridge University Press. Scott, M. B., & Lyman, S. M. (1968). Accounts. American Sociological Review, 3, 46-62. Shaver, K. G. (1970). Defensive attribution: Effects of severity and relevance on the responsibility assigned for an accident. Journal of Personality and Social Psychology, 14, 101-113. Sillar, A. L. (1980). The sequential and distributional structure of conflict interactions as a function of attributions concerning the locus of responsibility and stability of conflicts. In D. Nimmo (Ed.), Communication yearbook 4 (pp. 217-235). New Brunswick, NJ: Transaction Books.
201 Sitkin, S. B., & Bies, R. J. (1993). Accounts in conflict situations: Using explanation to manage conflict. Human Relations, 46, 349-370. Sykes, G. M., & Matza, D. (1957). Techniques of neutralization: A theory of delinquency. American Sociological Review, 22, 664-670. Takaku, S. (2000). Culture and status as influences on account giving: A comparison between the United States and Japan. Journal of Applied Social Psychology, 30, 371-388. Tedeschi, J. T., & Reiss, M. (1981). Verbal strategies in impression management. In C. Antaki (Ed.), The psychology of ordinary explanations of social behaviour (pp. 271-309). London: Academic Press. Tennen, H., & Affleck, G. (1990). Blaming others for threatening events. Psychological Bulletin, 108, 209-232. Thomas, E. A. C., & Parpal, M. (1987). Liability as a function of plaintiff and defendant fault. Journal of Personality and Social Psychology, 53, 843-857. Thomas, K. W. (1976). Conflict and conflict management. In M. Dunnette (Ed.), Handbook of industrial and organizational psychology (pp. 889-935). Chicago: Rand McNally. Thomas, K. W. (1992). Conflict and conflict management: Reflections and update. Journal of Organizational Behavior, 13, 265-274. Ting-Toomey, S., & Oetzel, J. G. (2002). Cross-cultural face concerns and conflict styles. In W. B. Gudykunst & B. Mody (Eds.), Handbook of international and intercultural communication (2nd ed., pp. 143-164). Thousand Oaks, CA: Sage. Walster, E. (1960). Assignment of responsibility for an accident. Journal of Personality
202 and Social Psychology, 3, 73-79. Wang, Q. (2006). Linking goals to avoidance in interpersonal conflict situations: A cognitive approach. Unpublished dissertation, University of Maryland, College Park. Weiner, B. (1986). An attribution theory of motivation and emotion. New York: Springer Verlag. Weiner, B. (1992). Excuses in everyday interaction. In M. L. McLaughlin, M. J. Cody, & S. J. Read (Eds.), Explaining one’s self to others: Reason-giving in a social context (pp. 131-146). Hillsdale, NJ: Lawrence Erlbaum Associates. Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New York: Springer-Verlag. Weiner, B. (2000). Intrapersonal and interpersonal theories of motivation from an attributional perspective. Educational Psychology Review, 12, 1-14. Weiner, B. (2006). Social motivation, justice, and the moral emotions: An attributional approach. Mahwah, NJ: Lawrence Erlbaum. Weiner, B., Amirkhan, J., Folkes, V. S., & Verette, J. A. (1987). An attributional analysis of excuse giving: Studies of a naïve theory of emotion. Journal of Personality and Social Psychology, 52, 316-324. Weiner, B., Figueroa-Muñoz, A., & Kakihara, C. (1991). The goals of excuses and communication strategies related to causal perceptions. Personality and Social Psychology Bulletin, 17, 4-13. Weiner, B., & Handel, S. (1985). Anticipated emotional consequences of causal communications and reported communication strategy. Developmental Psychology, 21, 102-107.
203 Wilmot, W. W., & Hocker, J. L. (2005). Interpersonal conflict (7th ed.). New York: McGraw Hill. Wilson, R. J., & Jonah, B. A. (1988). Assignment of responsibility and penalties for an impaired driving incident. Journal of Applied Social Psychology, 18, 564-583. Wilson, S. R. (2005). Seeking and resisting compliance: Why people say what they do when trying to influence others. Thousand Oaks, CA: Sage. Wrench, J. S., Thomas-Maddox, C., Richmond, V. P., & McCroskey, J. C. (2008). Quantitative research methods for communication: A hands-on approach. New York: Oxford University Press. Yirmiya, N., & Weiner, B. (1986). Perceptions of controllability and anticipated anger. Cognitive Development, 1, 273-280.
doc_270369076.pdf