THE ENCODING AND RETRIEVAL OF NUMERICAL DATA FOR DECISION MAKING IN ACCOUNTING CONTEXTS

Description
This paper presents a model that concerns the encoding and retrieval of numerical data in accounting
decision-making contexts. The model identibes the forms of encoded representations used to capture
numerical data in memory structures, indicates the relative retrievability of those encoded representations,
and discusses factors that may affect encoding and retrieval processes.

Pergamon Accounttng, Organtzatfons and Society, Vol. 20, No. 7/B. pp. 585-610, 1995
Copyri& 0 1995 Elsevier Science Ltd
Printed in Great Britain. Au tights reserved.
0361-36B2/95 $9.50+0.00
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA FOR DECISION
MAKING IN ACCOUNTING CONTEXTS: MODEL DEVELOPMENT*
THOMAS KIDA
University of Massachusetts at Amherst
and
JAMES F. SMITH
University of Massachusetts at Amherst
Abstract
This paper presents a model that concerns the encoding and retrieval of numerical data in accounting
decision-making contexts. The model identibes the forms of encoded representations used to capture
numerical data in memory structures, indicates the relative retrievability of those encoded representa-
tions, and discusses factors that may affect encoding and retrieval processes. We argue that affective or
evaluative reactions to numerical data are central to encoding and retrieval processes in accounting
decision contexts. The basic tenets of the model are presented in 12 propositions that are supported by
theoretical and empirical research in areas such as cognitive development, human memory, social
cognition and human information processing. The propositions identity issues that are empirically
testable, and as such, can provide a framework for research into the encoding and retrieval operations
utilized when accounting decision makers consider numerical data.
Much of the information produced by the
accounting function and subsequently used by
decision makers is in the form of numerical
data. However, very little is known about fun-
damental issues concerning decision making
with numerical information, such as how the
data are encoded into decision-makers’ knowl-
edge structures. The purpose of this paper is to
present a model of encoding and retrieving
numerical data that can aid our understanding
of decision making with accounting numbers
and provide a framework for future research.
Specifically, the model concerns the different
forms of encoded representations used to
capture numerical data in memory structures,
the relative retrievability of those encoded
representations, and the factors that affect
retrievability.
Accounting researchers are increasingly
aware of the importance of encoding and mem-
ory issues for both the preparers and users of
accounting information. Libby (1989) notes
that a complete understanding of accounting
decision making must place substantial weight
on memory issues. This is especially true since
working memory has a limited storage capa-
city. For example, Libby and Trotman (1993)
indicate that the volume of accounting evi-
dence recorded, and the time frame over
which data must be considered, results in
expert auditors having to rely on information
retrieved from long-term memory. Also, con-
* We would like to thank Dennis HaMO, Cindy Moeckel, Jeffrey Cohen, Mario Maletta, Brenda Anderson, Chris Agogha,
Sudip Bhattacharjee, Sue Machuga, Pam Trafford, Kim Moreno and Mary Curammeng for their helpful comments.
585
586 T. KIDA and I. F. SMITH
stant reference to external storage can be ments) are viewed as the most retrievable
costly, both financially and in terms of deci- form of encoded representation. In addition,
sion-maker time and effort. Therefore, the the model discusses issues of information
encoded representations stored in memory load, decision-makers’ objectives, confidence
structures, and the decision-maker’s ability to in recall, and the reconstruction of memory
retrieve that data, have important effects on traces as they relate to encoding and retrieving
accounting decision making. numerical data for decision making.
We argue that a central component of
numerical encoding and retrieval invohes
affective responses. Affect is specifically
defined here as evaluative reactions such that
the data are represented as a positive or nega-
tive valence in memory structures. Memory
models typically do not emphasize the under-
lying importance of affect and, therefore, the
implications of such affective responses for
memory and other related cognitive opera-
tions are not elucidated. Recent work sug-
gests, however, that affect plays an integral
role in memory and cognitive processes, neces-
sitating its explicit consideration in such mod-
els (Beach & Mitchell, 1987; Mitchell & Beach,
1990; Fiske & Taylor, 1991).
The basic tenets of the model are presented
in 12 propositions. The propositions are based
upon a review and synthesis of theoretical and
empirical research in a number of related areas,
including cognitive development, human mem-
ory, social cognition and human information
processing.’ As with any model of this nature,
several empirical investigations will be needed
to test its propositions. Consequently, the
model is meant to stimulate empirical research
into the encoding and retrieval of numerical
data in accounting decision contexts.
Consistent with fuzzy trace theory (Brainerd
St Reyna, 1993, 1990a), the model specifies that
decision makers derive gist from numerical
data. Specifically, we argue that, when indivi-
duals observe numerical data in accounting
judgment and decision-making contexts, they
draw comparisons among data items and deter-
mine affective reactions to the data. The model
posits that while numerical data may be initially
encoded as numerical values, comparisons, and
affective reactions, the affective reactions pro-
vide the most retrievable memory trace.
Further, affective reactions to broad underly-
ing constructs (e.g. variables that represent a
number of observed numerical measure-
The following section discusses the basic
characteristics of the model. A series of propo-
sitions are described, which are summarized in
Table 1, relevant research examined, and impli-
cations for decision making with numerical
data indicated. The major components of the
model, addressing the different forms of
encoded representations used to capture
numerical data in memory structures, the rela-
tive retrievability of those representations, and
the factors that inlhtence retrievability, are
schematically presented in Fig. 1.
MODEL CHARACTERISTICS
I ni ti al encoded representati ons
Proposition 1: Numerical data may be initially encoded
as numerical values, as comparisons between numerical
values, and as affective (evaluative) reactions to those
comparisons.
’ The concern of the model presented here is the type of encoded representations that result when numerical data are
observed, and not with the associative links between the numerical data. See Wyer and Sruh (1986), Collins and Loftus
(1975) Anderson (1976) and Raaijmakers and Shiffrin (1981) for examples and discussions of various associative memory
models. Considerable work has also been performed on the cognitive processes involved in numerical computations (e.g.
Frensch & Geary, 1993; Sokol et al., 1991; Widaman et al., 1989; McCloskey et al., 1991; Logic & Baddeley, 1987) That
research provides important evidence on arithmetic processes conducted on numbers (e.g. additive, multiplicative, etc.)
and arithmetic fact retrieval. However, the present study is primarily concerned with how numbers that represent
different constructs are encoded and retrieved in decision-making tasks, and thus, that research is outside the immediate
scope of interest.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 587
TABLE 1, Summary of the model’s propositions
Proposition 1:
Proposition 2a:
Proposition 2b:
Proposition 3:
Proposition 4a:
Proposition 4b:
Proposition 4c:
Proposition 5a:
Proposition 5b:
Proposition 6:
Proposition 7:
Proposition 8:
Numerical data may be initially encoded as numerical values, as comparisons between numercial
values, and as affective (evaluative) reactions to those comparisons.
The retrievability of encoded representations increases from numercial values, to comparisons
between numerical values, to affective reactions to those comparisons.
Affective reactions to underlying constrncts or broad categories of data will have greater retrievability
than affective reactions to a datum.
Memory for a numerical value, comparison or evaluative reaction will be greater as the similarity
between encoding and retrieval conditions increases.
Given differences in the retrievability of encoded representations, decision makers will reconstruct
numerical values and comparisons to be consistent with a recalled affect.
Reconstructing numerical and comparative data from affective representations is more likely to occur
when a decision maker has a strong positive or negative affect.
Reconstructing numerical and comparative data from affective representations is more likely to occur
as the time between encoding and recall increases.
The encoding and retrieval of numerical data will depend upon the decision-maker’s objectives for
that data when they are first observed.
The retrievability of numerical values will increase if it is believed that they will be needed for future
decisions.
Decision makers will be more confident in recalled affective responses than in recalled numerical
values or comparisons.
For most decision contexts, affective reactions to numerical data will be used to process (combine)
data, and such processing will contribute to the greater relative retrievability of affect in future tasks.
As the amount of numerical data increases, the relative difference in the retrievability of affective
reactions to underlying constructs or broad data categories will increase as compared to other forms
of encoded representations.
This proposition indicates that basic sensory
data are potentially encoded in different
forms.* It posits that encoding may take the
form of a numerical value, such as the actual
number perceived, a point estimate, or a range
between two numerical values.3 In most judg-
ment and decision-making contexts, however,
numbers take on significance when relation-
ships or comparisons are made, and are there-
fore encoded as such (see proposition 5b for an
exception). For example, a current year num-
ber may be compared to (a) prior years’ num-
bers, (b) an average, (c) another numerical
variable, or (d) an internal norm. Such compar-
isons result in categorical representations such
as above/below average, increasing/decreas
ing, etc. In addition, the proposition indicates
that an affective reaction to those relationships
is encoded. That is, the numerical information
is encoded in terms of a bipolar evaluative
a A number of nonnumerical encoding models developed in psychology reflect this position. For example, Wyer and SruR
(1986) note that encoding may be more general or abstract than the original data itself, and the concept of encoding
behavioral data as traits is fundamental to a number of impression formation models (Fiske & Taylor, 1991; Asch, 1946;
Srull & Wyer, 1989).
3 A point estimate refers to a numerical value that approximates the actual number observed. It may be used because it is
easier to encode and is not materially different than the actual number. For example, a net income of 3.142 million may be
encoded as around three million.
588 T. KIDA and J. F. SMITH
Determine and I
NUMERICAL COMPARISON EVALUATIVE
VALUE AFFI?Cf
Q 1
I
Rsonsrmcr number and
Fig. 1. The encoding and retrievability of numerical data
dimension. While a number of terms may be
used to represent the affective reaction (e.g.
favorable/unfavorable, good/bad), the basic
underlying feature is that the reaction repre-
sents a positive or negative valence in memory
structures.4
Such encoding traces are important to a
number of decision contexts in which account-
ing information is prepared and used. As a con-
sequence, the model is potentially relevant to
decisions faced by auditors, management
accountants, financial analysts, portfolio man-
agers, bank loan officers, financial and operat-
ing managers, etc. For example, research
indicates (Ameen & Strawser, 1994) that audi-
tors employ relatively simple analytical review
techniques, in which they observe current year
numerical values (e.g. account balances, ratio
values), make comparisons to prior years’
values and/or industry norms, and evaluate
* Atfect is commonly used as a generic term to refer to a range of feelings and emotional responses (e.g. joy, sadness,
anger, jealousy, etc.; see Fiske & Taylor (1991) and Isen (1984)). Fiske and Taylor (1991) note that the feelings most
frequently of interest are underlying evaluations (i.e. positive and negative reactions). The term affect is used in this paper
to refer to such evaluative reactions. Therefore, the terms affective and evaluative are used interchangeably. Note that the
broad range of emotional responses experienced by Individuals (e.g. joy, sadness, etc.) stih have an underlying positive or
negative valence associated with them (Westbrook, 1987; Izard, 1977), reflecting the primary importance of evaluative
affect. Abelson et al. (1982) report, for example, that factor analyses of a broad range of affects reveals two general factors,
which they term positive and negative affect (also see Fiske & Taylor, 1991). Also note that if other specific types of affect,
such as sadness, joy, anger, etc., are associated with a positive or negative valence in memory structures, the effects
proposed in this paper should be enhanced. That is, such affects suggest a personal reaction to the numerical data, which
the self-reference literature suggests would result in an enhanced ability to remember the positive or negative valence
(Rogers er al., 1977; HaIpin et al., 1984; Katz, 1987).
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 589
whether the comparisons indicate that account
balances are reasonable or unreasonable. To
assess and improve operating performance in
managerial contexts, managers often compare
actual operating data to budgets or standards,
evaluate the resulting variances, and determine
which variances to investigate. As part of their
investment decisions, portfolio managers
observe a firm’s financial statement numbers,
compare them to relevant benchmarks (e.g.
other firms’ numbers, industry averages, inter-
nal norms, etc.), and evaluate whether they
represent favorable or unfavorable firm charac-
teristics. And, when making capital budgeting
decisions, financial managers consider numeri-
cal information for a given project (e.g. net
present value, payback period, etc.), make
comparisons to other projects or some firm
specific norm, and evaluate whether the pro-
ject is beneficial to the firm. In effect, the mem-
ory traces indicated in proposition 1 are
relevant to numerous decision contexts faced
by both preparers and users of accounting
information.
Proposition 1 is consistent with basic con-
cepts of fuzzy trace theory (Brainerd &
Reyna, 1990a, 1993). Fuzzy trace theory is a
gistdriven approach to cognition that has
been applied to a number of common deci-
sion-making tasks, including framing effects
(Reyna & Brainerd, 1991), transitive inference
(Reyna & Brainerd, 1990), conjunction pro-
blems (Reyna, 1991) and class inclusion pro-
blems (Brainerd & Reyna, 1990b). As Brainerd
and Reyna (1990a, p.8) note, fuzzy trace theory
indicates that, as information items are
encoded, they are mined for their senses, pat-
terns and gists. Gist can relate to various levels
of abstraction. In effect, there are fuzzy-to-ver-
batim continua, where one end concerns fuzzy
traces, which are vague representations relat-
ing to the sense or pattern of the data
observed, while the other end concerns verba-
tim traces, which preserve the data encoun-
tered. Intermediate positions relate to traces
that vary in the degree to which they approx-
imate fuzzy and verbatim boundaries. Because
data are mined for their essence, fuzzy trace
theory argues that individuals will store more
information in memory than the specific verba-
tim data encountered. That is, memory may
hold both verbatim traces and traces of varying
levels of gist. For example, when subjects are
shown three objects (A, B, and C) arranged left
to right in a transitivity task, where the lengths
of the objects are given as A= 18cm, B= 17.5cm
and C= 17cm, subjects may encode more than
the actual lengths into memory structures. For
example, they may also encode that A is long, C
is short, and B is not long or short and, at a
further level of abstraction, they may encode
that things get longer to my left. Within fuzzy
trace theory, gist may take many different
forms, depending upon the characteristics of
the task. Considerable support exists for encod-
ing such various levels of gist (see, for example,
Brainerd & Reyna, 1990a , 1993; Brainerd &
Kingma, 1984; Chapman & Lindenberger,
1988; Reyna et al., 1990).
These data from fuzzy trace theory support
the view that multiple traces may be encoded
into memory, a basic feature of proposition 1.
In addition, we argue that, for many accounting
decision tasks, comparisons among numerical
data and evaluative reactions are particularly
relevant, and are, in essence, two forms of
gist that are encoded into memory structures.
In fact, one important aspect of the model pre-
sented here is that it specifies the type of gist
that decision makers will encode. Note that
these forms of gist apply to a wide variety of
accounting judgment and decision-making
tasks where individuals must determine their
personal preference (either expressed or
implied) for a decision alternative. These may
involve tasks where decision makers are asked
to select a preferred alternative from a set of
competing alternatives, or where they must
determine their degree of preference for a sin-
gle case. While not universal (see the section
on a decision-maker’s objectives for excep-
tions), the gists are relevant to a broad spec-
590 T. KIDA and J. F. SMITH
trum of decision tasks for, as Hogarth (1987)
notes, the issue of decision-maker preference is
“common to almost all choice situations” (p.
1>.5
The importance of comparisons and affect
has also been recognized in other research
streams. The significance of comparisons for
encoding operations can be seen in impression
formation models (e.g. Higgins & Lurie, 1983)
as well as in descriptive accounting research
using verbal protocol analysis. For example,
Bouwman et al . (1987) analyzed the protocols
of professional financial analysts who were
required to screen prospective financial invest-
ments. A substantial amount of decision activity
revolved around comparing numbers to prior
years’ amounts, internal norms or industry
averages. Bedard and Biggs (1991) analyzed
the protocols of auditors who attempted to
uncover an inventory overhead allocation
error using analytical review. Once again, the
protocols indicated a significant amount of
comparison decision behavior. As Bouwman
(1983) notes, decision makers typically trans-
late quantitative data into qualitative terms. In
fact, computer programs based upon protocols
often utilize operators that translate numerical
data into qualitative terms representing com-
parisons. Protocol analysis performed in
accounting contexts also reveals that a signifi-
cant portion of predecisional behavior is eva-
luative in nature. For example, Biggs et al .
(1988) uncovered a high percentage of evalua-
tive operators when auditors adjusted an audit
program based upon analytical review (also see
Biggs & Mock, 1983).
While affect has not been emphasized in
memory and cognitive models, some research-
ers are beginning to recognize its impact in
encoding and processing data. Concerning
affective states in general, Zajonc (1980,
p. 153) notes that “There are probably very
few perceptions and cognitions in everyday
life that do not have a significant affective com-
ponent.” That is, we do not just perceive data,
we evaluate it. Beach and Mitchell’s (1987)
recent image model employs affective features
(see also Mitchell & Beach, 1990), and some
impression formation models indicate that
behavioral data are encoded as traits and as
affective reactions (SruIl & Wyer, 1989). Fiske
and Taylor (1991) also recognize the underly-
ing importance of affect, noting that almost
anything a person remembers about behaviors
or traits has an affective reaction linked to it,
and that affect’s property of cutting across
essentially all domains suggests an affective
memory code (see also Rogers, 1983).
Note that the encoding of numerical data as
comparisons or affective reactions is poten-
tially a functional cognitive heuristic, given lim-
itations of the human memory system. It is, in
effect, a form of chunKng that allows a greater
amount of information to be stored in a given
space (Miller, 1956). For example, instead of
storing two different company sales figures
for the current and prior period, comparisons
can be made and a single qualitative term (e.g.
sales are increasing), or an affective reaction
(e.g. sales are favorable), can be stored. This
type of data should be easier to store and
should take less time to process in the future
since the number of chunks are reduced
(Newell & Simon, 1972). That is, it takes less
time to recall and use an affective reaction,
than to recall two numbers, compare them,
and evaluate the comparison. In addition,
such encoding traces may have other benefi-
cial effects. For exampIe, Reimers et aE.
(1993) present evidence which indicates
greater consensus when auditors make control
risk assessments using a comparative response
scale (e.g. low, medium, high) versus a numer-
ical response scale.
’ As is noted later, in some situations affective traces are not applicable given the objectives of the task. For example, affect
is not required if decision makers are asked to determine a rule or just observe relationships in data (e.g. the transitivity
task previously discussed). See the discussion of a decision maker’s objectives and footnote 11 for further elaboration of
this point.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 591
Retrievability of encoded representations
The first proposition indicates that numerical
data may be initially encoded in different
forms. More importantly, however, the model
also specifies that the retrievability of the mem-
ory trace of these different encoded represen-
tations will differ. The issue of retrievability is
addressed in the following propositions.
Proposition 2a: The retrievability of encoded represen-
tations increases from numerical values, to comparisons
between numerical values, to affective reactions to
those comparisons.
Proposition 2b: Affective reactions to underlying con-
structs or broad categories of data will have greater
retrievability than affective reactions to a datum.
The propositions indicate that decision
makers will be better able to retrieve memory
traces for numerical data that are in the form of
a descriptive, evaluative response, rather than
in the form that most closely corresponds to
the actual data observed (i.e. numerical
values). The inability to retrieve a memory
trace may be due to a permanent loss of the
trace from memory or to the use of inappropri-
ate retrieval mechanisms (Loftus & Loftus,
1980; Squire, 1987).6 Both are relevant to the
retrievability propositions listed above. That is,
we argue that affect will enhance retrievability
because affective traces will be retained in
memory longer than other forms of encoded
representations, and that, of the traces existing
in memory at a given time, affect will be the
easiest to retrieve.
Relevant research. Research investigating
related issues in cognitive development, mem-
ory, and impression formation provide support
for these propositions. The cognitive develop-
ment literature indicates a greater availability
for fuzzy versus verbatim traces (Brainerd &
Kingma, 1984, 1985; Brainerd & Reyna, 1988,
1990a, 1990b; Reyna, 1988; Reyna & Brainerd,
1987; Reyna et al., 1990). As Brainerd and
Reyna (1990a, p. 19) note, complex, well-articu-
lated memory structures are likely to disinte-
grate more quickly than simple structures.
Detailed information is not only memorially
unstable, it is less likely to be fully encoded
(Bransford & Franks, 1971). In effect, fuzzy-to-
verbatim continua indicate a susceptibility to
forgetting traces, as well as the precision of
memory traces (Brainerd & Reyna, 1990a).
Also, as Brainerd and Reyna (1993, p.48)
note, gist has a greater retrieval advantage
because it can be accessed by a broader range
of retrieval cues as compared to verbatim data.
As we argued in proposition 1, comparisons
and affective reactions are forms of gist that
accounting decision makers extract from
numerical data. Also, affect reflects a greater
level of abstraction and is closer to the fuzzy
end of a fuzzy-verbatim continuum. As a conse-
quence, research from fuzzy trace theory pro-
vides support for our contention that a
comparison and affect will have a retrievability
advantage over a numerical trace.
Support for the relative retrievability of the
different encoded representations also is found
in Wyer and Srull’s (1986) model of human
cognition in social contexts. They note that,
“
. when the processing of information
requires several stages, the material involved
’ While some researchers have argued that all long-term memory traces are permanent (Penfield, 1969) and that forgetting
is therefore caused by the use of inappropriate retrieval cues, Loftus and Loftus (1980) present compelling evidence that
the inability to recall data can be due to an actual loss of the memory trace as well as the use of inapproporiate retrieval
mechanisms. (Also see Squire (1987) for neurobiological arguments for the permanent loss of memory traces.) Whatever
the underlying reason, the central concern of the model presented here is the relative retrievability of the different
encoded representations. Also note that the conceptual distinction between short and long-term memory, while not
essential to the present model, is not inconsistent with the model’s propositions. That is, the model argues for differences
in the retrievability of encoded representations, and that all representations may be lost over time. This can fit a memory
model with multiple levels, or a model with a short-term/long-term dichotomization, where data may be lost quickly
because it is not transferred from shortterm to lcng-term memory, or data may be lost from long-term memory over a
longer period of time (Lynch & Srull, 1982).
592 T. KIDA and J. F. SMITH
in earlier stages of processing is more likely to
be displaced than that involved in later stages”
(p. 326). As will be seen in proposition 7, we
argue that affective reactions generally are used
to process or combine different information
items, and are therefore used in the later stages
of information processing. Research also indi-
cates that verbatim traces fade more quickly
than gist because of retroactive interference
from subsequently encoded traces (Brainerd
& Reyna, 1993, 1989; Brainerd et al ., 1990)’
In addition, some empirical evidence from
memory and social cognition research points
to the enhanced retrievability of affect. For
example, Graf and Mandler (1984) and Graf et
al . (1982) found better recall when subjects
initially rated how much they liked or disliked
words than when they assessed the internal
characteristics (e.g. number of vowels) of
words, and Posner and Snyder (1975) found
shorter response times when subjects matched
words to “emotional tones” than when they
matched words to words, suggesting greater
ease of access for affective responses (also
see Hyde &Jenkins, 1969).*
Our arguments concerning the enhanced
retrievability of affective memory traces
encompass the nature of affect itself.’ Zajonc
(1984) points to the primacy of affect evident
in research examining the genetic develop-
ment of individual organisms (i.e. ontogeny)
and groups of related organisms (i.e. philo-
geny) (also see Donohew et al ., 1988). Zajonc
notes, for example, that the limbic system of
the brain, which controls emotional reactions,
was present before humans evolved language
and the cognitive capacities dependent upon
language. In addition, infants cry and smile, dis-
playing affective reactions, long before they
acquire any semblance of verbal skills (Izard,
1978, 1979). This evidence suggests that, to
some extent, affective responses are “hard-
wired” in the human system. Therefore, it is
likely that affect is central to human cognitive
processes, and that affective responses are
readily accessed from memory. In addition,
Zajonc et al . (1982) note that the original cog-
nitive bases of certain emotions can be forgot-
ten or disassociated from an affective memory
trace (p. 214). For example, when reminded of
a movie seen in the past, a person may readily
remember that he/ she liked or disliked it, but is
often unable to remember much of the factual
information about the movie that may have
caused that response, once again demonstrat-
ing the greater retrievability of affect.
’ Research on certain memory processes is also related in that it indicates differences in the retrievability of different types
of data from memory structures (see, for example, Craik & Lockhart, 1972a,b; Craik & Tulving, 1975; Fisher & Craik, 1977;
Parkin, 1984; Horton &Mills, 1984; Koriat & Melkman, 1987; Craik & Jacoby, 1979; Anderson & Reder, 1979; Walker etal .,
1983; Klein & Loftus, 1990), and similar effects have been found in the impression formation literature (Hamilton, 1981;
Hamilton er al., 1980b).
s Note that there is a great deal of research on the effect of general affective states (e.g. moods) on memory and judgment
(for reviews see Bower, 1981; Isen, 1984, 1987; Fiske & Taylor, 1991). For example, it has been found that individuals
often remember material whose valence is consistent with their present mood (e.g. Bower et al., 1981; Salovey & Singer,
1988; Clark & Teasdale, 1985; see Blaney, 1986; Mayer, 1986; Isen, 1987 for reviews). In addition, a general positive mood
results in more positive judgments across a variety of contexts (e.g. Clark 8t Williamson, 1989; Isen, 1984, 1987; Mayer &
Salovey, 1988; Mayer, 1986; Fiedler et al ., 1986). General affect has also been found to impact decision-making style, with
positive moods resulting in more expansive, inclusive and quicker decision styles (Fiske & Taylor, 1991; Fiedler, 1988;
Isen, 1987; Isen & Means, 1983). While this research is of interest in that it points to an impact of affect on cognition, it
typically relates to general affective states (e.g. moods). The affect of concern in the present model differs in that it relates
to affective reactions that are specific to data observed within a decision-making context.
9 While we specifically focus on evaluative reactions to numerical data in the present model, research on other affective
states is pertinent because, as prior work indicates (e.g. Westbrook, 1987; Izard, 1977) the broad range of affective states
have an underlying positive or negative valence associated with them.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 593
Note that while Zajonc’s work points to the
importance of affect, he has argued that, in
some situations, affect may be experienced
with little cognitive processing (i.e. it may
operate in a system separate from cognition).
This has been a controversial position that has
been criticized by a number of researchers. For
example, Lazarus (1982, 1984, 1990) argues
that cognition, defined as appraisal, is neces-
sary for affect, Epstein (1983, 1984) argues
that preconscious cognitions precede affect
(i.e. cognition need not be conscious), while
others suggest a resolution given different
meanings of cognition (e.g. Averill, 1990).
While still under debate, the work of all these
researchers and theorists emphasizes the
underlying importance of affect. It also is
important to note that while Zajonc put forth
the position that affect may be separate from
cognition in some instances, he did not mean
to imply that such was always the case. In fact,
Zajonc et al . (1982, p. 211) indicate that, in
perhaps most instances, cognition is an impor-
tant factor in determining affect. We argue that,
when applied to numerical decision-making
contexts, affect is, almost by necessity, a reac-
tion to comparisons among numerical values,
and therefore likely to involve cognitive
processing.
In summary, there is an increasing recogni-
tion of the fundamental role of affect in human
cognition, and there is evidence to suggest that
affective responses are essential in the encod-
ing and retrieval processes in memory. Further,
a considerable amount of research investigat-
ing differing psychological issues supports the
proposed retrievability of the encoded repre-
sentations set forth in the above propositions;
memory traces for data comparisons will be
more retrievable than those for numerical
values, and the most retrievable memory trace
will result from affective reactions (see propo-
sition 5b for an exception).
Underl yi ng constructs. The above proposi-
tions indicate that, among evaluative represen-
tations, affective responses to underlying
constructs or broad categories of data will be
more retrievable than affective responses to
specific numerical items. An underlying con-
struct refers to a dimension that, while not
directly measurable, may be captured by one
or more measured variables. For example,
when portfolio managers make buy/ sell deci-
sions, auditors make going-concern decisions,
or bank loan officers make lending decisions,
they examine underlying dimensions such as a
firm’s profitability, leverage, and liquidity. The
underlying dimension “profitability” is not
directly measurable. Rather, the construct is
evaluated by observing a single financial mea-
surement (e.g. net income), or by combining a
number of observed measurements (e.g. sales,
gross profit, net income, return on investment,
earnings per share, etc.). In this context, pro-
position 2b posits that the resulting evaluation
of the construct “profitability” will be more
readily recalled from memory than evaluations
of the data measuring that construct.
Underlying constructs are a type of broad
data category. However, the term broad data
category is more generally defined here to
refer to any related combination of cues. For
example, in a choice context, a decision maker
must choose an alternative from a set of com-
peting alternatives. When making capital bud-
geting decisions, a financial manager may
choose a project from a number of proposed
projects. The overall evaluation of a project is,
in essence, an affective response to a broad
category that represents all information rele-
vant to that project. Proposition 2b indicates
that the memory trace for the overall evalua-
tion is more retrievable than affective
responses to specific decision variables. Given
that assessments of broad data categories occur
later in the sequence of processing than evalua-
tions of the individual data items from which
they are comprised (i.e. decision makers first
evaluate individual cues and then combine
them to form an overall evaluation of a con-
struct or broad data category), the retrievabil-
ity of broad data categories will be enhanced
given the arguments previously set forth (i.e.
Wyer & Srull, 1986; Brainerd & Reyna, 1989,
1993).
I mpl i cati ons. The proposition that affective
594 T. KlDA and J. F. SMITH
responses are more readily retrieved from
memory than other forms of data representa-
tions can have important implications for deci-
sions made when data are retrieved from
memory. The importance of memory increas-
ingly is recognized by accounting and auditing
researchers. Reliance on memory is essential in
accounting decision making because, even if
data are available for review, it is costly to con-
tinually refer back to external storage (Birnberg
& Shields, 1984), and limitations on working
memory necessitate reliance on long-term
memory (Libby & Trotman, 1993). While the
use of affective responses may serve as an effi-
cient cognitive heuristic, decision quality may
be impaired if numerical information is lost and
only an affective response can be subsequently
retrieved from memory for a future decision.
The effects of encoded representations on
subsequent judgments have been examined in
social cognition and memory research, and the
results suggest that the retrieval of specific sti-
mulus information is based, at least in part, on
how it was initially categorized in memory,
especially after a considerable delay between
the initial encoding and retrieval. A number
of studies have found that subjects base cur-
rent recall, judgments, and behavior on prior
categorizations of information, even though
such categorizations may not be appropriate
to the decision presently being made (e.g. Hig-
gins & Lurie, 1983; Carlston, 1980; Higgins &
Rholes, 1978; Lingle & Ostrom, 1979; Sherman
et al., 1978). As Higgins and Lurie (1983, p.
528) note, “. subjects are using their prior
categorizations for subsequent recall and infer-
ences without taking sufficient account of the
context in which the categorization was
made.”
Consider, for example, that portfolio man-
agers may make investment decisions on a con-
tinuing basis, where they assess sets of
investment alternatives that change over time.
If the net income of a particular firm is com-
pared initially to three other firms being con-
sidered, and is markedly higher than the
others, it may be represented affectively in
memory as “excellent net income”. Given the
greater retrievability of affective responses, if
that firm’s income is subsequently retrieved
from memory for comparison with a second
set of firms in the future, and the affective
response (but not the numerical value) is acces-
sible, the manager may conclude again that the
net income is “excellent,” even though the
income measure may not compare as favorably
with the new firms. In fact, the manager may
“infer” the original firm’s income value to be
higher than it actually was because the base of
comparison (i.e. the incomes of the set of new
firms) is higher (Higgins & Lurie, 1983). The
issue of inference or reconstruction in memory
is more fully discussed in propositions 4a-c.
Similar effects may occur if data initially
encoded for one type of task are subsequently
retrieved from memory for another. Thus, the
relative retrievability of affective representa-
tions may, in certain situations, impair the qual-
ity of subsequent decisions.
Encoding specijkity
The previous propositions discussed the rela-
tive retrievability of encoded representations
without considering the impact of differences
in encoding and retrieval conditions. Given the
specificity principle in memory research,
which indicates that retrieval of a memory
trace is more likely to occur if retrieval condi-
tions are similar to the conditions of encoding
(Tulving & Thomson, 1973; Tulving, 1983), the
following proposition is advanced:
Proposition 3: Memory for a numerical value, compar-
ison or affective reaction will be greater as the similarity
between encoding and retrieval conditions increases.
This proposition stresses the importance of
viewing access to memory as a function not
only of the type of encoded representation,
but also of the match between the conditions
that exist at encoding and at retrieval. Support
for the importance of the similarity of condi-
tions can be found in numerous studies inves-
tigating different types of encoding and
retrieval operations (Thomson & Tulving,
1970; Fisher & Craik, 1977; Morris et al.,
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 595
1977; Bransford et al ., 1979; Smith et al ., 1978;
Geiselman & Glenny, 1977; Tulving & Pearl-
stone, 1966). For example, in a study examin-
ing word recall, Fisher and Craik (1977) found
that semantic retrieval cues provided better
recall than rhyming cues for words that were
initially encoded using semantically associated
word-pairs, but that rhyming cues provided
better recall than semantic cues for words
that were initially encoded using word-pairs
that rhymed. Similar results are reported by
Thomson and Tulving (1970), who varied the
strength of association between word-pairs in
encoding and retrieval tasks, and Barclay et al .
(1974) who emphasized different properties of
words in encoding and retrieval. As Schacter
(1986) notes, there is considerable evidence
indicating an interactive effect between encod-
ing and retrieval conditions on the retrievabil-
ity of various encoded representations in
memory.
The similarity between encoding and retrie-
val conditions may be influenced by a number
of variables. Empirical evidence indicates
encoding specificity effects for task specific
variables such as the nature of encoding and
retrieval tasks, for “broader” contextual vari-
ables such as the setting in which the task is
performed (Smith et al ., 1978) and for the
general cognitive state of the individual
(Bower, 1981; also see Eich, 1980 for a
review). For example, Smith et al . (1978)
found that individuals who were tested in the
same surroundings in which they had learned
materials, and by the same experimenter, per-
formed significantly better than individuals
tested in new surroundings by a new experi-
menter. Thus, the degree of retrievability of
information appears to be affected by a range
of task and environmental conditions.
It should be noted that proposition 3 does
not preclude or diminish the importance of
the form of encoded representations on the
relative retrievability of the memory trace
(proposition 2a). Rather, it indicates that the
retrievability of any memory trace will be
enhanced by conditions of high encoding-
retrieval similarity, and diminished by low simi-
larity conditions. For example, while affective
responses and numerical values are posited to
be more retrievable the greater the similarity
between the encoding and retrieval condi-
tions, within any given state of encoding-retrie-
val compatibility (e.g. high or low), affective
responses will remain relatively more retrieva-
ble than numerical values. This position is sup-
ported by research that has found differences
in recall performance for different forms of
encoding, when the degree of similarity
between encoding and retrieval conditions
was held constant (Morris et al ., 1977; Fisher
& Craik, 1977, exps. 1, 2 and 3; Nelson et al .,
1974).
Reconstructi on of recal l ed encoded
representati ons
Proposition 2a indicates that not all forms of
encoded representations will remain in mem-
ory over the same length of time. This suggests
that as the time between initial encoding and
subsequent retrieval increases, attempts to
recall encoded representations that are not
accessible may result in the retrieval (con-
sciously or unconsciously) of a related accessi-
ble representation, and the “reconstruction” of
inaccessible traces to be consistent with the
information retrieved. Since affective reactions
are posited to be the most accessible memory
trace, there will be a greater likelihood that
numerical values and comparisons will be
reconstructed to be consistent with the
recalled affect as the delay between encoding
and retrieval increases. These issues are pre-
sented in the following propositions:
Proposition 4a: Given differences in the retrievability of
encoded representations, decision makers will recon-
struct numerical values and comparisons to be consis-
tent with a recalled affect.
Proposition 4b: Reconstructing numerical and compara-
tive data from affective representations is more likely to
occur when a decision maker has a strong positive or
negative affect,
Proposition 4c: Reconstructing numerical and compara-
tive data from affective representations is more likely to
596 T. KIDA and J. F. SMITH
occur as the time between encoding and recall
increases.
These propositions suggest that memory for
numerical values and comparisons may be
influenced by the affective responses with
which they are associated. That is, while
numerical values and comparisons may be initi-
ally encoded as accurate copies of actual
observed data, over time they may be replaced
by memory traces that are reconstructed based
upon related affective responses. lo Further,
data never in fact observed may be erro-
neously “remembered” (i.e. inferred) because
they are expected to be present given a certain
recalled affective response, and such recon-
structive behaviors are more probable with a
strongly held affect. Note that these proposi-
tions are not meant to imply that all memory
contains errors. We often can accurately recall
data even when inconsistencies are evident
(Matlin, 1989). They do, however, indicate a
directional reconstruction which increases
over time. Note also that these propositions
do not imply that there is a relationship
between the accuracy of memory for the differ-
ent types of representations. In fact, since
actual numerical values will be quickly lost,
there may be no correlation between the accu-
racy of recalled evaluative traces and the accu-
racy of verbatim traces. As such, these
propositions are not inconsistent with the
representational independence principle of
fuzzy trace theory.
cept, along with how the variables are related
and what are acceptable values for those vari-
ables. Schemata can assist cognition by using
the general case to fill in for a specific case
(Fiske & Taylor, 1991, p. 171). That is, the
process of remembering details of a particular
credit assessment may be facilitated by a
schema that provides an “inventory” of vari-
ables generally considered in such an assess-
ment (see Tesser, 1978; Bransford 8z Johnson,
1972; Smith et al ., 1978; Arkes & Freedman,
1984; and Bransford, 1979 for support).
Evidence for reconstruction behavior can be
found in research investigating the effect of
schemata in decision making. A schema is an
organized base of knowledge about particular
concepts (e.g. persons, events, things, etc.)
that has been constructed based on prior
experience. For example, for a bank loan off-
cer, the concept “credit-worthiness”
may
immediately bring to mind a set of variables
or measures that typically comprise that con-
While schemata can greatly aid decision pro-
cesses, evidence also suggests that decision
makers sometimes “remember” information
related to a general schema that was never
actually present in a specific case. That is, indi-
viduals infer or reconstruct information
because, within a particular context, it is
expected or highly plausible. As Choo (1989,
p. 481) notes, schemata not only direct atten-
tion to relevant data, they also allow for infer-
ences to be made when information is missing
or ambiguous. For example, Brewer and
Treyens (1981) administered a surprise recall
test of office contents to subjects who had
been previously waiting in an office. The data
indicated that subjects recalled items typically
found in an office (e.g. books) that were not
actually present in this instance. Arkes and
Freedman (1984) had subjects listen to pas-
sages about baseball games and complete a
recognition test that included key words in
the passages and items that were schema-con-
sistent (i.e. typical to baseball games), but not
in the passages. A number of subjects believed
such items were, in fact, present. Support for
reconstruction behaviors is also evident in the
impression formation literature, where subjects
erroneously recognized ‘new” traits as having
been presented before, if they were evalua-
tively similar to traits in the original observed
set (Tsujimoto, 1978).
Investigation into the existence of recon-
.^
‘I’ Of course, considering proposition 2b, affective responses to individual numerical values or comparisons may also be
retrieved to be consistent with an overall affective response to an underlying construct or broad data category.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 597
struction behavior also has been made in an
audit judgment context. Moeckel (1990) asked
professional auditors to review working papers
containing contradictions that would have
been discovered in the absence of reconstruc-
tion behaviors. Her findings indicated that
auditors (especially experienced auditors)
reconstructed memory traces by altering men-
tal representations of data so that they are con-
sistent with existing knowledge or memory
(see also Christ, 1993; Harris, 1981; Arkes &
Harkness, 1980; Sulin & Dooling, 1974; Chase
& Simon, 1973 for related findings in other
fields). In effect, a number of studies have
revealed that inferred items often are falsely
remembered to have been observed. In fact,
some schemata models predict lower overall
accuracy for items typical to a schema because
of the greater number of false recalls (infer-
ences) for typical as oppossed to atypical items
(e.g. Graesser & Nakamura, 1982). Birnberg
and Shields (1984) also note that reconstruc-
tive behavior may occur if data that were not
initially attended to subsequently becomes
important. Decision makers may fill in the unat-
.tended detail with a value consonant with
those data actually retrieved.
Evidence also suggests that reconstruction is
more likely to occur the greater the time
between initial encoding and recall, as indi-
cated in proposition 4c. For example, Spiro
(1980) provided subjects with stories about
individuals’ behavior, and information relating
to the potential outcomes of the behavior. Sub-
jects were subsequently asked to recall the
story, but at different time intervals (i.e. either
2 days, 3 weeks or 6 weeks later). He found
that in the face of outcome-story inconsisten-
cies, subjects confidently reported recollec-
tions of the story that were outcome-
consistent but grossly inaccurate, and that
more subjects made this type of error for
each of the two longer time intervals than for
the shorter interval. Higgins and McCann
(1984, p. 28) also note that individuals have
an increasing tendency over time to use prior
judgments as a basis for reconstruction, instead
of relying solely on details of the stimulus infor-
mation (see Higgins, 1981; Lingle & Ostrom,
1979; Higgins & Rholes, 1978).
I mplications. The use of inference in mem-
ory suggests that the past is not simply repro-
duced, but is reconstructed using knowledge-
based principles of coherence (Spiro, 1980).
This broad-based support for inference has
led many cognitive psychologists to view the
memory trace not as a literal “copy” or “snap-
shot” of an event, but rather as a fragmentary
and often distorted representation (Schacter,
1986).
Considering the role of affect proposed in
the model, and propositions 2a and b concern-
ing the relative retrievability of different forms
of encoded representations, we argue that
reconstruction will be manifested primarily by
altering memory traces of numerical values and
comparisons to be consistent with related
affective responses. That is, retrieved evalua-
tive affect will activate expectations concern-
ing underlying or related numerical values and
comparisons, and to the extent that memory
traces for such data are inaccessible, they will
be reconstructed to be affect-consistent (see
Lingle & Ostrom, 1981; Ostrom et al., 1980).
In effect, the propositions indicate a directional
reconstruction, where numbers will be recon-
structed to be consistent with recalled evalua-
tions when the verbatim traces are no longer
available.
As indicated in proposition 4c, reconstruc-
tion is more likely to occur as the time
between encoding and recall increases. Conse-
quently, susceptibility to reconstruction should
be greater in decision contexts in which a final
decision is reached some time after initial expo-
sure to basic numerical data. While this limits
its occurrence to some extent, many decision
contexts exhibit an extended time frame.
Accounting decision makers often operate in
a multitasking environment, where they must
attend to a number of different tasks over a
period of time. In addition, they often assess
many different alternatives when faced with a
choice context before a decision is reached. A
large amount of data may have to be attended
to and, therefore, the final decision can occur a
598 T. KIDA and J. F. SMITH
considerable time after the decision maker is
exposed to the information first encoded. For
example, financial managers may assess a num-
ber of capital projects proposed by various
divisions over a period of time before final
approval is given to one or a few select pro-
jects. Portfolio managers may make investment
decisions on a continuing basis, assessing sets
of investment alternatives that are changing
over time. In such circumstances, data
observed at different times in the decision pro-
cess may have to be combined or compared.
Propositions 4a-c suggest that the decision-
makers’ affective reactions to data observed
early in the decision process may be retrieved
later and used to reconstruct numerical values
when comparisons to the later alternatives’
numerical values must be made. As indicated
previously, such reconstructed numbers may
not be appropriate because the affective
responses used to reconstruct data were
based upon comparisons made early in the
decision process. For example, a net income
value considered excellent because it was con-
siderably higher than the values of other firms
assessed early in the decision process may not
compare as favorably to the income values of
firms assessed later, but it may be recon-
structed to be higher because the decision
maker retrieves the evaluation of net income
as excellent, as opposed to retrieving the
actual number.
Note that a numerical value may be recon-
structed from an affect encoded for the same
variable (e.g. an income value may be recon-
structed to be consistent with the encoded
affect, excellent income), and that numerical
values may be reconstructed based upon less
direct associations. Consider, for example, a
loan officer’s decision to grant credit. A net
income amount may have been encoded along
with other variables (e.g. amount of outstand-
ing debt, a loan applicant’s deposit balances at
the bank, etc.) to assess a customer’s “credit
worthiness”. In this instance, what is remem-
bered concerning specific variables (e.g. net
income) may be influenced by the overall eva-
luation of credit worthiness. That is, the
model posits that an overall evaluation has
greater retrievability than the evaluations of
the specific variables on which the overall
affect is based (see proposition 2b). Proposi-
tion 4a indicates that retrieving (consciously
or unconsciously) this prior evaluation of
credit worthiness will activate a set of expec-
tations concerning the variables typically asso-
ciated with this evaluation, and that specific
variables will be reconstructed to be schema-
consistent if not presently available in memory
structures.
The effect of decision-makers’ objectives
The importance of a decision-maker’s objec-
tives on different aspects of cognitive proces-
sing has been recognized in a number of areas.
Wyer and Srull’s (1986) model of social infor-
mation processing points to the effects of task
objectives on how information is encoded,
organized and retrieved (also see Srull &
Wyer, 1986). For example, studies examining
cognition in social contexts indicate memory
differences for information initially encoded
in a recall exercise versus an evaluative task.
For strict memory tasks, subjects appear to
retrieve information largely in the order in
which it is presented, while for impression
formation tasks they retrieve information as it
relates to aspects of the person being evaluated
(Srull, 1983; Srull & Brand, 1983; Hartwick,
1979; Wyer et al., 1984; Hamilton et al.,
1980a,b; Wyer & Gordon, 1982, 1984). Simi-
larly, evidence from consumer behavior
research indicates that task goals affect how
consumers recall information about products.
For example, the recall of information about
different product brands is based primarily on
brand names for strict memory tasks, whereas
for the task of choosing a brand, recall is largely
based on attributes relating to brand prefer-
ence (Biehal & Chakravarti, 1982; Lynch &
Srull, 1982). Waller and Felix (1984) also recog-
nize the impact of one’s goals on the data
abstracted in memory for audit decision mak-
ing. In essence, the objectives of a task can
have important effects on cognitive operations
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 599
(see also Payne (1982) for a discussion of simi-
lar effects).
Of prime importance for the encoding and
retrieval characteristics of the model devel-
oped here is the specific objective that a deci-
sion maker has for a given datum, as postulated
in the following proposition.
Proposition 5a: The encoding and retrieval of numerical
data will depend upon the decision-maker’s objectives
for that data when they are first observed.
A primary implication of this proposition
concerns the retrievability of numerical data.
While the model posits that numerical values
are not, in general, as readily retrievable as
other memory traces, they are obviously
retained in memory for long periods of time
in some circumstances (e.g. telephone num-
bers, addresses, social security numbers, etc.).
We argue that the objectives that a decision
maker has for the numerical data will affect
the retrievability of that data. For example,
numerical data whose purpose is to serve as
an identifier or label, as a benchmark for com-
parison, or as a factual report, are likely to be
retained longer in memory as numerical values
if they are needed for future reference. This
leads to the following proposition:
Proposition 5b: The retrievability of numerical values
will increase if it is believed that they will be needed
for future decisions.
This suggests that while numerical values are
posited to be generally the least retrievable
form of encoded representation, under certain
circumstances their retrievability may be
enhanced. Research indicates that specific
task and formatting manipulations can affect
the types of representations processed, and
that decision makers may sometimes be aware
of exact numerical data (see, for example,
Reyna & Brainerd, 1991; Brainerd & Reyna,
1988, 1990b).
Numerical values not viewed as essential to
future decisions at the time of encoding are
likely to be lost in memory, even though asso-
ciated comparisons and evaluations may be
retained longer, while numerical values that
are repeatedly referenced in the future are
more likely to be retained, without regard to
the existence of associated comparisons or
evaluations. Note, however, that the vast
majority of numerical data observed by deci-
sion makers in judgment contexts will not be
of the nature described above, but rather, will
primarily concern a judgment presently being
made. In these more common instances, the
objective for most numerical data is to make
comparisons and evaluations for the present
decision and, as such, the data are likely to
be lost in memory more quickly than related
evaluative responses, as indicated in proposi-
tion 2a.
ConBdence in affect
While we can be proven incorrect or inac-
curate in what we believe, it is difficult to be
wrong about how we feel (Zajonc, 1980). For
example, the amount by which the debt of a
particular firm is above or below a given
benchmark is readily observable, and there-
fore subject to reasonably “objective” verilica-
tion. How “good” or “bad” this information is,
given a particular decision context, is not
observable, and may be based upon a com-
plex set of factors that differ for individual
decision makers. Thus, affective responses
are less likely to be considered in error than
other forms of data, leading to the following
proposition:
Proposition 6: Decision makers will be more confident
in recalled affective responses than in recalled numer-
ical values or comparisons.
Birnberg and Shields (1984) note that if deci-
sion makers are sufficiently confident in data
retrieved, they may use such information with-
out referring back to external storage. In fact,
Moeckel and Plumlee (1989) operationalized
confidence as a willingness to rely on recogni-
tions rather than reexamine source documents.
They found that auditors sometimes confused
their own inferences drawn from observed
data with evidence actually observed, and
600 T. KIDA and J. F SMITH
were generally at least as confident in incom-
plete and inaccurate memories as they were in
accurate recognitions. Such confidence may
lead to judgment errors (Libby, 1989). With
respect to the immediate model, this points
to the importance of the implications dis-
cussed concerning propositions 2 and 4. If
prior affective responses are retrieved with
confidence, decision makers will be less likely
to review the previously observed data under-
lying that affect. This may result in a suscept-
ibility to judgment errors to the extent that
prior affective responses are not appropriate
to the immediate judgment, but are neverthe-
less used to reconstruct numerical data.
The effect of processing on encoded
representations
While an essential aspect of the model pre-
sented here concerns the role of affect in
encoding and retrieval operations, its impor-
tance to data processing (i.e. how data are
combined to form a judgment) is readily appar-
ent. We argue that decision makers primarily
use affect in processing operations, even in
instances in which the data are immediately
accessible from external sources. This is consis-
tent with fuzzy trace theory’s fuzzy processing
preference, which indicates that individuals
reason by processing traces that are as global
as possible (Brainerd & Reyna, 1993, 1990a;
Reyna, 1988). Support for this is found in
both the cognitive development literature
(Brainerd & Reyna, 1990b) and the judgment
and decision-making Iiterature (Arkes, 1991).
Note also that the use of affect in processing
provides further support for the proposed
retrievability of affect previously discussed.
That is, combining data to form an overall judg-
ment concerns the latest stage in decision mak-
ing, and the use of affective responses at this
stage of processing would further enhance
affect’s future retrievability. This is supported
by research indicating that verbatim represen-
tations fade more rapidly than gist because of
retroactive interference from subsequently
encoded traces (Brainerd & Reyna, 1993,
1989; Brainerd et al., 1990). As Wyer and Srull
(1986) state, data in early stages of encoding
and processing are likely to be displaced by
data involved in the later stages. This leads to
the following proposition:
Proposition 7: For most decision contexts, affective
reactions to numerical data will be used to process
(combine) data, and such processing wilI contribute
to the greater relative retrievability of affect in future
tasks.
In most contexts, the data and the judgment
variable represent more than one construct;
no common underlying dimension is present.
In such instances, the data must be trans-
formed to a common scale in order to effec-
tively combine fundamentally different
constructs. We propose that the scale used
typically is affective in nature. For example,
an auditor must often consider a number of
different constructs (e.g. profitability, lever-
age, liquidity, etc.) when deciding whether a
client’s audit opinion should be qualified for
going-concern reasons. At some point in the
decision process, these constructs must be
combined and consolidated onto a single
dimension (i.e. the decision to qualify). In
such instances, auditors are not likely to trans-
form the constructs directly to the decision
dimension (e.g. liquidity is low, qualify; profit-
ability is high, don’t qualify). Rather, they are
likely to combine affective responses to con-
tructs, either consciously or unconsciously,
and then transform a resulting overall affec-
tive reaction to the decision variable. For
example, the auditor may observe liquidity
measures as below average and evaluate the
construct as unfavorable, evaluate above aver-
age profitability measures as favorable, and
combine these affective responses to arrive
at an overall evaluation, which is then trans-
formed to the judgment response (e.g. the
firm’s continuity status is favorable; there-
fore, do not qualify). In effect, representing
different types of data in terms of affective
responses provides an efficient heuristic in
situations where a mechanism must be used
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 601
to collapse a number of different constructs to
a single judgmental response. ’ ’
In summary, for most judgment and decision-
making contexts, affective reactions are likely
to be combined when determining a judgment
response. Since the use of affect in such pro-
cessing activities relates to later stages of the
decision process, these encoded representa-
tions are likely to displace the representations
found in early stages (Wyer & Srull, 1986) rein-
forcing the enhanced retrievability of affect
stated in proposition 2a.
The effect of i nformati on l oad on
encoded representati ons
As was previously stated (proposition 2b), in
general, affective reactions to underlying con-
structs and broad categories provide the most
retrievable memory trace. Proposition 8
extends this argument by positing that the
enhanced relative retrievability of such repre-
sentations will be accentuated as the amount of
data increases. This will occur primarily
because the encoded representations that are
closer in form to the observed data will be lost
more quickly with a large data set than with a
small data set, leaving affective reactions to
constructs and broad categories in memory.
In addition to the reasons for affective retrie-
vability presented in the discussion of proposi-
tions 2a and b, a basis for the present
proposition is the consistent finding from
human information processing research indi-
cating that decision makers utilize cognitive
heuristics to simplify decision tasks (Tversky
& Kahneman, 1974; Hogarth, 1987). Such heur-
istics are especially likely to be used as task
complexity increases, and a major component
of complexity is the amount of data to which a
decision maker must attend. This aspect of
complexity can be manifested in the number
of cues considered when judging a single
case, or in the types of cues or the number of
alternatives that must be considered when
choosing one alternative from a set of compet-
ing alternatives. For example, considerable
research has indicated the use of simplifying
decision strategies, such as elimination by
aspects, when a large number of alternatives
must be considered in a choice context
(Payne, 1976; Paquette & Kida, 1988; see
Ford et al ., 1989 for a review). So too, the
use of affective reactions to underlying con-
structs and broad data categories act as a sim-
plifying heuristic to make a complex task
manageable. For example, as the number of
cues in a judgment task increases, related,
observed cues are more likely to be aggregated
into an affective response to a relevant under-
lying construct. Similarly, a larger number of
alternatives in a choice context is more likely
to result in memory traces for the overall affec-
tive reactions to the different alternatives, as
opposed to affective responses to the indivi-
dual cues. As previously noted, a broad data
category may be detined as the data set com-
prising an alternative. These memory traces for
broad data categories concern data utilized clo-
ser to the end of processing (i.e. when the final
decision is made), and are posited to be more
retrievable than data observed early in proces-
sing operations (e.g. numerical values). Related
work on fuzzy trace theory also supports this
contention, indicating that verbatim memory
traces appear to be more sensitive than gist
traces to retroactive interference from subse-
.
” In contexts in which the data and the decision variable represent a common dimension, affective responses are not
essential. For example, to judge whether a firm’s profitability is above or below average, a decision maker may combine a
number of observed measures (e.g. net income, earnings per share, etc.) aLI of which measure the same unobservable
construct. In this circumstance, the decision maker may directly transform each variable onto a profitability dimension
(e.g. net income indicates below average profitability; earnings per share is welI below average, etc.) and combine them to
arrive at a final judgment (e.g. profitability is well below average). Note, however, that if this profitability assessment is the
basis for an action (e.g. buy/don’t buy the firm’s stock), the profitability assessment or the variables underlying that
assessment must ultimately be transformed onto an action dimension. In this instance, an affective dimension is again
likely to be used at some point io the decision process since the cues and final judgment represent different dimensions.
602 T. KIDA and J. F. SMITH
quent processing (Brainerd & Reyna, 1993).
Therefore, differences in relative retrievability
will increase as stated in the following proposi-
tion.
Proposition 8: As the amount of numerical data
increases, the relative difference in the retrievability
of affective reactions to underlying constructs or broad
data categories will increase as compared to other
forms of encoded representations.
The implication of the foregoing argument is
that the effects noted in previous propositions
concerning the enhanced retrievability of eva-
luative affect will be accentuated as the infor-
mation load increases, and will primarily
concern affective responses to constructs or
data categories. If data must be retrieved for
future decisions, there is a greater likelihood
that overall affective reactions will be primar-
ily retrievable; if numerical values and compar-
isons must be recalled, they are more likely to
be inferred from this general affect, increasing
the potential for errors of reconstruction.
CONCLUDING REMARKS
The foregoing sections outlined the proposi-
tions of a model that concerns the encoding
and retrieval of numerical data for decision
making. We argued that evaluative affect plays
a primary role in encoding and retrieval opera-
tions, and we provided a review of research
from a number of related areas in support of
the propositions advanced. As indicated, when
decision makers observe numerical data, they
typically draw comparisons among the data
and determine evaluative reactions. Accord-
ingly, numerical data may be initially encoded
in these different forms.” Further, within the
affective reaction form, individuals determine
and encode affective reactions to underlying
constructs or broad data categories, where
applicable. While the data may be encoded in
these different forms, the resulting memory
traces are not considered to be equally retrie-
vable. The model indicates that evaluative
affect is the most retrievable form of memory
trace. Also, affective reactions to broad data
categories are viewed as the most accessible
type of affective reaction. In addition, the
model states that confidence in recall is great-
est for evaluative affect. These general charac-
teristics can result in the reconstruction of
memory traces. Given evaluative affect’s
greater retrievability, and decision-makers’ con-
fidence in the accuracy of that affect, the
model indicates directional reconstruction,
positing that numerical values and compari-
sons will be reconstructed from recalled eva-
luative affect. Further, reconstruction can be
affected by a number of decision factors. Spe-
cifically, greater reconstruction is expected (1)
as the time between encoding and retrieval
increases, (2) as the strength of an individual’s
positive/ negative affect increases, and (3) as
information load increases. The model also
addresses additional factors that can affect
retrievability of encoded representations, stat-
ing, for example, that a decision-maker’s objec-
tive for a datum, and the similarity of encoding
and retrieval conditions, can impact the recall
of a memory trace.
Underlying reasons for the importance of
affect include the basic nature of affective
responses and the stage in the decision-making
process in which the reaction occurs. Given
that affective responses are primary human
reactions to external stimuli (Zajonc, 1980)
and that affective reactions can potentially
represent numerous data points in one aggre-
gated response, the use of evaluative affect is,
in effect, one of the most basic simplifying
heuristics that decision makers can utilize to
make decision tasks more manageable. While
this heuristic can facilitate the decision-making
._
“ While this model is specifically concerned with decision-makers’ reactions to numerical data in accounting contexts, we
believe that the underlying issues concerning affective encoded representations discussed in the model can also be
applicable to nonnumerical data.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 603
process, especially in complex tasks, it can also
result in greater susceptibility to judgment
errors. Its effects on decision quality is a pro-
ductive area for future research.
The model developed here concerns the
encoded representations that are used to cap-
ture numerical data in accounting decision-
makers’ memory structures. As such, it
involves basic, underlying reactions to data
that are applicable to decision makers of vary-
ing experience levels. That is, the relevance
and implications of the propositions set forth,
concerning the underlying issues of numerical,
comparative and affective encoded representa-
tions, apply to both experts and novices. How-
ever, experts typically have well-developed
schemata and scripts (Choo, 1989; Christ,
1993; Frederick, 1991). Differences in how
comparisons and affective reactions are deter-
mined by experts as compared to novices may
therefore occur. An experienced decision
maker may be more likely to arrive at affective
responses by comparing numerical values to
internal norms, while a novice, who does not
have the benefit of well-developed schemata,
may be more likely to arrive at affective
responses by drawing relationships and com-
parisons among the data available in the judg-
ment task. The well-developed knowledge
structures of experts can obviously enhance
the quality of decision making (Libby & Freder-
ick, 1990; Bonner, 1990). In some instances,
however, they may be conducive to decision
error (see, for example, Nelson et al., 1995).
Bimberg and Shields (1984) note that accoun-
tants use their expertise to retrieve facts from
memory. If expertise results in greater reliance
on internal schemata previously developed,
there may be a greater likelihood of reconstruc-
tive behavior. This is reflected in Moeckel’s
(1990) finding that experienced auditors
appeared to exhibit more reconstructive beha-
vior when they reviewed a set of workpapers.
As suggested by the model, such reconstruc-
tion may also be more likely if an expert is
more confident in his or her knowledge base.
While many of the model’s propositions are
based upon findings in related psychological
research, most of the propositions, as they spe-
cifically relate to numerical encoding and
retrieving, have yet to be examined. As such,
the model is meant to stimulate and provide a
framework for future research in this area.
Given that much of the information in account-
ing contexts is numerical, the model’s proposi-
tions are potentially relevant to a wide range of
decisions made by both preparers and users of
accounting information. Initially, basic
research must be conducted to examine
whether accounting decision makers encode
numerical information in different representa-
tional forms, recall those representations differ-
entially, and make memorial reconstructions,
as set forth in the model. For example, tests
of recall may be conducted to examine
whether evaluative reactions to basic numeri-
cal data or underlying constructs are retained
in memory longer than the comparisons and
numerical values on which they are based
when auditors assess the financial viability of
a firm, portfolio managers make buy/ sell deci-
sions, or bank loan officers make credit deci-
sions. In addition, recognition tasks may be
conducted to test whether these decision
makers reconstruct numerical values and com-
parisons to be consistent with a recalled affect.
After the basic tenets of the model have been
examined, an important issue to investigate is
the extent to which the encoding and recall
behaviors of accounting decision makers are
functional versus dysfunctional. However, spe-
cific recommendations concerning research
into this issue, and the resulting benefits for
practice, must wait until the results of the
basic research are known. As noted, encoding
numerical data as affective reactions may be an
efficient cognitive heuristic that can aid deci-
sion processes in complex tasks. However,
the retrieval mechanisms specified in the
model may, in some instances, result in mem-
ory errors and possibly flawed decisions. If
basic research uncovers memorial representa-
tions and reconstruction as posited in the
model, future work can examine whether
such behaviors affect decision quality in
accounting contexts, and if so indicated,
604 T. KIDA and J. F. SMITH
decision aids may be developed. For example,
the evidence may suggest a need to focus deci-
sion-makers’ attention on the raw numerical
data related to affective reactions to alterna-
tives observed early in the decision process if
comparisons with alternatives considered later
in the process are required. Basic research may
also provide evidence on the relative benefits
of qualitative versus quantitative response
scales in accounting and auditing contexts
(e.g. the effects of numerical and comparative
risk assessments (see Dilla & Stone, 1991)).
Also note that while basic research on the
encoding and retrieval mechanisms discussed
in this paper is essential at this stage, that
research may affect future work investigating
other judgment and decision-making topics.
For example, the effect that encoded memory
traces have on the decision strategies (e.g. addi-
tive compensatory, eliminations by aspects,
etc.) employed by accounting decision makers
under varying levels of task complexity may be
examined (Payne, 1976,1982; Paquette & Kida,
1988).
Obviously much work needs to be per-
formed to empirically test, further refine and
extend the model. The propositions represent
testable hypotheses that may be examined in a
number of accounting decision-making con-
texts. We believe such work can yield fruitful
results and ultimately lead to a more complete
understanding of accounting decision pro-
cesses involving numerical information.
BIBLIOGRAPHY
Abelson, R. P., Kinder, D. R., Peters, M. D. & Fiske, S. T., Affective and Semantic Components in Political
Person Perception, J ournal of Personal i ty and Soci al Psychol ogy (1982) pp. 619-630.
Ameen, E. C. & Strawser, J. R., Investigating the Use of Analytical Procedures: an Update and Extension,
Audi ti ng: A J ournal of Practi ce and 7beory (1994) pp. 69-76.
Anderson, J, R., Language Memoty and Thought (Hillsdale, NJ: Lawrence Erlbaum Associates, 1976).
Anderson, J, R. & Reder, L., An Elaborative Processing Explanation of Depth in Processing, in Cermak, L.
S. and Craik, F. I. M. (eds), Level s of Processi ng i n Human Memory (Hillsdale, NJ: Lawrence Erlbaum
Associates, 1979).
Arkes, H. R., Costs and Benefits of Judgment Errors: Implications for Debiasing, Psychol ogi cal Bul l eti n
(1991) pp. 486-498.
Arkes, H. R. & Freedman, M. R., A Demonstration of the Costs and Benefits of Expertise in Recognition
Memory, Memory and Cogni ti on (1984) pp. 84-89.
Arkes, H. R. & Harkness, A. R., Effect of Making a Diagnosis on Subsequent Recognition of Symptoms,
Journal of Experi mental Psychol ogy: Human Learni ng and Memory (1980) pp. 568-575.
Asch, S. E., Forming Impressions of Personality, J ournal of Abnormal and Social Psychol ogy (1946) pp.
258-290.
Averill, J, R., Emotions as Episodic Dispositions, Cognitive Schemas, and Transitory Social Roles: Steps
toward an Integrated Theory of Emotion, in Ozer, D. J., Healy, J. M., Jr, & Stewart, A. J. (eds),
Perspecti ves i n Personal i ty: Vol . 3a. Sel fand Emoti on, pp. 137-165 (Greenwich, CT: JAI Press, 1990).
Barclay, J. R., Bransford, J. D., Franks, J. J., McCarreU, N. S. & Nitsch, K., Comprehension and Semantic
Flexibility, J ournal of Verbal Learni ng and Verbal Behavi or (1974) pp. 471-481.
Beach, L. R. & Mitchell, T. R., Image Theory: Principles, Goals, and Plans in Decision Making, Acta
Psycbol ogi ca (1987) pp. 201-220.
Bedard, J. C. & Biggs, S. F., Pattern Recognition, Hypotheses Generation, and Auditor Performance in an
Analytical Task, The Accounti ng Revi ew (1991) pp. 622-642.
Biehal, G. & Chakravarti, D., Information Presentation Format and Task Goals as Determinants of Con-
sumers’ Memory-retrieval and Choice Processes, J ournal of Consumer Research (1982) pp. 431-441.
Biggs, S. F. & Mock, T. J., An Investigation of Auditor Decision Processes in the Evaluation of Internal
Controls and Audit Scope Decisions, J ournal of Accounti ng Research (1983) pp. 234-255.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 605
Biggs, S. F.. Mock, T. J, & Watkins, P. R., Auditor’s Use of Analytical Review in Audit Program Design
(1988) pp. 148-161.
Bimberg, J, & Shields, M., The Role of Attention and Memory in Accounting Decisions, Accounting,
Organfzatfons and Society (1984) pp. 365-382.
Blaney, P. H., Affect and Memory: a Review, Psychological Buffetin (1986) pp. 229-246.
Banner, S. E., Experience Effects in Auditing: the Role of Task Specific Knowledge, The Accounting
Review (1990) pp. 72-92.
Bouwman, M. J., Human Diagnostic Reasoning by Computer: an IIlustration from Financial Analysis,
Management Science (1983) pp. 653-672.
Bouwman, M. J., Frishkoff, P. A., & Frishkoff, P., How Do Financial Analysts Make Decisions? A Process
Model of the Investment Screening Decision, Accounting, Organizations and Socfety (1987) pp. l-29.
Bower, G. H., Mood and Memory, American Psycbologfsst (1981) pp. 129-148.
Bower, G. H., Gilligan, S. G. & Monteiro, K. P., Selectivity of Learning Caused by Affective States,Joumal
of Experimental Psychology: General (1981) pp. 451-473.
Brainerd, C. J. & KIngma, J., Do Children have to Remember to Reason? A Fuzzy-trace Theory of
Transitivity Development, Developmental Review (1984) pp. 3 1 l-377.
Brainerd, C. J. & Kingma, J., On the Independence of Short-term Memory and Working Memory in
Cognitive Development, Cognitive Psychology (1985) pp. 210-247.
Brainerd, C. J. & Reyna, V. F., Generic Resources, Reconstructive Processing, and Children’s Mental
Arithmetic, Developmental Psychology (1988) pp. 324-334.
Brainerd, C. J. & Reyna, V. F., Output-interference Theory of Dual-task Deficits in Memory Development,
J ournal of Experimental Child Psychology (1989) pp. 1-18.
Brainerd, C. J. & Reyna, V. F., Gist Is the Grist: Fuzzy-trace Theory and the New lntuitionism, Develop
mental Review (1990a) pp. 3-47.
Brainerd, C. J. & Reyna, V. F., Inclusion Illusions: Fuzzy-trace Theory and Perceptual Salience Effects in
Cognitive Development, Developmental Review (1990b) pp. 365-403.
Brainerd, C. J. & Reyna, V. F., Memory Independence and Memory Interference in Cognitive Develop
ment, Psycbologfcal Review (1993) pp, 42-67.
Brainerd, C. J., Reyna, V. F., Howe, M. L. & Kingma, J., The Development of Forgetting and Reminiscence,
Monographs of the Society for Research in Child Development No. 222 (1990) pp. 3-4.
Bransford, J. D., Human Cognition: Learning, Understanding and Remembering (Belmont, CA: Wads-
worth, 1979).
Bransford, J. D. & Franks, J. J., The Abstraction of Linguistic Ideas, Cognitfve Psychology (1971) pp. 331-
350.
Bransford, J. D., Franks, J. J., Morris, C. D. & Stein, B. S., Some General Constraints on Learning and
Memory Research, in Cermak, L. S. & Craik, F. I. M. (eds), Leve2.s ofProcessing in Human Memory, pp.
331-354 (Hillsdale, NJ: Lawrence Erlbaum Associates, 1979).
Bransford, J. D. & Johnson, M. K., Contextual Prerequisites for Understanding: Some Investigations of
Comprehension and Recall, J ournal of Verbal Learning and Verbal Behavior (1972) pp. 717-726.
Brewer, W. F. & Treyens, J. C., Role of Schemata in Memory for Places, Cognitive Psychology (1981) pp.
207-230.
Carlston, D. E., The Recall and Use of Traits and Events in Social Inference Processes, J ournal of
Experimental Social Psychology (1980) pp. 303-329.
Chapman, M. & Lindenberger, U., Functions, Operations, and Decalage in the Development of Transi-
tivity, Developmental Psychology (1988) pp. 542-551.
Chase, W. G. & Simon, H. A., Perception in Chess, Cognitive Psychology (1973) pp. 55-81.
Choo, F., Cognitive Scripts in Auditing and Accounting Behavior, Accounting, Organizations and
Society (1989) p. 481.
Christ, M. Y., Evidence on the Nature of Audit Planning Problem Representations: an Examination of
Auditor Free Recalls, The Accounting Review (1993) pp. 304-322.
Clark, D. M. & Teasdale, J. D., Constraints on the Effects of Mood on Memory, J ournal of Personality and
Social Psychofogy (1985) pp. 1595-1608.
Clark, M. S. & Williamson, G. M., Mood and Social Judgments, in Wagner, H. L. & Manstead, A. S. R. (eds),
Handbook of Psychophysiology: Emotion and Social Behavior, pp. 347-370 (Chichester, U.K.: Wiley,
1989).
CoIlins, A. M. & Loftus, E. F., A Spreading-activation Theory of Semantic Memory, Psychofogicaf Review
(1975) pp. 407-428.
606 T. KIDA and J. F. SMITH
Craik, F. I. M. & Jacoby, L. L., Elaboration and Distinctiveness in Episodic Memory, in Nilsson, L. G. (ed.),
Perspectives on Memory Research (Hillsdale, NJ: Lawrence Erlbaum Associates, 1979).
Craik, F. 1. M. & Lockhart, R. S., Levels of Processing: a Framework for Memory Research, J ournal of
Experi mental Psychol ogy. Human Learni ng and Memory (1972a) pp. 671-684.
Craik, F. I. M. & Lockhart, R. S., Levels of Processing: a Framework for Memory Research, J ournal of
Verbal Learni ng and Verbal Behavi or (1972b) pp. 671-676.
Craik, F. I. M. & Tubing, E., Depth of Processing and Retention of Words in Episodic Memory, J ournal of
Experi mental Psychol ogy: General (1975) pp. 268-294.
Dilla, W. N., & Stone, D. N., Information Representation, Scaling, and Experience in Inherent Risk
Judgments, Working Paper, University of Illinois (1991).
Donohew, L., Sypher, H. E. & Cook, P. L., Communication and Affect: Themes and Issues, Ameri can
Behavi oral Sci enti st (1988) pp. 287-295.
Eich, J. E., The Cue-dependent Nature of Statedependent Retrieval, Memory and Cogni ti on (1980) pp.
157-173.
Epstein, S., The Unconscious, the Preconscious, and the Selfconcept, in Suls, J. & Greenwald, A. (eds),
Psychol ogi cal Perspecti ves on the Sev, Vol. 2, pp. 220-247 ( Hi&dale, NJ: Erlbaum, 1983).
Epstein, S., Controversial Issues in Emotion Theory, in Shaver, P. (ed.), Revi ew of Personal i ty and
Soci al Psychol ogy: Vol . 5. Emoti ons, Rel ati onshi ps, and Heal th, pp. 64-88 (Beverly Hi&, CA: Sage,
1984).
Fiedler, K., Emotional Mood, Cognitive Style, and Behavior Regulation, in Fiedler, K. & Forgas, J. P. (eds),
Affect, Cogni ti on, and Soci al Behavi or, pp. 100-l 19 (Toronto, Canada: Hogrefe, 1988).
Fiedler, K., Pampe, H. & Scherf, U., Mood and Memory for Tightly Organized Social Information, Eur-
opean J ournal of Soci al Psychol ogy (1986) pp. 149-164.
Fisher, R. P. & Craik, F. I. M., Interaction Between Encoding and Retrieval Operations in Cued Recall,
J ournal of Experi mental Psychol ogy: Human Learni ng and Memory (1977) pp. 701-711.
Fiske, S. T. & Taylor, S. E., Soci al Cogni ti on (New York: McGraw-Hill, Inc., 1991).
Frederick, D., Auditors’ Representation and Retrieval of Internal Control Knowledge, The Accounti ng
Revi ew (1991) pp. 240-258.
Ford, J. K., Schmitt, N., Schechtman, S. L., Hults, B. M. & Doherty, M. L., Process Tracing Methods:
Contributions, Problems, and Neglected Research Questions, Organi zati onal Behavi or and Human
Deci si on Processes (1989) pp. 75- 117.
Frensch, P. A. & Geary, D. C., Effects of Practice on Component Processes in Complex Mental Addition,
J ournal of Experi mental Psychol ogy: Learni ng, Memory, and Cogni ti on (1993) pp. 433-456.
Geisehnan, R. E. & Glenny, J., Effects of Imagining Speakers’ Voices on the Retention of Words Presented
Visually, Memory and Cogni ti on (1977) pp. 499-504.
Graesser, A. & Nakamura, G., The Impact of a Schema on Comprehension and Memory, Psychol ogy of
Learni ng and Moti vati on (1982) pp. 60-109.
Graf, P. & Mandler, G., Activation Makes Words More Accessible, but Not Necessarily More Retrievable,
J ournal of Verbal Learni ng and Verbal Behavi or (1984) pp. 553-568.
Graf, P., Mandler, G. & Haden, P., Simulating Amnesic Symptoms in Normal Subjects, Science (1982) pp.
1243-1244.
Halpin, J. A., Puff, C. R., Mason, H. F. & Marston, S. P., Self-reference and Incidental Recall by Children,
Bul l eti n of the Psycbonomi c Soci ety (1984) pp. 87-89.
Hamilton, D. L., Cognitive Representations of Persons, in Higgins, E. T., Herman, C. P. & Zanna, M. P.
(eds), Soci al Cogni ti on; The Ontari o Symposi um, Vol 1, pp. 135-159 (Hillsdale, NJ: Lawrence Erlbaum
Associates, 1981).
Hamilton, D. L., Katz, L. B. & Leirer, V. O., Cognitive Representation of Personality Impressions: Orga-
nizational Processes in First Impression Information, J ournal of Personal i ty and Soci al Pvcbol ogy
(1980a) pp. 1050-1063.
Hamilton, D. L., Katz, L. B. & Leirer, V. O., Organizational Processes in Impression Information, in Hastie,
R., Ostrom, T. M., Ebbesen, E. B., Wyer, R. S., Hamilton, D. L. & Carlston, D. E. (eds), Person Memory:
Z’be Cognitive Basi s of Soci al Percepti on, pp. 121-153 (H&dale, NJ: Lawrence Erlbaum Associates,
1980b).
Harris, R. J., Inferences in Information Processing, in Bower, G. H. (ed.), The Psychol ogy of Learni ng and
Moti vati on, Vol . 15, (New York: Academic Press, 1981).
Hartwick, J., Memory for Trait Information: a Signal Detection Analysis, J ournal of Experi mental Soci al
Psychol ogy (1979) pp. 533-552.
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA 607
Higgins, E. T., The “Communication Game”: Implications for Social Cognition, in Higgins, E. T., Herman,
C. P. & Zanna, M. P. (eds), Soci al Cogni ti on: The Ontari o Symposi um, Vol 1, pp. 343-392, (Hillsdale,
NJ: Lawrence Erlbaum Associates, 1981).
Higgins, E. T. & Lurle, L., Context, Categorization, and Recall: The “Change-of-Standard” Effect, Cognftfve
Psychol ogy (1983) pp. 525-547.
Higgins, E. T. & McCaM, C. D., Social Encoding and Subsequent Attitudes, Impressions, and Memory:
“Context-driven” and Motivational Aspects of Processing, J ournal of Personal i ty and Soci al Psychol -
ogy (1984) pp. 26-39.
Higgins, E. T. & Rholes, W. S., “Saying is Believing”: Effects of Message Modification on Memory and
Liking for the Person Described, J ournal of Experfmeni al Soci al Psychol ogy (1978) pp. 363-378.
Hogarth, R., J udgment and Choi ce, 2nd edn, (New York: John Wiley and Sons, 1987).
Horton, D. L. & Mills, C. B., Human Learning and Memory, Annual Revi ew ofPsychol ogy (1984) pp. 361-
394.
Hyde, T. W. &Jenkins, J. J., The Differential Effects of Incidental Tasks on the Organization of Recall of a
List of Highly Associated Words, J ournal of Experfmenl al Psychol ogy (1969) pp. 472-481.
Isen, A. M., Toward Understanding the Role of Affect in Cognition, in Wyer, R. S. & Srull, T. K. (eds),
Handbook of Soci al CognfUon, pp. 179-236 (Hillsdale, NJ: Lawrence Erlbaum Associates, 1984).
Isen, A. M., Positive Affect, Cognitive Processes, and Social Behavior, in Berkowitz (ed.), Advances i n
Experi mental Soci al Psychol o~, Vol . 20, pp. 203-253 (New York: Academic Press, 1987).
Isen, A. M. & Means, B., The Influence of Positive Affect on Decision-making Strategy, Soci al Cogni ti on
(1983) pp. 18-31.
Izard, C. E., Human Emoti ons (New York: Plenum Press, 1977).
Izard, C. E., On the Development of Emotions and Emotioncognition Relationships in Infancy, in Lewis,
M. & Rosenblum, L. (eds), The Devel opment of Affect (New York: Plenum Press, 1978).
Izard, C. E., Emotions as Motivations: An Evolutionary Developmental Perspective, in Dienstbier, R. (ed.),
Nebraska Symposi um on Moti uai i on, Vol . 27, (Lincoln: University of Nebraska Press, 1979).
Katz, A., Self-reference in the Encoding of Creative-relevant Traits, J ournal of Personal i ty (1987) pp. 97-
120.
Klein, S. B. & Loftus, J., Rethinking the Role of Organization in Person Memory: an Independent Trace
Storage Model, J ournal of Personal i ty and Soci al Psychol ogy (1990) pp. 400-410.
Koriat, A. & Melkman, R., Depth of Processing and Memory Organization, Psychol ogi cal Research (1987)
pp. 183-188.
Lazarus, R. S., Thoughts on the Relations between Emotion and Cognition, Ameri can Psychol ogi st (1982)
pp. 1019-1024.
Lazarus, R. S., On the Primacy of Cognition, Ameri can Psychol ogi st (1984) pp. 124-129.
Lazarus, R. S., Constructs of the Mind in Adaptation, in Stein, N. L., Leventhal, B. & Trabasso, T. (eds),
Psychol ogi cal and Bi ol ogi cal Approaches to Emoti on, pp. 3-20 (Hillsdale, NJ: Erlbaum, 1990).
Libby, R., Experimental Research and the Distinctive Features of Accounting Settings, University of
Illinois Golden Jubilee Symposium (1989).
Libby, R. & Frederick, D., Experience and the Ability to Explain Audit Findings, J ournal of Accounti ng
Research (1990) pp. 348-367.
Libby, R. & Trotman, K., The Review Process as a Control for Differential Recall of Evidence in Auditor
Judgments, Accounti ng, Otgani zatfons and Soci ety (1993) pp. 559-574.
Lingle, J. H. & Ostrom, T. M., Retrieval Selectivity in Memory-based Judgments, J ournal of Personal i ty
and Soci al Psychol ogy (1979) pp. 180-194.
Lingle, J. H. & Ostrom, T. M., Principles of Memory and Cognition in Attitude Formation, in Clark, M. S. &
Fiske, S. T. (eds), Affect and Cogni ti on: The Seventeenth Annual Carnegi e Symposi um on Cogni Non
(Hi l l sdal e, NJ: Lawrence Erlbaum Associates, 1981).
Loftus, E. F. & Loftus, G. R., On the Permanence of Stored Information in the Human Brain, Ameri can
Pvcbol ogfst (1980) pp. 409-420.
Angie, R. H. & Baddeley, A. D., Cognitive Processes in Counting, J ournal of Experi mental Psychol ogy:
Learni ng, Memory, and Cogni ti on (1987) pp. 310-326.
Lynch, J. G., Jr, & Srull, T. K., Memory and Attentional Factors in Consumer Choice: Concepts and
Research Methods, J ournal of Consumer Research (1982) pp. 18-37.
Matlin, M. H., Cogni ti on (New York: Holt, Rinehart and Winston, Inc., 1989).
Mayer, J. D., How Mood Intluences Cognition, in Sharkey, N. E. (ed.), Advances i n Cognitive Sci ence, pp.
290-314 (Chichester, U.K.: Ellis Horwood Limited, 1986).
608 T. KIDA and J. F. SMITH
Mayer, J. D. & Salovey, P., Personality Moderates the Interaction of Mood and Cognition, in Fiedler, K. &
Forgas, J. (eds), Affect, Cognition, and Social Behavior, pp. 87-99 (Toronto, Canada: Hogrefe, 1988).
McCloskey, M., Harley, W. & Sokol, S. M., Models of Arithmetic Fact Retrieval: an Evaluation in Light of
Findings From Normal and Brain-damaged Subjects, JoumaE of Experimental Psychology: Learning,
Memory, and Cognition (1991) pp. 377-397.
Miller, G. A., The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing
Information, Psychological Review (1956) pp. 81-97.
Mitchell, T. R. & Beach, L. R., “. Do I Love Thee? Let Me Count .” Toward an Understanding of
Intuitive and Automatic Decision Making, Organizational Behavior and Human Decision Processes
(1990) pp. l-20.
Moeckel, C., The Effect of Experience on Auditors’ Memory Errors, Journal ofilccounti~g Research
(1990) pp. 368-387.
Moeckel, C. L. & Plumlee, D., Auditors’ Confidence in Accurate and Inaccurate Recognition of Audit
Evidence, Accounting Reutew (1989) pp. 653-666.
Morris, C. D., Bransford, J. D. & Franks, J. J., Levels of Processing Versus Transfer Appropriate Processing,
J ournal of Verbal Learning and Verbal Behavior (1977) pp. 519-533.
Nelson, D. L., Wheeler, J. W., Jr, Borden, R. C. & Brooks, D. H., Levels of Processing and Cuing: Sensory
Versus Meaning Features, J ournal of Experimental Psychology: General (1974) pp. 971-977.
Nelson, M., Libby, R. & Bonner, S. E., Knowlwedge Structures and the Estimation of Conditional Prob-
abilities in Audit Planning, The Accounting Review (1995) pp. 27-48.
Newell, A. & Simon, H. A., Human Problem Solving (Englewood Cliffs, NJ: Prentice-Hall, 1972).
Ostrom, T. M., Lingle, J. H.. Pryor, J. 8. & Geva, N., Cognitive Organization of Person Impressions, in
Hastie, R., Ostrom, T. M., Ebbesen, E. B., Wyer, R. S., Hamilton, D. L. & Carlston, D. E. (eds). Person
Memory: The Cog&the Basis ofsocial Perception (Hillsdale, NJ: Lawrence Erlbaum Associates, 1980).
Paquette, L. & Kida, T., The Effect of Decision Strategy and Task Complexity on Decision Performance,
Organizational Behavior and Human Decision Processes (1988) pp. 128-142.
Parkin, A. J., Levels of Processing, Context, and Facilitation of Pronunciation, Acta Psycbologica (1984)
pp. 19-29.
Payne, J. W., Task Complexity and Contingent Processing in Decision Making: an Information Search and
Protocol Analysis, Organizational Behavior and Human Performance (1976) pp. 366-387.
Payne, J, W., Contingent Decision Behavior, Psychological Bulletin (1982) pp. 382-402.
Pentield, W., Consciousness, Memory, and Man’s Conditioned Reflexes, in Pribram, K. (ed.), On the
Biology of Learning (New York: Harcourt, Brace and World, 1969).
Posner, M. 1. & Snyder, C. R. R., Facilitation and Inhibition in the Processing of Signals, in Rabbitt, P. M. A.
& Domic, S. (eds), Attention and Performance V, pp. 669-682 (New York: Academic Press, 1975).
Raaijmakers, J. G. W. & Shiffrin, R. M., Search of Associative Memory, Psychological Review (1981) pp.
93-134.
Reimers, J., Wheeler, S. & Dusenbury, R., The Effect of Response Mode on Auditors’ Control Risk
Assessments, Auditing: A J ournal of Practice and Theory (1993) pp. 62-78.
Reitman, J. S., Skilled Perception in Go: Deducing Memory Structures from Inter-response Times, Cog-
nitive Psychology (1976) pp. 336-356.
Reyna, V. F., Reiterative versus Reconstructive Memory in Opinion Change, Presented at the 1st Annual
Nags Head Conference onJ udgement and Decision Making (Kill Devil Hill, NC, 1988).
Reyna, V. F., Class inclusion, the Conjunction Fallacy, and Orher Cognitive Illusions, Developmental
Review (1991) pp. 317-336.
Reyna, V. F. & Brainerd, C. J., On the Relationship between Memory and Judgement in Opinion Change,
Presented at the 28th Annual Meeting of the Psycbonomic Society (Seattle, WA, 1987).
Reyna, V. F. & Brainerd, C. J., Fuzzy Processing in Transitivity Development, Annals of Operations
Research (1990) pp. 37-63.
Reyna, V. F. & Brainerd, C. J., Fuzzy-trace Theory and Framing Effects in Choice: Gist Extraction,
Truncation, and Conversion, J ournaI of Behavioral Decision Making (1991) pp. 249-262.
Reyna, V. F., Brainerd, C. J. & Connolly, T., Just the Bottom Line, Please: A Fuzzy-trace Theory of Framing
Effects in Choice, Presented at the Fifth I nternational Conference on the Foundation and Applica-
tions of Utility, Risk, and Decision Theories (Duke University, Durham, NC, 1990).
Rogers, T. B., Emotion, I magery, and Verbal Codes: A Closer Look at an Increasingly Complex Interaction,
in YuiIle, J. (ed.), Imagery, Memory, and Cognition, pp. 285-365 (U.K.: Open University Press, 1983).
THE ENCODING AND RETRIEVAL OF NUMERICAL DATA
609
Rogers, T. B., Kuiper, N. A. St Kirker, W. S., Self-reference and the Encoding of Personal Information,
J ournal of Personal i ty and Soci al Psychol ogy (1977) pp. 677-688.
Salovey, P. & Singer, J. A., Mood Congruency Effects in Recall of Childhood versus Recent Memories,
J ournal of Soci al Behavi or and Personal i ty (1988) pp. 1-22.
Schacter, D. L., The Psychology of Memory, in LeDoux, J. E. & Hirst, W. (eds), Mi nd and Brai n:
Di al ogues i n Cogni ti ve Neurosci ence, pp. 189-214 (New York: Cambridge University Press,
1986).
Sherman, S. J., Ahlm, K., Berman, L. & Lynn, S., Contrast Effects and Their Relationship to Subsequent
Behavior, J ournal of Experi mental Soci al Psychol ogy (1978) pp. 340-350.
Smith, E. E., Adams, N. & Schorr, D., Fact Retrieval and the Paradox of Interference, Cogni ti ve PsychoZ-
ogy (1978) pp. 438-464.
Smith, S. M., Glenberg, A. & Bjork, R. A., Environmental Context and Human Memory, Memory and
Cogni ti on (1978) pp. 342-353.
Sokol, S. M., McCloskey, M., Cohen, N. J, & Aliminosa, D., Cognitive Representations and Processes in
Arithmetic: Inferences From the Performance of Brain-damaged Subjects, J oumaI of Experi mental
Psychol ogyc Learni ng, Memory, and Cogni ti on (19’91) pp. 355-376.
Spiro, R. J., Accommodative Reconstruction in Prose Recall, J ournal of VerbaI Learni ng and Verbal
Behavi or (1980) pp. 84-95.
Squire, L., Memory and Brai n (New York: Oxford University Press, 1987).
Srull, T. K., Organizational and Retrieval Processes in Person Memory: an Examination of Processing
Objectives, Presentation Format, and the Possible Role of Self-generated Retrieval Cues, J ournal of
Personal i ty and Soci al Psychol ogy (1983) pp. 115‘7- 1170.
Srull, T. K. & Brand, J. F., Memory for Information About Persons: the Effect of Encoding Operations on
Subsequent Retrieval, J ournal of Verbal Learni ng and Verbal Behavi or (1983) pp. 219-230.
Srull, T. K. & Wyer, R. S., Jr, The Role of Chronic and Temporary Goals in Social Information Processing,
in Sorrentino, R. M. & Higgins, E. T. (eds), Handbook of Moti vati on and Cogni ti on, pp. 503-549
(New York: Guilford Press, 1986).
Sruil, T. K. & Wyer, R. S., Jr, Person Memory and Judgment, Psychol ogi cal Revi ew (1989) pp. 58-83.
Sulin, R. A. & Dooling, D. J., Intrusion of a Thematic Idea in Retention of Prose, J ournal OfExperi mental
Psychol oey (1974) pp. 255-262.
Tesser, A., Self-generated Attitude Change, Advances i n Experi mental Soci al Psychol ogy (1978) pp. 289-
338.
Thomson, D. M. & Tubing, E., Associative Encoding and Retrieval: Weak and Strong Cues, J ournal of
Experi mental Soci al Psychol ogy (1970) pp. 255-262.
Tsujimoto, R. N., Memory Bias Toward Normative and Novel Trait Prototypes, J ournal of Personal i ty
and Soci al Psychol ogy (1978) pp. 1391-1401.
Tubing, E., El ements of Epi sodi c Memory (New York: Oxford University Press, 1983).
Tubing, E. & Pearlstone, Z., Availability Versus Accessibility of Information in Memory for Words,
J ournal of Verbal Learni ng and Verbal Behavi or ~(1966) pp. 381-391.
Tubing, E. & Thomson, D. M., Encoding Specificity and Retrieval Processes in Episodic Memory, Psy-
chol ogi cal Revi ew (1973) pp. 352-373.
Tversky, A. & Kahneman, D., Judgment Under IJncertainty: Heuristics and Biases, Science (1974) pp.
1124-1131.
Walker, N., Jones, J. P. & Mar, H. H., Encoding Process,es and the Recall of Text, Memory and Cogni ti on
(1983) pp. 275-282.
Wailer, W. & Felix, W., The Auditor and Learning from Experience: Some Conjectures, Accounti ng,
Organi zati ons and Soci ety (1984) pp. 383-408.
Westbrook, R. A., Product/ Consumption-based Affective Responses and Postpurchase Processes, J ournal
of Marketi ng Research (1987) pp. 258-270.
Widaman, K. F., Geary, D. C., Cormier, P. & Little, T. D., A Componential Model for Mental Addition,
J ournal of Experi mental Psychol ogy: Learni ng, Memory, and Cogni ti on (1989) pp. 898-919.
Wyer, R. S. & Gordon, S. E., The Recall of Information About Persons and Groups, J ournal of Experi -
mental Soci al Psychol ogy (1982) pp. 128164.
Wyer, R. S. & Gordon, S. E., The Cognitive Representation of Social Information, in Wyer, R. S. & Srull, T.
K. (eds), Handbook of Soci al Cogni ti on, pp. 73-150 (Hillsdale, NJ: Lawrence Erlbaum Associates,
1984).
610 T. KIDA and J. F. SMITH
Wyer, R. S., Bodenhausen, G. V. & SrnlI, T. K., The Cognitive Representation of Persons and Groups and
Its Effect on Recall and Recognition Memory, J ournal ofExperimental Social Psychology (1984) pp.
445-469.
Wyer, R. S., Jr, & SrnR, T. K., Human Cognition in Its Social Context, Psycbologicul Review (1986) pp.
322-359.
Zajonc, R. B., Feeling and Thinking: Preferences Need No Inferences, American Psychologist (1980) pp.
151-175.
Zajonc, R. B., On the Primacy of Affect, American Psychologist (1984) pp. 117-123.
Zajonc, R. B., Pietrmonaco, P. & Bargh, J., Independence and Interaction of Affect and Cognition, in
Clark, M. S. and Fiske, S. T. (eds), Affect and Cognition: The Seventeenth Annual Carnegie Sytnpo-
sium on Cognition, pp. 21 l-227 @BUsdale, NJ: Lawrence Erlbaum Associates, 1982).

doc_405733775.pdf