Description
Sound symbolism, the linguistic process in which the sounds of a word provide cues about the word's meaning, is not a new phenomenon. Plato first described the effect in the dialogues of Cratylus, and authors throughout time have used the sounds in words to describe people.
43
? 2004 by JOURNAL OF CONSUMER RESEARCH, Inc. ? Vol. 31 ? June 2004
All rights reserved. 0093-5301/2004/3101-0004$10.00
A Sound Idea: Phonetic Effects of Brand
Names on Consumer Judgments
ERIC YORKSTON
GEETA MENON*
In this article we examine a phenomenon known as sound symbolism, where the
sound of a word conveys meanings. Speci?cally, brand names are composed of
individual sounds called phonemes and we investigate how this phonetic structure
of brand names affects a consumer’s evaluation of products and their underlying
attributes. We demonstrate that consumers use information they gather from pho-
nemes in brand names to infer product attributes and to evaluate brands. We also
demonstrate that the manner in which phonetic effects of brand names manifest
is automatic in as much as it is uncontrollable, outside awareness and effortless.
S
ound symbolism, the linguistic process in which the
sounds of a word provide cues about the word’s mean-
ing, is not a new phenomenon. Plato ?rst described the effect
in the dialogues of Cratylus, and authors throughout time
have used the sounds in words to describe people (e.g., the
miniature Lilliputians and the giant Brobdingnagians in
Swift’s Gulliver’s Travels) and objects (e.g., the large, dan-
gerous Bludgers, the big, round Quaf?e, and the small, fast
Golden Snitch in the Quidditch game in Rowling’s Harry
Potter and the Sorcerer’s Stone) represented by those words.
When presented with ?ctitious or unfamiliar words, indi-
viduals consistently use sound symbolism to interpret mean-
ings from the name about the referenced object (Jacobson
and Waugh 1987; Sapir 1929). The effect is extremely well
documented whether the tested individual’s native language
is English (Klink 2000, 2001), Chinese (Klank, Huang, and
Johnson 1971; Lapolla 1994), or even Navajo (Sapir 1929).
In fact, sound symbolism has been observed to exist in
native languages in North America, Latin America, Asia,
Australia, and Africa, as well as more developed languages
such as English, Finnish, French, German, Modern Greek,
and Japanese (for a complete discussion of these examples
see Hinton, Nichols, and Ohala 1994).
*Eric Yorkston is assistant professor of marketing, University of South-
ern California, Los Angeles, CA 90089-0443 (e-mail: yorkston@marshall
.usc.edu). Geeta Menon is Harold MacDowell Faculty Fellow and associate
professor of marketing, New York University, New York, NY 10012-1126
(e-mail: [email protected]). This article is based on the doctoral dis-
sertation of the ?rst author under the supervision of the second. The authors
thank members of the dissertation committee, Adamantios Gafos, Eric
Greenleaf, and Vicki Morwitz, for their helpful comments during the dif-
ferent stages of this project. The authors acknowledge the data collection
assistance provided by Sucharita Chandran, Andrea Morales, and Manoj
Thomas. In addition, the authors are especially grateful to the JCR editors,
Dawn Iacobucci and David Mick, the associate editor, and three reviewers
for their constructive feedback during the review process.
Past research on sound symbolism has focused on two
aspects: the range of the effect (i.e., its universality; Huang,
Pratoomaraj, and Johnson 1969) and the aspects of meaning
it affects (Klink 2000, 2001; Ohala 1994). Less is known
about the sound symbolism process itself. For instance, re-
search has examined how linguistic classi?ers affect cate-
gorization schema (Schmitt and Zhang 1998), and how an
individual’s formal writing system(i.e., logographic vs. pho-
netic) affects processing (Schmitt, Pan, and Tavassoli 1994;
Tavassoli and Han 2001). Yet we do not know how or why
individuals use sound symbolismas a source of object mean-
ing. Current research has neither revealed whether this is a
controlled strategy or an automatic process nor has it ex-
amined conditions when sound symbolism is utilized. The
effect is well established, but we still lack understanding of
the process by which it occurs.
Sound symbolism has been recently recognized as an im-
portant factor in howindividuals infer speci?c meaning from
unfamiliar brand names (Klink 2000, 2001). A brand name
is composed of individual sounds called phonemes. These
phonemes serve two purposes. First, they are the basic build-
ing blocks of language and are combined to form syllables
and, in turn, words. These words and the syllables from
which they are composed supply what we traditionally think
of as the meaningful units of a brand name. Second, pho-
nemes can provide meaning themselves through sound sym-
bolism. These sounds provide cues about how the brand
may perform on particular attribute dimensions. We theorize
that if a brand name has phonemes that represent attributes
a consumer desires, consumers will hold more positive at-
titudes and exhibit higher purchase intentions toward that
brand. The current article provides a demonstration of the
process by which sound symbolism manifests in consumer
judgments. We demonstrate that the process is uncontrol-
lable, outside of awareness, and effortless, making it au-
44 JOURNAL OF CONSUMER RESEARCH
tomatic (Bargh 1989). By understanding the underlying, the-
oretical processes of sound symbolism, we achieve greater
understanding of how consumers interpret brand names and
use them to evaluate brands.
Like Klink (2000, 2001), we examine the effects of
sounds symbolism on attribute perceptions, but in addition,
in our research we examine their effects on brand evalua-
tions as well, using a more subtle between-subjects design.
Unlike Klink (2000, 2001), our primary objective is to de-
lineate the automaticity of the underlying process by which
phonetic effects occur.
CONCEPTUAL FRAMEWORK
Sounds, and the resulting aural frequencies, are based
upon the position and curvature of the tongue in the mouth,
ranging from a high-front to low-back position (see Klink
2000 for a review). The affective meanings generated by
sound symbolism follow a similar pattern. These vowel
sounds roughly form an ordered sound-symbolic list: [e¯],
, [e], [a¯], [a], [o¯], [o], [a¨], , and [u¨] (e.g., beat, bit, bet,
bait, bat, boat, bought, posh, but, put, and boot). The ex-
istence of this hierarchy is fairly consistent across languages
(Makino, Nakada, and Ohso 1999). Sound symbolism con-
veys information such that high-front vowels (e.g., ee in
?ea and i in ?y) represent associations with smaller size and
less power than low-back vowels (e.g., the ow in bout and
oo in boot), which, in turn, connote greater size, and more
power (Hinton et al. 1994; Makino et al. 1999). In an em-
pirical demonstration, Klink (2000) showed that the use of
front vowels (as opposed to the back vowels) in brand names
conveys attribute qualities of smallness, lightness, mildness,
thinness, fastness, coldness, bitterness, femininity, weak-
ness, lightness, and prettiness. Furthermore, Klink (2001)
demonstrated the interactive effects of sound symbolism
with semantic meaning in affecting consumers’ evaluations
of various brands.
Consistent with Klink (2000, 2001), we expect that when
consumers encounter a brand name, they infer attribute
meaning. In testing our theory, we used the longer, broader
[a¨] sound (as in bother and chop) from the middle of the
phonetic vowel scale and contrasted it against the shorter,
tighter sound (as in kiss and nymph) from the lighter,
sharper, smaller, end of the sound-symbolism scale. The
derivation of the pair of sounds that we use in our experi-
ments is a function of the theory associated with the position
of the tongue in the mouth (O’Grady, Dobrovolsky, and
Katumba 1997). The chosen pair of vowel sounds we test
( and [a¨]) are not the extremes on either end of this scale.
For example, the sounds ee and oo lie nearer the two re-
spective endpoints of the phonetic-symbolic spectrum. Thus,
the vowel pairing we use (i.e., and [a¨]) may be considered
a more conservative test of the sound symbolismhypothesis.
An additional consideration in our decision to test only one
sound pair was our strong desire to control for latent se-
mantic associations that could interfere with our studies (a
discussion of how we controlled for semantic associations
is below). Since the goal of our research was not merely to
provide further support for the phenomenon of sound sym-
bolism but to understand the process by which sound sym-
bolism affects meaning, we decided that the advantages
gained by ruling out this alternative hypothesis outweighed
the costs of limiting the range of tested vowels. Formally,
our baseline hypothesis is
H1: Consumers will evaluate the individual attributes
of brand names with the phonetic [a¨] sound heav-
ier than attributes of brand names with the pho-
netic sound.
Stated simply, sound symbolism affects attribute percep-
tions. This raises the possibility that consumers use the brand
name as a phonetic cue regardless of perceived value of this
information. We are now left with the question of when
exactly consumers are affected by the phonetics of brand
names. We posit that consumers attempt to strategically use
the brand name as information when they perceive it to
supply diagnostic information. Alternately, consumers will
attempt to ignore, and adjust for, the phonetic meaning in
situations where the brand name is perceived as less diag-
nostic. Thus,
H2: The effects of brand names on attribute percep-
tions will be moderated by the perceived diag-
nosticity of the brand name, such that, the pho-
netic effects of brand names will manifest only
when the brand name is described as the true ver-
sus the test name.
It has been documented recently that a large number of
consumer decisions are nonconscious (see Bargh 2002 for
a review). Bargh (1989) asserts that a process may have one
or more of the four automatic criteria to be differentiated
from a conscious or controlled process; that is, a process is
automatic if it is effortless, unintentional, used outside of
awareness, or uncontrollable. We investigate the automatic-
ity of sound symbolism effects in this article.
If sound symbolism manifests automatically and is used
in an uncontrollable fashion, then participants who are told
that the phonetic information is nondiagnostic after they
have experienced the brand name information should not
be able to completely discount the phonetic information.
This is because if the process is partly automatic, people
are either unaware of or unable to control for the incor-
poration of phonetic effects in their judgments (Bargh 1989,
2002). On the other hand, if the experienced effects of sound
symbolism are discounted at the time it is felt, then it should
not enter into the judgment. Speci?cally,
H3: The timing of information about the diagnosticity
of brand names will moderate the effects of sound
symbolism on attribute perceptions, such that:
a) When this information is provided at the same
time as the brand name information (i.e., sound
symbolism can be discounted while being ex-
perienced), the phonetic effects will be erased
in the low diagnosticity condition, and will
manifest in the high diagnosticity condition (as
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 45
predicted in hypothesis 2).
b) When this information is provided after the
brand name information (i.e., sound symbolism
has been experienced), the basic phonetic effects
will manifest in both the low and high diag-
nosticity conditions (as predicted in hypothesis
1).
It is also possible that phonetic effects manifest effort-
lessly. Gilbert (1989) suggests that the automatic process in
judgment formation is a two-stage process comprising of
the initial anchor based on an automatic input, and the sub-
sequent correction (e.g., incorporating ignored inputs, or
correcting weights) performed in a more controlled manner.
For example, in experiments on stereotype activation and
use, Gilbert and Hixon (1991) found that the application of
an activated stereotype was moderated by the availability
of cognitive resources. Those with capacity constraints were
less likely to engage in corrective adjustments to the au-
tomatically activated stereotypes.
If an automatic source of information is used (i.e., the
sound symbolism of the brand name) together with a con-
trolled input (i.e., information about the diagnosticity of the
brand name), then the automatic input will have a greater
effect when cognitive resources are constrained than when
they are abundantly available (Bargh and Thein 1985; Gil-
bert, Pelham, and Krull 1988). With increased cognitive
load, information that is automatically processed will have
a proportionately greater impact on judgments as many of
the conscious, effort-requiring adjustments will not be pos-
sible (Bargh and Thein 1985; Gilbert 1989). In the current
study, we use Gilbert’s load paradigm to reverse the mod-
erating effect of diagnosticity and replicate the basic pho-
netic effect when cognitive constraints are imposed. Thus,
H4: Cognitive capacity will moderate the effects of
sound symbolism on attribute perceptions, such
that:
a) Under conditions of normal cognitive capacity,
the phonetic effects will be erased in the low
diagnosticity condition, and will manifest in
the high diagnosticity condition (as predicted
in hypothesis 2).
b) Under conditions of impaired cognitive capac-
ity, the diagnosticity of the brand name will
not have an effect and the basic phonetic ef-
fects will manifest (as predicted in hypothesis
1).
We now present two studies that were designed to test
these hypotheses.
STUDY 1
Method
Choice of Stimuli. We chose ice cream as the product
category for the studies reported in this article, as it produced
high interest and involvement among the student population,
was likely to be purchased in the near future, and elicited
product evaluations that incorporated a small number of
attributes. An initial pretest ( ) established that ice n p48
cream attributes consisted of three main factors: taste, cost,
and calorie content. Since taste was reported as the most
primary attribute and is a fairly ambiguous one, we further
explored the underlying factors of taste. When participants
were asked to elaborate upon taste, the attributes of smooth-
ness, creaminess, richness, and sweetness emerged. Our
?ndings were corroborated by a Consumer Reports (1994)
article on ice cream that discussed taste in their overall brand
comments using the terms “creamy,” “rich,” “smooth,” and
“sweet.” Our pretest, however, indicated that although peo-
ple, in general, preferred creamier, richer, and smoother ice
creams, preference for sweetness was more divided. There-
fore, the ?nal attributes we used were smoothness, cream-
iness, and richness, three attributes that consumers prefer
more of, and that may be communicated by the sound sym-
bolism of the brand name.
The ?rst goal in designing the brand names for the studies
was to develop two names that were ?ctitious and either
avoided or controlled for other linguistic complexities. Since
we were examining the symbolic differences between the
[a¨] and sounds, we held all consonants in the words
constant, avoiding hard consonants, since their rough sound
symbolizes attributes that contradict those desired in our
product category of ice cream (Schloss 1981; Vanden Bergh
et al. 1984).
The second goal was to have two names that were con-
sidered similar. Besides balancing the names phonetically,
we pretested ?ve pairs of names that met the above phonetic
requirements. Forty-four undergraduate participants ranked
the 10 names, with one being the name most likely to be
an ice cream name and 10 being the name least likely to
be an ice cream. Frosh and Frish best met the criterion of
being equally likely to be an ice cream name (M’s p
and 3.55, respectively, NS). These were also the most 3.33
preferred mean rankings among all the product names.
Based on this pretest and our research goals, we chose
the names Frosh and Frish differing only on the phonetic
sounds [a¨] and . In general, the [a¨] sound is associated
with perceptions of objects being bigger, heavier, slower,
and duller. The sound, on the other hand, is associated
with perceptions of the object being smaller, lighter, livelier,
and sharper. Therefore, in the context of ice creams, we
expect the [a¨] sound in Frosh to communicate a smoother,
creamier, and richer ice cream than the sound in Frish.
To ensure that a phonetic distinction manifesting through
affective meaning drove the evaluations of the two brand
names, we examined the denotative meaning evoked by
these names through another pretest. We asked 56 partici-
46 JOURNAL OF CONSUMER RESEARCH
pants to list the ?rst three words that came to their mind
when they thought of the ice cream name. Acontent analysis
was performed to ensure that there were no common word
associations that could be moderating the brand name eval-
uations. Two independent coders blind to the hypotheses
classi?ed the resulting 167 words as positive, negative, or
neutral when describing an ice cream. Coders’ classi?cations
matched in 84.4% of the cases. Disagreement between cod-
ers was settled through discussion. Frosh and Frish gener-
ated equivalent numbers of positive (30 vs. 28), negative
(10 vs. 13), and neutral (43 vs. 43) semantic associations
( , NS). Therefore, as desired, the two names were
2
x ! 1
well matched in terms of semantic associations. In general,
of all the semantic associations the names generated, 34.7%
were positive (e.g., frosty and fresh), 51.5% were neutral
(e.g., name and Irish), and 13.8% were negative (e.g., frog
and ?sh).
Design. We used a 2 (sound symbolism of brand name)
# 2 (diagnosticity of brand name) # 2 (timing of diag-
nosticity information) between-subjects design. Sound sym-
bolism was manipulated using Frish and Frosh. We manip-
ulated the diagnosticity of the brand name by informing
participants that the brand name was either a ‘true’ name
(high diagnosticity) or a ‘test’ name (low diagnosticity) at
the time they ?rst encountered the brand name. Participants
in the true name (i.e., high diagnosticity) condition were
informed that the brand name in the press release was the
name that would eventually be used once the product came
to market. Participants in the test name (i.e., low diagnos-
ticity) condition were informed that the brand name in the
press release was only for testing purposes and would not
be the name of the ice cream when it was released to the
public. Further, in order to manipulate the timing of the
diagnosticity information, this information was provided ei-
ther simultaneously with the names of the ice cream or
afterward.
Procedure. One hundred and twenty-six undergraduate
students in a large northeastern university participated in the
study for partial course credit. Participants read a paragraph
stating that a new ice cream was to be introduced into the
area and were asked to read a press release (containing our
manipulations) describing the planned promotional activities
that were to accompany the ice cream debut. After reading
the press release, participants evaluated the ice cream on its
richness, smoothness, and creaminess. They then reported
their overall evaluation of the ice cream, and their intentions
to purchase it. At the end of the questionnaire, participants
reported how much the brand name reminded them of an
ice cream, listed their ice cream consumption, and assessed
the dif?culty of the brand evaluation task. Finally, they were
asked how involved they were in the study and whether they
were aware of using the information gathered fromthe brand
name. The participants were then debriefed and dismissed.
Results
Manipulation and Confound Checks. In order to con-
?rm that our manipulation of the diagnosticity of the ice
cream name worked as intended, we conducted 2 (brand
name) # 2 (diagnosticity) # 2 (timing) ANOVA on the
likelihood that the ice cream name revealed to participants
would be used as the brand name when ice creamwas ?nally
launched. This ANOVA indicated a main effect of diag-
nosticity such that participants who were told that the name
of the ice cream was a test name believed that the brand
name in the press release was less likely to be an ice cream
name than individuals in the true name condition (M’s p
vs. 4.10; , ). No other effect 2.60 F(1, 117) p17.56 p ! .01
was signi?cant ( ). p 1 .05
A similar analysis on ratings of involvement revealed that
participants reported equal involvement in the study irre-
spective of condition (overall , all contrast mean p4.39
), indicating that our manipulation of the diagnosticity F ! 1
of the brand name was not confounded with consumer in-
volvement. Finally, the reported frequency of ice creamcon-
sumption did not vary across experimental conditions.
Hypotheses Tests. We created an Attribute Perception
Index using the attribute evaluations of creaminess, richness,
and smoothness evaluations (Cronbach’s ) and a a p.86
Brand Evaluation Index using an average of the overall
evaluation and purchase intention (Cronbach’s ). a p.81
We used a 2 (sound symbolism of brand name: Frish vs.
Frosh) # 2 (diagnosticity of brand name: test vs. true) #
2 (timing: simultaneous vs. after) ANCOVA (with the level
of involvement in purchase decisions in the product category
as a covariate) to test the hypotheses. For our hypotheses
to be supported, we should obtain a signi?cant three-way
interaction (hypothesis 3), made up of a two-way interaction
between brand name and diagnosticity in the simultaneous
condition (hypothesis 3a), and a main effect of brand name
in the after condition (hypothesis 3b). Note that support for
hypothesis 3a is also support for hypothesis 2, and support
for hypothesis 3b is support for our baseline hypothesis.
(i) Attribute Perception Index. We conducted a 2 #
ANCOVA on the Attribute Perception Index that re- 2 #2
vealed the predicted three-way interaction (F(1, 117) p
, ). In addition, the main effect of brand name 3.84 p ! .05
was signi?cant ( , ). Means are F(1, 117) p9.51 p ! .01
graphically presented in ?gure 1.
In order to investigate the signi?cant three-way interac-
tion, we ?rst examined the simple effects of the 2 (sound
symbolism) # 2 (diagnosticity of the brand name) effects
for the condition when the diagnostic information was si-
multaneously provided with the brand names. This
ANCOVA revealed a signi?cant two-way interaction
( , ), accompanied by a signi?cant F(1, 121) p4.34 p ! .05
main effect of brand name ( , ). F(1, 121) p10.07 p ! .01
When the name was described as true, participants utilized
the sound symbolism and evaluated the associated attributes
for Frosh higher than Frish ( vs. 4.25; contrast M’s p5.06
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 47
FIGURE 1
STUDY 1 RESULTS
NOTE.—Attribute ratings as a function of timing of diagnostic information (Attribute Perception Index).
, ). Individuals presented with a test F(1, 122) p4.43 p ! .05
brand name discounted the diagnosticity of the name and
evaluated Frish equivalent to Frosh on the attributes (Frish
vs. Frosh ; contrast ). These re- M p4.36 M p4.08 F ! 1
sults support hypothesis 2 and hypothesis 3a.
Next, we conducted a similar analysis of the simple 2 #2
effects when diagnostic information was provided after the
sound symbolism was encountered. Consistent with hy-
pothesis 3b, analysis revealed only a main effect of sound
symbolism ( , ). Frosh ( ) F(1, 118) p9.94 p ! .01 M p5.00
was rated higher than Frish ( ) on the Attribute M p4.09
Perception Index regardless of whether the name was de-
scribed as a true or test name. The results support hypothesis
3a. Hypothesis 1 is also further supported as the baseline
sound symbolism effect again manifests.
(ii) Brand Evaluation Index. We predicted a similar
set of results using brand evaluations as the dependent mea-
sure. A ANCOVA on the Brand Evaluation In- 2 #2 #2
dex revealed the predicted three-way interaction
( , ). As with the Attribute Percep- F(1, 118) p4.78 p ! .05
tion Index, the main effect of brand name was signi?cant
( , ); no other effect was signi?cant. F(1, 118) p9.39 p ! .01
Given the signi?cant three-way interaction, we examined
the simple effects of the 2 (sound symbolism) # 2 (diag-
nosticity of the brand name) design separately by the timing
of the diagnostic information. In the condition where the
diagnostic information was provided after the sound sym-
bolism was encountered, analysis revealed a signi?cant in-
teraction ( , ), accompanied by a sig- F(1, 121) p7.77 p ! .01
ni?cant main effect of brand name ( , F(1, 121) p10.83
). When the name was described as true, participants p ! .01
used the sound symbolism and evaluated the associated at-
tributes for Frosh higher than Frish ( vs. 3.97; M’s p5.12
contrast , ). Individuals presented F(1, 122) p8.14 p ! .01
with a test brand name discounted the diagnosticity of the
name and evaluated Frish equivalent to Frosh on the attrib-
utes (Frish vs. Frosh ; contrast ). M p4.50 M p4.16 F ! 1
These results support hypotheses 2 and 3a.
A similar analysis of the simple effects completes 2 #2
the replication of our attribute perception ?ndings for overall
brand evaluations. When the diagnostic information was
provided simultaneously with the sound symbolism of the
brand names, the main effect of sound symbolism was sig-
ni?cant ( , ), such that Frosh F(1, 121) p9.29 p ! .01
( ) was rated higher than Frish ( ) on the M p4.84 M p4.00
Brand Evaluation Index regardless of whether the name was
described as true or test, supporting hypothesis 1 and hy-
pothesis 3b.
Awareness of In?uence of Phonetic Effects. We also
examined participants’ ratings of whether the brand name
affected their evaluations by conducting a 2 #2 #2
ANOVA. These ratings were elicited on a seven-point, se-
mantic differential scale anchored at “Brand names did not
in?uence my evaluations at all” and “Brand names in?u-
enced my evaluations a lot.” This analysis revealed no sig-
ni?cant effect, with the mean for the sample at a 2.68 on a
scale of seven ( ), with a higher number in- midpoint p4
dicating a higher level of awareness. We then computed a
correlation between these ratings and the Attribute Percep-
tion Index (see Wegener, Petty, and Dunn 1998). The cor-
relation was nonsigni?cant ( ), indicating unawareness p 1 .6
on the part of the participants of the in?uence of brand names
on judgments.
Discussion
The results of this study provide evidence that sound
symbolism in?uences attribute perceptions (hypothesis 1).
However, the phonetic effects of brand names can be un-
dercut by alternate information at the time at which brand
48 JOURNAL OF CONSUMER RESEARCH
names are encountered that discredits the effects of sound
symbolism. Our ?ndings suggest that when information re-
garding the diagnosticity of the brand name is provided at
the time the name is encountered, consumers can control
whether or not to process the underlying sound symbolism
(hypothesis 2). However, the use of sound symbolism in
brand perceptions is not completely in the individuals’ con-
trol. The individuals’ inability to discount a brand name that
is demonstrated to be nondiagnostic after their initial en-
counter with the name attests to the uncontrollable manner
in which phonetic effects manifest (hypothesis 3). This study
also provides initial evidence that participants are unaware
of using sound symbolism as an input in judgments. Our
next study tests whether the use of sound symbolism is
effortless, a third element of an automatic process.
STUDY 2
Method
Design. To test hypothesis 4, we used a 2 (sound sym-
bolism of brand name) # 2 (diagnosticity of brand name)
# 2 (cognitive capacity) between-subjects design. Sound
symbolism was again manipulated using Frish and Frosh.
The diagnosticity of the brand name was manipulated as
before by informing participants that the name was a true
versus a test name at the time the brand name information
was encountered (i.e., the simultaneous condition in study
1). Finally, participants were read a list of numbers and had
to count the number of times the experimenter said the word
?ve in the condition in which cognitive capacity was im-
paired. Participants in the normal cognitive capacity con-
dition did not undertake this task.
Procedure. One hundred and eleven undergraduate stu-
dents in a large West Coast university participated in the
study for partial course credit. Unlike the previous study,
this study was administered on computer. As before, par-
ticipants read a paragraph stating that a new ice cream was
to be introduced in the area and were asked to read a press
release describing the planned promotional activities that
were to accompany the ice cream debut. Again, participants
in the true name condition were informed that the brand
name in the press release was the name that would eventually
be used once the product came to market, while the partic-
ipants in the test name condition were informed that the
brand name in the press release was only for testing purposes
and would not be the name of the ice cream when it was
released to the public. Unbeknownst to participants, the time
taken to read this initial press release was monitored. After
reading the press release, participants evaluated the ice
cream overall, rated the ice cream on its richness, smooth-
ness, and creaminess and stated their purchase intention.
Finally, attribute importance ratings were collected. All eval-
uations were elicited on 100-point sliding rule scales. Par-
ticipants were then debriefed and dismissed.
Results
Manipulation Checks. To test the success of the ma-
nipulation impairing cognitive capacity, we used the time
taken to read the article about the ice cream as an indication
of task dif?culty. A ANOVA on the reading time 2 #2 #2
indicated a signi?cant main effect of the cognitive capacity
manipulation ( , ) with participants F(1, 103) p7.09 p ! .01
in the normal cognitive capacity condition taking less time
( sec.) than those in the cognitive capacity im- M p30.33
paired condition ( sec.).
1
Thus, our manipulation M p37.45
to impair cognitive capacity was successful.
Hypotheses Tests. We used a 2 (sound symbolism of
brand name: Frish vs. Frosh) # 2 (diagnosticity of brand
name: test vs. true) # 2 (cognitive capacity: normal vs.
impaired) ANOVA to test hypothesis 4. As in study 1, we
created an Attribute Perception Index using the attribute
evaluations of creaminess, richness, and smoothness (Cron-
bach’s ). For hypothesis 4 to be supported, we a p.82
should obtain a signi?cant three-way interaction, comprising
of a two-way interaction between brand name and diag-
nosticity in the normal cognitive capacity condition (hy-
pothesis 2), and a main effect of brand name in the low
cognitive capacity condition (hypothesis 1).
A ANOVA on the Attribute Perception Index 2 #2 #2
revealed the predicted three-way interaction (F(1, 103) p
, ; see ?g. 2). In addition, the main effect of 3.84 p ! .05
brand name was signi?cant ( , ). No F(1, 103) p8.03 p ! .01
other effect was signi?cant.
Given the signi?cant three-way interaction, we investi-
gated the simple effects separately in the two cognitive load
conditions. A 2 (sound symbolism) # 2 (diagnosticity of
the brand name) analysis in the normal cognitive capacity
condition revealed a signi?cant interaction (F(1, 107) p
, ) accompanied by a signi?cant main effect of 4.84 p ! .05
brand name ( , ). An examination of F(1, 107) p7.68 p ! .01
the means indicates that under normal cognitive capacity
and when the name was described as true participants util-
ized the sound symbolism and evaluated the associated at-
tributes for Frosh higher than Frish ( vs. 49.2; M p63.3
contrast , ). Individuals presented F(1, 108) p5.47 p ! .05
with a test brand name discounted the diagnosticity of the
name and evaluated Frish equivalent to Frosh on the attrib-
utes (Frish vs. Frosh ; contrast ). M p54.9 M p50.3 F ! 1
These results are consistent with hypothesis 2 and support
hypothesis 4a.
Analysis of the simple effects for the impaired 2 #2
cognitive capacity condition revealed only a signi?cant main
effect of sound symbolism ( , ). Re- F(1, 107) p7.81 p ! .01
gardless of whether the name was described as true or test,
when respondents were cognitively impaired, Frosh (M p
) was rated higher than Frish ( ) on the At- 62.4 M p50.0
1
These results were replicated through a self-reported measure of task
dif?culty.
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 49
FIGURE 2
STUDY 2 RESULTS
NOTE.—Attribute ratings as a function of cognitive capacity (Attribute Perception Index).
tribute Perception Index; this is consistent with hypothesis
1 and supports hypothesis 4b.
2
Discussion
The results of this study indicate that participants were
unable to actively discount the brand name information
when they were cognitively impaired. We thus obtain evi-
dence for the use of sound symbolism being effortless such
that even when cognitively impaired, the effects ?lter
through to judgments. The results of this study provide cor-
roborative support to the prior study’s ?ndings that the in-
ferences of attribute information from brand names and the
subsequent evaluations of the brand are automatic processes.
GENERAL DISCUSSION
In this article, we provide support for the proposition that
consumers gather and process information from brand
names in an automatic manner. However, consumers can
control for this provided they process information that dis-
credits the brand name at the time the name is encountered.
An important aspect of sound symbolism established in this
article is how it affects consumers’ attitudes at the attribute
level (hypothesis 1). Consumers may knowingly control for
the process when they have access to information that credits
or discredits the brand name at the time at which they en-
counter the brand name (hypothesis 2). Yet, consumers are
not fully aware of this process and cannot fully control for
the effect; we show evidence for the uncontrollability of the
effect unless the name is discredited at the time of encoding
(hypothesis 3) and the consumer is not cognitively impaired
2
A similar analysis using the Brand Evaluation Index (Cronbach’s
) revealed an identical pattern of results; we do not report these a p.75
results here due to space constraints.
(hypothesis 4). These results provide evidence for the au-
tomatic processing of brand names and their phonetic
effects.
Our purpose in this article was to understand the process
underlying sound symbolism. The basic phenomenon of
sound symbolism has been demonstrated for a wide range
of sounds across multiple dimensions (Klink 2000). In our
article, we focused on the internal validity of our studies in
order to build theory about the sound symbolism process.
We therefore tested one set of brand names and used a
between-subjects test of sound symbolism. This strengthens
the existing case for sound symbolism since participants are
not aware that sounds are being manipulated or that one
brand name should provide different information than an-
other. Further, we increase the internal validity of the ex-
periments by testing for phonetic effects when the brand
name is not the only information provided to the subject.
Additionally, we show that this effect carries through at-
tribute perceptions to the overall evaluation of the brand.
Demonstrating these effects of sound symbolism increases
our con?dence in the robustness of our theory.
Automatic processes in social cognition domains have
been demonstrated empirically in various contexts (see
Bargh 1989). This article adds to this literature by dem-
onstrating the automatic use of sound symbolism. This is
an important issue to examine, given that if phonetic effects
manifest even partially automatically, then this automatic
effect will exert a more consistent in?uence over attitudes
and behavior over time (see Bargh 2002). Since sound sym-
bolism is a fairly common naming device (Begley 2002), it
is possible that phonetic links are highly accessible. Atti-
tudes relating to the phonetic aspects of the brand name
would be automatically activated. Considering the limited
capacity consumers usually devote to product evaluation
(Hoyer 1984), the results of studies 1 and 2 indicate that
50 JOURNAL OF CONSUMER RESEARCH
under this diminished cognitive capacity, the phonetic ef-
fects of brand names may be activated and in?uence choice.
If brand name phonetic cues are chronically accessible, con-
gruency of brand name and product would be of even greater
importance.
Our studies tested names in which only one phoneme was
manipulated. In most categories, where multisyllabic words
are the norm, sound symbolism may occur in different po-
sitions and/or on multiple positions in the same word. This
?exibility would not only increase the complexity of the
message conveyed by sound symbolism but also may in-
crease the strength of the effect. This article has focused
primarily on the sound symbolism provided by the vowels
of a brand name. The role consonants play in brand names
and their effects on product evaluations is an underresear-
ched area. Past research has shown that brand names that
begin with hard consonants (i.e., Kodak and Pepsi), elicit
higher recognition (Vanden Bergh et al. 1984), and occur
more frequently than we would expect from the English
lexicon (Schloss 1981). Although linguistic anthropologists
have noted phonetic patterns for consonants (i.e., harder
sounding consonants tend to represent harder, sharper ob-
jects than softer sounding consonants), these ?ndings have
not been directly tested on brand names. A systematic study
of consonants and their interaction with multiple vowels is
an avenue worthy of future research.
Our studies demonstrated that sound symbolism, although
evaluated on the attribute level, affects overall evaluations.
We demonstrated that sound symbolism affected the attrib-
utes of creaminess, sweetness, and richness: three category
speci?c attributes. It appears that consumers will take the
broader dimensions that sound symbolism operates upon
(e.g., smallness, lightness, mildness, fastness, etc.) and ex-
trapolate these ?ndings to the speci?c properties of ice
cream. Creating a successful brand name depends not only
upon the creation of a name that is congruent with the prod-
uct category, but one that phonetically ?ts the positioning
of the brand within that product category. Future research
could examine whether it is possible to phonetically ma-
nipulate meaning over a wide range of attributes in a broad
set of product categories.
The automatic nature of the sound symbolism leads to
important implications of this research upon the use of per-
sonal names as brand names. For example, if the ice cream
was named for its maker, Joseph Frish, then the brand name
should be seen as less diagnostic, and the phonetic effect
should not appear. This assumes that the information as to
the name’s origin is known at the time of encoding and that
processing capabilities are not constrained. Telling the con-
sumer that the ice cream was named for the maker after the
ice cream has been tasted should have little effect upon
consumers’ use of sound symbolism in accordance with our
?ndings. An interesting further study would be where we
manipulate the diagnosticity of the name through the use of
a personal name versus a purely ?ctitious name.
Over 2,000 years ago, Plato supported the onomatopoetic
theory of language by arguing that words were formed by
the imitation of ideas in sounds (Jowett 1953). The argument
still holds. Sound symbolism shows that the sounds of words
still convey meaning. More important, consumers recognize
this phonetic meaning and will use it as a source of product
information. The effects of time, the in?uence of foreign
languages, the desire of euphony, and elements of chance
all distill and, at times, distort the effects of sound sym-
bolism. Yet even with these caveats, attending to phonetic
meanings and leveraging the attribute associations when cre-
ating a brand image is sound advice when creating a new
brand name.
[David Glen Mick and Dawn Iacobucci served as editors
and Punam Anand Keller served as associate editor for
this article.]
REFERENCES
Bargh, John A. (1989), “Conditional Automaticity: Varieties of
Automatic In?uence in Social Perception and Cognition,” in
Unintended Thought, ed. James S. Uleman and John A. Bargh,
New York: Guilford, 3–51.
——— (2002), “Losing Consciousness: Automatic In?uences on
Consumer Judgment, Behavior, and Motivation,” Journal of
Consumer Research, 29 (September), 280–85.
Bargh, John A. and Roman D. Thein (1985), “Individual Construct
Accessibility, Person Memory, and the Recall-Judgment Link:
The Case of Information Overload,” Journal of Personality
and Social Psychology, 49 (November), 1129–46.
Begley, Sharon (2002), “StrawBerry Is No BlackBerry: Building
Brands Using Sound,” Wall Street Journal, August 26, B1.
Consumer Reports (1994), “Ice Cream: What’s the Scoop?” Con-
sumer Reports, June, 373–77.
Gilbert, Daniel T. (1989), “Thinking Lightly about Others: Auto-
matic Components of the Social Inference Process,” in Un-
intended Thought, ed. James S. Uleman and John A. Bargh,
New York: Guilford, 189–211.
Gilbert, Daniel T. and J. Gregory Hixon (1991), “The Trouble of
Thinking: Activation and Application of Stereotypic Beliefs,”
Journal of Personality and Social Psychology, 60 (4), 509–17.
Gilbert, Daniel T., Brett W. Pelham, and Douglas S. Krull (1988),
“On Cognitive Busyness: When Person Perceivers Meet Per-
sons Perceived,” Journal of Personality and Social Psychol-
ogy, 54 (May), 733–40.
Hinton, Leanne, Johanna Nichols, and John J. Ohala, eds. (1994),
Sound Symbolism, Cambridge: Cambridge University Press.
Hoyer, Wayne D. (1984), “An Examination of Consumer Decision
Making for a Common Repeat Purchase Product,” Journal of
Consumer Research, 11 (December), 822–29.
Huang, Yau-Huang, Sawat Pratoomaraj, and Ronald C. Johnson
(1969), “Universal Magnitude Symbolism,” Journal of Verbal
Learning and Verbal Behavior, 8 (1), 155–56.
Jacobson, Roman and Linda R. Waugh (1987), The Sound Shape
of Language, 2d ed., New York: Mouton de Gruyter.
Jowett, Benjamin (1953), The Dialogues of Plato: Translated into
English with Analyses and Introductions, London: Oxford
University Press.
Klank, Linda J. K., Yau-Huang Huang, and Ronald C. Johnson
(1971), “Determinants of Success in Matching Word Pairs in
Tests of Phonetic Symbolism,” Journal of Verbal Learning
and Verbal Behavior, 10 (2), 140–48.
Klink, Richard R. (2000), “Creating Brand Names with Meaning:
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 51
The Use of Sound Symbolism,” Marketing Letters, 11 (1),
5–20.
——— (2001), “Creating Meaningful New Brand Names: AStudy
of Semantics and Sound Symbolism,” Journal of Marketing:
Theory and Practice, 9 (Spring), 27–34.
Lapolla, Randy J. (1994), “An Experimental Investigation into
Phonetic Symbolism as It Relates to Mandarin Chinese,” in
Hinton et al., eds.
Makino, Seiichi, Seiichi Nakada, and Mieko Ohso (1999), Ko-
dansha’s Basic English-Japanese Dictionary, Tokyo: Kodan-
sha Int.
O’Grady, William, Michael Dobrovolsky, and Francis Katumba
(1997), Contemporary Linguistics: An Introduction, 3d. ed.,
New York: St. Martin’s.
Ohala, John J. (1994), “The Frequency Code Underlies the Sound-
Symbolic Use of Voice Pitch,” in Hinton et al., eds.
Sapir, Edward (1929), “A Study in Phonetic Symbolism,” Journal
of Experimental Psychology, 12 (3), 225–39.
Schloss, Ira (1981), “Chickens and Pickles,” Journal of Advertising
Research, 21 (6), 47–49.
Schmitt, Bernd H., Yigang Pan, and Nader T. Tavassoli (1994),
“Language and Consumer Memory: The Impact of Linguistic
Differences between Chinese and English,” Journal of Con-
sumer Research, 21 (December), 419–31.
Schmitt, Bernd H. and Shi Zhang (1998), “Language Structure and
Categorization: A Study of Classi?ers in Consumer Cogni-
tion, Judgment, and Choice,” Journal of Consumer Research,
25 (September), 108–22.
Tavassoli, Nader T. and Jin K. Han (2001), “Scripted Thought:
Processing Korean Hancha and Hangul in a Multimedia Con-
text,” Journal of Consumer Research, 28 (December),
482–93.
Vanden Bergh, Bruce G., Janay Collins, Myrna Schultz, and Keith
Adler (1984), “Sound Advice on Brand Names,” Journalism
Quarterly, 61 (4), 835–40.
Wegener, Duane T., Richard E. Petty, and Meghan Dunn (1998),
“The Metacognition of Bias Correction: Naive Theories of
Bias and the Flexible Correction Model,” in Metacognition:
Cognitive and Social Dimensions, ed. Guy Lories, Benoit
Dardenne, and Vincent Y. Yzerbyt, Thousand Oaks, CA: Sage,
202–27.
doc_800567711.pdf
Sound symbolism, the linguistic process in which the sounds of a word provide cues about the word's meaning, is not a new phenomenon. Plato first described the effect in the dialogues of Cratylus, and authors throughout time have used the sounds in words to describe people.
43
? 2004 by JOURNAL OF CONSUMER RESEARCH, Inc. ? Vol. 31 ? June 2004
All rights reserved. 0093-5301/2004/3101-0004$10.00
A Sound Idea: Phonetic Effects of Brand
Names on Consumer Judgments
ERIC YORKSTON
GEETA MENON*
In this article we examine a phenomenon known as sound symbolism, where the
sound of a word conveys meanings. Speci?cally, brand names are composed of
individual sounds called phonemes and we investigate how this phonetic structure
of brand names affects a consumer’s evaluation of products and their underlying
attributes. We demonstrate that consumers use information they gather from pho-
nemes in brand names to infer product attributes and to evaluate brands. We also
demonstrate that the manner in which phonetic effects of brand names manifest
is automatic in as much as it is uncontrollable, outside awareness and effortless.
S
ound symbolism, the linguistic process in which the
sounds of a word provide cues about the word’s mean-
ing, is not a new phenomenon. Plato ?rst described the effect
in the dialogues of Cratylus, and authors throughout time
have used the sounds in words to describe people (e.g., the
miniature Lilliputians and the giant Brobdingnagians in
Swift’s Gulliver’s Travels) and objects (e.g., the large, dan-
gerous Bludgers, the big, round Quaf?e, and the small, fast
Golden Snitch in the Quidditch game in Rowling’s Harry
Potter and the Sorcerer’s Stone) represented by those words.
When presented with ?ctitious or unfamiliar words, indi-
viduals consistently use sound symbolism to interpret mean-
ings from the name about the referenced object (Jacobson
and Waugh 1987; Sapir 1929). The effect is extremely well
documented whether the tested individual’s native language
is English (Klink 2000, 2001), Chinese (Klank, Huang, and
Johnson 1971; Lapolla 1994), or even Navajo (Sapir 1929).
In fact, sound symbolism has been observed to exist in
native languages in North America, Latin America, Asia,
Australia, and Africa, as well as more developed languages
such as English, Finnish, French, German, Modern Greek,
and Japanese (for a complete discussion of these examples
see Hinton, Nichols, and Ohala 1994).
*Eric Yorkston is assistant professor of marketing, University of South-
ern California, Los Angeles, CA 90089-0443 (e-mail: yorkston@marshall
.usc.edu). Geeta Menon is Harold MacDowell Faculty Fellow and associate
professor of marketing, New York University, New York, NY 10012-1126
(e-mail: [email protected]). This article is based on the doctoral dis-
sertation of the ?rst author under the supervision of the second. The authors
thank members of the dissertation committee, Adamantios Gafos, Eric
Greenleaf, and Vicki Morwitz, for their helpful comments during the dif-
ferent stages of this project. The authors acknowledge the data collection
assistance provided by Sucharita Chandran, Andrea Morales, and Manoj
Thomas. In addition, the authors are especially grateful to the JCR editors,
Dawn Iacobucci and David Mick, the associate editor, and three reviewers
for their constructive feedback during the review process.
Past research on sound symbolism has focused on two
aspects: the range of the effect (i.e., its universality; Huang,
Pratoomaraj, and Johnson 1969) and the aspects of meaning
it affects (Klink 2000, 2001; Ohala 1994). Less is known
about the sound symbolism process itself. For instance, re-
search has examined how linguistic classi?ers affect cate-
gorization schema (Schmitt and Zhang 1998), and how an
individual’s formal writing system(i.e., logographic vs. pho-
netic) affects processing (Schmitt, Pan, and Tavassoli 1994;
Tavassoli and Han 2001). Yet we do not know how or why
individuals use sound symbolismas a source of object mean-
ing. Current research has neither revealed whether this is a
controlled strategy or an automatic process nor has it ex-
amined conditions when sound symbolism is utilized. The
effect is well established, but we still lack understanding of
the process by which it occurs.
Sound symbolism has been recently recognized as an im-
portant factor in howindividuals infer speci?c meaning from
unfamiliar brand names (Klink 2000, 2001). A brand name
is composed of individual sounds called phonemes. These
phonemes serve two purposes. First, they are the basic build-
ing blocks of language and are combined to form syllables
and, in turn, words. These words and the syllables from
which they are composed supply what we traditionally think
of as the meaningful units of a brand name. Second, pho-
nemes can provide meaning themselves through sound sym-
bolism. These sounds provide cues about how the brand
may perform on particular attribute dimensions. We theorize
that if a brand name has phonemes that represent attributes
a consumer desires, consumers will hold more positive at-
titudes and exhibit higher purchase intentions toward that
brand. The current article provides a demonstration of the
process by which sound symbolism manifests in consumer
judgments. We demonstrate that the process is uncontrol-
lable, outside of awareness, and effortless, making it au-
44 JOURNAL OF CONSUMER RESEARCH
tomatic (Bargh 1989). By understanding the underlying, the-
oretical processes of sound symbolism, we achieve greater
understanding of how consumers interpret brand names and
use them to evaluate brands.
Like Klink (2000, 2001), we examine the effects of
sounds symbolism on attribute perceptions, but in addition,
in our research we examine their effects on brand evalua-
tions as well, using a more subtle between-subjects design.
Unlike Klink (2000, 2001), our primary objective is to de-
lineate the automaticity of the underlying process by which
phonetic effects occur.
CONCEPTUAL FRAMEWORK
Sounds, and the resulting aural frequencies, are based
upon the position and curvature of the tongue in the mouth,
ranging from a high-front to low-back position (see Klink
2000 for a review). The affective meanings generated by
sound symbolism follow a similar pattern. These vowel
sounds roughly form an ordered sound-symbolic list: [e¯],
, [e], [a¯], [a], [o¯], [o], [a¨], , and [u¨] (e.g., beat, bit, bet,
bait, bat, boat, bought, posh, but, put, and boot). The ex-
istence of this hierarchy is fairly consistent across languages
(Makino, Nakada, and Ohso 1999). Sound symbolism con-
veys information such that high-front vowels (e.g., ee in
?ea and i in ?y) represent associations with smaller size and
less power than low-back vowels (e.g., the ow in bout and
oo in boot), which, in turn, connote greater size, and more
power (Hinton et al. 1994; Makino et al. 1999). In an em-
pirical demonstration, Klink (2000) showed that the use of
front vowels (as opposed to the back vowels) in brand names
conveys attribute qualities of smallness, lightness, mildness,
thinness, fastness, coldness, bitterness, femininity, weak-
ness, lightness, and prettiness. Furthermore, Klink (2001)
demonstrated the interactive effects of sound symbolism
with semantic meaning in affecting consumers’ evaluations
of various brands.
Consistent with Klink (2000, 2001), we expect that when
consumers encounter a brand name, they infer attribute
meaning. In testing our theory, we used the longer, broader
[a¨] sound (as in bother and chop) from the middle of the
phonetic vowel scale and contrasted it against the shorter,
tighter sound (as in kiss and nymph) from the lighter,
sharper, smaller, end of the sound-symbolism scale. The
derivation of the pair of sounds that we use in our experi-
ments is a function of the theory associated with the position
of the tongue in the mouth (O’Grady, Dobrovolsky, and
Katumba 1997). The chosen pair of vowel sounds we test
( and [a¨]) are not the extremes on either end of this scale.
For example, the sounds ee and oo lie nearer the two re-
spective endpoints of the phonetic-symbolic spectrum. Thus,
the vowel pairing we use (i.e., and [a¨]) may be considered
a more conservative test of the sound symbolismhypothesis.
An additional consideration in our decision to test only one
sound pair was our strong desire to control for latent se-
mantic associations that could interfere with our studies (a
discussion of how we controlled for semantic associations
is below). Since the goal of our research was not merely to
provide further support for the phenomenon of sound sym-
bolism but to understand the process by which sound sym-
bolism affects meaning, we decided that the advantages
gained by ruling out this alternative hypothesis outweighed
the costs of limiting the range of tested vowels. Formally,
our baseline hypothesis is
H1: Consumers will evaluate the individual attributes
of brand names with the phonetic [a¨] sound heav-
ier than attributes of brand names with the pho-
netic sound.
Stated simply, sound symbolism affects attribute percep-
tions. This raises the possibility that consumers use the brand
name as a phonetic cue regardless of perceived value of this
information. We are now left with the question of when
exactly consumers are affected by the phonetics of brand
names. We posit that consumers attempt to strategically use
the brand name as information when they perceive it to
supply diagnostic information. Alternately, consumers will
attempt to ignore, and adjust for, the phonetic meaning in
situations where the brand name is perceived as less diag-
nostic. Thus,
H2: The effects of brand names on attribute percep-
tions will be moderated by the perceived diag-
nosticity of the brand name, such that, the pho-
netic effects of brand names will manifest only
when the brand name is described as the true ver-
sus the test name.
It has been documented recently that a large number of
consumer decisions are nonconscious (see Bargh 2002 for
a review). Bargh (1989) asserts that a process may have one
or more of the four automatic criteria to be differentiated
from a conscious or controlled process; that is, a process is
automatic if it is effortless, unintentional, used outside of
awareness, or uncontrollable. We investigate the automatic-
ity of sound symbolism effects in this article.
If sound symbolism manifests automatically and is used
in an uncontrollable fashion, then participants who are told
that the phonetic information is nondiagnostic after they
have experienced the brand name information should not
be able to completely discount the phonetic information.
This is because if the process is partly automatic, people
are either unaware of or unable to control for the incor-
poration of phonetic effects in their judgments (Bargh 1989,
2002). On the other hand, if the experienced effects of sound
symbolism are discounted at the time it is felt, then it should
not enter into the judgment. Speci?cally,
H3: The timing of information about the diagnosticity
of brand names will moderate the effects of sound
symbolism on attribute perceptions, such that:
a) When this information is provided at the same
time as the brand name information (i.e., sound
symbolism can be discounted while being ex-
perienced), the phonetic effects will be erased
in the low diagnosticity condition, and will
manifest in the high diagnosticity condition (as
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 45
predicted in hypothesis 2).
b) When this information is provided after the
brand name information (i.e., sound symbolism
has been experienced), the basic phonetic effects
will manifest in both the low and high diag-
nosticity conditions (as predicted in hypothesis
1).
It is also possible that phonetic effects manifest effort-
lessly. Gilbert (1989) suggests that the automatic process in
judgment formation is a two-stage process comprising of
the initial anchor based on an automatic input, and the sub-
sequent correction (e.g., incorporating ignored inputs, or
correcting weights) performed in a more controlled manner.
For example, in experiments on stereotype activation and
use, Gilbert and Hixon (1991) found that the application of
an activated stereotype was moderated by the availability
of cognitive resources. Those with capacity constraints were
less likely to engage in corrective adjustments to the au-
tomatically activated stereotypes.
If an automatic source of information is used (i.e., the
sound symbolism of the brand name) together with a con-
trolled input (i.e., information about the diagnosticity of the
brand name), then the automatic input will have a greater
effect when cognitive resources are constrained than when
they are abundantly available (Bargh and Thein 1985; Gil-
bert, Pelham, and Krull 1988). With increased cognitive
load, information that is automatically processed will have
a proportionately greater impact on judgments as many of
the conscious, effort-requiring adjustments will not be pos-
sible (Bargh and Thein 1985; Gilbert 1989). In the current
study, we use Gilbert’s load paradigm to reverse the mod-
erating effect of diagnosticity and replicate the basic pho-
netic effect when cognitive constraints are imposed. Thus,
H4: Cognitive capacity will moderate the effects of
sound symbolism on attribute perceptions, such
that:
a) Under conditions of normal cognitive capacity,
the phonetic effects will be erased in the low
diagnosticity condition, and will manifest in
the high diagnosticity condition (as predicted
in hypothesis 2).
b) Under conditions of impaired cognitive capac-
ity, the diagnosticity of the brand name will
not have an effect and the basic phonetic ef-
fects will manifest (as predicted in hypothesis
1).
We now present two studies that were designed to test
these hypotheses.
STUDY 1
Method
Choice of Stimuli. We chose ice cream as the product
category for the studies reported in this article, as it produced
high interest and involvement among the student population,
was likely to be purchased in the near future, and elicited
product evaluations that incorporated a small number of
attributes. An initial pretest ( ) established that ice n p48
cream attributes consisted of three main factors: taste, cost,
and calorie content. Since taste was reported as the most
primary attribute and is a fairly ambiguous one, we further
explored the underlying factors of taste. When participants
were asked to elaborate upon taste, the attributes of smooth-
ness, creaminess, richness, and sweetness emerged. Our
?ndings were corroborated by a Consumer Reports (1994)
article on ice cream that discussed taste in their overall brand
comments using the terms “creamy,” “rich,” “smooth,” and
“sweet.” Our pretest, however, indicated that although peo-
ple, in general, preferred creamier, richer, and smoother ice
creams, preference for sweetness was more divided. There-
fore, the ?nal attributes we used were smoothness, cream-
iness, and richness, three attributes that consumers prefer
more of, and that may be communicated by the sound sym-
bolism of the brand name.
The ?rst goal in designing the brand names for the studies
was to develop two names that were ?ctitious and either
avoided or controlled for other linguistic complexities. Since
we were examining the symbolic differences between the
[a¨] and sounds, we held all consonants in the words
constant, avoiding hard consonants, since their rough sound
symbolizes attributes that contradict those desired in our
product category of ice cream (Schloss 1981; Vanden Bergh
et al. 1984).
The second goal was to have two names that were con-
sidered similar. Besides balancing the names phonetically,
we pretested ?ve pairs of names that met the above phonetic
requirements. Forty-four undergraduate participants ranked
the 10 names, with one being the name most likely to be
an ice cream name and 10 being the name least likely to
be an ice cream. Frosh and Frish best met the criterion of
being equally likely to be an ice cream name (M’s p
and 3.55, respectively, NS). These were also the most 3.33
preferred mean rankings among all the product names.
Based on this pretest and our research goals, we chose
the names Frosh and Frish differing only on the phonetic
sounds [a¨] and . In general, the [a¨] sound is associated
with perceptions of objects being bigger, heavier, slower,
and duller. The sound, on the other hand, is associated
with perceptions of the object being smaller, lighter, livelier,
and sharper. Therefore, in the context of ice creams, we
expect the [a¨] sound in Frosh to communicate a smoother,
creamier, and richer ice cream than the sound in Frish.
To ensure that a phonetic distinction manifesting through
affective meaning drove the evaluations of the two brand
names, we examined the denotative meaning evoked by
these names through another pretest. We asked 56 partici-
46 JOURNAL OF CONSUMER RESEARCH
pants to list the ?rst three words that came to their mind
when they thought of the ice cream name. Acontent analysis
was performed to ensure that there were no common word
associations that could be moderating the brand name eval-
uations. Two independent coders blind to the hypotheses
classi?ed the resulting 167 words as positive, negative, or
neutral when describing an ice cream. Coders’ classi?cations
matched in 84.4% of the cases. Disagreement between cod-
ers was settled through discussion. Frosh and Frish gener-
ated equivalent numbers of positive (30 vs. 28), negative
(10 vs. 13), and neutral (43 vs. 43) semantic associations
( , NS). Therefore, as desired, the two names were
2
x ! 1
well matched in terms of semantic associations. In general,
of all the semantic associations the names generated, 34.7%
were positive (e.g., frosty and fresh), 51.5% were neutral
(e.g., name and Irish), and 13.8% were negative (e.g., frog
and ?sh).
Design. We used a 2 (sound symbolism of brand name)
# 2 (diagnosticity of brand name) # 2 (timing of diag-
nosticity information) between-subjects design. Sound sym-
bolism was manipulated using Frish and Frosh. We manip-
ulated the diagnosticity of the brand name by informing
participants that the brand name was either a ‘true’ name
(high diagnosticity) or a ‘test’ name (low diagnosticity) at
the time they ?rst encountered the brand name. Participants
in the true name (i.e., high diagnosticity) condition were
informed that the brand name in the press release was the
name that would eventually be used once the product came
to market. Participants in the test name (i.e., low diagnos-
ticity) condition were informed that the brand name in the
press release was only for testing purposes and would not
be the name of the ice cream when it was released to the
public. Further, in order to manipulate the timing of the
diagnosticity information, this information was provided ei-
ther simultaneously with the names of the ice cream or
afterward.
Procedure. One hundred and twenty-six undergraduate
students in a large northeastern university participated in the
study for partial course credit. Participants read a paragraph
stating that a new ice cream was to be introduced into the
area and were asked to read a press release (containing our
manipulations) describing the planned promotional activities
that were to accompany the ice cream debut. After reading
the press release, participants evaluated the ice cream on its
richness, smoothness, and creaminess. They then reported
their overall evaluation of the ice cream, and their intentions
to purchase it. At the end of the questionnaire, participants
reported how much the brand name reminded them of an
ice cream, listed their ice cream consumption, and assessed
the dif?culty of the brand evaluation task. Finally, they were
asked how involved they were in the study and whether they
were aware of using the information gathered fromthe brand
name. The participants were then debriefed and dismissed.
Results
Manipulation and Confound Checks. In order to con-
?rm that our manipulation of the diagnosticity of the ice
cream name worked as intended, we conducted 2 (brand
name) # 2 (diagnosticity) # 2 (timing) ANOVA on the
likelihood that the ice cream name revealed to participants
would be used as the brand name when ice creamwas ?nally
launched. This ANOVA indicated a main effect of diag-
nosticity such that participants who were told that the name
of the ice cream was a test name believed that the brand
name in the press release was less likely to be an ice cream
name than individuals in the true name condition (M’s p
vs. 4.10; , ). No other effect 2.60 F(1, 117) p17.56 p ! .01
was signi?cant ( ). p 1 .05
A similar analysis on ratings of involvement revealed that
participants reported equal involvement in the study irre-
spective of condition (overall , all contrast mean p4.39
), indicating that our manipulation of the diagnosticity F ! 1
of the brand name was not confounded with consumer in-
volvement. Finally, the reported frequency of ice creamcon-
sumption did not vary across experimental conditions.
Hypotheses Tests. We created an Attribute Perception
Index using the attribute evaluations of creaminess, richness,
and smoothness evaluations (Cronbach’s ) and a a p.86
Brand Evaluation Index using an average of the overall
evaluation and purchase intention (Cronbach’s ). a p.81
We used a 2 (sound symbolism of brand name: Frish vs.
Frosh) # 2 (diagnosticity of brand name: test vs. true) #
2 (timing: simultaneous vs. after) ANCOVA (with the level
of involvement in purchase decisions in the product category
as a covariate) to test the hypotheses. For our hypotheses
to be supported, we should obtain a signi?cant three-way
interaction (hypothesis 3), made up of a two-way interaction
between brand name and diagnosticity in the simultaneous
condition (hypothesis 3a), and a main effect of brand name
in the after condition (hypothesis 3b). Note that support for
hypothesis 3a is also support for hypothesis 2, and support
for hypothesis 3b is support for our baseline hypothesis.
(i) Attribute Perception Index. We conducted a 2 #
ANCOVA on the Attribute Perception Index that re- 2 #2
vealed the predicted three-way interaction (F(1, 117) p
, ). In addition, the main effect of brand name 3.84 p ! .05
was signi?cant ( , ). Means are F(1, 117) p9.51 p ! .01
graphically presented in ?gure 1.
In order to investigate the signi?cant three-way interac-
tion, we ?rst examined the simple effects of the 2 (sound
symbolism) # 2 (diagnosticity of the brand name) effects
for the condition when the diagnostic information was si-
multaneously provided with the brand names. This
ANCOVA revealed a signi?cant two-way interaction
( , ), accompanied by a signi?cant F(1, 121) p4.34 p ! .05
main effect of brand name ( , ). F(1, 121) p10.07 p ! .01
When the name was described as true, participants utilized
the sound symbolism and evaluated the associated attributes
for Frosh higher than Frish ( vs. 4.25; contrast M’s p5.06
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 47
FIGURE 1
STUDY 1 RESULTS
NOTE.—Attribute ratings as a function of timing of diagnostic information (Attribute Perception Index).
, ). Individuals presented with a test F(1, 122) p4.43 p ! .05
brand name discounted the diagnosticity of the name and
evaluated Frish equivalent to Frosh on the attributes (Frish
vs. Frosh ; contrast ). These re- M p4.36 M p4.08 F ! 1
sults support hypothesis 2 and hypothesis 3a.
Next, we conducted a similar analysis of the simple 2 #2
effects when diagnostic information was provided after the
sound symbolism was encountered. Consistent with hy-
pothesis 3b, analysis revealed only a main effect of sound
symbolism ( , ). Frosh ( ) F(1, 118) p9.94 p ! .01 M p5.00
was rated higher than Frish ( ) on the Attribute M p4.09
Perception Index regardless of whether the name was de-
scribed as a true or test name. The results support hypothesis
3a. Hypothesis 1 is also further supported as the baseline
sound symbolism effect again manifests.
(ii) Brand Evaluation Index. We predicted a similar
set of results using brand evaluations as the dependent mea-
sure. A ANCOVA on the Brand Evaluation In- 2 #2 #2
dex revealed the predicted three-way interaction
( , ). As with the Attribute Percep- F(1, 118) p4.78 p ! .05
tion Index, the main effect of brand name was signi?cant
( , ); no other effect was signi?cant. F(1, 118) p9.39 p ! .01
Given the signi?cant three-way interaction, we examined
the simple effects of the 2 (sound symbolism) # 2 (diag-
nosticity of the brand name) design separately by the timing
of the diagnostic information. In the condition where the
diagnostic information was provided after the sound sym-
bolism was encountered, analysis revealed a signi?cant in-
teraction ( , ), accompanied by a sig- F(1, 121) p7.77 p ! .01
ni?cant main effect of brand name ( , F(1, 121) p10.83
). When the name was described as true, participants p ! .01
used the sound symbolism and evaluated the associated at-
tributes for Frosh higher than Frish ( vs. 3.97; M’s p5.12
contrast , ). Individuals presented F(1, 122) p8.14 p ! .01
with a test brand name discounted the diagnosticity of the
name and evaluated Frish equivalent to Frosh on the attrib-
utes (Frish vs. Frosh ; contrast ). M p4.50 M p4.16 F ! 1
These results support hypotheses 2 and 3a.
A similar analysis of the simple effects completes 2 #2
the replication of our attribute perception ?ndings for overall
brand evaluations. When the diagnostic information was
provided simultaneously with the sound symbolism of the
brand names, the main effect of sound symbolism was sig-
ni?cant ( , ), such that Frosh F(1, 121) p9.29 p ! .01
( ) was rated higher than Frish ( ) on the M p4.84 M p4.00
Brand Evaluation Index regardless of whether the name was
described as true or test, supporting hypothesis 1 and hy-
pothesis 3b.
Awareness of In?uence of Phonetic Effects. We also
examined participants’ ratings of whether the brand name
affected their evaluations by conducting a 2 #2 #2
ANOVA. These ratings were elicited on a seven-point, se-
mantic differential scale anchored at “Brand names did not
in?uence my evaluations at all” and “Brand names in?u-
enced my evaluations a lot.” This analysis revealed no sig-
ni?cant effect, with the mean for the sample at a 2.68 on a
scale of seven ( ), with a higher number in- midpoint p4
dicating a higher level of awareness. We then computed a
correlation between these ratings and the Attribute Percep-
tion Index (see Wegener, Petty, and Dunn 1998). The cor-
relation was nonsigni?cant ( ), indicating unawareness p 1 .6
on the part of the participants of the in?uence of brand names
on judgments.
Discussion
The results of this study provide evidence that sound
symbolism in?uences attribute perceptions (hypothesis 1).
However, the phonetic effects of brand names can be un-
dercut by alternate information at the time at which brand
48 JOURNAL OF CONSUMER RESEARCH
names are encountered that discredits the effects of sound
symbolism. Our ?ndings suggest that when information re-
garding the diagnosticity of the brand name is provided at
the time the name is encountered, consumers can control
whether or not to process the underlying sound symbolism
(hypothesis 2). However, the use of sound symbolism in
brand perceptions is not completely in the individuals’ con-
trol. The individuals’ inability to discount a brand name that
is demonstrated to be nondiagnostic after their initial en-
counter with the name attests to the uncontrollable manner
in which phonetic effects manifest (hypothesis 3). This study
also provides initial evidence that participants are unaware
of using sound symbolism as an input in judgments. Our
next study tests whether the use of sound symbolism is
effortless, a third element of an automatic process.
STUDY 2
Method
Design. To test hypothesis 4, we used a 2 (sound sym-
bolism of brand name) # 2 (diagnosticity of brand name)
# 2 (cognitive capacity) between-subjects design. Sound
symbolism was again manipulated using Frish and Frosh.
The diagnosticity of the brand name was manipulated as
before by informing participants that the name was a true
versus a test name at the time the brand name information
was encountered (i.e., the simultaneous condition in study
1). Finally, participants were read a list of numbers and had
to count the number of times the experimenter said the word
?ve in the condition in which cognitive capacity was im-
paired. Participants in the normal cognitive capacity con-
dition did not undertake this task.
Procedure. One hundred and eleven undergraduate stu-
dents in a large West Coast university participated in the
study for partial course credit. Unlike the previous study,
this study was administered on computer. As before, par-
ticipants read a paragraph stating that a new ice cream was
to be introduced in the area and were asked to read a press
release describing the planned promotional activities that
were to accompany the ice cream debut. Again, participants
in the true name condition were informed that the brand
name in the press release was the name that would eventually
be used once the product came to market, while the partic-
ipants in the test name condition were informed that the
brand name in the press release was only for testing purposes
and would not be the name of the ice cream when it was
released to the public. Unbeknownst to participants, the time
taken to read this initial press release was monitored. After
reading the press release, participants evaluated the ice
cream overall, rated the ice cream on its richness, smooth-
ness, and creaminess and stated their purchase intention.
Finally, attribute importance ratings were collected. All eval-
uations were elicited on 100-point sliding rule scales. Par-
ticipants were then debriefed and dismissed.
Results
Manipulation Checks. To test the success of the ma-
nipulation impairing cognitive capacity, we used the time
taken to read the article about the ice cream as an indication
of task dif?culty. A ANOVA on the reading time 2 #2 #2
indicated a signi?cant main effect of the cognitive capacity
manipulation ( , ) with participants F(1, 103) p7.09 p ! .01
in the normal cognitive capacity condition taking less time
( sec.) than those in the cognitive capacity im- M p30.33
paired condition ( sec.).
1
Thus, our manipulation M p37.45
to impair cognitive capacity was successful.
Hypotheses Tests. We used a 2 (sound symbolism of
brand name: Frish vs. Frosh) # 2 (diagnosticity of brand
name: test vs. true) # 2 (cognitive capacity: normal vs.
impaired) ANOVA to test hypothesis 4. As in study 1, we
created an Attribute Perception Index using the attribute
evaluations of creaminess, richness, and smoothness (Cron-
bach’s ). For hypothesis 4 to be supported, we a p.82
should obtain a signi?cant three-way interaction, comprising
of a two-way interaction between brand name and diag-
nosticity in the normal cognitive capacity condition (hy-
pothesis 2), and a main effect of brand name in the low
cognitive capacity condition (hypothesis 1).
A ANOVA on the Attribute Perception Index 2 #2 #2
revealed the predicted three-way interaction (F(1, 103) p
, ; see ?g. 2). In addition, the main effect of 3.84 p ! .05
brand name was signi?cant ( , ). No F(1, 103) p8.03 p ! .01
other effect was signi?cant.
Given the signi?cant three-way interaction, we investi-
gated the simple effects separately in the two cognitive load
conditions. A 2 (sound symbolism) # 2 (diagnosticity of
the brand name) analysis in the normal cognitive capacity
condition revealed a signi?cant interaction (F(1, 107) p
, ) accompanied by a signi?cant main effect of 4.84 p ! .05
brand name ( , ). An examination of F(1, 107) p7.68 p ! .01
the means indicates that under normal cognitive capacity
and when the name was described as true participants util-
ized the sound symbolism and evaluated the associated at-
tributes for Frosh higher than Frish ( vs. 49.2; M p63.3
contrast , ). Individuals presented F(1, 108) p5.47 p ! .05
with a test brand name discounted the diagnosticity of the
name and evaluated Frish equivalent to Frosh on the attrib-
utes (Frish vs. Frosh ; contrast ). M p54.9 M p50.3 F ! 1
These results are consistent with hypothesis 2 and support
hypothesis 4a.
Analysis of the simple effects for the impaired 2 #2
cognitive capacity condition revealed only a signi?cant main
effect of sound symbolism ( , ). Re- F(1, 107) p7.81 p ! .01
gardless of whether the name was described as true or test,
when respondents were cognitively impaired, Frosh (M p
) was rated higher than Frish ( ) on the At- 62.4 M p50.0
1
These results were replicated through a self-reported measure of task
dif?culty.
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 49
FIGURE 2
STUDY 2 RESULTS
NOTE.—Attribute ratings as a function of cognitive capacity (Attribute Perception Index).
tribute Perception Index; this is consistent with hypothesis
1 and supports hypothesis 4b.
2
Discussion
The results of this study indicate that participants were
unable to actively discount the brand name information
when they were cognitively impaired. We thus obtain evi-
dence for the use of sound symbolism being effortless such
that even when cognitively impaired, the effects ?lter
through to judgments. The results of this study provide cor-
roborative support to the prior study’s ?ndings that the in-
ferences of attribute information from brand names and the
subsequent evaluations of the brand are automatic processes.
GENERAL DISCUSSION
In this article, we provide support for the proposition that
consumers gather and process information from brand
names in an automatic manner. However, consumers can
control for this provided they process information that dis-
credits the brand name at the time the name is encountered.
An important aspect of sound symbolism established in this
article is how it affects consumers’ attitudes at the attribute
level (hypothesis 1). Consumers may knowingly control for
the process when they have access to information that credits
or discredits the brand name at the time at which they en-
counter the brand name (hypothesis 2). Yet, consumers are
not fully aware of this process and cannot fully control for
the effect; we show evidence for the uncontrollability of the
effect unless the name is discredited at the time of encoding
(hypothesis 3) and the consumer is not cognitively impaired
2
A similar analysis using the Brand Evaluation Index (Cronbach’s
) revealed an identical pattern of results; we do not report these a p.75
results here due to space constraints.
(hypothesis 4). These results provide evidence for the au-
tomatic processing of brand names and their phonetic
effects.
Our purpose in this article was to understand the process
underlying sound symbolism. The basic phenomenon of
sound symbolism has been demonstrated for a wide range
of sounds across multiple dimensions (Klink 2000). In our
article, we focused on the internal validity of our studies in
order to build theory about the sound symbolism process.
We therefore tested one set of brand names and used a
between-subjects test of sound symbolism. This strengthens
the existing case for sound symbolism since participants are
not aware that sounds are being manipulated or that one
brand name should provide different information than an-
other. Further, we increase the internal validity of the ex-
periments by testing for phonetic effects when the brand
name is not the only information provided to the subject.
Additionally, we show that this effect carries through at-
tribute perceptions to the overall evaluation of the brand.
Demonstrating these effects of sound symbolism increases
our con?dence in the robustness of our theory.
Automatic processes in social cognition domains have
been demonstrated empirically in various contexts (see
Bargh 1989). This article adds to this literature by dem-
onstrating the automatic use of sound symbolism. This is
an important issue to examine, given that if phonetic effects
manifest even partially automatically, then this automatic
effect will exert a more consistent in?uence over attitudes
and behavior over time (see Bargh 2002). Since sound sym-
bolism is a fairly common naming device (Begley 2002), it
is possible that phonetic links are highly accessible. Atti-
tudes relating to the phonetic aspects of the brand name
would be automatically activated. Considering the limited
capacity consumers usually devote to product evaluation
(Hoyer 1984), the results of studies 1 and 2 indicate that
50 JOURNAL OF CONSUMER RESEARCH
under this diminished cognitive capacity, the phonetic ef-
fects of brand names may be activated and in?uence choice.
If brand name phonetic cues are chronically accessible, con-
gruency of brand name and product would be of even greater
importance.
Our studies tested names in which only one phoneme was
manipulated. In most categories, where multisyllabic words
are the norm, sound symbolism may occur in different po-
sitions and/or on multiple positions in the same word. This
?exibility would not only increase the complexity of the
message conveyed by sound symbolism but also may in-
crease the strength of the effect. This article has focused
primarily on the sound symbolism provided by the vowels
of a brand name. The role consonants play in brand names
and their effects on product evaluations is an underresear-
ched area. Past research has shown that brand names that
begin with hard consonants (i.e., Kodak and Pepsi), elicit
higher recognition (Vanden Bergh et al. 1984), and occur
more frequently than we would expect from the English
lexicon (Schloss 1981). Although linguistic anthropologists
have noted phonetic patterns for consonants (i.e., harder
sounding consonants tend to represent harder, sharper ob-
jects than softer sounding consonants), these ?ndings have
not been directly tested on brand names. A systematic study
of consonants and their interaction with multiple vowels is
an avenue worthy of future research.
Our studies demonstrated that sound symbolism, although
evaluated on the attribute level, affects overall evaluations.
We demonstrated that sound symbolism affected the attrib-
utes of creaminess, sweetness, and richness: three category
speci?c attributes. It appears that consumers will take the
broader dimensions that sound symbolism operates upon
(e.g., smallness, lightness, mildness, fastness, etc.) and ex-
trapolate these ?ndings to the speci?c properties of ice
cream. Creating a successful brand name depends not only
upon the creation of a name that is congruent with the prod-
uct category, but one that phonetically ?ts the positioning
of the brand within that product category. Future research
could examine whether it is possible to phonetically ma-
nipulate meaning over a wide range of attributes in a broad
set of product categories.
The automatic nature of the sound symbolism leads to
important implications of this research upon the use of per-
sonal names as brand names. For example, if the ice cream
was named for its maker, Joseph Frish, then the brand name
should be seen as less diagnostic, and the phonetic effect
should not appear. This assumes that the information as to
the name’s origin is known at the time of encoding and that
processing capabilities are not constrained. Telling the con-
sumer that the ice cream was named for the maker after the
ice cream has been tasted should have little effect upon
consumers’ use of sound symbolism in accordance with our
?ndings. An interesting further study would be where we
manipulate the diagnosticity of the name through the use of
a personal name versus a purely ?ctitious name.
Over 2,000 years ago, Plato supported the onomatopoetic
theory of language by arguing that words were formed by
the imitation of ideas in sounds (Jowett 1953). The argument
still holds. Sound symbolism shows that the sounds of words
still convey meaning. More important, consumers recognize
this phonetic meaning and will use it as a source of product
information. The effects of time, the in?uence of foreign
languages, the desire of euphony, and elements of chance
all distill and, at times, distort the effects of sound sym-
bolism. Yet even with these caveats, attending to phonetic
meanings and leveraging the attribute associations when cre-
ating a brand image is sound advice when creating a new
brand name.
[David Glen Mick and Dawn Iacobucci served as editors
and Punam Anand Keller served as associate editor for
this article.]
REFERENCES
Bargh, John A. (1989), “Conditional Automaticity: Varieties of
Automatic In?uence in Social Perception and Cognition,” in
Unintended Thought, ed. James S. Uleman and John A. Bargh,
New York: Guilford, 3–51.
——— (2002), “Losing Consciousness: Automatic In?uences on
Consumer Judgment, Behavior, and Motivation,” Journal of
Consumer Research, 29 (September), 280–85.
Bargh, John A. and Roman D. Thein (1985), “Individual Construct
Accessibility, Person Memory, and the Recall-Judgment Link:
The Case of Information Overload,” Journal of Personality
and Social Psychology, 49 (November), 1129–46.
Begley, Sharon (2002), “StrawBerry Is No BlackBerry: Building
Brands Using Sound,” Wall Street Journal, August 26, B1.
Consumer Reports (1994), “Ice Cream: What’s the Scoop?” Con-
sumer Reports, June, 373–77.
Gilbert, Daniel T. (1989), “Thinking Lightly about Others: Auto-
matic Components of the Social Inference Process,” in Un-
intended Thought, ed. James S. Uleman and John A. Bargh,
New York: Guilford, 189–211.
Gilbert, Daniel T. and J. Gregory Hixon (1991), “The Trouble of
Thinking: Activation and Application of Stereotypic Beliefs,”
Journal of Personality and Social Psychology, 60 (4), 509–17.
Gilbert, Daniel T., Brett W. Pelham, and Douglas S. Krull (1988),
“On Cognitive Busyness: When Person Perceivers Meet Per-
sons Perceived,” Journal of Personality and Social Psychol-
ogy, 54 (May), 733–40.
Hinton, Leanne, Johanna Nichols, and John J. Ohala, eds. (1994),
Sound Symbolism, Cambridge: Cambridge University Press.
Hoyer, Wayne D. (1984), “An Examination of Consumer Decision
Making for a Common Repeat Purchase Product,” Journal of
Consumer Research, 11 (December), 822–29.
Huang, Yau-Huang, Sawat Pratoomaraj, and Ronald C. Johnson
(1969), “Universal Magnitude Symbolism,” Journal of Verbal
Learning and Verbal Behavior, 8 (1), 155–56.
Jacobson, Roman and Linda R. Waugh (1987), The Sound Shape
of Language, 2d ed., New York: Mouton de Gruyter.
Jowett, Benjamin (1953), The Dialogues of Plato: Translated into
English with Analyses and Introductions, London: Oxford
University Press.
Klank, Linda J. K., Yau-Huang Huang, and Ronald C. Johnson
(1971), “Determinants of Success in Matching Word Pairs in
Tests of Phonetic Symbolism,” Journal of Verbal Learning
and Verbal Behavior, 10 (2), 140–48.
Klink, Richard R. (2000), “Creating Brand Names with Meaning:
PHONETIC EFFECTS OF BRAND NAMES ON JUDGMENTS 51
The Use of Sound Symbolism,” Marketing Letters, 11 (1),
5–20.
——— (2001), “Creating Meaningful New Brand Names: AStudy
of Semantics and Sound Symbolism,” Journal of Marketing:
Theory and Practice, 9 (Spring), 27–34.
Lapolla, Randy J. (1994), “An Experimental Investigation into
Phonetic Symbolism as It Relates to Mandarin Chinese,” in
Hinton et al., eds.
Makino, Seiichi, Seiichi Nakada, and Mieko Ohso (1999), Ko-
dansha’s Basic English-Japanese Dictionary, Tokyo: Kodan-
sha Int.
O’Grady, William, Michael Dobrovolsky, and Francis Katumba
(1997), Contemporary Linguistics: An Introduction, 3d. ed.,
New York: St. Martin’s.
Ohala, John J. (1994), “The Frequency Code Underlies the Sound-
Symbolic Use of Voice Pitch,” in Hinton et al., eds.
Sapir, Edward (1929), “A Study in Phonetic Symbolism,” Journal
of Experimental Psychology, 12 (3), 225–39.
Schloss, Ira (1981), “Chickens and Pickles,” Journal of Advertising
Research, 21 (6), 47–49.
Schmitt, Bernd H., Yigang Pan, and Nader T. Tavassoli (1994),
“Language and Consumer Memory: The Impact of Linguistic
Differences between Chinese and English,” Journal of Con-
sumer Research, 21 (December), 419–31.
Schmitt, Bernd H. and Shi Zhang (1998), “Language Structure and
Categorization: A Study of Classi?ers in Consumer Cogni-
tion, Judgment, and Choice,” Journal of Consumer Research,
25 (September), 108–22.
Tavassoli, Nader T. and Jin K. Han (2001), “Scripted Thought:
Processing Korean Hancha and Hangul in a Multimedia Con-
text,” Journal of Consumer Research, 28 (December),
482–93.
Vanden Bergh, Bruce G., Janay Collins, Myrna Schultz, and Keith
Adler (1984), “Sound Advice on Brand Names,” Journalism
Quarterly, 61 (4), 835–40.
Wegener, Duane T., Richard E. Petty, and Meghan Dunn (1998),
“The Metacognition of Bias Correction: Naive Theories of
Bias and the Flexible Correction Model,” in Metacognition:
Cognitive and Social Dimensions, ed. Guy Lories, Benoit
Dardenne, and Vincent Y. Yzerbyt, Thousand Oaks, CA: Sage,
202–27.
doc_800567711.pdf