See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/23547268

Self-Construal, Reference Groups, and Brand Meaning

Article  in  Journal of Consumer Research · December 2005

DOI: 10.1086/497549 · Source: RePEc

CITATIONS READS
1,560 7,576

2 authors:

Jennifer Edson Escalas James R Bettman

Vanderbilt University Duke University
37 PUBLICATIONS   6,332 CITATIONS    141 PUBLICATIONS   31,594 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by James R Bettman on 17 October 2017.

The user has requested enhancement of the downloaded file.

Self-Construal, Reference Groups, and
Brand Meaning
JENNIFER EDSON ESCALAS
JAMES R. BETTMAN*

We propose that consumers purchase brands in part to construct their self-

concepts and, in so doing, form self-brand connections. We focus on reference
groups as a source of brand meaning. Results from two studies show that brands
with images consistent with an ingroup enhance self-brand connections for all
consumers, whereas brands with images that are consistent with an outgroup have
a stronger negative effect on independent versus interdependent consumers. We
propose that this differential effect is due to stronger self-differentiation goals for
consumers with more independent self-concepts. We also find greater effects for
more symbolic than for less symbolic brands.

P eople engage in consumption behavior in part to con-

struct their self-concepts and to create their personal
identity (e.g., Belk 1988; Richins 1994). We examine one
THEORETICAL DEVELOPMENT
Brands, Meaning Transfer, and Self-Brand
aspect of this construction process, namely, the appropria- Connections
tion of brand associations derived from the usage of these
brands by reference groups. Building on McCracken’s Possessions can be used to satisfy psychological needs,
(1988) theory of meaning movement, we propose that the such as actively creating one’s self-concept, reinforcing and
symbolic properties of reference groups become associated expressing self-identity, and allowing one to differentiate
with the brands those groups are perceived to use. These oneself and assert one’s individuality (e.g., Ball and Tasaki
associations can then be transferred from reference groups 1992; Belk 1988; Kleine, Kleine, and Allen 1995). Posses-
to consumers as consumers select brands with meanings sions can also serve a social purpose by reflecting social
congruent with an aspect of their self-concept, in this article ties to one’s family, community, and/or cultural groups, in-
operationalized as independent versus interdependent self- cluding brand communities (Muniz and O’Guinn 2001).
construals. When brand associations are used to construct Consumer researchers have extended some of these pos-
the self or to communicate the self-concept to others, a session findings to brands (Escalas and Bettman 2003; Four-
connection is formed with the brand. Two studies provide nier 1998). For example, recent research indicates that con-
empirical support for the notion that brands used by ingroups sumers construct their self-identity and present themselves
to others through their brand choices based on the congru-
enhance consumers’ self-brand connections, whereas brands
ency between brand-user associations and self-image as-
used by outgroups detract from such connections. We show
sociations (Escalas and Bettman 2003).
further that this influence of ingroup versus outgroup brand Levy (1959) asserted that people do not buy products just
usage differs depending on whether the consumer has a for what they do, but also for what the product means; thus,
primarily independent or interdependent self-construal. Fi- brands can be symbols whose meaning is used to create and
nally, we show that these effects are moderated by the degree define a consumer’s self-concept. McCracken’s (1988)
to which the brands are symbolic, that is, communicate model of meaning transfer asserts that such meaning orig-
something about the user. inates in the culturally constituted world, moving into goods
via the fashion system, word of mouth, reference groups,
subcultural groups, celebrities, and the media. For example,
*Jennifer Edson Escalas is associate professor of management at the
Owen Graduate School of Management, Vanderbilt University, Nashville,
meanings “get into” a brand through advertising because
TN, 37203 (jennifer.escalas@owen.vanderbilt.edu). James R. Bettman is ads reference the general cultural symbols needed to provide
the Burlington Industries Professor at the Fuqua School of Business, Duke meaning. Similarly, reference group usage of a brand pro-
University, Durham, NC 27708-0120 (jrb12@mail.duke.edu). The authors vides meaning via the associations consumers hold regard-
would like to thank the editor, associate editor, and three reviewers for
their constructive suggestions.
ing that group (Muniz and O’Guinn 2001). Next, meaning
moves from goods to consumers, as consumers construct
378

䉷 2005 by JOURNAL OF CONSUMER RESEARCH, Inc. ● Vol. 32 ● December 2005

All rights reserved. 0093-5301/2005/3203-0005$10.00
SELF-CONSTRUAL, REFERENCE GROUPS, AND BRAND MEANING 379

themselves through their brand choices based on congruency the outgroup symbolism in constructing one’s possible self.
between brand image and self-image. Thus, the meaning For example, if I am not a member of a fraternity (and do
and value of a brand is not just its ability to express the not desire to be a member) and see fraternity members wear-
self, but also its role in helping consumers create and build ing Polo clothing, I may specifically choose not to wear
their self-identities (McCracken 1989). Polo clothing in an attempt to distance myself from the
Reference groups can be a critical source of brand mean- fraternity symbolism of the Polo brand. Thus, the type of
ings. Consumers use others as a source of information for group associated with the brand (ingroup versus outgroup)
arriving at and evaluating one’s beliefs about the world, will moderate the effect of brand associations on self-brand
particularly others who share beliefs and are similar on rel- connections.
evant dimensions. Consumer research on reference groups Similarly, if a brand is not typically associated with an
has demonstrated congruency between group membership ingroup (e.g., its image is incongruent with the group), this
and brand usage (e.g., Bearden and Etzel 1982; Bearden, may negatively affect self-brand connections. The same
Netemeyer, and Teel 1989; Burnkrant and Cousineau 1975; identification processes that lead to a connection with a
Childers and Rao 1992; Moschis 1985) and has defined brand associated with an ingroup lead to rejection of a brand
several types of social influence (e.g., Bearden and Etzel with an image incongruent with the ingroup. However, if a
1982; Park and Lessig 1977). We elaborate on value- brand’s image does not match an outgroup, the prediction
expressive reference group influences, characterized by the is not quite as clear. The lack of match may actually be
need for psychological association with a group either to viewed favorably, based on balance-theory considerations
resemble the group or due to a liking for the group. We (Heider 1946), thus leading to enhanced self-brand connec-
provide an empirical demonstration of these ideas by dem- tions. Alternatively, the lack of match may simply be viewed
onstrating that brand use by reference groups is a source of as irrelevant, leading to no effect on self-brand connections.
brand meaning. Consumers form associations between ref- We hypothesize that the balance-theory view will be more
erence groups and the brands they use and transfer these descriptive and that brand associations that are incongruent
meanings from brand to self by selecting brands with mean- with an outgroup will have a favorable impact on self-brand
ings relevant to an aspect of their current self-concept or connections. In sum, we have:
possible self.
H1a: Brand associations consistent with an ingroup
A critical distinction in terms of such self-construction will have a favorable effect on self-brand con-
processes is that between the use of brand associations de- nections, whereas brand associations inconsis-
riving from one’s own group (an ingroup) versus groups to tent with an ingroup will have an unfavorable
which one does not belong (an outgroup). Consumers are effect on self-brand connections.
likely to accept meanings from brands associated or con-
sistent with an ingroup and reject meanings associated or H1b: Brand associations consistent with an outgroup
consistent with an outgroup. Consumers form connections will have an unfavorable effect on self-brand
to brands that become meaningful through this process; self- connections, whereas brand associations incon-
brand connections are intended to measure the extent to sistent with an outgroup will have a favorable
which individuals have incorporated brands into their self- effect on self-brand connections.
concept (Escalas and Bettman 2003). Therefore, given our
focus on the self-construction processes using brands, rather Note that hypotheses 1a and 1b predict a two-way interaction
than information processing in response to an advertisement, of ingroup versus outgroup and the degree to which brand
the primary dependent variable in our studies is a measure associations are consistent (brand image matches) or incon-
of the degree to which consumers have formed a self-brand sistent (brand image does not match) with the group.
connection.
If reference groups use and become associated with par-
ticular brands (i.e., the brand’s image is consistent with or
The Role of Self-Construal
matches the group), such meaning may be appropriated by Brands become linked to the self when a brand is able to
consumers as they construct their self-identities. For ex- help consumers achieve goals that are motivated by the self.
ample, if I consider myself to be an intellectual and my For example, brands can be used to meet self-expression
member group of intellectuals tends to drive Volvo auto- needs, publicly or privately. They can serve as tools for
mobiles, I also may choose to drive a Volvo car as a symbol social integration or connecting us to the past. They may
of how intellectual I am. As a result, consumers may form act as symbols of personal accomplishment, provide self-
self-brand connections to the brands used by reference esteem, allow one to differentiate oneself and express in-
groups to which they belong. Conversely, consumers may dividuality, and help people through life transitions. A major
avoid associations derived from groups to which they do focus of our article is how self-construction goals differ
not belong. When outgroup members use a brand, consum- depending upon a consumer’s self-construal. We propose
ers may form associations about the brand that they would that the influence of ingroup and outgroup brand usage will
not like to have transferred to themselves. Nevertheless, the differ depending upon whether the consumer has a primarily
brand becomes meaningful through the process of avoiding independent or interdependent self-construal.
380 JOURNAL OF CONSUMER RESEARCH

Social cognition research on the self has developed a (Kampmeier and Simon 2001), we can make more detailed
variety of theoretical constructs to explain the complex na- predictions about the combined influence of independent
ture of self-knowledge and self-related behavior. The self is versus interdependent self-construal, ingroup versus out-
conceptualized as consisting of multiple aspects (Linville group, and whether a brand matches the image of a group.
1987), including social roles and personality traits, the most The most clear predictions relate to the case of outgroup
important of which are schematic self-aspects (Markus brand associations. Kampmeier and Simon (2001) show that
1977), and possible selves, that is, individuals’ ideas of what when the focus is on a comparison to an outgroup, the
they might become, what they would like to become, and differentiation aspect of the individual self is emphasized.
what they are afraid of becoming (Markus and Nurius 1986). Thus, for more independent individuals, comparison to the
In this article, we focus on two aspects of self, independent outgroup should lead to a heightened need to differentiate
and interdependent selves, and examine the effects of dif- from the outgroup to create a unique self-concept. However,
ferences between individuals in the extent to which they are people with more interdependent self-construals should be
schematic on these two aspects. more immune to outgroup brand associations, as their pri-
Although the self-concept often is considered to be dis- mary motivation stems from forming relationships within
tinct from other people’s self-concepts, recent cross-cultural the ingroup. This implies that a brand associated with the
evidence suggests that individuals’ mental representations outgroup should lead to lower self-brand connections for
of self may depend on social aspects of self, such as rela- more independent individuals than for more interdependent
tionships with others and membership in social groups individuals. There is not a clear argument for differential
(Brewer and Gardner 1996; Markus and Kitayama 1991; effects for the ingroup related to degree of independence
Triandis 1989). Such research indicates that on average, versus interdependence. Thus, we hypothesize:
Westerners tend to focus on the personal self, thinking of
H2: Brand associations consistent with an outgroup
themselves in terms of unique personal traits and attributes
will lead to lower self-brand connections for in-
and de-emphasizing others (independent self-construal),
dependent self-construals than for interdependent
whereas Easterners tend to focus on the social self and
self-construals.
how the self is related to other people (interdependent self-
construal; Markus and Kitayama 1991). For example, Trian- Note that hypothesis 2 implies a three-way interaction of
dis (1989) argues that more individualistic cultures are char- ingroup versus outgroup, independent versus interdepen-
acterized by more focus on the private self and less emphasis dent, and brand image matching versus not matching that
on the collective self, with increased emphasis on the col- of the group on self-brand connections, and the specific
lective self for less individualistic cultures. interaction we propose relies on both self-construal differ-
These two aspects of self can coexist within the individual ences and ingroup versus outgroup considerations.
(Aaker and Lee 2001; Brewer and Gardner 1996) and can
vary across ethnocultural background within Western so-
ciety. That is, individuals may have both independent and Brand Symbolism
interdependent aspects of self but may differ in the relative Our basic premise is that consumers appropriate the mean-
strength of those aspects on a chronic basis, leading to in- ing of brands as they construct their self-identities, partic-
dividual differences in self-construal that can be assessed ularly brand meaning that arises from reference group use
(Singelis 1994). There are also likely to be chronic differ- and nonuse of brands. Thus far we have treated all brands
ences in the relative strength of independent versus inter- equally in our hypotheses. However, some brands are better
dependent self-construals across ethnocultural backgrounds able than others to communicate something about the person
within American society. For example, compared to Asian using them. For example, prior consumer research proposes
Americans and Hispanic Americans, whites will be rela- that publicly consumed (vs. privately consumed) and luxury
tively more independent and less interdependent in their self- (vs. necessity) products are better able to convey symbolic
construals (Aaker and Schmitt 2001; Lee, Aaker, and Gard- meaning about an individual (Bearden and Etzel 1982). Ad-
ner 2000; Markus and Kitayama 1991; Triandis 1989). In ditionally, a brand that is very popular and used by many
this article, we focus on differences in self-construals due different types of people (e.g., a Honda Accord automobile)
to ethnocultural background differences and individual dif- may not communicate specific associations about the person
ferences (i.e., chronic tendencies). These differences are im- who uses it. Therefore, we expect the basic effect postulated
portant, because independent self-construals can lead to in hypotheses 1a and 1b to be moderated by the degree to
motivations different from interdependent self-construals. which a brand is perceived to be symbolic, that is, able to
Independent self-construal goals include both independence communicate something about the individual using the
(i.e., self-determination) and differentiation (i.e., distinc- brand. Consumers will be more likely to form self-brand
tiveness), whereas interdependent self-construal goals focus connections to symbolic brands with appropriate associa-
on aspects of self shared with some subset of others, en- tions as they construct their self-identities than with brands
hancing maintenance of relationships (Aaker and Schmitt that do not communicate much about the self-identity of the
2001; Kampmeier and Simon 2001). user. Conversely, consumers will be more likely to reject
By considering different facets of the independent self forming a self-brand connection with symbolic brands with
SELF-CONSTRUAL, REFERENCE GROUPS, AND BRAND MEANING 381

inappropriate associations than with nonsymbolic brands. Method

Thus, we hypothesize a three-way interaction between group
type, brand image match, and brand symbolism.
Participants. Three hundred and eighty-eight under-
graduate students at a public university in the Southwest
H3: The effects on self-brand connections postulated participated in this study to meet an introductory marketing
in hypotheses 1a and 1b will be moderated by the course requirement. Data were collected over two sequential
degree to which a brand is perceived to be sym- semesters. One hundred participants had to be eliminated
bolic, that is, able to communicate something from the data set due to incomplete or improper responses,
about the user’s self-identity, with more symbolic leaving a total of 288 participants.1
brands having a more pronounced effect com-
pared to less symbolic brands. Procedure. This study used a Visual Basic program
that allowed us to customize the study based on participants’
responses. The program began with a short study introduc-
Note that we expect that symbolic brands will show stron- tion, after which participants entered a group to which they
ger effects for both independent and interdependent self- belonged (i.e., an ingroup; “In the box below, we would
construals, so we do not predict a four-way interaction. like you to type in the name of a group on campus that you
To test our hypotheses, we conduct two studies that ex- belong to and feel a part of. You should feel you are this
amine differences in the effect on self-brand connections type of person and that you fit in with these people. This
of brand associations arising from reference groups for group should be a tightly knit group, consisting of individ-
individuals with independent versus interdependent self- uals who are very similar to one another.”). Next they en-
concepts. In the first study, we compare Asian and Hispanic tered a group to which they did not belong (i.e., an outgroup;
American participants to white American participants; “In this box, we would like you to type in the name of a
within American society, Asians and Hispanics tend to have group on campus that you do not belong to and do not feel
more interdependent self-construals, while whites have more a part of. You should feel you are not this type of person
independent self-construals (e.g., Aaker and Schmitt 2001; and that you do not fit in with these people. This group
we also check this below). In study 2, we examine inde- should be a tightly knit group, consisting of individuals who
pendent-interdependent differences independently of eth- are very similar to one another.”).
nicity by measuring chronic tendencies toward independent After each group, participants were asked to list one brand
and interdependent self-construals (Singelis 1994) and lim- that was consistent with the group and one brand that was
iting our participants to those who are high on either one not (“In the box below, we would like you to type in a brand
or the other type of self-construal, but not both. Study 2 that is consistent with the group that you belong to. This
also adds a measure of the degree to which our participants can be a brand that members of the group actually use or
believe the brands they listed communicate something about it can be a brand that shares the same image as the group.
the brand’s user; these measures enable us to test hypothesis A brand is considered to be a name or symbol that distin-
3. In addition, we include thought protocols and other mea- guishes one seller’s goods from another’s”; and “Now, we
sures in study 2 to provide some insight into the processes would like you to type in a brand that is NOT consistent
influencing the determination of self-brand connections. with the group you belong to. This can be a brand that
members of the group would never use or it can be a brand
that has the opposite image from the group.”). Thus, each
STUDY 1 participant entered four brands, corresponding to four group-
brand pairs: ingroup-brand matches, ingroup-brand does not
match, outgroup-brand matches, and outgroup-brand does
In this study, we examine the influence of ingroups and not match. Next, participants completed a series of scale
outgroups on self-brand connections due to differences in questions indicating the degree to which they fit with each
independent and interdependent selves that arise from eth- group. After a short, unrelated filler task designed to reduce
nocultural background differences within the United States. potential demand effects, participants rated the degree to
Specifically, we compare Asian and Hispanic consumers to which they had self-brand connections with these four
white consumers, based on previous research that shows that brands (see details below). This was followed by some
Asian and Hispanic Americans generally have more inter- additional manipulation checks and demographic informa-
dependent self-construals, while white Americans have more tion, including ethnicity: African American, Asian, His-
independent self-construals (Aaker and Schmitt 2001; Lee panic, white, and “other.” The program ended with a de-
et al. 2000; Markus and Kitayama 1991; Triandis 1989). We briefing statement. The entire procedure took approximately
also test the degree to which these groups differ on inde- one half hour.
pendence and interdependence. Although there is less var-
iation on independence and interdependence within Amer- Independent Variables. Participants were divided into
ican ethnocultural groups than between cross-cultural groups
(Triandis 1993), this limitation makes our study a conser- 1
Examples of improper responses include listing the same brand twice,
vative test of our hypotheses. listing adjectives rather than brands, or typing in “I don’t know.”
382 JOURNAL OF CONSUMER RESEARCH

two groups based on their self-reported ethnicity. Asians ments” and “It is important for me to respect decisions made
and Hispanics make up the interdependent ethnocultural by the group,” a p .58) and two of the independent scale
group, while whites form the independent ethnocultural items (“I’d rather say ‘No’ directly than risk being misun-
group. Participants who checked African American or derstood” and “I enjoy being unique and different from
“other” (or left the question blank) were eliminated from others in many respects,” a p .44 ). These items were cho-
the data set, leaving a total of 243 participants. As noted sen based on their having the highest item-to-total corre-
above, during the study each participant entered two groups, lations with the corresponding twelve-item Singelis (1994)
an ingroup and an outgroup. Additionally, for each group, scale in previous research. Further checks measured the de-
participants entered a brand consistent with the image of the gree to which the participant belongs to the ingroup and the
group and a brand not consistent with the image of the group outgroup he/she entered, assessed using the average of three
(for examples of groups and brands listed, see table 1), items (“I consider myself to be this type of person,” “I
resulting in a set of four brands. The groups and brands are belong to this group,” and “I fit in with this group of people,”
idiosyncratic to each participant and are not of interest in anchored by strongly disagree [0]/strongly agree [100];
the analysis; the data are coded only by group type (ingroup a p .96).
vs. outgroup) and brand image match (image matches versus
image does not match). Thus, we have a 2 # 2 # 2 mixed
design in our study, with ethnocultural group (independent Results
vs. interdependent) as a between-subjects variable and group The model used in the analyses to predict self-brand con-
type and brand image match as within-subjects variables. nections is a mixed ANOVA model, with ethnocultural
Dependent Variable. Self-brand connections were group (white vs. Asian/Hispanic, or independent vs. inter-
measured using seven items (anchored by strongly disagree dependent) as a between-subjects factor and group type (in-
[0] to strongly agree [100]; see Escalas and Bettman 2003), group vs. outgroup) and brand image match (match vs. no
averaged to form one self-brand connection score per par- match) as within-subject factors. The dependent variable
ticipant per brand (a p .96): was self-brand connections.

1. This brand reflects who I am. Manipulation Checks. White participants exhibited
2. I can identify with this brand. significantly stronger independent selves on the two
3. I feel a personal connection to this brand. Singelis scale items compared to the Asian/Hispanic
4. I use this brand to communicate who I am to other participants (white p 65.69, Asian/Hispanic p 58.41,
people. F(1, 241) p 8.43, p ! .01). However, the Asian/Hispanic
5. I think this brand helps me become the type of person participants were only directionally higher on the inter-
I want to be. dependent-self measure compared to the white participants
6. I consider this brand to be “me” (it reflects who I (Asian/Hispanic p 61.02, white p 57.08, F(1, 241) p
consider myself to be or the way that I want to present 1.54, p p .22), making our tests more conservative. Par-
myself to others). ticipants considered themselves to belong to the ingroup
7. This brand suits me well. significantly more than they felt they belonged to the out-
group (ingroup p 82.74, outgroup p 15.90, F(1, 241) p
1687.31, p ! .001).
Manipulation Checks. To check whether our ethno-
cultural groups vary on the degree to which they have in- Hypotheses 1a and 1b. These hypotheses predict
dependent versus interdependent self-construals, we had par- a two-way interaction of group type and brand image
ticipants complete two of the Singelis (1994) interdependent match, such that whether or not brand associations match
scale items (“I often have the feeling that my relationships a group’s image has a different effect on self-brand con-
with others are more important than my own accomplish- nections depending on whether the group is an ingroup or

TABLE 1

EXAMPLES OF GROUPS AND BRANDS LISTED BY PARTICIPANTS

Participant Brand with associations Brand with associations

number Type of group Group listed that match group that do not match group

1 Ingroup Conservatives Polo Pacific Sunwear

Outgroup Hippies Birkenstock Banana Republic
2 Ingroup Chi-Omega Gap Prada
Outgroup Kappa Kappa Gamma Burberry Old Navy
3 Ingroup Academic group Guess Levi’s
Outgroup Athletic group Nike Gucci
4 Ingroup Business school IBM Martha Stewart
Outgroup Basketball team Nike Doc Martens
SELF-CONSTRUAL, REFERENCE GROUPS, AND BRAND MEANING 383

outgroup. We find a significant interaction of group type by FIGURE 2

brand image match on self-brand connections (F(1, 723) p
SELF-BRAND CONNECTIONS ACROSS TYPE OF GROUP BY
394.93, p ! .001; see fig. 1). As suggested by hypothesis BRAND IMAGE MATCH BY SELF-CONSTRUAL IN STUDY 1
1a, brands consistent with the ingroup have more positive (HYPOTHESIS 2; 0–100 SCALE)
self-brand connections than brands that do not match the
image of the ingroup (match p 63.59, no match p 17.31;
a priori contrast F(1, 723) p 191.20, p ! .001). As sug-
gested by hypothesis 1b, brands consistent with the outgroup
have less positive self-brand connections than brands that
do not match the image of the outgroup (match p 24.05,
no match p 45.02; a priori contrast F(1, 723) p 117.26,
p ! .001). Ethnocultural group, group type, and brand im-
age match also have significant main effects that are qual-
ified by the significant interaction (ethnocultural group:
white p 36.15, Asian/Hispanic p 40.30, F(1, 241) p
4.84, p ! .05; group type: ingroup p 40.17, outgroup p
34.17, F(1, 723) p 11.96, p ! .001; and brand image
match: match p 43.39, no match p 30.96, F(1, 723) p
66.68, p ! .001).
Hypothesis 2. Hypothesis 2 implies a three-way inter-
action of ethnocultural group by group type by brand image
match. In particular, we hypothesize that self-brand connec-
tions to brands with associations that match the outgroup will
be lower for white (more independent) participants compared
to Asian/Hispanic (more interdependent) participants. We find
a marginally significant three-way interaction of ethnocultural
group by group type by brand image match (F(1, 723) p
3.13, p p .08; this interaction and the two-way interaction
reported above are the only significant interactions in the
model). Figure 2 shows the pattern of results, which supports
hypothesis 2, as do preplanned comparisons. The contrast
comparing the white individuals’ to the Asian/Hispanic in-
dividuals’ self-brand connections in the outgroup, brand im-
age match conditions is significant (white p 20.94, Asian/
Hispanic p 30.59; F(1, 723) p 11.59, p ! .001), but there (F(1, 723) ! 1.0), ingroup brand image match (F(1, 723) p
are no significant differences between white and Asian/His- 1.63, NS), or ingroup brand image does not match conditions
panic individuals in the outgroup brand image does not match (F(1, 723) p 2.49, p p .12).

FIGURE 1 Discussion of Results from Study 1

SELF-BRAND CONNECTIONS ACROSS TYPE OF GROUP BY
BRAND IMAGE MATCH IN STUDY 1 (HYPOTHESES 1A AND
As hypothesized, we find that consumers report stronger
1B; 0–100 SCALE) self-brand connections for brands with images that are con-
sistent with the image of an ingroup compared to brands
with images that are not consistent with an ingroup (hy-
pothesis 1a). The positive effect of ingroup brand associa-
tions on self-brand connections is consistent with both
assimilation goals for individuals with interdependent self-
construals (here, the Asian and Hispanic participants) and
self-determination goals for those with independent self-
construals (here, the white participants). Also as hypothe-
sized, brands consistent with an outgroup are less likely to
show a self-brand connection than brands with images in-
consistent with the outgroup, and brands not consistent with
the outgroup have a positive effect on self-brand connections
(hypothesis 1b), empirically demonstrating that consumers
use brand symbolism in a manner consistent with balance
theory predictions (Heider 1946). The negative effect of
384 JOURNAL OF CONSUMER RESEARCH

outgroup brand associations on self-brand connections is from the data set due to incomplete or improper responses,
stronger for white (independent) consumers than for Asian/ leaving a total of 161 participants.
Hispanic (interdependent) consumers. As Kampmeier and
Simon (2001) argue, differentiation needs are more predom- Procedure. This study is virtually identical to that used
inant for the independent self when outgroups are the focus; in study 1, with four exceptions. First, half of the subjects
thus, whites, who tend to be more independent, show a completed the Singelis (1994) independent and interdepen-
stronger effect (hypothesis 2). dent self-construal scales at the beginning of the study, while
In study 1, we examined self-construal differences due half the subjects completed these scales at the end of the
to ethnocultural background. In study 2, we examine chron- study.2 Second, after participants rated the degree to which
ic tendencies toward interdependent and independent self- they fit within each group, they were asked to rate the extent
construals without reference to ethnicity, using the scales to which an association with each group type would com-
developed by Singelis (1994) as a measure of these two municate something positive or negative about them. Third,
types of self-construals. We expect the same general pattern after participants rated the degree to which they had formed
of results for this different way of categorizing chronic dif- self-brand connections with the four brands they had entered
ferences in independence and interdependence. We also ex- earlier (generated by the four combinations of ingroup vs.
amine hypothesis 3, that effects on self-brand connections outgroup and brand image matches vs. does not match),
are stronger for brands that are symbolic and communicate they were also asked to rate the brands on a number of
something about the user. Finally, we use thought protocols dimensions, including the degree to which the brand was
to examine our findings about outgroups in more detail. able to communicate something symbolic about the brand’s
user. This was followed by the collection of thought pro-
tocols, where participants were asked to retrospectively re-
STUDY 2 port the thoughts they had when they rated the degree to
which each brand reflected “you” versus “not you.” With
In this study, we again examine the influence of ingroups the exception that the filler task was slightly different, in all
and outgroups on self-brand connections, with the expec- other respects the study was the same as study 1. The entire
tation that the effects will differ for people with chronically procedure took approximately one half hour.
independent versus interdependent self-concepts. We ex-
amine individual differences in independent and interde- Independent Variables. Participants completed the
pendent self-construals without regard to ethnic background, entire Singelis (1994) scales for independent (12 items,
thus providing a test of the effect of independence versus a p .64) and interdependent (12 items, a p .62) chronic
interdependence unconfounded by ethnicity. We use the two self-concepts. Based on median splits, participants were
Singelis scales (1994) for independence and interdepen- divided into high and low groups for each self-construal
dences to identify participants who are high on one type of type. Participants who were high in independent and low in
self-construal, but low on the other. Additionally, we ask interdependent were considered to be schematic on inde-
participants to rate the degree to which the brands they have pendence, while participants who were high in interdepen-
listed are able to communicate something about one’s self- dence and low in independence were considered to be sche-
identity to test hypothesis 3. We propose that the extent to matic on interdependence. Participants who were high on
which a brand has such symbolic characteristics will mod- both or low on both scales were eliminated from the data
erate the basic effects postulated in hypotheses 1a and 1b, set, leaving a total of 75 participants.3 By construction, the
with stronger effects for more symbolic brands (i.e., those interdependent participants scored significantly higher on
brands better able to communicate about one’s self-identity). the mean of the interdependence items (69.37 vs. 54.35,
In addition, we collect thought protocols and other measures F(1, 743) p 102.62, p ! .001) and significantly lower on
to explore the processes underlying our findings. In partic- the mean of the independence items (57.21 vs. 73.65,
ular, we examine the degree to which participants explicitly F(1, 73) p 127.52, p ! .001) compared to the independent
report negative and positive links to the self for outgroups participants. Participants also rated the extent to which each
and the degree to which being associated with a group re- of their four brands was symbolic with two 100-point scale
flects positively or negatively on an individual. We expect items (“to what extent does this brand communicate some-
that we will find additional evidence that independents are thing specific about the person who uses it?” anchored by
more concerned with differentiating themselves from out- “does not communicate a lot/communicates a lot”; and “how
groups than are interdependents. much does this brand symbolize what kind of person uses
it?” anchored by “not at all symbolic/highly symbolic”;
Method 2
No order effects were found as a result of this counterbalancing measure.
3
We also ran the analyses reported for this study using a continuous
Participants. Two hundred and fourteen undergradu- index of the degree of independence relative to interdependence for each
individual. This index was constructed using the Singelis scales as (in-
ate students at a public university in the Southwest partic- dependent ⫺ interdependent)/(independent + interdependent). Using this
ipated in this study to meet an introductory marketing course index allowed us to retain all participants in the analysis. The results were
requirement. Fifty-three participants had to be eliminated virtually identical to those reported below.
SELF-CONSTRUAL, REFERENCE GROUPS, AND BRAND MEANING 385

a p .89). These two items were averaged. The remaining FIGURE 3

independent variables are based on the instructions to par-
SELF-BRAND CONNECTIONS ACROSS TYPE OF GROUP BY
ticipants to enter groups and brands during the computer BRAND IMAGE MATCH IN STUDY 2 (HYPOTHESES 1A AND
program and are identical to those in study 1 (group type 1B; 0–100 SCALE)
and brand image match).
Dependent Variable. Self-brand connections were
again measured using seven 0–100 sliding scale items (see
above). The seven items were averaged to form one self-
brand connection score per participant per brand (a p .96).
Process Measures. The thought protocols were coded
by two coders blind to the experimental manipulations and
hypotheses for the purpose of examining links to the self
made with respect to outgroups. Thoughts were coded with
respect to whether they expressed a positive link between
the self and the brand or a negative link. Intercoder reliability
was high; there were 17 disagreements for the positive self-
link codes (11.3%) and eight for the negative self-link cod-
ing (5.3%). Disagreements were resolved by the authors, or outgroup, thus predicting a two-way interaction of group
blind to experimental conditions. Additionally, participants type and brand image match. Study 2 replicates study 1:
completed three items to measure the degree to which par- the interaction of group type by brand image match on
ticipants felt being associated with the group would reflect self-brand connections is significant (F(1, 211) p 110.23,
either negatively or positively on them (“how would being p ! .001; see fig. 3). Brands that match the ingroup have
associated with this group reflect on someone?” anchored more favorable self-brand connections than those that do
by very negatively/very positively; “how much would you not match (ingroup match p 59.55, ingroup no match p
like to be identified with this group and what they repre- 19.12, F(1, 211) p 150.34, p ! .001), whereas brands that
sent?” anchored by not at all/very much; and “to what extent match the outgroup have lower self-brand connections than
would you like being linked to this group and what they brands that do not match the outgroup (outgroup match p
stand for?” anchored by “definitely disliked being linked/ 27.74, outgroup no match p 36.06, F(1, 211) p 6.36, p !
definitely like being linked”; a p .93). .01). Group type and brand image match also have sig-
Manipulation Checks. The degree to which the par- nificant main effects that are qualified by the significant
ticipant feels that he/she belongs to each group was assessed interaction; however, self-construal does not (group type:
using the same average of three items as in study 1 (a p ingroup p 39.34, outgroup p 31.90, F(1, 211) p 7.66, p !
.96). .01; brand image match: match p 43.65, no match p
27.59, F(1, 211) p 46.06, p ! .001).
Results Hypothesis 2. Hypothesis 2 proposes that self-brand
connections to brands with associations that match the out-
The model used in the analyses to predict self-brand group will be lower for independent individuals compared
connections is a general linear model with measured self- to interdependent individuals. We find a significant three-
construal (independent vs. interdependent) as a between- way interaction of self-construal by group type by brand
subjects factor, group type (ingroup vs. outgroup) and brand image match (F(1, 211) p 5.20, p ! .05). The pattern of re-
image match (match vs. no match) as within-subjects factors, sults shown in figure 4 replicates the pattern found in study
and brand symbolism included as a continuous variable. All 1, supporting hypothesis 2, as do preplanned comparisons.
the two-way, three-way, and four-way interactions of self- The contrast comparing independent to interdependent in-
construal, group type, brand image match, and brand sym- dividuals’ self-brand connections in the outgroup, brand im-
bolism were included in the model used to test all three age match conditions is significant (independent p 21.66,
hypotheses, and the dependent variable was self-brand con- interdependent p 34.69; F(1, 211) p 13.03, p ! .01), but
nections. there are no significant differences between independent and
Manipulation Checks. Participants considered them- interdependent individuals in the outgroup brand image does
selves to belong to the ingroup significantly more than they not match condition (F(1, 211) ! 1.0) or in any of the in-
felt they belonged to the outgroup (ingroup p 87.09, out- group conditions (F’s(1, 211) ! 1.0).4
group p 17.06, F(1, 193) p 1,931.63, p ! .001). 4
We also included a three-item measure of aspiration level for both
Hypotheses 1a and 1b. Once again, these hypotheses ingroups and outgroups (for items, see Escalas and Bettman 2003). We
find a significant interaction of group type by self-construal on aspiration
assert that whether or not brand associations match a level, with independents aspiring to belong to the outgroup less than in-
group’s image will have different effects on self-brand terdependents. This finding supports the idea that independents do not want
connections depending on whether the group is an ingroup to identify with the outgroup, consistent with our differentiation hypothesis.
386 JOURNAL OF CONSUMER RESEARCH

FIGURE 4 is more negative when a brand is considered to be symbolic,

compared to when it is not (t(212) p ⫺2.71, p ! .01).
SELF-BRAND CONNECTIONS ACROSS TYPE OF GROUP BY
BRAND IMAGE MATCH BY SELF-CONSTRUAL IN STUDY 2 Additional Process Measures. We focus on the
(HYPOTHESIS 2; 0–100 SCALE)
thought protocol coding for self-links related to the out-
groups to shed further light on the findings above regarding
greater differentiation from outgroups among independents.
In the outgroup conditions, the proportion of individuals
explicitly stating a self-link is affected by image match for
independents, but not for interdependents (see fig. 6). The
simple main effect of brand image match versus mismatch
within independent self-construals is significant for both
positive self-links and negative self-links (F’s(1, 120) 1
6.0, p’s ! .02) and is not significant for interdependent self-
construals in either thought type (F’s(1, 120) ! 1.0, NS). The
difference between independent and interdependent self-
construals on positive self-links within the brand image does
not match condition is significant (independent p .28,
interdependent p .08, F(1, 120) p 6.17, p p .01). Simi-
larly, the difference in negative self-links within the brand
image matches condition between independent and inter-
dependent self-construals is significant (independent p .61,
interdependent p .35, F(1, 120) p 6.95, p ! .01). These
outgroup findings are consistent with the notion that inde-

FIGURE 5

SELF-BRAND CONNECTIONS ACROSS TYPE OF GROUP BY

BRAND IMAGE MATCH BY BRAND SYMBOLISM IN STUDY 2
(HYPOTHESIS 3; 0–100 SCALE)

Hypothesis 3. In this hypothesis, we propose that the

effects of ingroup and outgroup brand associations on self-
brand connections will be stronger when the brand is con-
sidered to have symbolic characteristics, that is, when the
brand is perceived to communicate something about the
person who uses it. In support of hypothesis 3, we find a
significant three-way interaction of group type by brand
image match by brand symbolism (F(1, 212) p 20.52,
p ! .001; the four-way interaction of group type by brand
image match by brand symbolism by self-construal is not
significant). For ease of interpretation, we present the results
after dichotomizing the brand symbolism measure; however,
the results reported stem from the continuous variable anal-
ysis. The pattern of results supports hypothesis 3, as shown
in figure 5. The effect of a brand image mismatch with the
outgroup is more positive when the brand is considered to
be symbolic (t(212) p 3.15, p ! .01; the difference between
the match conditions is not significant; t(212) p ⫺.36, NS).
Within the ingroup the effect of a brand association match
is significantly stronger when a brand is considered to be
symbolic, compared to when it is not (t(212) p 8.88, p !
.001), and the effect of a brand mismatch with an ingroup
SELF-CONSTRUAL, REFERENCE GROUPS, AND BRAND MEANING 387

FIGURE 6 outgroup is perceived of as more negative for individuals

PROPORTION EXPRESSING SELF-LINKS IN THOUGHT COD-
characterized by independent, compared to interdependent,
ING ACROSS SELF-CONSTRUAL BY BRAND IMAGE MATCH self-construals.
FOR OUTGROUPS IN STUDY 2 (0–1.0 SCALE)
Discussion of Results from Study 2
In this study, as in study 1, consumers report higher self-
brand connections for brands with associations that are con-
gruent with an ingroup compared to those that are incon-
gruent (hypothesis 1a). Additionally, brands that are
associated with the image of an outgroup are less likely to
show a self-brand connection than brands with images not
associated with an outgroup (hypothesis 1b). The negative
effect of outgroup brand associations on self-brand connec-
tions only occurs for independent consumers, and not for
interdependent consumers (hypothesis 2), which we believe
is due to the strong differentiation needs of independents
when outgroups are focal (Kampmeier and Simon 2001).
Our thought protocols and measure of the degree to which
being associated with a group reflects positively or nega-
tively on the individual also provide evidence for this dif-
ferentiation-based explanation. Independent individuals’
thoughts contain explicit self-links for outgroups, and they
believe that an association with an outgroup communicates
something more negative about them than interdependent
individuals do.
The pattern of results supports the general theory pro-
posed in this article: consumers use brands to create or com-
municate their self-concept. These motivations are different
for different chronic aspects of self, as measured by the
Singelis scales (1994). Furthermore, these goals are better
achieved with symbolic brands than nonsymbolic brands, as
evidenced by the support for hypothesis 3. Only symbolic
brands are used to differentiate oneself from an outgroup.
In both of our studies, the positive effect of ingroups on
self-brand connections is stronger than the negative effect
of outgroups. However, even in the case of powerful ingroup
associations, the effect is stronger for those brands that are
pendent self-construals are concerned with differentiating perceived to communicate something symbolic about the
themselves from the outgroup, while interdependent self- brand’s user compared to those brands that do not. This
construals are less concerned with outgroup brand associa- provides additional evidence that consumers use brands to
tions. communicate their self-concept, utilizing brand associations
Finally, we find a significant interaction of group type that are the result of reference group brand usage.
by self-construal on our measure of the degree to which
participants felt being associated with a group would re-
flect either negatively or positively on them (F(1, 193) p CONCLUSION
6.75, p ! .01). Both independent and interdependent indi- General Discussion
viduals believe an association with an ingroup communi-
cates something favorable about them (independent p In this article, we argue that consumers appropriate brand
79.87, interdependent p 80.71). However, independent meanings emerging from associations of brands with ref-
self-construals believe an association with the outgroup erence groups to construct their self-concepts. In making
communicates more negative (i.e., less favorable) infor- this claim, we integrate literature from the sociological/an-
mation about them than the interdependent individuals thropological tradition in marketing and social psychology
do (independent p 30.33, interdependent p 42.16; con- research. Our studies show that consumers report higher self-
trast F(1, 193) p 17.26, p ! .001). This is also consistent brand connections for brands with images that are consistent
with Kampmeier and Simon’s (2001) theory that indepen- with the image of an ingroup compared to brands with im-
dent individuals are concerned with differentiating them- ages that are inconsistent with the image of an ingroup. This
selves from outgroups: we find that an association with an finding agrees with the brand congruency findings of pre-
388 JOURNAL OF CONSUMER RESEARCH

vious consumer research on value-expressive social influ- listed a social group to which they did not belong, resulting
ence, that is, that consumers use brands whose images match in wide variance in the nature and specificity of these out-
reference groups to which they belong to establish a psy- groups (see table 1). An interesting issue for future research
chological association with those groups. Further, we find would be to explore the effects of different types of out-
that self-brand connections are lower for brands with images groups. It is very likely that there are both outgroups that
that are consistent with the image of an outgroup compared one does not belong to but are relatively unimportant and
to brands with images that are inconsistent with an outgroup. groups to which one does not belong but wants to be clearly
This finding that consumers reject the social meanings of distanced from. Different types of outgroups may have vary-
brands that arise from outgroup brand usage is an important ing effects on the extent to which consumers reject the social
contribution of this article. meaning arising from outgroup brand usage. In addition, the
We also show that independent versus interdependent self- degree to which a brand is strongly or weakly associated
construals interact with our congruency findings to deter- with an ingroup or outgroup may moderate our effects.
mine the level of self-brand connections using two different
approaches to operationalizing chronic differences in self- [Dawn Iacobucci served as editor and Joseph Priester
construal: ethnicity and Singelis’s (1994) scales. We find served as associate editor for this article.]
that the negative effect of outgroup brand associations on
self-brand connections is stronger for independent consum- REFERENCES
ers than for interdependent consumers. We argue and pro-
vide some empirical support for the notion that this is due Aaker, Jennifer L. and Angela Y. Lee (2001), “‘I’ Seek Pleasures
to the stronger needs of more independent consumers to and ‘We’ Avoid Pains: The Role of Self-Regulatory Goals in
Information Processing and Persuasion,” Journal of Con-
differentiate themselves from outgroups (Kampmeier and sumer Research, 28 (June), 33–49.
Simon 2001). We also find that our effects are moderated Aaker, Jennifer L. and Bernd Schmitt (2001), “Culture-Dependent
by brand symbolism, such that brands that communicate Assimilation and Differentiation of the Self: Preferences for
something about the user yield stronger effects than brands Consumption Symbols in the United States and China,” Jour-
that do not. In the case of ingroup associations, the positive nal of Cross-Cultural Psychology, 32 (September), 561–76.
effect of image congruency is stronger for those brands that Ball, A. Dwayne and Lori H. Tasaki (1992), “The Role and Mea-
are perceived to communicate something symbolic about surement of Attachment in Consumer Behavior,” Journal of
the brand’s user compared to those brands that do not. In Consumer Psychology, 1 (2), 155–72.
the case of outgroup associations, only symbolic brands are Bearden, William O. and Michael J. Etzel (1982), “Reference
used to differentiate oneself from outgroup associations; the Group Influence on Product and Brand Purchase Decisions,”
Journal of Consumer Research, 9 (September), 183–94.
effects disappear for brands that are considered to not com- Bearden, William O., Richard G. Netemeyer, and Jesse E. Teel
municate anything about the brand’s user. (1989), “Measurement of Consumer Susceptibility to Inter-
personal Influence,” Journal of Consumer Research, 15
Limitations and Directions for Future Research (March), 473–81.
Belk, Russell W. (1988), “Possessions and the Extended Self,”
Our article focuses on the constructs of independent ver- Journal of Consumer Research, 15 (September), 139–68.
sus interdependent self-construals, which bear a close re- Brewer, Marilynn B. and Wendi Gardner (1996), “Who Is This
semblance to the individualism-collectivism dimension of ‘We’? Levels of Collective Identity and Self Representations,”
culture proposed by Hofstede (1980) and others (e.g., Trian- Journal of Personality and Social Psychology, 71 (July),
dis 1989). We find that independents are motivated to dif- 83–93.
ferentiate themselves from outgroups, whereas interdepen- Burnkrant, Robert E. and Alain Cousineau (1975), “Informational
dents are relatively unaffected by outgroups. However, and Normative Social Influence in Buyer Behavior,” Journal
of Consumer Research, 2 (December), 206–15.
research on individualism versus collectivism has found that Childers, Terry L. and Akshay R. Rao (1992), “The Influence of
collectivists may behave antagonistically toward outgroup Familial and Peer-Based Reference Groups on Consumer De-
members (e.g., by pursuing conflict with outgroup members; cisions,” Journal of Consumer Research, 19 (September),
Leung 1988). Although many of these studies use strangers 198–211.
rather than well-defined outgroups, this finding appears to Escalas, Jennifer Edson and James R. Bettman (2003), “You Are
be inconsistent with our finding that interdependents pay What They Eat: The Influence of Reference Groups on Con-
little attention to outgroups. Further research could explore sumer Connections to Brands,” Journal of Consumer Psy-
a number of possible ideas for how to reconcile this apparent chology, 13 (3), 339–48.
conflict, such as contrasting interpersonal behavior toward Fournier, Susan (1998), “Consumers and Their Brands: Developing
outgroup members with consumption behavior or exploring Relationship Theory in Consumer Research,” Journal of Con-
sumer Research, 24 (March), 343–73.
other constructs (e.g., interpersonal orientation, self-moni- Heider, Fritz (1946), “Attitudes and Cognitive Organization,” Jour-
toring, or social identity) that may underlie the independent- nal of Psychology, 21 (January), 107–12.
interdependent differences we find. Hofstede, Geert (1980), Culture’s Consequences: International
As mentioned above, many of the studies examining in- Differences in Work-Related Values, Beverly Hills, CA: Sage.
dividualism versus collectivism use strangers as a proxy for Kampmeier, Claudia and Bernd Simon (2001), “Individuality and
well-defined outgroups. In our research, each participant Group Formation: The Role of Independence and Differen-
 SELF-CONSTRUAL, REFERENCE GROUPS, AND BRAND MEANING 389

tiation,” Journal of Personality and Social Psychology, 81 McCracken, Grant (1988), Culture and Consumption, Blooming-
(September), 448–62. ton: Indiana University Press.
Kleine, Susan Schultz, Robert E. Kleine III, and Chris T. Allen ——— (1989), “Who Is the Celebrity Endorser? Cultural Foun-
(1995), “How Is a Possession ‘Me’ or ‘Not Me’? Character- dations of the Endorsement Process,” Journal of Consumer
izing Types and an Antecedent of Material Possession At- Research, 16 (December), 310–21.
tachment,” Journal of Consumer Research, 22 (December), Moschis, George P. (1985), “The Role of Family Communication
327–43. in Consumer Socialization of Children and Adolescents,”
Lee, Angela Y., Jennifer L. Aaker, and Wendi L. Gardner (2000), Journal of Consumer Research, 11 (March), 898–913.
“The Pleasures and Pains of Distinct Self-Construals: The Muniz, Albert M. and Thomas C. O’Guinn (2001), “Brand Com-
Role of Interdependence in Regulatory Focus,” Journal of munity,” Journal of Consumer Research, 27 (March), 412–32.
Personality and Social Psychology, 78 (June), 1122–34. Park, C. Whan and V. Parker Lessig (1977), “Students and House-
Leung, Kwok (1988), “Some Determinants of Conflict Avoidance,” wives: Differences in Susceptibility to Reference Group In-
Journal of Cross-Cultural Psychology, 19 (1), 125–36.
fluence,” Journal of Consumer Research, 4 (September),
Levy, Sidney J. (1959), “Symbols for Sale,” Harvard Business
102–10.
Review, 37 (July–August), 117–24.
Linville, Patricia W. (1987), “Self-Complexity as a Cognitive Richins, Marsha L. (1994), “Valuing Things: The Public and Pri-
Buffer against Stress-Related Illness and Depression,” Journal vate Meanings of Possessions,” Journal of Consumer Re-
of Personality and Social Psychology, 52 (April), 663–76. search, 21 (December), 504–21.
Markus, Hazel (1977), “Self-Schemata and Processing Information Singelis, Theodore M. (1994), “The Measurement of Independent
about the Self,” Journal of Personality and Social Psychology, and Interdependent Self-Construals,” Personality and Social
35 (August), 63–78. Psychology Bulletin, 20 (October), 580–91.
Markus, Hazel and Shinobu Kitayama (1991), “Culture and the Triandis, Harry C. (1989), “The Self and Behavior in Differing
Self: Implications for Cognition, Emotion, and Motivation,” Cultural Contexts,” Psychological Review, 96 (July), 506–20.
Psychological Review, 98 (April), 224–53. ——— (1993), “Collectivism and Individualism as Cultural Syn-
Markus, Hazel and Paula Nurius (1986), “Possible Selves,” Amer- dromes,” Cross-Cultural Research: The Journal of Compar-
ican Psychologist, 41 (September), 954–69. ative Social Science, 27 (August–November), 155–80.

View publication stats