Vous êtes sur la page 1sur 14

Brand Congruence in Interpersonal Relations: A Social Network Analysis

Author(s): Peter H. Reingen, Brian L. Foster, Jacqueline Johnson Brown, Stephen B. Seidman
Source: The Journal of Consumer Research, Vol. 11, No. 3 (Dec., 1984), pp. 771-783
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2489067
Accessed: 28/09/2009 10:49

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucpress.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The
Journal of Consumer Research.

http://www.jstor.org
Brand Congruence in Interpersonal
Relations: A Social Network Analysis

PETER H. REINGEN
BRIAN L. FOSTER
JACQUELINEJOHNSON BROWN
STEPHEN B. SEIDMAN*

Previous studies dealing with the notion of brand congruence suffer from
questionable methods of group determination, suspect demonstrations of brand
congruence effects, and inadequate attention paid to types of social relation. To
overcome these shortcomings, the present study uses graph-theoretic social
network techniques to examine interpersonal relationships and brand choice
behavior in natural environments. The brand choices of individuals in a social
relationship were compared to those of unrelated individuals across various
products, types of social relation, and types of basic sociological structure (dyad,
clique, and 2-plex). While significant brand congruence effects were obtained,
they were clustered in a few products mediated by types of social relation.
Conspicuousness of the product, as traditionally defined, was found to be
insufficient to account for these findings.

T hat humans are influenced by others has long Although most research on social influence focuses
been a subject of interest to sociologists and on primary groups, manifestations of social influence
social psychologists. Since the classical work by Asch do not appear to be confined to them. For what can
(1953), much evidence has been produced that groups be construed as secondary groups, Burnkrant and
influence members' behavior (see Shaw 1976 for a Cousineau (1975), Cohen and Golden (1972), and
review), and it is not surprising that students of Reingen (1982) showed that the mere presentation of
consumer behavior have focused on social influence information about others' brand evaluations or behav-
to explain consumer product/brand choices (e.g., Ven- ior influenced subjects' responses to stimuli. Nor does
katesan 1966). The importance attached to the social social influence appear to be restricted to membership
influence of relevant others is also reflected in Fishbein groups. For example, Cocanougher and Bruce (1971)
and Ajzen's (1975, Chapter 7) behavioral intention found that socially distant groups (i.e., groups with
paradigm in its subjective norm component (see Ba- which a consumer does not regularly interact) can
gozzi 1981; Miniard and Cohen 1981; and Ryan 1982 influence consumers if consumers hold favorable at-
for recent research), and in marketing practitioners titudes toward the members or activities of that group,
who make substantial use of interpersonal influence and Carmon and Mannheim (1979) demonstrated the
scenarios in communications (see Hawkins, Best, and influence of nonmembership groups on the self-image
Coney 1983, Chapter 7; Lessig and Park 1978; and development of architects. The overall evidence there-
Zaltman and Wallendorf 1983, Chapter 6 for specific fore suggests that consumers may orient themselves
examples). to membership groups (primary or secondary groups)
as well as to nonmembership groups (e.g., aspiration
groups) in shaping their behaviors. These various
* Peter H. Reingen is Associate Professor of Marketing at the types of groups are commonly referred to in texts of
College of Business Administration, Brian L. Foster is Professor consumer behavior as reference groups when they
and Department Chair of Anthropology, and Jacqueline Johnson significantly influence an individual's behavior (e.g.,
Brown is a doctoral candidate in Marketing, all at Arizona State Engel and Blackwell 1982, Chapter 6; Hawkins et al.
University, Tempe, AZ 85287. Stephen B. Seidman is Associate 1983, Chapter 7).1
Professor of Mathematics, George Mason University, Fairfax, VA
22030. The research was supported by a Faculty Grant-in-Aid to
the first author. We appreciate the helpful comments by William
' In many cases, reference groups lack the characteristics usually
Bearden, Richard Burdick, Larry Crosby, Mike Hutt, Jerome Ker-
nan, and an anonymous JCR reviewer. attributed to groups in the social-psychological literature. The term
(Continued p. 2.)
771
? JOURNAL OF CONSUMERRESEARCHe Vol. 11 December 1984
772 THE JOURNAL OF CONSUMER RESEARCH

REFERENCE GROUP INFLUENCE Using the framework originally proposed by Bourne


AND BRAND CHOICE (1957), Bearden and Etzel (1982) expanded upon Park
and Lessig's work in an excellent study which inves-
Much of the extant literature on reference groups' tigated when reference groups exert influence. For
influence on consumer brand choice falls into two publicly consumed products, it was hypothesized that
major research streams, which can be distinguished in reference group influence for the brand of the product
terms of methodology and conceptual foci: informal should be strong because brand consumption will be
group studies and perceived influence studies. seen by others, whereas for privately consumed prod-
ucts, influence should be weak because brand con-
Informal Group Studies sumption will be inconspicuous. As in the Park and
Lessig (1977) study, subjects were asked to indicate
The informal group studies by Stafford (1966), Witt their agreement with statements manifesting the types
(1969), and Witt and Bruce (1970, 1972) have in of reference group influence as they applied to product
common an attempt to relate group cohesiveness to or brand selection decisions. The results supported
brand choice congruence within groups. In these stud- Bearden and Etzel's hypothesis about differences in
ies, group membership was determined by first asking perceived reference group influence between publicly
a set of subjects to nominate others (e.g., friends, and privately consumed products.
relatives, or neighbors), and then by obtaining partic- Finally, Moschis (1976) employed Festinger's (1954)
ipation of the nominees. These studies assert that theory of social comparison to explain why informal
informal groups exert influence toward congruence groups influence purchasing behavior.2 Group influ-
on member brand choices (e.g., Stafford 1966) and ence was determined by having subjects indicate their
that the extent of influence varies with group cohe- extent of agreement with statements such as "Many
siveness (Witt 1969) and the nature of the product cosmetic items I own are similar to those of my
(Witt and Bruce 1970, 1972). However, the evidence friends" (p. 242). Although Moschis found that "in-
in support of the informal group influence hypothesis 'formalgroups exert a great influence on their members'
is contradictory. For example, Hansen (1969) was purchasing behaviors" (p. 240), it is apparent that no
unable to support the existence of a relationship actual informal groups were examined in his study.
between group cohesiveness and brand congruence,
and Ford and Ellis (1980) were unable to replicate the Limitations of Previous Studies
Stafford (1966) findings.
These studies have yielded important insights into
reference group influences on consumer behavior and
Perceived Influence Studies have identified an important area for research. Nev-
The perceived influence studies by Bearden and ertheless, they suffer from several limitations which
Etzel (1982) and Park and Lessig (1977) employed render suspect certain aspects of the produced evi-
survey methodology to assess various types of perceived dence.
reference group influence across products. Based on Group Definition. One major problem with the
the work by Deutsch and Gerard (1955), Jahoda informal group studies by Stafford (1966), Witt (1969),
(1972), and Kelman (1961), Park and Lessig developed and Witt and Bruce (1970, 1972) centers around
a scale to measure informational, utilitarian, and group definition. Since a group consisted of a nomi-
value-expressive reference group influence. Subjects nator and several nominees in those studies, it is
were asked to assume that an individual had decided possible that the selection procedure yielded groups
to purchase a product but had not yet decided which which were diverse in basic sociological structure. For
brand of the product to buy. They then were given example, in one case a group might consist of individ-
verbal descriptions of situations representing the dif- uals, each of whom is tied to every other in a social
ferent kinds of reference group influence. Finally, they relation (e.g., friendship), while in another instance
were asked to indicate for each product to what extent the nominator might be tied to every nominee, but
each situation was relevant to his or her brand choice. the nominees may not be tied to each other. These
Overall, the Park and Lessig (1977) findings suggest two structures have completely different sociological
greater susceptibility to reference group influence for meanings. The first case illustrates a basic structure
students than for housewives. widely discussed in sociology-the clique. For some
purposes, the second example may not qualify as an
"reference group" has become established by long usage, so it is informal group at all in the sociological sense: most
employed here. However, for the informal group studies examined persons, for instance, are either not interdependent or
later, the groups should have characteristics similar to more tradi- completely unaware of each other's existence.
tionally defined groups. Although there is a lack of agreement about
what does and does not constitute a group, it is generally suggested
2
that several persons can be considered a group when they are Since the Moschis (1976) study did not examine actual brand
interdependent or are aware of each other (Ford and Ellis 1980). choices, it is classified here as a perceived influence study.
NETWORKANALYSIS OF BRAND CONGRUENCE 773

That the second case is not an unlikely phenomenon important information on various types of influence
is suggested by Ford and Ellis (1980) who, in their and on when/why reference groups may affect buyer
replication study of Stafford (1966), found that at least behavior, by their very nature they only suggest the
four of their 10 groups had members who did not possibility of reference group influence, since no direct
know the other group members (p. 131).3 Also, Witt evidence is provided.
(1969, note 2) points out that his group selection
procedure may have yielded "groups" that were ref- Types of Social Relations. There is an additional
shortcoming in the literature which is common to
erents for the nominator but not for the other mem-
both research streams: the investigators assumed that
bers. It is unlikely that meaningful data are obtained
when there is substantial heterogeneity in the nature certain types of social relationships form the basis for
reference group influence. Friends, relatives, neighbors,
of the units due to the concept of group being poorly
or fellow students may or may not be the referents of
defined. In fact, in these previous informal group
an individual employed in brand decision-making for
studies, some groups may not have been groups at
the products examined in the Hansen (1969), Moschis
all-even under very relaxed conditions for what
(1976), Stafford (1966), Witt (1969), and Witt and
constitutes such a structure. A more appealing strategy,
Bruce (1970, 1972) studies. Similarly, an individual's
both conceptually and statistically (in the sense of
minimizing extraneous sources of systematic variance fellow workers may or may not be the appropriate
referents for pool tables in Bearden and Etzel's work
and error variance arising from an imprecise group
(1982, p. 187). It would seem advisable, therefore, to
definition), would be to assess brand congruence in
groups whose sociological and formal structural prop- examine brand congruence in basic sociological struc-
erties are systematically and precisely controlled. tures that arise from different types of social relation-
ships. One may find brand congruence for bread
Influence Assessment. A second major problem among family members, for example, but such con-
with several of the informal group studies relates to gruence is less likely in a group of acquaintances who
how informal group influence is assessed. Stafford play touch football. Thus, the type of social relation-
(1966), for instance, suggests that informal group ships under investigation may be an important pre-
influence is demonstrated when within-group brand dictor of brand congruence.
choices are similar, but when the groups vary among A second reason why an examination of the effect
themselves (p. 71). This may not hold in general. of type of social relation on brand congruence may
Consider the following scenario. All group members be important concerns the notion of brand visibility.
and all groups agree on the same brand "X." According In the traditional approach, a private product is op-
to the Stafford criterion, there would be no informal erationally defined as one used at a person's residence
group influence. Missing, however, are comparisons (Bearden and Etzel 1982, p. 186). Since brand con-
between persons who are in a group and those who sumption of such a product would not be socially
are in no group. In a second scenario, imagine that conspicuous, it is argued that purchasing behavior is
such controls are available. Suppose further that all largely governed by attributes rather than by influences
persons who are not in a group choose brand "Y." of others (Bourne 1957). However, there is a potential
The inferences about group influence resulting from problem with this traditional argument. It fails to take
the two scenarios would be diametrically opposed. We into account the effect of social interaction on brand
suggested that a more convincing case can be made visibility. Social relationships may channel opportu-
for or against the informal group hypothesis by com- nities for observation of brand choice, and they may
paring the brand choices within a group against the provide a stimulus for brand-related verbal commu-
choices of individuals who are in a different group nications. That is, social relations may make brand
and in no group. In short, informal group studies choice behavior visible even for "private" products.
which examine choice behavior only within/across For example, one may find brand congruence among
groups contain a serious design flaw: group influence members of a group of frequently interacting friends
is an intragroup phenomenon, so its presence or even for "private" products such as laundry detergent,
absence is not directly testable through intergroup kitchen soap, and refrigerators, which the literature to
methods (Ford and Ellis 1980). date claims are not subject to group influence (e.g.,
Assael 1981, pp. 332-333). Thus conspicuousness, as
Perceived Influence. Although the perceived influ-
traditionally defined, may not necessarily be a good
ence studies by Bearden and Etzel (1982), Moschis
predictor of brand congruence.
(1976), and Park and Lessig (1977) have yielded

OVERVIEW
3Perhaps this is a reason why Ford and Ellis (1980) failed to
replicate the Stafford data. Stafford (1966, p. 70) claims that his
groups consisted of women who were friendly toward each other Although group influences on consumer behavior
and who interacted among each other, but no direct data are are commonly judged to be pervasive (e.g., Zaltman
furnished. and Wallendorf 1983, p. 197), the previous discussion
774 THE JOURNAL OF CONSUMER RESEARCH

suggests that the evidence produced to date is less structure is finding a precise and general way of
than compelling and is sometimes even contradictory. representing properties of social structure. Some of
To summarize the arguments presented thus far, the the most promising results in this area have come to
previous literature suffers from: be called "social network analysis," which is being
1. Imprecisedefinition (determination)of groups carried out by researchers in sociology, anthropology,
psychology, communications, mathematics, and other
2. Questionablemeasurementof brand congruenceef- fields.
fects The basic ideas behind social network analysis have
3. Lack of a systematicattempt to examine the brand been present in social anthropology and sociology for
congruence notion across different types of basic many years in the writings of Radcliffe-Brown (1965),
structure (e.g., "true cliques" and other "clique- Nadel (1957), and others. In general, network analysis
like" structures) provides a formal and precise way of defining terms
4. Inadequate attention to the way congruence may like "relation," "clique," and "group""so that they
vary across differenttypes of social relationships can be applied unambiguously to data on populations
of individuals. Several formalisms have been adopted
5. A tendencyto see perceivedreferencegroupinfluence by social scientists, including various statistical and
as evidence for direct influence. algebraic techniques (see Burt 1980 and Knoke and
The most general goal of the present study is to Kuklinski 1982 for recent reviews). The most widely
address these shortcomings by examining brand con- used, however, is graph-theoretic.
gruence among individuals in empirical systems of In Foster and Seidman's (1981, 1983) graph-theo-
social relationships-what Radcliffe-Brown (1965, p. retic method-the one adopted here-graphs are used
190) called "the network of actually existing rela- to represent relationships among individuals. A graph
tions"-using recently developed techniques for social is simply a set of points, some or all of which are
network analysis. The study is exploratory in nature. joined by lines. In general, points represent individuals,
Although it was expected to yield important substantive and each line represents a relationship between a pair
insights, the emphasis of the research was on illustrat- of individuals it joins. Each graph represents a single
ing how the limitations in previous investigations can relation. A series of graphs may be used to represent
be effectively addressed with social network analysis. different relations by having the lines in each graph
Our primary goal was to make a methodological correspond to a specified kind of dyadic relationship
contribution to the study of consumer sociology (e.g., each line represents a friendship tie in one graph,
through an exemplary application of formal algorithms an employer/employee relation in another, and so
for network analysis. on). It is in these graphs that features of social structure
The overall research strategy was as follows: stated (including cliques and other groups) are defined using
brand choices (and preferences) of individuals in a concepts from the mathematical theory of graphs.
specified social relationship were compared with those To use this formalism in social research, it is
of individuals who were not in that social relation. necessary to identify patterns in graphs which corre-
The comparison was carried out across various types spond to meaningful sociological properties of the
of social relation (e.g., friendship, roommate) and society represented by graphs. There are many ways
across a series of tangible/intangible products (e.g., to do so; one could, for instance, see if a graph is
shampoo, TV show). Moreover, analysis focused on transitive or not, whether it is centralized or not, and
brand congruence within several kinds of basic socio- so on. The theoretical problem of interest here, how-
logical structures or groups. Before this analysis is ever, concerns whether or not membership in social
described in more detail, some general comments on groups such as cliques is associated with greater brand
the network analysis of social structure will be helpful. choice similarity, and the analytical task is to identify
subsets of graphs which correspond to meaningful
social groups.
Network Analysis of Social Structure In the Foster and Seidman (1981) method, such
groups and other social units are called basic structural
Anthropologists, sociologists, and other social sci- units (or BSUs); in their general, abstract form, they
entists have been concerned with precisely the kinds are called basic structures. An example of a BSU is
of methodological problems that arise in the consumer suggested by a widely used notion in sociology, the
behavior literature reviewed here. In fact, a classic clique. The first attempt to make a formal, graph-
problem in social science concerns the relationship theoretic definition of clique was by Luce and Perry
between individual behavior (e.g., individuals' social (1949), who defined a clique as a maximal complete
decisions) and social structure (e.g., the nature of the subgraph of a graph-that is, the largest possible set
social networks in which individuals find themselves). of points of a graph such that each point is tied by a
In many diverse contexts, a prerequisite to addressing line to every other point. Such basic structures-
the relationship between individual behavior and social maximal complete subgraphs-have several important
NETWORKANALYSIS OF BRAND CONGRUENCE 775

FIGUREA FIGURE B
ILLUSTRATIONOF MAXIMALCOMPLETESUBGRAPHS ILLUSTRATIONSOF 2-PLEXES
A G F C D E

C D

B C D E A B A B
NOTE: Maximal complete subgraphs are ABC, CD, and DEFG. These subgraphs are
NOTE: In the four-point structure on the left, each point is adjacent to at least two others.
complete, since each point is tied directly to every other. They are maximal, since the For the five-point structure on the right, each point is connected to at least three others.
completeness criterion would be violated by adding any other point in the graph (e.g., if C
If C were adjacent to B in the structure on the left and E to A in the structure on the right,
were added to E, F, or G).
both of the structures would be cliques (i.e., 1-plexes).

mathematical properties that correspond to properties erties of k-plexes can be found in Seidman and Foster
of the sociological concept of clique, and representa- (1978a, 1978b). Most important, both the redundancy
tions of these basic structures in graphs of social and robustness of true cliques are retained, which
relations can thus be regarded as BSUs. makes it reasonable to see k-plexes as formalizing a
The properties of such structures that are of par- controlled weakening of the critical sociological prop-
ticular interest are, first, extreme robustness (i.e., erties of cliques.
removal of a point changes only the size of the unit, The present investigation examined brand congru-
not its structural properties), and second, great redun- ence within "true cliques" (called cliques from now
dancy of ties among individuals (i.e., an individual on) and 2-plexes in various social relations, with the
can be reached through different paths)-a fact of restriction that basic structures of either kind had at
obvious sociological interest in relation to social con- least three members. In addition to these basic struc-
trol, flow of information among members, and so on. tures, analysis was performed on dyadic relationships
From a practical analytical standpoint, it is important (called dyads from now on), whether or not they
to note that, using computers, rapid algorithms are existed within the cliques or 2-plexes. Although a
now available for identifying cliques in graphs. Figure reference group has been defined as "a person or
A illustrates maximal complete subgraphs in a larger group" (Bearden and Etzel 1982, p. 184), the literature
graph. extant is largely silent on brand congruence within
This concept of clique is extremely widely used, mere dyads.
but a number of problems have arisen in its application
(see Seidman and Foster 1978a, 1978b). One such
problem is that the restrictiveness of the completeness Expectations
criterion often means that few interesting cliques are Although the study was exploratory in nature and
found even in relatively dense networks. Seidman and no specific hypotheses were advanced for each cell of
Foster developed one way of relaxing this property to the overall design, it was guided by several general
provide for "weaker" clique structure by formalizing expectations. First, significant brand congruence effects
the principle that each point is connected to almost were expected; while the previous evidence was judged
every other point.4 They define a structure called a to be less than compelling on several grounds, it points
k-plex, which is a basic structure with n points such in that direction. Second, as already discussed, the
that each point is adjacent to at least n - k others. presence or absence of brand congruence depends on
Thus in a 2-plex with five points, each point must be the type of social relation under investigation, so
connected to at least three others, but each point is brand congruence effects were expected to occur in
not connnected to every other point. The k-plex appropriate relations for "private" as well as "public"
notion is a generalization of the clique concept, since products, with "public" and "private" being tradition-
a "true clique" (i.e., complete subgraph) is a special ally determined (e.g., by Bearden and Etzel 1982)
case of k-plexes (where k = 1). Figure B illustrates through a pretest.
2-plexes. A full treatment of the mathematical prop-
METHOD
4Another attemptat "loosening"the cliquestructurewasproposed To address the problems we have identified in the
by Alba (1973) in his diameter-nclique scheme. See Seidman and
Foster(1978a) for importantsociologicaland mathematicaldiffer- brand congruence literature, an unusual kind of design
ences betweenthat scheme and the one employed here. is necessary, and this requires some tradeoffs among
776 THE JOURNAL OF CONSUMER RESEARCH

different methodological considerations. First, an ex- EXHIBIT


perimental design is probably not possible, since the SOCIAL RELATIONSHIPS
present state of knowledge is insufficient to produce a
compelling design for the issues we have identified. Description Code Word
Second, a survey would not be very useful, since it is
1. A person's roommate. Roommate
not possible to get fine-grained data on most social
network properties through samples. This is so because 2. A person's neighbor(s) who lives adjacent
we would be sampling relationships, not individuals. to her. Adjacent
At the present time, data on individual choices can
3. A person's neighbor(s) across the hall
support inferences about only the crudest network who live(s) no farther away than the
properties, not the properties of interest to us here. neighbor(s) living in rooms adjacent to
What is needed is very detailed information on a her. Neighbor
small, natural community of some sort: this will help
4. With whom a person usually plays sports. Sports
us to avoid some of the pitfalls in the earlier studies,
albeit at the cost of experimental controls and of a 5. With whom a person usually studies. Study
large universe that would allow us to generalize to a
6. With whom a person usually eats her
variety of other cases. Meals
meals (breakfast, lunch and/or dinner).

Subjects and Procedure 7. With whom a person usually shares a


bathroom at the same time. Bathroom
Subjects were 49 members of a sorority at a large
8. With whom a person usually parties or
Southwestern state university. Of the total number of Party
goes out with.
subjects, 37 (75.5 percent) lived in the sorority house
and 35 (71.4 percent) were at least sophomores. The 9. A person's closest friends. Friend
subjects were enrolled in various majors, with business 10. With whom a person usually discusses a
constituting the largest single category (30.5 percent). very personal problem. Problem
The sorority was paid one dollar for each member's
participation.
Subjects responded to a self-administered question- These were chosen to cover a broad spectrum of
naire at one of their monthly meetings. Since sorority
important relations arising from spatial proximity,
policy prohibited nonmembers from attending the
special occasions, and affective bonds. The subjects
meeting, some control over questionnaire administra-
were instructed to list names of individuals who were
tion had to be sacrificed. The sorority president had,
either within or outside of the sorority, since con-
however, been carefully instructed to follow certain
straining subjects to list only sorority members would
procedures to minimize contamination of data (e.g.,
produce "artificial" social networks. A total of 568
due to interactions between subjects while completing
different individuals was named.
the task). The response quality of data obtained was
Finally, for each of the products listed in Figure C,
judged to be at least satisfactory.
subjects were asked to furnish the name both of the
brand they usually buy/consume and of the brand
Instrument they prefer whether or not they buy/consume the
The instrument first provided some general instruc- product. Information on preferred brands was obtained
tions and a brief cover for the study. This was followed in an effort to control indirectly for gifts and for
by a strongly stated promise of confidentiality, since consumption patterns developed prior to joining the
the subject was required to reveal not only her own sorority. Under such conditions, congruence on brands
name but also those of individuals she interacted with actually consumed may not be found, while congru-
in various types of social relationships. ence on preferred brands may be observed.
Subjects were asked to furnish names of individuals The products listed in Figure C were selected from
for the social relationships shown in the Exhibit.6 a preliminary list of 55 products.7 A pretest was
conducted on a convenience sample of 198 subjects
'For example, the roommate, adjacent, and neighbor names
similar to those in the study. These subjects rated
furnished by the subjects allowed for an internal check. The each product on a five-point "private/public con-
information on roommate and adjacent had complete inter-subject sumption" scale based on Bearden and Etzel's (1982)
agreement, and the information on neighbor (the operationalization
of which was more difficult to comprehend-see Exhibit) had 88
percent agreement. Furthermore, no overlap of membership across The data were generatedduring the sorority's monthly meeting;
cliques should have occurred for bathroom, and no cliques should thus the informationon conversationwas judged to be less mean-
have existed for neighbor and adjacent. These results were obtained. ingful.
6 Names of persons with whom a subject had carried on a lengthy 7 Due to high degreesof nonusageamong subjects,data obtained
conversation during the previous 24 hours were also obtained. on home computers,cars, tennis racquets,and video gamesare not
However, this social relation was eliminated from consideration. reported.
NETWORK ANALYSIS OF BRAND CONGRUENCE 777

FIGUREC
SUMMARY OF RESULTS

Products #1 #2 #3 #4 #5 #6 #7 #8 #9 10 #1 #12 #13 #14 #1 5 16


Relations Shampoo Tootipaste Detrgeret Magazirne TVShow Soda Pop Calculator Camera WrlstWatch Pizza Beer Sport TenniaShoe Jeans Restaurant Shirt/Blouse
Relations D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C D P C

Roommate AV./ v L
13,0,01 P X
Adjacent A X XX
[23,2(12),0] P XA _Z__Z_I __Z ?1/__ -2 11 v-r7- ,11
Neighbor A _ _ x
E35,23(49),0] p _ xx Z- A A
Sports A _ A- - I- k_Z- PV VV PV kV
(I17,0,0] p _zz_zz_zZ_ _ z/ [/_z/ z // 1l l // I/ V / V/ 1l
Study A X __z_Z I1 _ A I / M
VI _
[21,0,3(9)] P z__Z__Z-_ZC Z- A _Z-_Z / I/ IA I I/ I1 V1 1
Meals A _ _ X x
[96,92(183),29(68)] P -
_ I I I I I
Bathroom A X I xx xx xx _ xx
[32,2(12),5(21)] p x XX
Party A
(62,15(59),16(51)J p _ x
Friend A IXXXX
[57,4(25),7(24)] P_
Problem A
[47,3(16),4(15)] P _

x =p C 0.0005
xx = P C 0.0001
NOTE: Products #1 through #6
were "private" products; products #7 through #16 were "public" products. Based on the pretest data, products are listed from most private to most
public. Products #1 through #4
and #12 through #16 deviated from the scale midpoint by at least 0.5. Numbers in brackets under relations show the number of dyads, 2-plexes, and
cliques, respectively. Numbers in parentheses within brackets show the number of pairs for 2-plexes and cliques, respectively. Empty cells indicate nonsignificant findings (p > 0.0005).
Diagonal lines in cells indicate no 2-plex or clique substructures. D = dyad, P = 2-plex, C = clique, A = actual brand, P = preferred brand.

formulations. The final selection of products was least one chooses the other) for several reasons. First,
governed first by the requirement that brand con- it is commonly found (as in the present study) that
sumption of each was likely to be visible in at least few if any social relations in a clique, for example,
one of the social relationships (e.g., bathroom/sham- are symmetric (i.e., that every person chooses each
poo, TV/roommate, beer/party). Second, the tradi- other). This may lead to unwarranted conclusions by
tional argument that greater brand congruence is ignoring structure present in a social system. Second,
expected for "publicly" than for "privately" consumed there is no compelling theoretical reason to assume
products was incorporated into the study by including that choices have to be symmetric for influence to
products which clearly fell (>3 ? 0.5) in either the occur.
private or public direction of the scale in the pretest The MATCOMP output was then submitted to
(see note for Figure C). It bears repeating that the STRUCT, which searches for cliques and 2-plexes.
traditional scaling of a product's conspicuousness was The details are unimportant for our present purposes,
undertaken primarily to illustrate its shortcomings as but for cliques the algorithm searches for sets of points
a predictor of brand congruence in the absence of such that each point is connected to every other point,
controlling social relations. As already discussed, effects whereas for 2-plexes it determines basic structures
were expected for "public" as well as for "private" with n points such that each point is adjacent to at
products, as determined by the pretest. least n - 2 others but not to every other point. The
data are output in a manner that allows each person
Data Analysis to be matched with the cliques (2-plexes) of which she
is a member. This information is then entered in an
After coding the data, analysis was performed with "attribute file" along with some basic demographics
SONET-1 (Foster and Seidman 1978; Seidman and and brand choices. These attribute files, along with
Foster 1979), a package of computer programs for an adjacency matrix of all ones, are input to a program
graph-theoretic analysis of social structure. The data called CAT3 which does a series of pairwise searches
for each social relation were first submitted to MAT- to determine whether or not each pair of individuals
COMP, which constructs a sociomatrix to represent is in a clique (2-plex) together, makes the same brand
the graph for each relation. MATCOMP then searches choice, and so on.
for connected components (i.e., largest possible The output of CAT3 is a subset of the adjacency
subgraphs such that any two of its points are joined matrix, containing ones only in cells where, say,
by a path) in the graphs. Relations were symmetrized common brand choice occurs. These subset adjacency
(e.g., persons A and B are in a social relation if at matrices are input to a program called STAT which
778 THE JOURNAL OF CONSUMER RESEARCH

transforms the data to make them suitable for cate- significant results were in the predicted direction. The
gorical data analysis. Our data did not lend themselves average raw effect (percent brand congruence for per-
to multivariate approaches to categorical data analysis sons in a BSU minus percent brand congruence for
(e.g., logit-type modeling) that would have enabled us persons not in one) was 28.9 percent across the
to draw inferences beyond what is reported here and significant findings (range: 11.8 percent for study/
to subject the brand congruence hypothesis to a stron- shampoo/dyad to 50.5 percent for friend/pizza/2-
ger test. The data were strongly nonindependent and plex). Overall, these findings support the expectation
unbalanced-e.g., many pairs were in the same BSU of significant brand congruence effects for several of
for more than one type of social relation, dyads in the cells in Figure C, even when stringent statistical
cliques formed a subset of the entire set of dyads, and criteria are applied.
pairs were in a clique for one social relation but in a Also in accord with expectations, significant product
2-plex for another. A large number of zero counts results appear to be dependent on the type of social
(when no cliques and/or 2-plexes were found) posed relation and type of basic structure (e.g., pizza/neigh-
an additional limitation. bor/2-plexes). Although we cannot be certain on the
It was therefore decided to separately examine the basis of the present data, this suggests an explanation
nonempty cells of Figure C through a series of 2 X 2 for the tendency of the brand congruence efforts to be
contingency tables, one dimension being "in a com- clustered in only a few products, e.g., TV show, pizza,
mon BSU/not in a common BSU" and the other and restaurant.
dimension being "common brand choice/not common
brand choice." The frequencies in the tables were Determinants of Brand Congruence Effects:
based on pairwise comparisons. Thus, the same unit Exploratory Analysis
of analysis was employed for each of the types of basic
structure, making statistical analysis more feasible. To address these questions, attention was focused
Statistical tests employing Chi-square were run sepa- on the extent to which the product and type of social
rately for each cell in the 10 (relations) X 2 (actual/ relationship affected the findings. Although formal
preferred brand choices) X 3 (dyad, clique, 2-plex) statistical procedures had to be largely abandoned at
X 16 (products) overall data matrix. A total of 960 this point in favor of a more descriptive approach due
tests was possible, but 736 were actually performed, to the data limitations already specified, this less
since for five of the 10 social relations no cliques or formal analysis yielded several important insights.
2-plexes existed. In the case of clique or 2-plex overlap
(when a pair of individuals was in more than one Conspicuousness Criterion. As expected, the tra-
ditional prediction that the most pervasive brand
clique or 2-plex for a social relation), a pair was
congruence would occur with the "public" products
counted only once.
does not appear to be consistent with our findings.
The previous literature suggests that brand congruence
RESULTS effects would be especially strong for "public" products
Figure C summarizes the (corrected for continuity) such as clothing and beer (Assael 1981, pp. 332-333)
Chi-square results of comparing the brand congruence as well as for wristwatches (Bearden and Etzel 1982,
of co-members and non-co-members in a specified p. 190). Yet, none of our findings were statistically
type of BSU in a specified social relation. Data dis- significant for these products: on the contrary, signif-
cussion and interpretation are restricted to those tests icant findings were obtained for "private" products
which were significant at p c 0.0005 or p < 0.0001 to such as shampoo and toothpaste. It would be difficult
guard against the increased probability of finding to argue that products such as shampoo and toothpaste
significant results due to the large number of tests.8 are more conspicuous across the various types of
social relationships than are products such as running
Brand Congruence Effects shoes, jeans, shirt/blouse, wristwatch, and beer, but
this is not to suggest that conspicuousness is unnec-
Of the 736 statistical tests, 26 were significant at p essary for brand congruence to occur. Products were
c 0.0005 and 13 of them were significant at p c 0.0001. selected so that their consumption was likely to be
Percentage differences were also examined; all of the visible in at least one of the types of relation (e.g.,
shampoo/bathroom). However, the conspicuousness
criterion appears to be insufficient to account for the
8 Due study's findings, even when the effect of type of social
to nonindependenceacross tests, an exact p = value for
the familyof tests could not be determined.The level of significance relation on brand visibility is taken into account.
of p ? 0.0005 was judged to be a reasonablecompromisebetween Overlap. That being so, attention was directed at
the exploratorynature of the study on one hand, and extremely
conservativestandardson the other. Note that for many of the how cliques (or other BSUs) were joined together by
cells in Figure C significanteffects should not be expected on a overlapping membership and how BSUs in different
priorigrounds. social relations intersected. Discussion is limited to
NETWORK ANALYSIS OF BRAND CONGRUENCE 779

FIGURE D
SIGNIFICANTBRAND AGREEMENTIN FRIENDSHIPCLIQUES

45 10 -44
385

10

13 ~10

10o10 92 190

0
~~~~~~10 6

299 379 399 / 408

10

401 511
340 18
NOTE: The number along a solid line indicates actual brand agreement in a clique pair for product #10 (pizza). The following are cliques: (2, 13, 75, 94); (2, 13, 94, 529); (18, 340, 529);
(44, 45, 75); (92, 190, 385); (299, 379, 401); and (399, 408, 511).

cliques, because the sheer number of dyads made for account for all of the data. For example, bathroom
difficult interpretation and graphical depiction, and cliques did not overlap, yet significant findings were
because an analysis of overlap for 2-plexes yielded obtained.
essentially the same conclusions as for cliques. Over- Friendship cliques (2, 13, 75, 94) and (2, 13, 94,
lapping clique membership in a given relation is 529) together form a 2-plex, and these cliques were
important because it provides insights into how brand further examined to obtain insights into the effects on
congruence within cliques is transmitted outside the brand choice of intersections across different social
cliques. And intersection of cliques within different relations. The substructure (2, 13, 94, 529) was found
relations is important sociologically because such to be a clique for the social relations friend, meal,
''overlap" may suggest especially cohesive groupings party, and problem and the substructure (2, 13, 75,
of individuals-a property which previous research 94) was found to be a clique for friend, meals, and
suggests is related to brand congruence (e.g., Stafford party. Figure E shows this clique overlap across rela-
1966; Witt 1969). tions graphically, along with information on actual
Figure D shows overlap of individuals in different brand congruence for each pair. Congruence is shown
friendship cliques along with the members' pairwise for all products to generate exploratory insights. Pair-
congruent brand choices for pizza (raw effect = 36.2 wise brand agreement varied from one product to
percent). Four of the seven cliques overlap (i.e., at another even though these cliques were probably co-
least one member belongs to more than one clique); hesive. At one extreme, there was complete within-
overlapping cliques account for 100 percent of the pair and across-pair agreement for the products pizza
pairwise congruence data for pizza. Figure D is an and running/tennis shoe. At the other extreme, not a
extreme example in which overlap of individuals single pair agreed on actual brand choices for shampoo,
across cliques completely spanned the brand congru- magazine, and camera. Clearly, the relationship be-
ence effect for one product. In general, clique overlap tween clique co-membership and congruent brand
of individuals for a type of social relation could not choice depends heavily on the product.
780 THE JOURNAL OF CONSUMER RESEARCH

FIGUREE
BRANDAGREEMENTS
OVERALL INTHREEMULTIPLEX
CLIQUES

3 6 10 13 316
529

2 5 10 13 14| / \ p2 8 10 13 14|
.___ --
....._.._.._._._._... ------ _------- ------
~~~~f131~~~~1>.7
3 10 11 13 15| j 0

L10 13 |;2 .._...__.__::__.__ 94 10 11 13 15|


|910 12 13 | 2 3 7 10 13 14

Party relations in clique (2, 13, 75, 94, 529)


Friend n Meal n Party n Problem relations in clique (2, 13, 94, 529)
*-----------------
-------- Friend n Meal n Party relations in clique (2, 13, 94, 75)

NOTE: Numbers in a rectangle indicate actual brand agreements in a clique pair for the relations as specified above. These numbers correspond to the products listed in Figure C. For
example, the pair 529/75 agrees on brand choices for products 10 and 13, the pair 2/75 agrees on brand choices for products 10 and 13, etc.

It should be emphasized, however, that clique in- pared between multiplex cliques (i.e., intersection
tersection across several social relations was relatively across at least three social relations) and cliques that
unusual. There were only three other cliques which were relation-specific. The five cliques previously
showed overlap across at least three social relations. identified met the criterion of multiplex cliques, and
Nor is it suggested that intersection across different there were seven cliques which were not only relation-
relations is necessarily important. Counter examples specific but which also had pairs that were pairs for
are easy to find; for instance, bathroom cliques showed only that specific relation. For both actual and pre-
virtually no overlap across relations, but significant ferred brands, there was significantly greater overall
brand congruence effects. brand congruence for multiplex cliques (X2 = 10.99,
This sort of intersection has been widely discussed p < 0.01, x2 = 10.42, p < 0.01, respectively). It should
(sometimes called multiplexity, sometimes many- be noted, however, that this overall result is dispro-
strandedness), and we thought it important to further portionately accounted for by cliques (2, 13, 94, 529)
examine the data with an eye to such intersections. and (2, 13, 75, 94) and their complete agreement for
Pairwise brand agreement for all products was com- the products pizza and running/tennis shoe.
NETWORK ANALYSIS OF BRAND CONGRUENCE 781

DISCUSSION this would have increased the complexity of the data


structure greatly, thus limiting its use as illustration.
In the interpretation and discussion of these findings, Overall, the results described here strongly suggest
it is necessary to carefully balance the study's strengths that brand congruence depends heavily on a complex
and limitations, since it is exploratory and had a non- mix of type of product, type of social relation, and
experimental design. On the one hand, the sorority type of social structural unit. In general, of course,
surely has characteristics that differentiate it from the fact that we found significant brand congruence
other social systems, such as neighborhoods, colleagues effects confirms previous findings of studies dealing
at work, or bridge clubs. For example, brand choice with the impact of informal groups on member brand
behavior may be unusually visible, and there may be choices (Stafford 1966; Witt 1969; Witt and Bruce
more joint consumption behaviors across a wider 1970; Witt and Bruce 1972).1o This is important
range of products than would be expected for other because we used a very different and more valid
social systems. On the other hand, this limitation is measure of social structure than did previous studies.
compensated for by maintaining much greater control It is also important because previous studies have not
on the kinds of relations and social groups than is systematically examined types of social relation and
usually present. The sorority setting limits the basic structure, and their methodologies have been
straightforward application of results to other cases in found to be suspect on several other grounds.
much the same way and for analagous reasons that
Our results also support the more specific, earlier
experimental results are difficult to generalize. Yet the findings of a correlation between group cohesiveness
real issue here is the controlled selection of relations, and member brand choice, in that we found signifi-
products, and social structures. It is in this sense that cantly greater overall brand congruence for multiplex
the study is most profoundly exploratory: its major cliques than for relation-specific cliques. Although we
substantive contribution is that variation in types of did not directly measure cohesiveness, it seems plau-
relations and group is important and appears to inter- sible that multiplex groups-i.e., cliques which occur
act with type of product to produce brand congruence. in several different relations-are mQre cohesive than
It may well be possible to return to broader-based those which are specific to a single relation. Still, not
survey methodology to address the problem of gener- all multiplex cliques showed brand congruence, and
alizability or to engage in experimental work to deal not all brand congruence occurred in multiplex cliques.
with causality issues when the fundamental dynamics And in any case, congruence varied by product. Clearly
relating different kinds of relations, structures, and there remains a great deal to explain.
products are better understood, but a great deal more Much of the uncertainty revolves around the nature
exploratory work is needed in this area before that of product and the notion of product visibility. The
will be productive.9 traditional conspicuousness hypothesis appears to be
It is also worth emphasizing that the method of
analysis employed here was exploratory in the sense
that it is not wholly sufficient in itself to address the
brand congruence question. Rather, it illustrates the 10No judgment is made on the pervasiveness of the brand
point that other analyses have missed important di- congruencephenomenonfound in the study. Persuasivearguments
could be advanced on both sides of the issue. On one hand, the
mensions of brand congruence by ignoring crucial findingscould be interpretedas suggestingthat brand congruence
complexities of social structure. To draw inferences is not as pervasiveas previous studies might imply. The current
beyond what is reported here and to subject the brand researchfocused on primaryinformal groups-the type of group
congruence hypothesis to a stronger and generally generally suggested to have the strongest influence on member
more sufficient test through multivariate approaches behavior.The subjectswere students, who have been found to be
more susceptible to reference group influence (Park and Lessig
to categorical data analysis would require a more 1977). Note further that (based on previous studies) the social
complex design. For example, types of social relation system environmentwas one which allowed for strong visibility of
and product could be randomized across different but brandconsumptionacross a wide arrayof products.Yet relatively
matched social systems (e.g., several similar neighbor- few of the tests were significant.
On the other hand, it could be argued that the current study
hoods) to deal with the problem of non-independence understatesthe case of brand congruencefor socially related indi-
which limited our approach to data analysis. Of course, viduals.The statisticalstandardswere quite conservative,yet many
tests survivedthem. (For the readerwho wishes to apply a more
conventional level of statisticalsignificance, 127 of the cells were
significantat p ? 0.05.) It could also be pointed out that many
9 Althoughthe study shareswith previousinformalgroup studies cells in the design should not be significanton a priori grounds
the problems of generalizabilityand causality, it should be re- (e.g., bathroom/calculator,sports/detergent,etc.) and that brand
emphasizedthat its basic goal was to illustrate how other major choice behaviorwas perhapsinfluencedby previoussocial interac-
limitations in previous investigationscan be effectivelyaddressed tion. Finally, the insignificanteffects for many products do not
with social networkanalysis.Social networkstudies typicallysuffer necessarilyimply that referencegroup influencewas not operating.
from these problems because social systems tend to have some A person may have many referencegroups (Shibutani 1955), and
unique characteristicsand members in voluntary social systems the study may not have captured the relevant ones in its list of
tend to have more similarpersonalattributes. social relations.
782 THE JOURNAL OF CONSUMER RESEARCH

insufficient and too imprecise to account for our congruence; thus it may be an alternative explanation
findings: significant effects were sometimes observed, to the results obtained here and in previous informal
but in many instances they were not found (e.g., group studies.
jeans, shirt/blouse, tennis/running shoe). Perhaps This alternative interpretation cannot be entirely
worse, the conspicuousness hypothesis in its original discounted, but several aspects of the data render it
form may be virtually untestable in correlational less plausible. First, it would be difficult to explain by
studies, since type of relation is uncontrolled and a similarity argument why congruence effects were
some social relations may allow for brand identification limited to only certain cells of the overall data matrix.
for products generally judged to be private. Even when For example, significant results were obtained in the
one defines a public product as one that is public friend/pizza cells. If similarity in brand decisions
across many social relations, the present data do not resulted primarily from shared personal attributes,
appear to support the proposition that brand congru- why were the results for all the other products insig-
ence will necessarily follow. nificant? Second, the previous analysis on multiplex
There are other ways in which the existing literature cliques revealed no brand agreements for several prod-
seems to fail to explain the present findings. What, ucts. The extent of shared personal attributes should
for example, may explain the fact that in our study, be especially strong for members of multiplex cliques.
the products TV show, restaurant, and pizza accounted Finally, it is important to recognize that brand
for the majority of the significant findings? It is congruence effects were limited to certain products,
difficult to account for this in the literature. In one types of social relation, and types of basic structure
discussion that addresses this kind of issue, Robertson interdependencies. Although the present data do not
(1971) maintains that products low in visibility, com- allow us to provide specific insights into the working
plexity, and perceived risk and high in testability are of these interdependencies, they suggest that types of
less susceptible to personal influences than are products social relations may channel joint/shared consumption
high in visibility, complexity, and perceived risk and processes (e.g., roommate/TV show; neighbor/pizza).
low in testability. Yet it would be difficult to argue What is common to all of these observations is the
that TV show choice (for which significant brand difficulty encountered in addressing them without the
congruence effects were found) is higher in visibility, use of sensitive, precise, and computable measures of
complexity, and perceived risk and lower in testability social structure such as those used in network analysis.
than is wristwatch brand choice (for which no signif- At least three capabilities are needed: (1) to work with
icant brand congruence effects were obtained). many relations-singly and in combination, (2) to
We suggest something more like the following. It is work with different types of social units whose prop-
likely that for products such as TV show, restaurant, erties have solid justification in social theory, and (3)
and pizza, joint brand consumption in interpersonal to relate both of these analytically and conceptually
settings is a more frequent phenomenon than it is for to the properties of products under investigation. Of
products such as watch, jeans, shirt/blouse, and so course, this program is demanding in its data require-
on. Joint brand consumption may be preceded by ments, and ingenuity in research design and data
interpersonal decision making ("What do you want collection are necessary. We suggest that social network
to watch?" "Where do you want to go?") among the analysis can provide valuable conceptual and com-
individuals who comprise a certain basic structure for putational foundations for carrying further the program
a type of social relation. As far as the significant we have outlined in this study.
findings for shampoo and magazine are concerned,
shared brand consumption is also a possibility in the [Received November 1983. Revised August 1984.]
social system under investigation. l
Perhaps joint consumption settings (many of which
may be preceded by interpersonal decision making) REFERENCES
are particularly important loci of influence. Before Alba, Richard D. (1973), "A Graph-Theoretic Definition of
accepting that possibility, however, it should be noted a Sociometric Clique," Journal of Mathematical Soci-
that interpersonal brand agreement is not necessarily ology, 3, 113-126.
due to influence. Although previous investigators have Asch, S. E. (1953), "Effects of Group Pressure Upon the
attributed brand congruence to informal group influ- Modification and Distortion of Judgments," in Group
ence, it is conceivable that brand congruence may be Dynamics, eds. D. Cartwright and A. Zander, New
due to interpersonal similarity. This similarity among York: Harper & Row.
Assael, Henry (1981), Consumer Behavior and Marketing
group members may manifest itself in greater brand
Action, Boston, MA: Kent.
Bagozzi, Richard P. (1981), "Attitudes, Intentions, and
Behavior: A Test of Some Key Hypotheses," Journal
II We examinedthe distributionsof brandchoicesfor the products of Personality and Social Psychology, 41, 607-627.
and found the shape of the distributionsunable to account for the Bearden, William 0. and Michael J. Etzel (1982), "Reference
effects. Group Influence on Product and Brand Purchase De-
NETWORKANALYSIS OF BRAND CONGRUENCE 783

cisions," Journal of Consumer Research, 9 (September), Lessig, V. Parker and C. Whan Park (1978), "Promotional
183-194. Perspectives of Reference Group Influence: Advertising
Bourne, Francis, S. (1957), "Group Influence in Marketing Implications," Journal of Advertising, 7, 41-47.
and Public Relations," in Some Applications of Behav- Luce, R. D. and A. Perry (1949), "A Method of Matrix
ioral Research, eds. R. Likert and S. P. Hayes, Paris: Analysis of Group Structure," Psychometrika, 14, 94-
UNESCO. 116.
Burnkrant, Robert E. and Alain Cousineau (1975), "Infor- Miniard, Paul W. and Joel B. Cohen (1981), "Examination
mational and Normative Social Influence in Buyer of the Fishbein-Ajzen Behavioral Intention Model's
Behavior," Journal of Consumer Research, 2 (Decem- Concepts and Measures," Journal of Experimental So-
ber), 206-215. cial Psychology, 17, 309-339.
Burt, Ronald S. (1980), "Models of Network Structure," in Moschis, George P. (1976), "Social Comparison and Informal
Annual Review of Sociology, Vol. 6, Palo Alto, CA: Group Influence," Journal of Marketing Research, 13
Annual Reviews, Inc., 79-141. (August), 237-244.
Carmon, Naomi and Bilha F. Mannheim (1979), "Reference Nadel, Siegried F. (1957), The Theory of Social Structure,
Groups and Professional Self-Image: The Case of Ar- London: Cohen and West.
chitects," Journal of Vocational Behavior, 14, 169-180. Park, C. Whan and V. Parker Lessig (1977), "Students and
Cohen, Joel B. and Ellen Golden (1972), "Informational Housewives: Differences in Susceptibility to Reference
Social Influence and Product Evaluation," Journal of Group Influences," Journal of Consumer Research, 4
Applied Psychology, 56 (February), 54-59. (September), 102- 110.
Cocanougher, A. Benton and Grady D. Bruce (1971), "So- Radcliffe-Brown, A. R. (1965), Structure and Function in
cially Distant Reference Groups and Consumer Aspi- Primitive Society, New York: The Free Press.
rations," Journal of Marketing Research, 8 (August), Reingen, Peter H. (1982), "Test of a List Procedure for
379-381. Inducing Compliance With a Request to Donate
Deutsch, M. and H. B. Gerard (1955), "A Study of Nor- Money," Journal of Applied Psychology, 67, 110-118.
mative and Informational Social Influences Upon In- Robertson, Thomas S. (1971), Innovative Behavior and
dividual Judgment," Journal of Abnormal and Social Communication, New York: Holt, Rinehart & Winston.
Psychology, 51, 624-636. Ryan, Michael J. (1982), "Behavioral Intention Formation:
Engel, James F. and Roger D. Blackwell (1982), Consumer The Interdependency of Attitudinal and Social Influence
Behavior, Hinsdale, IL: Dryden. Variables," Journal of Consumer Research, 9 (Decem-
Festinger, Leon (1954), "A Theory of Social Comparison ber), 263-278.
Processes," Human Relations, 7 (May), 117-140. Seidman, Stephen B. and Brian L. Foster (1978a), "A
Fishbein, Martin and Icek Ajzen (1975), Belief, Attitude, Graph-Theoretic Generalization of the Clique Concept,"
Intention and Behavior: An Introduction to Theory and Journal of Mathematical Sociology, 6, 139-154.
Research, Reading, MA: Addison-Wesley. and Brian L. Foster (1978b), "A Note on the
Ford, Jeffrey D. and Elwood A. Ellis (1980), "A Reexami- Potential for Genuine Cross-Fertilization between An-
nation of Group Influence on Member Brand Prefer- thropology and Mathematics," Social Networks, 1, 65-
ence," Journal of Marketing Research, 17 (February), 72.
125-132. and Brian L. Foster (1979), "SONET-I: Social
Foster, Brian L. and Stephen B. Seidman (1978), SONET- Network and Modeling System," Social Networks, 2,
I. Social Network Analysis and Modeling System, Vol. 85-90.
1, Binghamton, NY: Center for Social Analysis, State Shaw, Marvin E. (1976), Group Dynamics, New York:
University of New York. McGraw-Hill.
and Stephen B. Seidman (1981), "Network Structure Shibutani, Tamotsu (1955), "Reference Groups as Perspec-
and the Kinship Perspective," American Ethnologist, 8 tives," American Journal of Sociology, 60, 562-569.
(May), 329-355. Stafford, James E. (1966), "Effects of Group Influence on
and Stephen B. Seidman (1983), "A Strategy for the Consumer Brand Choice Preference," Journal of Mar-
Dissection and Analysis of Social Structures," Journal keting Research, 3 (February), 68-75.
of Social Biological Structure, 6, 49-64. Venkatesan, M. (1966), "Experimental Study of Consumer
Hansen, Flemming (1969), "Primary Group Influence and Behavior Conformity and Independence," Journal of
Consumer Conformity," Proceedings of the American Marketing Research, 3 (November), 384-387.
Marketing Association's Educators Conference, ed. Phi- Witt, Robert E. (1969), "Informal Social Group Influence
lip R. McDonald, Chicago: American Marketing As- on Consumer Brand Choice," Journal of Marketing
sociation, 300-305. Research, 6 (November), 473-477.
Hawkins, Del I., Roger J. Best, and Kenneth A. Coney and Grady D. Bruce (1970), "Purchase Decisions
(1983), Consumer Behavior: Implications for Marketing and Group Influence," Journal of Marketing Research,
Strategy, Plano, TX: Business Publications. 7 (November), 533-535.
Jahoda, Marie (1972), "Conformity and Independence: A and Grady D. Bruce (1972), "Group Influence and
Psychological Analysis," in Behavioral Science Foun- Brand Choice Congruence," Journal of Marketing Re-
dations of Consumer Behavior, ed. Joel Cohen, New search, 9 (November), 440-443.
York: Free Press, 339-354. Zaltman, Gerald and Melanie Wallendorf (1983), Consumer
Kelman, Herbert C. (1961), "Processes of Opinion Change," Behavior: Basic Findings and Management Implications,
Public Opinion Quarterly, 25 (Spring), 57-78. New York: John Wiley.
Knoke, David and James H. Kuklinski (1982), Network
Analysis, Beverly Hills, CA: Sage.

Vous aimerez peut-être aussi