Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Minority opinion influence: the role of issue-involvement and similarity Venkatasubramaniam, Ramesh 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1993_fall_phd_venkatasubramanian_ramesh.pdf [ 5.87MB ]
JSON: 831-1.0098868.json
JSON-LD: 831-1.0098868-ld.json
RDF/XML (Pretty): 831-1.0098868-rdf.xml
RDF/JSON: 831-1.0098868-rdf.json
Turtle: 831-1.0098868-turtle.txt
N-Triples: 831-1.0098868-rdf-ntriples.txt
Original Record: 831-1.0098868-source.json
Full Text

Full Text

MINORITY OPINION INFLUENCE:THE ROLE OF ISSUE-INVOLVEMENT AND SIMILARITYbyRAMESH VENKATASUBRAMANIAM (VENKAT)B.Com., Madurai-Kamaraj University, India 1983M.B.A., Madurai-Kamaraj University, India 1985M.B.A., Simon Fraser University, Canada 1988A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(Commerce and Business Administration)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAMay 1993© Ramesh Venkatasubramaniam, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) Department of COMMERCE AND BUSINESS ADMINISTRATIONThe University of British ColumbiaVancouver, CanadaDate  AUGUST 12, 1993DE-6 (2/88)ABSTRACTSocial influence in marketing has generally been conceptualized in terms ofconformity, where the individual's attitudes and behaviour are influenced by real orimagined group pressure. This is a one-way influence process where the group(majority) influences the individual. This research extends this conceptualization ofsocial influence to include the influence of minority or deviant opinions. Asimultaneous social influence paradigm is adopted, in which individuals may not onlyexperience conformity pressure from the majority, but may also be subject topersuasion by minority opinions in the group. Such situations may arise in consumergroups as such organizational buying committees or families.Several conditions that may determine the extent of conformity or minorityinfluence were delineated. It was hypothesized that the extent of social identificationwith a minority or majority source, i.e., source -similarity, would determine the extentof its influence. It was proposed that issue-involvement would play an important rolein determining conformity versus minority influence effects, as well as interactsignificantly with source-similarity. The role of other mediating variables in this socialinfluence process, such as source credibility and source feelings, were also explicated.An empirical test of the theory was undertaken through a 2 (high/low similarity)X 2 (high/low involvement) factorial design. Subject were exposed to persuasivecommunication from both a majority and a minority source, who advocated contraryiipositions. The two sources always assumed opposite social identities, and thus whenone source was similar to the subject, the other was dissimilar. As anticipated, theminority opinion was more persuasive when the minority was similar, rather thandissimilar. However, this effect was dependent on the level of involvement. Theresults were generally consistent with the proposed model, with both similarity andinvolvement playing a crucial role in determining the extent of minority influence.Source credibility and feelings towards the source were both significant mediators inthe social influence process. This research indicates a further need to explore the roleof involvement in such simultaneous influence contexts using other consumercontexts, and it opens several avenues for future research.iiiTABLE OF CONTENTSABSTRACT^ iiTABLE OF CONTENTS^ ivLIST OF APPENDICES xLIST OF TABLES xiLIST OF FIGURES^ xiiACKNOWLEDGEMENTS xiiiI. INTRODUCTION^ 11.0 Overview 11.1^Background^ 11.2 Proposed Research^ 51.2.1 Summary of Proposed Research^ 81.3 The Study^ 81.4^Potential Contribution of Research^ 91.5^Organization of Thesis^ 11II. BACKGROUND^ 122.0 Overview 122.1^The Conformity Paradigm^ 122.1.1 Conformity Effect In Marketing^ 142.2 The Minority Influence Paradigm 162.2.1 Minority Influence and Marketing^ 182.3^Perspectives on Majority-Minority Influence 18iv2.3.1 Theory of Conversion Behaviour^ 192.3.2 Model of Idiosyncrasy Credits 232.3.3 Commonality Between the Two Perspectives^242.4 Summary^ 25III.^LITERATURE REVIEW AND HYPOTHESES^ 263.0^Overview^ 263.1^Multiple Source Model^ 263.2^Source-Recipient Similarity 293.2.1 Social Identification and Social Categorization^303.2.2 Evidence Regarding Social Categorization Effect^333.2.3 Source Credibility and Source Similarity^343.2.4 Source-Related Feelings and Persuasion 363.2.5 Source Similarity Hypotheses^ 373.3^Issue-Involvement and Persuasion 393.3.1 Elaboration Likelihood Model and Minority Influence^403.3.2 Parallels Between ELM and TCB^ 423.3.3 Issue-Involvement and Source-Similarity^453.3.4 Issue-Involvement Hypotheses^ 463.4 Response-Involvement^ 473.4.1 Issue versus Response-Involvement^ 483.4.2 Response Involvement and Source-Similarity^49v3.4.3 Response-Involvement Hypotheses^ 51IV. PRETESTS^ 524.0 Overview^ 524.1 Selection of Group Discussion Topic^ 524.2 Pretest: Cafeteria Topic^ 544.2.1 Design and Subjects 554.2.2 Experimental Procedure^ 554.2.3 Stimulus Material^ 564.2.4 Manipulations 574.2.5 Measures^ 584.2.6 Demand Assessment and Debriefing^ 584.2.7 Results^ 584.3 Revision of Manipulations and Stimulus^ 594.4 Pilot Study^ 614.4.1 Reliability of Measures^ 614.4.2 Manipulation Check 634.4.3 Hypotheses Tests^ 654.4.4 Discussion^ 674.4.5 Steps Undertaken^ 724.5 Pretest: Joint Venture Topic 744.5.1 Pretest Summary and Overview of Final Design^76v iV. METHODOLOGY 785.0 Overview 785.1 Methodological Issues 785.1.1^Real versus Nominal Groups 785.1.2^Single Source versus Dual Sources 805.1.3 Attitude versus Attitude Change 805.1.4^Stimulus Material and Format 825.2 Main Experiment 825.2.1^Experimental Design 835.2.2^Subjects 835.2.3^Procedure 845.2.4^Manipulations 855.2.5^Stimulus 875.2.6^Measures 885.2.7^Analysis 895.2.8 Summary 90VI. RESULTS 916.0 Overview 916.1 Reliability Assessment 916.1.1^Cronbach's Alpha 916.1.2^Confirmatory Factor Analysis 92vii6.2 Validity Assessment^ 936.2.1 Convergent Validity^ 946.2.2 Discriminant Validity 946.3 Manipulation Checks^ 966.3.1 Involvement Manipulation Check^ 966.3.2 Similarity Manipulation Check 976.3.3 Minority-Majority Manipulation Check^ 996.4 Hypotheses Tests: Similarity and Involvement 996.4.1 Main Effect^ 1006.4.2 Interaction Effect 1006.4.3 Comparison of Means^ 1006.5 Hypotheses Tests: Credibility and Feelings^ 1016.5.1 Source Credibility^ 1016.5.2 Source Related Feelings 1026.6 Mediating Effects^ 1036.6.1 Methodology 1046.6.2 CREDIBILITY and FEELINGS as Criterion Variables^1066.6.3 ATTITUDE as the Criterion Variable^ 1086.6.4 Test of SIMILARITY-F and INVOLVEMENT-FInteraction^ 1096.7^Additional Analysis 1106.8 Summary^ 112viiiVII. CONCLUSION^ 1137.0 Overview 1137.1 Key Research Issues^ 1137.2 Summary of Results 1147.2.1 Measurement^ 1147.2.2 Effectiveness of Manipulations^ 1157.2.3 Hypotheses^ 1157.3 Discussion^ 1167.3.1 Similarity^ 1177.3.2 Involvement 1197.3.3 Mediating Effects^ 1217.3.4 Support for Theory 1227.4 Limitations of Research^ 1237.4.1 Generalizability and External Validity^ 1237.4.2 Design and Experimental Execution 1247.4.3 Measures^ 1257.5 Implications of Research 1257.5.1 Theoretical Implications^ 1257.5.2 Managerial Implications 1277.6 Future Research^ 128REFERENCES^ 131APPENDICES 142TABLES 174FIGURES^ 184ixLIST OF APPENDICES1. Pilot Study: Treatments, Stimulus Material and Measures^1422. Main Experiment: Treatments, Stimulus and Measures 158xLIST OF TABLES1. Test of Main Effects for Pretest 1^ 1742. Pilot Study - Reliability of Measures 1753. Pilot Study - Rotated Factor Matrix^ 1764. Pilot Study - Similarity X Involvement X Response ANCOVA^1775. Main Study - Reliability of Measures^ 1786. Parameter Estimates of Reliability Models 1797. Summary of Manipulation Checks for Main Study^ 1808. Main Study - Similarity X Involvement ANCOVA 1819. Path Analysis - CREDIBILITY and FEELINGS as Criterion Variables^18210. Path Analysis - ATTITUDE as the Criterion Variable^ 183xiLIST OF FIGURES1. Conformity Influence Model 1842. Dual Influence Model 1853. Simultaneous Social Influence Model 1864. General Reliability Model (True-Scores) 1875. Equal Units Reliability Model 1886. Parallel Form Reliability Model 1897. Single Factor Confirmatory Factor Model 1908. Main Hypotheses and Results 1919. Mediating Effects of Credibility and Feelings 192xiiACKNOWLEDGEMENTSI wish to thank Dr. Richard Pollay, my Research Supervisor, for the adviceand motivation he provided throughout this research project. I would also like toexpress my deep appreciation to Dr. David Tse for his encouragement during mytenure as a Ph.D student as well as for his insightful feedback on many aspects ofthis dissertation. This dissertation had gained a lot from discussions with variousfaculty members at The University of British Columbia, including Professors GerryGorn, Robert Kelly, Jong Won Park and Craig Pinder. I am very thankful to all ofthem. I would also like to express my gratitude to Ruby Visser and Rosalea fortheir support during my stay at the University of British Columbia. Last but notleast, I am greatly indebted to my mother and my brother for their constantsupport and their faith in my ability.I. INTRODUCTION1.0 OverviewThis chapter outlines the motivation behind this research as well as the scope of thisresearch. At the outset, within the field of marketing, a knowledge gap in the areaof social influence is identified and a theoretical model intended to fill this gap ispresented. The chapter also provides a brief overview of the theory on which thisresearch is founded. Finally, a description of the empirical work conducted to test theproposed model of social influence is presented. The chapter ends with a briefsummary of the contributions of this research.1.1 Background Social or group influence in marketing, it has been reported, extends to a wide rangeof products such as automobiles (Grubb and Stern 1971), cosmetics (Moschis 1976),illicit consumption of drugs and alcohol (Rose, Bearden and Teel 1992) and a varietyof private and public consumer products (Bearden and Rose 1990). Furthermore, theinfluence of reference groups is felt in diverse groups in the society such as physicians(Coleman, Katz and Menzel 1966), students and housewives (Park and Lessig 1977).Given that our decisions are often influenced by others, a thorough understanding ofthe social influence process can be useful theoretically and managerially.1Many consumer decisions are made in group settings and many others are madeafter directly or indirectly incorporating the opinions of close friends or family. Inconsumer groups (such as organizational buying committees, families or informalsocial groups) it is not uncommon to see members disagreeing on goals or actions.Thus, often multiple opinions may exist on the same issue. When multiple opinionsexist, marketers have believed that individuals in the group will conform to the will ofthe majority opinion or the group norm rather than risk siding with a minority or lesspopular opinion. Such a perspective is based on the works of social psychologistssuch as Asch (1951), Deutsch and Gerard (1955), and Sherif (1953), which has beenconsidered a fundamental axiom of social influence - individuals yield to groupinfluence and conform to group norms. The conformity perspective dominates thestudy of social influence to such an extent that the term conformity is oftenconsidered to be equivalent to social influence (Maass and Clark 1984).The reference group literature in marketing is generally based on the conformityperspective (e.g., Bearden and Rose 1990; Hansen 1969; Rose, Bearden and Teel1992; Stafford 1966; Venkatesan 1966; Witt 1969, Witt and Bruce 1970). Whileconsumer researchers have acknowledged that susceptibility to social influence islikely to vary across individuals (e.g., Bearden, Netemeyer and Teel 1989; ), meaningthat not all consumers are likely to submit to conformity pressures to the sameextent, the marketing literature is silent on the issue of whether individuals whodecide to maintain their individuality and deviate from group norms hold any influentialor persuasive power over other individuals in the group. Thus, social influence is still2viewed as a one-way process, where the influence flows from the larger group tosubgroups or individuals.Instead of assuming that an individual will gravitate towards the group norm (ormajority point of view), in recent years social psychologists have recognized deviatingor dissenting minorities within groups as agents of innovation (Moscovici 1976). Inother words, individuals in consumer groups could be the source of original ideas (cf.Nemeth 1985; Folkes and Kiesler 1991) and can be quite influential. This point isunderscored by the fact that adoption of new products and new patterns ofconsumption are often started by a few who dared to disregard the popular wisdom.Thus, the view that individuals in groups succumb to the influence of the majority(i.e., group norms) without having any influence on other group members is anincomplete view of social influence at best.Before explicating the proposed social influence model further and identifyingits relevance to consumer behaviour and marketing, it may be worthwhile defining theterms "minority" and "majority" in the context of this research. A minority or amajority can be defined in terms of two characteristics: (a) Size - numerically, aminority is smaller than the majority. The larger the strength of the majority, thegreater its influential power on other group members. (b) Position - a minority notonly opposes the group norms (majority view), but it also actively defends its ownpreference, and hence the minority is antinomic or represents a position contrary tothe norm (Moscovici 1980). Thus, in this context the usage of the term "minority"does not refer to ethnic minorities.3Within groups why does the majority or the minority have any influentialpower? When there is a majority opinion in a group, that opinion becomes acceptablefor several reasons. The fact that "so many people hold this view" often leads us tobelieve that the majority must be correct. The majority, due to its size, acquiresgreater credibility and hence persuasive power (Moscovici 1976). The minorityopinion, on the other hand, lacks the numerical strength. But individuals who aredeviant (i.e., minority) are known to be persuasive when they maintain their positionconsistently and unflinchingly. The consistency of the minority leads observers (othergroup members) to conclude that the minority is confident and certain, which in turnmakes the minority opinion acceptable (Moscovici 1980).The mounting evidence supporting minority influence in groups indicates thatin the field of marketing, where study of group influences on buyer behaviour occupiesan important place, a reconceptualization of group influence is essential. This wouldentail extending our current conception of group influence, which is mainly based onthe notion of conformity, to one that will allow both majority and minority withingroups to be the source of influence (see Figure 3). Such a model of social influenceis termed the simultaneous social influence paradigm (Clark and Maass 1988a; Maassand Clark 1984). In other words, individuals may be recipients of influence from themajority (representing the group norm) as well as the minority (representing a deviantopinion), and the relative influence of either source may depend on a number offactors which are outlined next.41.2 Proposed Research Consumers often seek the opinions of several sources (e.g., friends, neighbours,colleagues, salespersons, TV spokespersons etc.) on decisions ranging from choiceof restaurants to choice of cars. In such situations consumers not only try to evaluatethe information obtained from various sources , but they also try to gauge the socialsupport available for different alternatives. Frequently, consumers get contradictoryinformation from different sources and they may be forced to make a choice betweenalternatives that have varying social support. When social influence simultaneouslyoccurs from multiple sources we have previously assumed that the position with thegreater social support (majority view) will be more influential. The objective of thisresearch is to offer an extended conceptualization of social influence that allows bothminorities and the majority to be the source of influence. The conditions facilitatingeither a minority or a majority influence are specified. An empirical test is proposedto test the predictions of the model. The important variables in the proposed modelare as follows.First, based on previous research in persuasion, it is proposed that at least partof the influence in a social setting can be attributed to the extent to which therecipient of influence identifies with the source of influence. In other words,similarity between the source and the recipient of influence has been found to be acritical factor in persuasion (e.g., Berscheid 1966; Feick and Higie 1992; Moschis1976). Even a less popular position may be acceptable if one shares a common bondwith the proponent of that viewpoint (e.g., Clark and Maass 1988a; Martin 1988a).5Second, the degree of issue-involvement (in terms of personal implications) islikely to affect the influence of the majority and the minority sources. This researchextends Petty and Cacioppo's (1986) Elaboration Likelihood Model to a socialinfluence context to explain conformity versus minority influence. It is posited thatan uninvolved person is more likely to join the majority using a "majority must beright" type of heuristic. On the other hand, someone with a high issue-involvementis more likely to engage in diligent consideration of each alternative before making achoice, which would make the characteristics of the source less important. Undersuch circumstances, the less popular of the two opinions (i.e., the minority opinion)should stand a better chance of being persuasive.Previously, there has been scant attention paid to the role of issue-involvementin determining conformity versus minority influence, even though many scholars havecalled for an examination of this issue (Maass and Clark 1984; Chaiken and Stangor1987). The limited attempts made in this direction indicate that this could be afruitful avenue of research (see Trost, Maass and Kenrick 1992). 1 Issue-involvementhas been recognized as having a very important role in persuasion (Johnson and Eagly1989). Hence issue-involvement or personal involvement may be a key determinantof conformity versus minority influence effects.Third, response-involvement or response-type (public versus private response)1 Trost, Maass and Kenrick (1992) was published after the empirical work on thisdissertation was completed. To the best of my knowledge there is no other publishedresearch on the role of issue-involvement in this context. Given the importance of thissubject, in appropriate places in this dissertation Trost et al.'s work is compared withthis research in terms of the theory, methodology as well as findings.6is considered to be another important determinant of the nature of social influence.Often one's attitudes and preferences may be biased by whether or not one's positionis expressed publicly. Social desirability or need for social approval may cause adifference between private and public attitudes (Bearden and Rose 1990; Clark andMaass 1988a). Thus response-type is a likely determinant of conformity versusminority influence.Fourth, other variables such as source credibility and feelings towards thesource were also expected to influence the relative influences of the majority andminority sources. Research in the area of persuasion (e.g., Dholakia and Sternthal1977; Hovland and Weiss 1951; McGinnies 1973) and in the area of minorityinfluence has identified credibility as a critical determinant of the persuasion (Clark andMaass 1988b; Moscovici 1980). Another source-related variable, liking or feelingstowards the source, is also known to affect the effectiveness of a persuasivecommunication (Roskos-Ewoldsen and Fazio 1992). This research examines themediating role of both these variables, source credibility and source-related feelings,in determining the extent of minority or majority influence.Finally, individual difference variables such as one's susceptibility to socialinfluence (Bearden, Netemeyer and Teel 1989) or Attention to Social ComparisonInformation (Bearden and Rose 1990) as well as degree of risk-aversion will also playa role in whether the individual accepts the group norm or chooses to dissent andeven persuade other group members. Individual's risk-taking ability is known toimpact the nature of decisions (Fagley and Miller 1990). In many group-oriented7decision situations, dissent or deviance from the group norm may have attached riskssuch as the majority excluding dissenting members from the group and the individuallosing self-esteem due to social rejection. In this research risk-aversion or otherpersonality factors do not play a central role. However, risk-aversion is used as acovariate in the analysis of the experimental data.1.2.1 Summary of Proposed ResearchIt is posited that group influence includes not only conformity, but also minorityopinion influence. The extent to which minority or majority opinion is influential in agroup is likely to be determined by similarity between source and recipient, as well asissue-involvement and response-involvement of the recipient. Further, the perceivedcredibility of the source and the feelings towards the source will have an impact onthe relative influences of the majority and the minority.1 .3 The Study Initially several pretests and a pilot study were conducted to gain further insightinto the minority influence paradigm. Consistent with previous research in this area,an laboratory experiment was adopted as the methodology. The main experimentinvolved the manipulation of two variables - similarity of the majority/minority andissue involvement. Each subject was thus in one of the four cells - 2 (similar-minorityor dissimilar minority) X 2 (high versus low issue-involvement).8Previously uncommitted individuals received a stimulus which contained theopinions of both the minority and the majority on a particular issue, with the minorityand the majority taking contrary positions on the issue. When the minority was"similar" to the subject, the majority was "dissimilar" and vice-versa. Thus, minority-majority was a within-subjects manipulation.After subjects read the stimulus material, which was presented as the text ofa group discussion, they responded to a series of dependent measures which includedattitude toward issue discussed by the two sources, source-credibility of similar anddissimilar minorities, feelings towards the source, willingness to take risk, confidencein decision etc.In the experimental design the group pressure was created through a "nominal"group method, where subjects did not directly come in contact with the groupmembers. While further details of the design are presented in a later chapter, itshould be noted here that the methodology used was consistent with the designsused in this area by other scholars (Maass and Clark 1983; Martin 1988; Mugny andPapastamou 1980).1.4 Potential Contribution of Research This research was motivated by the need to bridge the apparent knowledge gapin the marketing literature in the area of social influence. As elaborated earlier, theview of social influence as occurring in only one direction - from the majority in the9group to the individual or minority in the group - is very limiting. Sound theoreticalreasoning and empirical evidence is available for the existence of minority opinioninfluence in groups. This research heeds the call of other marketing scholars (e.g.,Folkes and Kiesler 1991) and attempts to incorporate minority influence in anextended model of consumer social influence, thus making a potentially significantcontribution.This research endeavours to make a contribution to the minority-majorityinfluence literature by examining how issue-involvement affects the degree ofconformity or minority influence. Even though issue-involvement has been recognizedas an important variable in the persuasion literature (see extensive review by Johnsonand Eagly 1989), its role in determining conformity versus minority influence remainsunknown, but for one exception (Trost, Maass and Kenrick 1992).This research also hopes to enhance our understanding of the decision processin group contexts. With a exception of few scholars (such as Bearden and hiscolleagues), very little attention is being paid to group decision process in themarketing literature Many, if not most, consumer decisions are made in groupcontexts where there are direct or indirect influences by reference others. Purchasedecisions by families, informal groups of friends as well by organizations areexamples. This research attempts to provide further insight into the social influenceprocesses affecting consumer behaviour.101 .5 Organization of Thesis The rest of this thesis is organized as follows. Chapter Two provides the backgroundliterature drawn from social psychology and examines the differences betweenconformity and minority influence perspectives. In Chapter Three a literature reviewis undertaken leading to the delineation of the conditions facilitating minority ormajority influence. Specific hypotheses are proposed. The details of two pretests anda pilot study are contained in Chapter Four. The research methodology used in themain experiment is presented in Chapter Five. In Chapter Six a discussion of theresults of the main experiment is provided. And finally, Chapter Seven incorporatesan overview of the key findings, a general discussion of the results and theimplications of the research, limitations of the research, as well as directions for futureresearch.11II. BACKGROUND2.0 OverviewThis chapter includes a discussion of the conformity paradigm - the predominantview of social influence in consumer behaviour (see Figure 1), as well as the minorityinfluence paradigm (see Figures 2 and 3). While discussing each paradigm, first thegeneral theory from social psychology and other allied disciplines is presented,followed by a discussion of the relevant marketing literature. Alternative theoreticalperspectives on minority influence effect are discussed. The objectives of this chapterare (i) to explore the theory behind majority influence (i.e., conformity) as well asminority influence, and (ii) to further highlight the knowledge gap in marketing andemphasize the need for current research.2.1 The Conformity ParadigmThe Conformity Paradigm examines the "change in an individual's behaviour orattitudes towards those advocated by a group as result of real or imagined grouppressure" (Davis 1984). Asch (1951) studied the social and personal conditions thatinduce an individual to resist or yield to group pressure. In a typical experiment underAsch's paradigm, the majority (consisting of confederates) would repeatedly disagreewith an individual on a simple task such as judging the length of a series of lines.12Over repeated trials, many individuals were found to abandon their position and jointhe majority, even when the latter's opinion was contrary to fact. However, asignificant number of individuals did resist the majority influence and displayedindependence in their judgements (Asch 1951). More recent research indicates thatthere might be different mechanisms operating to produce conformity effect,depending on the nature or extremity of the norm. When the norm is clearly incorrect,it has been found that conformity occurs through a normative influence mechanism,and when the norm represents a factually correct position, conformity seems to occuras a result of an informational influence by the majority (Campbell and Fairey 1989).Based on the works of several social psychologists, including Asch, we know that thedegree of majority influence or conformity may be dependent on many factors suchas: (1) The character of the stimulus situation - where diminishing clarity of thestimulus condition increases the majority influence; (2) The character of group forces- where unanimity and larger (versus smaller) majority size lead to greater majorityinfluence; and (3) The character of the individual - including degree of confidence inone's judgement; susceptibility to social influence etc.Conformity is known to affect a wide variety of behavioral situations andoutcomes. Impact of conformity on many issues such as organizational performance(McGill 1990), alcoholism (Savoni 1989), sexual satisfaction (Wilson and Reading1989), and eating disorders (O'Brien and Bankston 1984) have been investigated.Pertinent to the present research are the studies which have established conformityeffects within consumer decision making contexts (e.g., Davis 1984; Hansen 1969;13Rose, Bearden and Teel 1992; Venkatesan 1966).2.1.1 Conformity Effect In Marketing LiteratureA study by Venkatesan (1966) was the first to test conformity effects in aconsumer setting. The differential effects of compliance (Kelman 1961) versusreactance (Brehm 1966) were examined in a situation where subjects had to selectthe "best" suit from identical ones under different forms of group pressure. The studydid find strong conformity effect in the absence of any objective standard. However,in the reactance condition, when the subjects perceived that the group pressure wasleading to the restriction of choice, subjects seemed indifferent or deliberately chosean alternative that would negate the group pressure. Thus, it seems individualssuccumb to group pressure when the pressure is somewhat subtle, but not when itis seen as a threat to their freedom of choice or action.A similar consumer judgemental problem was investigated by Davis (1984) -conformity on judgements of fashionability by women. The study involved two levelsof judgement ambiguity - low (i.e., present fashions) and high (i.e., future fashions).The subjects ranked six women's suits and then read an essay on fashions (whichincorporated opinions of other individuals that were discrepant from the subject'sopinions) and then subjects were asked to rank the garments a second time to detectany conformity effect. The evidence was suggestive of greater conformity underhigher ambiguity of the judgement task. Subsequent studies of group influence inconsumer behaviour have at least indirectly suggested conformity effects.14More recently, Rose, Bearden and Teel (1992) have further demonstratedconformity influence for illicit alcohol and drug consumption. The main contributionof this research is the integration of the attributional analysis with the group pressurephenomenon. These authors, based on earlier work by Ross, Gunter and Hoffman(1976), proposed that conformity effects, such as evidenced in Asch's studies, couldbe understood by undertaking an attributional analysis. When individuals face astrong majority opinion stacked against them, the attributional problem faced by theindividual can be stated as follows: (i) why are other group members expressing thesejudgements or performing these behaviours, (ii) what would my dissent from thegroup norm imply about me and my perception of the group, and (iii) what would mydissent imply to me about myself (i.e., self-perception).In studies conducted using high school students, these authors found that suchattributional thinking played a critical role in whether or not individual conformed togroup pressure. When individuals were able to provide an external explanation for agroup's behaviour it provided a mechanism for reducing conformity pressure, whileinternal attributions (where the locus of causality is with the actors) led to greaterconformity pressure. This work takes us a step further in understanding conformityeffects within groups as well as explaining the underlying mechanism behind deviancefrom group norm.There is also documented evidence that conformity effect might very muchdepend on individual personality characteristics (Bearden and Rose 1990). Relatedvariables such as Self-consciousness (Davis 1984), Susceptibility to Social Influence15(Bearden et al. 1989) and Attention to Social Comparison Information (Bearden andRose 1990) play an important role in determining the extent of conformity. Individualsscoring high on these personality related measures have been found to be more likelyto yield to group pressure. Since marketers are interested in grouping similarconsumers (for market segmentation and other purposes) attention to such individualdifference factors seems to be a fruitful line of investigation.2.2 The Minority Influence ParadigmMoscovici (1976) has challenged the "functionalist" perspective of theconformity literature, which assumes the capitulation of the minorities to the majorityand ignores the role of the minority as agents of social change or innovation.Moscovici and his colleagues have demonstrated that in group decisions minorities canindeed be persuasive (Moscovici 1976; Moscovici, Lage and Naffrechoux 1969;Moscovici and Personnaz 1980).The earliest evidence supporting such a proposition came from Moscovici, Lageand Naffrechoux (1969), who demonstrated that a consistent minority can exertinfluence over the majority. In an experiment using Asch's paradigm, these authorsasked their subjects to perform a simple colour perception task. The fact that allsubjects had full visual capacity was made known to all participants. Then, thesubjects were asked to judge the colour of "blue" slides that varied only in theirluminance. When the two confederates consistently labelled the slides as "green",168.42% of all responses were green and 32% of subjects reported having seen "green"at least once (even though all slides were in fact "blue"). However, when theconfederates were inconsistent they did not influence the rest of the group. Eventhough these results are quite modest, they generated a lot of interest which has ledto the formulation of minority influence theories (Moscovici 1976), more formalmodels (Latane 1981; Tanford and Penrod 1984) and a plethora of research activity(see extensive review by Maass and Clark 1984).An important determinant of minority influence is considered to be its behavioralstyle, which includes behavioural consistency, autonomy, fairness, rigidity andinvestment (Moscovici 1976). Among these different aspects of behavioural style,consistency has been the most commonly studied (see Maass and Clark 1984). It hasbeen suggested that a consistent behaviour by the minority (which includesunflinching maintenance of one's position in the face of conformity pressure) wouldlead other group members to make favourable attributions regarding the minority (i.e.,the minority is certain and confident in its position), making the minority influential.On the other hand, an inconsistent minority attracts less favourable reaction, therebydiminishing minority influence.Minorities have been found to be persuasive in many contexts. Initially,minority influence was established in experimental studies involving colour perceptions(e.g., Moscovici, Lage and Naffrechoux 1969; Moscovici and Faucheux 1972). Later,the minority influence paradigm was applied to situations involving more complexsocial judgements. These studies offer further evidence for the influence of minority17opinions in contexts such as jury decisions (Nemeth and Wachtler 1973), and groupdiscussions of a variety of social issues such as air pollution (Mugny and Papastomou1980), gay rights (Maass and Clark 1983), feminism (Paicheler 1976), andmilitarization (Mugny 1975). In examining minority influence in these diverse socialtopics, researchers have used subjects from different age groups such as teenagers(Mugny 1975) and adults (Moscovici et al. 1969).2.2.1 Minority Influence and MarketingFolkes and Kiesler (1991), in a review paper on social cognition research inconsumer behaviour, briefly alluded to role that minority opinions may play inconsumer group settings. However, even these authors did not elaborate on thissubject at any length. They speculated, on the basis of Nemeth's work (Nemeth1988), that the presence of minority opinions within groups may lead to moreinnovative decisions. To date in the marketing literature there has not been anyempirical or theoretical examination of this issue or other issues surrounding theexistence of minority opinions in group settings. Thus, the void in the marketingliterature is obvious. There is indeed a need for more theory and empiricalexamination of the minority influence paradigm in a marketing context.2.3 Perspectives on Majority-Minority InfluenceIn the social psychology literature, two major theoretical views exist as explanations18of the minority influence phenomenon. These two views are discussed briefly here.2.3.1 Theory of Conversion BehaviourMoscovici (1980) proposed the Theory of Conversion Behaviour (TCB) whichargues that the underlying processes leading to conformity or minority influence areessentially different. Briefly, when an individual faces a strong majority that disagreeswith him/her, a comparison process is triggered, where the individual is motivated toreduce the disagreement. Individuals operating under such a motive do not criticallyexamine the validity of the position taken by the majority or challenge the majority.This interpretation is consistent with the findings of studies using the Asch (1951)paradigm (such as Bearden and Rose 1990; Venkatesan 1966), where the subjectsoften accept a blatantly incorrect position expressed by a strong majority. On theother hand, if a minority presents a "different" view in a group, one feels free tochallenge this view without any fear of social exclusion or condemnation. Hence,messages presented by the minority lead to a validation process, where argumentsand counter-arguments are raised. The comparison process is similar to the peripheralprocess and the validation process is similar to the central process in Petty andCacioppo's (1986) Elaboration Likelihood Model (ELM). Just as in the case of theperipheral process, the comparison process also involves reduced attention to themessage and increased attention to other cues such as the source. The validationprocess, on the other hand, parallels the central process because it involves criticalprocessing of the message.19A more detailed comparison between the TCB and the ELM is presented in the nextchapter.Tests of Moscovici's theory have shown that those conforming to the majorityoften do so only in public and often go back to their own positions privately. On theother hand, those accepting a minority position seem to exhibit internal attitudechange (see Maass and Clark 1984). Thus, majority influence is often limited to onlya public compliance, whereas minority influence tends to be a private conversion.According to Moscovici, it is the "behavioural style" of the minority, throughthe maintenance of a resolutely nonconformist position, that leads to a favourableimpression about the minority. In addition to this characteristic of consistency, otherattributes like competence, certainty, autonomy and conviction are also important inenhancing the minority's influence (Moscovici and Nemeth 1974).Similar to the attributional analysis used by Rose et al. (1992) to explain aconformity effect, Moscovici and his colleagues too have used attribution theory toexplain the minority influence effect (see Moscovici and Nemeth 1974). Accordingto this explanation, when a minority consistently disagrees with the majority over time(behavioural style), such behaviour will produce a person-attribution (i.e., the causeof disagreement will be attributed some characteristic of the minority). When aminority within a group maintains its position with consistency, in the face of strongmajority pressure, a negative or a positive person-attribution could be made. Theminority may be seen as obdurate or uncooperative or not team-oriented, and if suchan attribution is made then the minority is unlikely to be effective. On the other hand,20if the minority's consistent deviation from the majority opinion is attributed tocharacteristics such as courage, confidence or expertise in the opinion issue, thensuch a minority will be influential.Nemeth, Swedlund and Kanki (1974) found that the perception of minoritybeing consistent and possessing knowledge in the subject matter (a positive person-attribution) occurred not merely due to repeated disagreement of the minority with themajority, but because the minority articulated a well-defined position. Further, studiesshow that as the numerical strength the minority increases, the perception of theminority's competence became more favourable (Nemeth, Wachtler and Endicott1977). When the minority consists of more than one individual it is difficult to makea negative person-attribution (such as the minority is crazy or dogmatic) and theminority's opinions are closely examined leading to greater minority influence (cf.Moscovici 1976).Other researchers have raised objections to Moscovici's attributional accountof behavioural consistency. Maass and Clark (1984) raise the question of why shouldminority's consistency lead to a positive attributions like certainty or confidence ratherthan negative attributions such as "craziness or dogmatism". Chaiken and Stangor(1987) raise more serious concerns regarding the attributional account by pointing outthat the central task of the perceiver in Kelly's (1972) theory is not to infer thecommunicator's (i.e., minority's) dispositions (such as certain, confident), but to inferthe causes for the communicator's message. This questions the validity of theattributional analysis. Chaiken and Stangor (1987) pointed out that in the attribution21and persuasion literature, inferences about the communicator's dispositions have beenfound to have, at best, an indirect relation to persuasion. The lack of process-orientedmethodologies and the lack of attention to motivational variables in previous researchlimits further understanding of how exactly the consistency of the minority contributesto its increased persuasiveness (for a more complete discussion of this issue, seeChaiken and Stangor 1987).Further challenge to the TCB comes from social impact theory (Latane 1981;Latane and Wolf 1981) as well as from the social influence model by Tanford andPenrod (1984). These researchers treat both minority and majority influence as partof the same process. The social impact theory, for instance, views the degree ofsocial influence on individuals as a multiplicative function of the number of sources,their strength (e.g., expertise) and their immediacy (e.g., proximity). In other words,holding other factors constant, the difference in the minority versus majority influenceis attributable to differences in their numerical strength alone. Latane and Wolf(1981) used their mathematical model (which assumes a negatively acceleratingpower function relating group size to observed influence) to reinterpret previouslypublished studies in both conformity as well as minority influence and found that theirmodel explained a substantial amount of variance in both cases.Chaiken and Stangor (1987), in a review of the literature, conclude that thereis a distinct possibility that more than one process may operate for both majority andminority influence conditions, depending on the individual's motives. If motives ofindividuals facing the majority versus the minority are similar (e.g., seeking correct22information), then Chaiken and Stangor (1987) think that a qualitative difference inthe processes underlying majority-minority influence is meaningless. But such adifference may be operative when individuals facing the majority versus the minorityhave different motives (e.g., seeking social approval versus seeking a valid opinion).Empirical results are mixed at this point. Some researchers have found essentially nodifference in the amount of cognitive processing undertaken in the majority andminority conditions, but have found qualitative differences in the nature of thoughts(e.g., Maass and Clark 1983; Trost, Maass and Kenrick 1992). On the other hand,as observed earlier, quantitative models of social influence (e.g., Latane 1981) seemto account for the observed influence in both conditions without resorting to a dualinfluence perspective, thus providing a more parsimonious approach (Latane and Wolf1981) .2. 3. 2 Model of Idiosyncrasy CreditsThe crux of Hollander's model (1958, 1964) is based on the assumption thata deviant individual, a minority within a group, can gain acceptance from other groupsmembers only if this individual attained a sufficiently high status within the group.According to Hollander (1958, 1964), initially all members of a group conform togroup norms. By exhibiting competence, an individual is able to build up "idiosyncrasycredits" or favourable impressions held by others in the group, and then such anindividual will be permitted not to conform, to innovate and to even exert influence.Thus, according to Hollander's model, no ordinary member of a group can exert23influence as a minority within a group.Hollander's work has shown that factors such as perceived competence of anindividual as well as length of membership in a group contribute to the individualbecoming more influential, if he/she chooses to deviate from the group norms(Hollander 1961). Further, Hollander held that early nonconformity, when theindividual has not accumulated sufficient idiosyncrasy credits, will make that individualless influential as compared to when such deviance occurs when the individual hasa favourable reputation in the group.2.3.3 Commonality Between the Two PerspectivesMoscovici's model suggests that a minority's influence is enhanced byconsistent and staunch opposition to the majority right from the outset, whereasHollander's model argues that the minority will be effective only by conforming initiallyand showing competence in the subject matter, before deviating from the group norm.Thus, both models use nonconformity as a means of achieving influence is groupcontexts, however they differ in terms of when such conformity should occur in thegroup decision process (see Bray, Johnson and Chilstrom 1982).The common theme between the two models is that the minority should beperceived in a positive light (positive person-attribution) in order for the minority to besuccessful. Thus, it seems the disagreement between the two models is not assignificant as it might first appear.242.4 SummaryThe findings within psychology as well as the marketing literature have providedextensive support for the conformity effect under a variety of settings using variedsample of subjects, pointing to the robustness of the conformity effect. In themarketing literature, it is apparent that there has been no direct attempt to incorporatethe role of minority opinion influence, although studies have examined and providedexplanations for why individuals or an individual may deviate from group norms (e.g.,Bearden et al. 1989; Rose et al. 1992). This does underscore the need for extendingthe conceptualization of social influence in consumer/marketing settings. While theprocess underlying the minority influence is still being debated (see subsection 2.3.1),as Hollander's theory suggests, knowledge about the minority's ability or competencecould be important in determining its influence. As detailed in the next chapter,source credibility, an attributional variable, is explicitly incorporated in this research.25III. LITERATURE REVIEW AND HYPOTHESES3.0 Overview This chapter provides a discussion of the proposed model and the relatedtheory. An attempt is made to integrate theory from psychology as well as marketingand a simultaneous social influence model that considers both the conformity as wellas minority influence effects within a single framework is presented (see Figure 3).Similarity (based on social identity), issue-involvement and response-involvement areproposed as the critical variables determining the nature of the social influence. Inaddition, the role of other intervening variables in the social influence process is alsoidentified. Based on this theory and related literature review a set of testablehypotheses is derived.3.1 Multiple Sources Model Previously, marketing scholars have paid little attention to how individualsintegrate information from multiple sources. While there are exceptions to thisstatement (e.g., Moore and Reardon 1987), most studies in consumer researchexamine information processing or persuasion using a single source, such as anadvertisement. In reality, it is not at all uncommon for consumers to be exposed tocongruent or incongruent information from many different sources. For instance, a26consumer may hear the opinions of many sources on a product. Sometimes themultiple sources may present the same opinion leading to a unanimous view,but often minority opinions exist, and may be even more commonplace thanunanimity. Within consumer groups like families or other informal social groups, suchdivergent views can exist in group members' reactions to a movie, tastes in food orfashions, experiences with car dealers or mechanics etc. Thus, a consumer seekingthe opinion of his/her reference group members is more often than not likely to findmultiple opinions.There has been some research on how individuals integrate information frommultiple sources (e.g., Harkins and Petty 1987; Moore and Reardon 1987). Thesestudies have usually manipulated the number of sources and argument quality (as inMoore and Reardon 1987) or have also varied the similarity of the multiple sources (asin Harkins and Petty 1987). In a marketing application, Moore and Reardon (1987)created a single source versus multiple source ad by presenting just one versus foursatisfied customer/s in the ad for a consumer product.In Harkins and Petty (1987) as well in the marketing study by Moore andReardon (1987) found that multiple sources, as opposed to a single source, triggergreater cognitive processing and this in turn mediates persuasion. It seems that whenconsumers perceive the multiple sources as being independent and representingdivergent perspectives, the motivation to think about the issue is enhanced. In thecase of information from multiple sources, the perception that the opinions representmore than one person's knowledge may lead to a more diligent examination of such27information.It is important to note that in these studies, the multiple sources represent thesame side of the issue, for instance, in the Moore and Reardon study all fourconsumers extolled the virtues of the same product. This is akin to creating amajority influence situation or conformity pressure. The individual facing multiplesources gets different arguments from different sources, all in favour of or opposedto an issue.The present research can be viewed as a study of multiple sources (see Figure3). However, it differs from previous research in one important respect. In thisresearch, while individuals receive information from multiple sources (i.e., differentmembers of a group discussion), the opinions expressed by the sources are notunanimous and two different opinions are expressed each with a different degree ofsocial support (i.e., a majority or a minority opinion).In a consumer context, one does not merely acquire information from the socialenvironment, but one may also try to evaluate the popularity (social support) of eachalternative. A consumer may also encounter influence attempts by different sourceswhose positions constitute a majority or a minority. This research examines theprocessing of information in such a context. Thus, the previous work on multiplesources in extended by this research.In this research, as outlined earlier, the objective is to examine the conditionsunder which minority and majority influences occur. A previously uncommittedindividual is given two opinions on an issue, one representing the majority view and28the other a minority view. At this stage the variables that determine the nature andextent of social influence in such a multiple source context are examined.3.2 Source-Recipient Similarity The social comparison theory (Festinger 1954) states that individuals will preferto compare with "similar" others when they wish to verify the validity of theiropinions. In consumption decisions, which involve one's tastes or values, knowingthe preferences of similar others is more relevant than knowing the preferences ofdissimilar others (cf. Goethals and Darley 1977). Furthermore, similarity between thesource of communication and the recipient of the message is known to increase thepersuasiveness of the message (e.g., Berscheid 1966).Evidence regarding the persuasive power of a similar source can be obtainedfrom the diffusion of innovations literature as well. Studies in rural sociology (seeRogers 1983) lend credence to the notion that similarity or homophily between aninnovator and a non-innovator leads to greater influence by the innovator. Homophilyis the extent to which two individuals are similar in terms of age, sex, education,social status etc. (Rogers 1983). In a consumer behaviour study, Brown and Reingen(1987) operationalized homophily through occupation, age, sex and education.Research in the area of advertising has also incorporated source similarity, and theevidence generally supports the contention that a similar-source is more persuasivethan a dissimilar-source (e.g., Feick and Higie 1992).29Research in the area of minority influence also indicates support for the greaterpersuasiveness of a similar, as opposed to dissimilar, source (e.g., Clark and Maass1988a, 1988b; Martin 1988a). In these studies, frequently social categorization(Tajfel and Turner 1979) is used to create perceptions of similarity or dissimilarity bydenoting the source as a either a member of an ingroup or an outgroup. Variablesused to create perceptions of similarity or dissimilarity include gender (a.g., Martin1988b; Perez and Mugny 1987), school affiliation (e.g., Clark and Maass 1988a;Martin 1988a), sexual-orientation (Clark and Maass 1988b) among others.'Much of the research incorporating "similarity" in minority influence researchis based on the Social Identification Theory. This theory offers further insight intowhy a similar source may be more persuasive.3.2.1 Social Identification and Social CategorizationTajfel and his colleagues (Tajfel, Flament, Billing and Bundy 1971; Tajfel andTurner 1979; Turner 1982) showed that even a "minimal group" situation, wheresubjects are categorized for ad hoc or administrative reasons, was sufficient to createingroup favouritism and discrimination of outgroup subjects. This mere categorizationeffect led Tajfel and his colleagues to suggest that individuals over-evaluate theingroup and under-evaluate the outgroup. Tajfel and Turner (1979) argued thatindividuals are motivated to enhance their self-esteem through the acquisition of a1 In this research, the terms similarity and dissimilarity mean the same as the termsingroup and outgroup respectively.30"positive social identity", where social identity is defined as an "individual'sknowledge of his membership of certain social groups and the emotional andevaluative meaning resulting from their membership" (Tajfel 1972; see also Martin1988a).Individuals achieve a positive social identity by engaging in self-favouring socialcomparisons so that one's own group is perceived more favourably than the othergroup. Such self-favouring social comparison is often accomplished by attributingnegative characteristics to the outgroup.In a later work Tajfel (1982, Ch.16) commented that there is a distinctionbetween a social group and a social category. An example of the latter would be"all people using Brand A toothpaste", and members of this category are highlyunlikely to become a group. The social category in the above example would turn intoa social group if it was known that Brand A contained a dangerous side-effect, whichwould bring all users of Brand A together in a common cause (cf. Rabbie and Horwitz1988). The common fate will bind all users of Brand A together and they may nowpursue common actions (e.g., a class-action suit), engage in direct or indirectconversations and may, thus, become interdependent. Rabbie and Horwitz (1988)have argued persuasively that such interdependence between individuals is necessaryfor a "social group" and it is such interdependence that causes ingroup favouritismand outgroup discrimination to occur.Interpreting Tajfel's work, Mugny (1982) suggested that individuals attributestereotypical characteristics to members of different social groups. If an individual is31influenced, then he/she will not only have adopted the position advocated by thesource, but may also assume the stereotypical characteristics attributed to that source(which arises from that source's group membership). This will involve self-attributionof stereotypical characteristics associated with the influence source's social group.It is apparent that there is a "psychological cost" involved in redefining one'ssocial identity. The psychological cost may depend on: (i) strength of identificationwith one's own group, and (ii) desirability of characteristics of the influence source'sgroup (which could be an ingroup or an outgroup). Thus, if an individual is influencedby an outgroup (which is usually attributed negative characteristics), then this wouldrequire the individual to make greater changes in social identity than when theindividual is influenced by members of the ingroup (which is usually attributed positivecharacteristics).Explicating further on the social categorization analysis of social influence,Turner (1987) provided insight into why individuals may be more susceptible topersuasive attempts from a similar, rather than a dissimilar, source. Individuals willgenerally expect "similar others" to exhibit similar behaviours when exposed toidentical situations. Thus, it would make sense to anticipate agreement with "similarothers" on most cases. On the other hand, one would not normally expect to agreewith a member of an outgroup because of the a priori knowledge that such a personis dissimilar. According to Turner (1987), disagreement with dissimilar sources oroutgroups can easily be discounted on the basis of the differences in thecharacteristics between the recipient and the source. However, disagreement with32a similar source is more difficult to reconcile. The uncertainty caused by suchdisagreement (with a similar source) is often reduced by changing one's attitudetoward the disagreeing similar source.3.2.2 Evidence Regarding Social Categorization EffectWithin the area of minority-majority influence research, the results of studiesusing social categories or groups are generally consistent. These studies usuallycompare an in/outgroup minority with an out/ingroup majority or compare aningroup minority with an outgroup minority. The dependent measures include sourcecredibility, source image, as well as direct and indirect effects on attitude.Clark and Maass (1988b) manipulated the minority's status (either ingroup oroutgroup) and kept the majority's status constant (ingroup). On the focal topic of"abortion", the minority seemed to be more persuasive than the majority, but only inprivate measures not in public measures. This result can be attributed to the higher"conformity pressure" felt when having to respond publicly.Perez and Mugny (1987) found a more marked indirect persuasive effect whenan ingroup minority was matched against an outgroup majority. However, anoutgroup minority facing an outgroup majority exerted the greatest direct influence onattitude because in this situation the "psychological cost" of switching one's socialidentity is not an issue (as both sources belong to the outgroup and the individual mayfeel no direct involvement).Finally, when an ingroup minority is compared against an outgroup minority, the33former seems to be more persuasive on public measures, while there was nodifference on private measures (Martin 1988a). This is again consistent with theexplanation that in the case of public measures one would not risk altering one'ssocial identity by taking the siding with the outgroup.In summary, the cost of redefining one's social identity (when one agrees witha dissimilar person) and the a priori knowledge that a similar source is likely to possesssimilar views (or values or tastes) may account for the greater persuasiveness of asimilar source. In addition to these factors a similar source is also likely to be viewedas more credible. The issues relating to source credibility are considered next.3.2.3 Source Credibility and Source SimilarityThe social categorization framework discussed above would suggest that anoutgroup or a dissimilar source is seen as possessing less desirable characteristics andan ingroup is seen as possessing more desirable characteristics (cf. Mugny 1982).This leads to the inference that a dissimilar source would be perceived as being lesscredible than a similar source.Both Mugny's and Turner's perspectives lead to the conclusion that theingroup, which is likely to be seen as more credible than the outgroup, will be morepersuasive. The greater influence of a highly credible source is supported by previousresearch (e.g., Kelman and Hovland 1953). These studies examined the main effectof source credibility and the general consensus was that a highly credible source ismore persuasive than a less credible source.34Other researchers who have investigated the interaction of source credibilitywith other variables have often come to different conclusion (e.g., Dholakia andSternthal 1977; Sternthal, Dholakia and Leavitt 1978). Sternthal et al. (1978) haveshown that a low credible source might be more persuasive than a high crediblesource when the message is pro-attitudinal, and not counter-attitudinal. A crediblesource is also known to suppress the possibility of counter-argumentation, andminimization of negative thoughts increases the persuasiveness of the highly crediblesource (Dholakia, Sternthal and Leavitt 1978).Studies in majority-minority influence also take the position that less crediblesources will have less persuasive influences. In these studies, the majority is oftenportrayed as the credible source and the minority as lacking in credibility (Moscovici1980, p.214). Moscovici's Theory of Conversion Behaviour suggests that whenfacing a strong majority with an opposing viewpoint, the individual engages in asocial comparison process, which calls for little information processing. On the otherhand a minority's viewpoints will be criticized without fear, and in the process anexamination of issues will take place, which is known as the validation process(Moscovici 1980). Thus, a minority source facilitates greater cognitive processing.Also, there is evidence that a minority may have a more indirect or latent influencethan a majority (e.g., Perez and Mugny 1987; Moscovici and Lage 1976).Taking the notion of source credibility effects further, Clark and Maass (1988b)have argued that if ingroup members are more credible than outgroup members, thena minority from an ingroup (similar source) should be more credible than a minority35from an outgroup (dissimilar source). Since the outgroup minority is less credible thanthe ingroup minority, Clark and Maass argue that the ingroup minority is likely to bemore persuasive in direct or public measures (where psychological costs of agreeingwith an outgroup minority may be a consideration). If ingroup minorities have morecredibility than outgroup minorities, then the outgroup minority (the less credible ofthe two) is more likely to generate greater cognitive activity and may stand a betterchance of being persuasive in indirect or private measures (where social identity is notthreatened).When both the majority and the minority came from the ingroup, the resultsregarding the difference in perceived source credibility is mixed. While some studiessuggest a higher credibility for the majority, others suggest no difference in credibilitybetween the two sources. However, a minority from an ingroup is not only seen asmore credible than an outgroup minority, but is also more persuasive (Clark and Maass1988a).3.2.4 Source Related Feelings and PersuasionIn recent years there has been an accumulation of evidence pointing to source-likability as an important determinant of a source's persuasiveness (e.g., Hamilton,Hunter and Burgoon 1990; Roskos-Ewoldsen and Fazio 1992, Wood and Kallgren1988). The likability of the source is a seen as a dimension that is distinct fromsource credibility or expertise (e.g., Wood and Kallgren 1988).Source credibility is often seen as a function of a source's expertise and36trustworthiness (cf. Dholakia and Sternthal 1977). Evaluating a source's credibilityis likely a cognitive or thought-oriented process. On the other hand, source likabilityis likely a result of an affective or feeling-oriented process. While the two may notbe independent dimensions, they are distinct from each other.In the context of minority influence research, source credibility or source imagehas been examined in previous research (e.g., Clark and Maass 1988b; Perez andMugny 1987), but not source-related feelings. It follows from the social identitytheory, which argues that stereotypically positive (negative) characteristics areassociated with a similar (dissimilar) source, that a similar source will lead to morepositive source related feelings than a dissimilar source. The research on persuasionwould suggest that such feelings should have a direct or indirect effect on thepersuasiveness of that source.3.2.5 Source -Similarity HypothesesMAIN EFFECT OF SIMILARITY: A minority represents a deviant opinion or theopinion with less social support in the group. Accepting such an opinion has risksattached to it. However, the probability of a such an opinion being accepted shouldbe enhanced when the minority is made up of ingroup and the majority is made up ofoutgroup members.Hypothesis 1: The minority will be more influential when it is similarto the subject, than when it is dissimilar.37It should be noted that while this hypothesis is not stated in terms of a directcomparison between the majority and minority source, it is nevertheless implied.Since each subject is exposed to both the minority and the majority sources, and sinceboth sources advocate contrary positions, greater influence of one source meansreduced influence of the other. Thus the majority will be less (more) influential whenthe minority is similar (dissimilar).SOURCE CREDIBILITY AND FEELINGS: First, the perceived credibility of thesource is a function of the perceived similarity of the source. In other words, themore similar a source, the more likely it will be perceived as very credible. Second,the perceived credibility of the minority is likely to have a mediating effect on theinfluential power of the minority.Hypothesis 2: A similar minority will be perceived as being more crediblethan a dissimilar minorityHypothesis 2a: Higher (lower) perceived credibility of the minoritywill be associated with greater (lesser) minority influence.A similar minority can be expected to generate more favourable feelings thana dissimilar minority based on the earlier discussion. Furthermore, the feelingstowards a minority source is likely to have an effect on the extent of its influence.Hypothesis 3: A similar minority will be generate more positivesource-related feelings than a dissimilar minority.Hypothesis 3a: Greater positive (negative) feelings towards theminority will be associated with greater (lesser) minorityinfluence.383.3 Issue-InvolvementPrevious research in minority influence has generally used topics with littlepersonal relevance to the participants of the study. The type of stimuli studied rangefrom simple perceptual ones (e.g., colour slides, geometric figures) to at bestmoderately involving political/social topics (e.g., abortion, death penalty, gay rights).Personal involvement has been found to be a critical variable in the persuasionliterature (e.g., Johnson and Eagly 1989; Petty and Cacioppo 1979, 1986; Petty,Cacioppo and Schumann 1983). Consumer researchers have also explored the roleof involvement and have generally found that the involvement level in a productmoderates the persuasiveness of the message such as advertising (e.g., Gardner,Mitchell and Russo 1985; Kardes 1988; Krugman 1965; Petty, Cacioppo andSchumann 1983). 2In studies of minority influence, which fall under the rubric of persuasionresearch, little attention has been paid to the role of personal relevance orinvolvement. It is worth noting that scholars reviewing the research in this area havecalled for a systematic incorporation of issue-involvement into minority influence2 In this research a distinction is made between issue-involvement (Petty andCacioppo 1979, 1986) and ego-involvement (Sherif and Hovland 1961). In the caseof issue-involvement, the personal relevance is based on a close tie between theattitude issue (or object) and a self-defining reference group. In the present context,an individual with no prior attitude towards an issue is expected to processinformation relating to that issue if the issue under consideration has implications forthe individual's future (see Leippe and Eskin 1987). According to the typology ofpersonal involvement provided by Johnson and Eagly (1989), the involvement in thisresearch context can be defined as outcome-relevant involvement. In this research,the terms issue-involvement, personal relevance and outcome-relevant involvementmean the same thing.39research (Chaiken and Stangor 1984). A very interesting question that has beenunexplored is whether conformity (majority influence) or minority influence is somehow dependent on an individual's level of involvement in the issue underconsideration.3.3.1 Elaboration Likelihood Model and Minority InfluenceThe basic tenet of Petty and Cacioppo's (1986) Elaboration Likelihood Model(ELM) is that the effectiveness of different methods of persuasion depend on"whether the elaboration likelihood of the communication issue (i.e., probability ofmessage-relevant or issue-relevant thoughts occurring) is high or low" (Petty,Cacioppo and Schumman 1983). According to the ELM, when an product or an issuebecomes personally relevant to an individual, it becomes more important to form areasoned opinion. When there is little personal relevance, individuals are lessconcerned about the accuracy of their attitudes.Under increased involvement the central route to persuasion is more likely tobe activated, and a more careful consideration of the true merits of the product or theissue takes place. Attitude formation under high involvement is a function of suchinformation processing activity. Under low involvement, however, individuals simplylook for simple acceptance or rejection cues in the message context and quality of thearguments are not carefully considered. The reduced personal relevance of theproduct or the issue is the cause of reduced attention to the message quality andattitude formation under low involvement more likely takes the peripheral route.40In a group context a simple peripheral cue may be the degree of supportavailable for a certain position. In other words, if a person under low involvementobserves many people supporting a certain position, this may provide the justificationfor accepting that position without any further questions. Since a low involvementperson has very little personal stake in the situation, agreeing with the majority is aquick way of reaching a solution with minimal effort. Thus, under low involvementone is more likely to observe conformity to the group norm (majority influence).On the other hand, under high involvement one would more critically examinethe message from the two sources, and the source itself is less important. A highlyinvolved consumer is more likely to give consideration to both the minority and themajority messages. Hence, the fact that one opinion is supported by more peoplemay not be a strong influence on the high involvement subjects. This line ofreasoning suggests that there is a greater likelihood of minority influence under highinvolvement.Contrary to this explanation, it is also conceivable that under high involvementindividuals may turn to the majority, rather than the minority. Under situationsinvolving very high personal/social risks or great uncertainty, taking the side of amajority may facilitate a diffusion of the risk. In such a case, adopting a minorityposition is less likely to resolve the uncertainty or perception of risk. While, this lineof reasoning provides a rival hypothesis to the ELM predictions, it is likely to bedependent on two things: (i) very high uncertainty or perceived risk, and (ii) individualdifferences in willingness to assume risk.41A recent study by Trost, Maass and Kenrick (1992) found support for theirprediction that a majority opinion is more influential under conditions of high personalrelevance of an issue, while a minority is persuasive only under low involvement.Under high involvement, they found that the minority source was derogated and wasnot influential.It seems the predictions based on the elaboration likelihood model (ELM) andthe Trost et al. (1992) results are in opposite directions. ELM suggests a focus on thequality of arguments under high involvement and reduced attention paid to theminority's or majority's status should make the minority more persuasive. On theother hand, Trost et al.'s (1992) result provides a rival hypothesis that higherinvolvement may diminish minority influence. It is important to note that while Trostet al. (1992) have examined the role of involvement, their predictions were not drawnfrom the ELM, but from Moscovici's Theory of Conversion Behaviour (TCB). Tofurther understand why ELM and TCB lead to different predictions, let us consider thesimilarities and differences between these two theories.3.3.2 Parallels Between ELM and TCBBoth ELM and TCB are dual process models. ELM suggests that a central orperipheral route to attitude formation is activated depending on whether the level ofinvolvement in that attitude issue/object is high or low (Petty and Cacioppo 1986).TCB is also a dual process framework. TCB suggests that exposure to majorityopinion leads to shallow processing or a comparison process and exposure to a42minority opinion leads to a careful examination of the arguments or a validationprocess.Both the TCB and the ELM frameworks suggest dual processing routes. In theTCB, exposure to majority triggers a process that is akin to the peripheral process andexposure to a minority triggers a process similar to the central process in the ELMNemeth's work sheds further light on the dual persuasion routes. According toNemeth (Nemeth 1985; Nemeth and Kwan 1985), majorities foster convergentthinking (through conformity pressure), while minorities facilitate divergent thinking(often by pointing out that the majority's solutions are not the only solutions). Thus,individuals exposed to minorities engage in greater cognitive elaboration.Moscovici's (1980) theory indicates that a majority opinion should triggerminimal cognitive elaboration and activate the comparison process, irrespective of thelevel of involvement in the issue. This led Trost et al. (1992) to hypothesize that thedegree ob personal relevance in an issue would play a role only when the persuasivemessage comes from the minority, but not the majority. Predictions based on ELMwould, on the other hand, suggest that under high level of involvement the sourcecharacteristics (e.g., minority/majority or source-similarity) become less important andmessage content becomes the central focus.It is conceivable that when facing a strong majority an individual, due toconformity pressure, may decide to accept the majority's position without criticalexamination, thereby essentially functioning in a low involvement mode. On the otherhand, when facing a minority, the pressure to conform is reduced and also now the43individual has two different viewpoints to consider (the majority and the minorityviews), this could lead to the individual being in a "decision making mode" as well asexperiencing greater involvement when facing a minority opinion. Thus, it seems, inthe TCB an individual's issue-involvement becomes secondary to the group pressureexperienced while facing a majority. On the other hand, the ELM would hold that theprocess (or depth of processing) is likely to be determined by the extent of individualinvolvement in the issue rather than by the characteristics of the source (i.e., majorityor minority).There are also other key differences between the TCB and the ELM that warrantexamination. First, the ELM specifically considers the "personal relevance" of theissue to the individual, which TCB does not. Second, while both the TCB and theELM consider an individual's motivation, they deal with different underlyingmotivations. In the ELM it is the presence or absence of personal relevance (wherethe outcome of a certain course of action does or does not have personal implications)and in the TCB it is merely the presence or absence of the need to conform. Thus theunderlying cognitive processes in the central/peripheral routes are not the same asthose in the comparison/validation conditions.Given the controversy surrounding the nature of the process underlying majorityand minority influence (see earlier discussion on dual and single process in Chapter 2),as stated earlier, this research has adopted the ELM perspective. It should also benoted that part of the conflict between the ELM and TCB predictions may have to beempirically resolved. Depending on which motive (i.e., concern for personal44consequences versus need for social approval) dominates, the nature of socialinfluence may differ.3.3.3 Issue-Involvement and Source-SimilarityIn this research, in addition to considering the impact of involvement on thedegree of conformity or minority influence, we also have to consider the other sourcecharacteristic, i.e., source-similarity. If the ELM framework is applied, both source-similarity and minority/majority support will be treated as peripheral cues under highinvolvement. Thus, similarity of source should have little effect in the highinvolvement condition, whereas for the low involvement individuals source-similaritymay be another peripheral cue for forming an attitudinal position without a great dealof cognitive elaboration.Research in the area of social comparisons, however, provides a possible rivalhypothesis here. When consumers make choices very often information from similarothers may be more relevant information than information from dissimilar others (cf.Goethals and Darley 1977). For instance, knowing the preferences and attitudes ofsimilar others may provide a better guide to how one should behave in an unfamiliarsituation. A marketing study by Moschis (1976) did lend credence to this viewpoint.Information from a similar source could have greater utility to highly involvedindividuals. This line of reasoning would suggest that highly involved individuals aremore likely to lean towards a similar-source (or a similar minority) than uninvolvedindividuals.45At this juncture, we again have two conflicting predictions. A ELM basedprediction would be that source-similarity would operate as a peripheral cue and wouldthus be a significant factor only in low involvement conditions. But social comparisonliterature, which suggests that information from similar others is more useful thaninformation from dissimilar others, would suggest the contrary. That is, source-similarity would be more important under high involvement.3.3.4 Issue-Involvement HypothesesMAIN EFFECT OF INVOLVEMENT: The Elaboration Likelihood Model would leadto the prediction that involved individuals would engage in more diligent considerationof information and would form their attitudes based on the respective quality of theminority and majority arguments. On the other hand, uninvolved individuals, not beingmotivated to process information, may look at the degree of "support" for eachposition within the group as a cue to decide their own attitudes, thus making amajority influence more likely.Hypothesis 4: Greater minority influence will be observed under high, ratherthan low issue-involvement.INTERACTION BETWEEN INVOLVEMENT AND SIMILARITY: Similarity betweenthe source and the recipient of the message is another peripheral cue. Hence, underlow involvement a similar minority can be more influential than a dissimilar one.However, when the involvement is high, source characteristics should not matter. Inother words, the similarity main effect hypothesis presented earlier is further qualified46by taking into account the level of issue-involvement. We can hypothesize that:Hypothesis 5: Similar and dissimilar minorities will vary in their level ofinfluence in the low issue-involvement condition, but not in the highissue-involvement condition.Hypothesis 5a: Under low involvement, a similar minority will be moreinfluential than a dissimilar minority.Hypothesis 5b: Under high involvement, similar and dissimilar minorities willnot differ in their influence.These hypotheses actually constitute a test of the Elaboration Likelihood Modelin a group influence setting. If the hypotheses are tenable, then this research wouldhave extended the ELM to group influence contexts.3.4 Response-InvolvementIn studies of consumer attitudes and preferences, there is very little appreciationof the fact that attitudes or preferences stated in public may not tally with onesprivately held position. For instance a consumer who attends a party may say thathe/she likes the food served in order not to offend the host. At other times one maysimply say positive things due to self-presentational motives. In these cases, theindividual may hold very different private views. Thus, it is very likely that the typeof social influence - majority versus minority influence - will depend on whether or notone's opinions are expressed in private or in public.Leippe and Elkin (1987) have examined the difference between private and47public responses. They use the term response-involvement because the relevance ofthe message can increase or decrease depending on whether or not the recipient'sattitudinal response will be presented for some sort of public scrutiny. Under highresponse-involvement (public response) individuals are less interested in resolving theissue, but rather the concern is with presenting a moderate or publicly acceptableposition. However, under low response-involvement (private response) one may bemore willing to reveal one's true attitude.Moscovici's (1980) Theory of Conversion Behaviour is consistent with theabove and states that majority influence leads to merely superficial compliance,whereas acceptance of minority position often leads to deeper internal change.Among others, Maass and Clark (1983) provided support to this theory when theyfound that greater minority influence occurred under private response, whereconfidentiality was assured, rather than under public response. Apparently, the lackof conformity pressure in the former case and the presence of such a pressure in thelatter case causes this difference in responses.3.4.1 Issue versus Response -InvolvementPreviously, the interaction between issue-involvement and response-involvement has not been tested in a consumer context. When motives clash whatwill be the outcome? Issue-involvement encourages systematic processing that issensitive to how well the message concurs with ones personal standards, whereasresponse-involvement encourages self-presentational needs (Leippe and Elkin 1987).48In a consumer setting one can visualize situations where one may be both issue andresponse involved. For instance, a consumer who is about to make a decision on acar purchase is likely to be issue-involved and if the consumer has to discuss his/heropinion with some friends (reference others) whose views are unknown thenresponse-involvement is also likely to be high. Work by Cialdini et al. (1976) suggeststhat when the two motives interact issue-involvement will dominate. Cialdini et al.(1976) note that "a person's concern for appearances should be dwarfed by outcomesconnected with the topic itself" (p.664).In other words, individuals who are issue-involved should carefully consider theminority and majority messages without paying much attention to the numerical sizeof the sources. Since these subjects are likely to reach a decision based on themessage, not the source strength, they are more likely to have the same position inprivate as well as public response conditions. On the other hand, when issue-involvement is low one may use the majority size as a cue in the public responsecondition, leading to conformity. But in private, when such conformity pressure isabsent they may deviate from the majority position.3.4.2 Response Involvement and Source-SimilaritySocial comparison theory (Festinger 1954; Goethals and Darley 1977) suggeststhat on matters involving personal values and tastes, one prefers to compare with asimilar other when other objective comparison standards are not available. Hence,similarity information is likely to be very salient and hence in both public as well as49private response conditions a person should prefer a similar and not a dissimilar other.In other words, irrespective of the mode of response a similar minority will be moreinfluential than a dissimilar minority. Thus, it seems any significant interaction effectbetween similarity and response-type is unlikely.Research in minority influence suggests that interaction between response-typeand source-similarity might be more complex. The outgroup (or dissimilar) minorityis often perceived as being less credible than the ingroup (similar) minority. Clark andMaass (1988b) have argued that the ingroup minority is likely to be more persuasivein direct or public measures (where psychological costs of agreeing with an outgroupminority may be a consideration). If ingroup minorities have more credibility thanoutgroup minorities, then consistent with the source credibility literature, the outgroupminority (the less credible of the two) is more likely to generate greater cognitiveactivity and may stand a better chance of being persuasive in indirect or privatemeasures (where social identity is not threatened).Kruglanski and Mayseless (1987) have also suggested that whether one seekssocial comparison information from a similar or a dissimilar person may depend on themotivations of the individual at that point. This is consistent with Clark and Maass'sarguments. In a public response condition, individuals may operate under a motiveto protect their social identity. In a private response condition, however, with nothreat to social identity individuals may be more likely to "seek the truth", thus payingmore attention to the views of dissimilar sources as well. The early literature on socialcomparison does suggest that dissimilar sources do not provide good basis for50evaluation of opinions (or sources of information). Thus, a test of two competinghypotheses is possible here.3.4.3 Response-Involvement HypothesesMAIN EFFECT OF RESPONSE-INVOLVEMENT:Hypothesis 6: Greater minority influence will occur under private,rather than public response condition.INTERACTION EFFECTS:Hypothesis 7: Extent of minority influence will differ due toresponse-involvement only in the low issue-involvement condition,but not in the high issue-involvement condition.Hypothesis 7a: Under low issue-involvement there will be a greaterminority influence in the low response-involvement condition(private response) as compared to the high response-involvementcondition (public response).Hypothesis 7b: Under high issue-involvement, the degree of minorityinfluence will not differ due to response-involvement.Hypothesis 8: In both the high and low response-involvement conditions,a similar minority will be more influential than a dissimilar minority.These hypotheses specify the conditions for minority and majority influence.An experiment was designed to test these hypotheses. In this empirical test anindividual faced a majority and a minority taking two opposing positions on an issue.Conditions facilitating conformity versus minority influence were varied. The pretestsconducted to refine the methodology are presented in the next chapter.51IV. PRETESTS4.0 OverviewIn addition to the initial investigations conducted to identify the focal topic anddetermine the format of the stimulus material, several pretests were conducted toobtain further insight. This chapter discusses three specific preliminary studies. First,a pretest conducted using the "cafeteria decision problem" is discussed. Second,a pilot study, which used a larger sample size, but the same decision problem ispresented. Third, since the results of the pilot study suggested substantialmodifications to the design and measures, another pretest was conducted. Thispretest used the "joint venture decision problem" as the setting for persuasion.4.1 Selection of Group Discussion Topic Initially, several potential topics were tested to identify a topic that would beinvolving to the student participants and at the same time will have a minimal chanceof the subjects having strong or extreme prior attitudes. The reason for choosing atopic with moderate or non-existent attitudes is elaborated in the next chapter.Several issues relevant to marketing were examined. These included: (i)environmental safety of products and who (consumer or marketer) should take thelead; (ii) whether or not abortion pills should be made available, (iii) whether or not52Sunday shopping should be restricted through legislation; (iv) whether a new cafeteriato be started on campus should be run by the University or by a private franchise, and(v) whether the university should capitalize on an "invention" by its faculty and enterinto a joint venture to manufacture and market a product. In each of these issues, thepretest subjects were presented with two options and were asked to state theirpreference on a seven-point scale (strongly agree - strongly disagree).At this stage, the objective was to find out the extent to which the two sidesof each issue were preferred by subjects. The means for the two opposing sides ofthe cafeteria problem were not significantly different (support for university = 4.0;support for private franchise = 3.7; n = 25).The Sunday shopping problem was pretested using a single scale (stronglyagree - strongly disagree), where the item read, "I think it is a good idea to restrictSunday shopping hours through legislation." The mean score obtained was 4.3 ona seven-point scale, however only 11 subjects were on the agree side of the scalecompared 17 who were on the disagree side, while 8 were in a neutral position.Thus, there seemed to be a bias towards one side of the issue (i.e., against Sundayshopping). A review of studies on the peoples attitudes towards Sunday shoppinglegislation indicated that there was strong support for Sunday shopping in BritishColumbia (71 % support versus only 18% opposition) and any restrictive legislationwould meet with stiff resistance, especially from younger members of the society(Halifax Chronicle-Herald 1990, p.A5). As the minority influence literature indicatesthat a strong zeitgeist limits the minority's influence when it is seen as anti-zeitgeist53(in this case, in favour of law restricting Sunday shopping), this topic would havepossibly limited the direction of minority influence to one direction (i.e., when it wasadvocating a pro Sunday shopping position). Hence this topic was not consideredfurther.'For the joint venture topic, the mean preference of 15 randomly selectedsubjects was 3.7 on a seven-point scale (where 1 indicated high preference for thejoint venture and 7 indicated low preference for the joint venture). The mean of 3.7was not significantly different from 4.0, which indicated a neutral position on thescale. Since the subjects in the pool seemed to have moderate attitudes towards thejoint venture and the cafeteria topics, these two topics were used in further research.Also, the joint venture and cafeteria topics were fictitious, hence minimizing anychance of subjects having prior attitudes.4.2 Pretest: Cafeteria Topic The preliminary tests, described above, facilitated the selection of anappropriate group discussion topic. Next, a pretest with all three independentvariables (issue involvement, response involvement and source similarity) wasconducted using a small sample. The main purpose of this pretest was to identify1 The zeitgeist was of particular importance because this research was intendedto be conducted using young adults in British Columbia (university students mostlybelow 23 years of age), where according to a national survey Sunday shopping wasvery popular (Halifax Chronicle Herald, August 9, 1990, p.A5).54potential weakness in the manipulations and also get feedback on the clarity ofinstructions provided to subjects.4.2.1 Design and SubjectsThe subjects for this pretest came from two third year marketing classes. Atotal of 42 subjects were used in this pretest. Subjects were randomly assigned toone of the eight cells in the design [Involvement (2) X Similarity (2) X Response Type(2)1. The subjects participated as part of a course requirement.4.2.2 Experimental ProcedureThe study involved the following steps. This pretest was conducted in classwith the data collection for all cells of the design completed simultaneously. First,subjects were informed that they were participating in a Faculty Research Project.Then the subjects were told that they would be presented with a decision problem andwill be provided some information relating to the problem, following which they willhave to express their opinion or preferred solution. Following this general instruction,the subjects were individually given a brief scenario (which contained a description ofa decision problem as well as both issue and response involvement manipulations).Then they were told that a group of students as well as non-students on campus had,at the request of the researchers, previously engaged in a group discussion on thesame topic and that a summary of that group discussion will be made available to thesubjects now. At this point the subjects were given a one page summary of this55ostensible group discussion. After reading the text the subjects responded todependent measures, followed by demand assessment and then a completedebriefing.4.2.3 Stimulus MaterialThe cafeteria problem was chosen as the focal topic for this pretest. Thestimulus material involved the subjects reading the text of a group discussion. In thegroup discussion text, the arguments were presented in two columns - one columnconsisting of support arguments for the university running the cafeteria and the othercolumn consisting of a arguments supportive of a private franchise running theoperation. Six arguments were presented on each side. (see Appendix 1 contains thestimulus material used in the Pilot Study described in section 4.4, which is identicalto the one used in this pretest))Earlier, a sample of 20 subjects from the same subject pool was drawn togenerate arguments in favour of both the "university operating the cafeteria" and the"private franchise operating the cafeteria." The arguments generated were thenshown to a group of 16 subjects who were asked to evaluate the strength of eachargument (after reading each argument they responded on a 7-point scale anchoredvery strong argument - very weak argument). The data from this exercise were1 In this pretest, the arguments in favour of the university running the cafeteriaalways appeared in the left column and the argument favouring the private franchisewere always on the right column. However, in the pilot study reported in section 5.2,however, the larger sample allowed for randomizing the order of presentation as well.56submitted to a t-test and there was no significant difference in the mean argumentstrength for the pro-university and pro-franchise arguments [t (df =15) = 1.01,p > 0.10].4.2.4 ManipulationsISSUE-INVOLVEMENT was manipulated in the initial stage where subjects werepresented with a decision scenario. One half of the subjects were informed that theiruniversity is about to make a decision regarding the operation of a new cafeteria,which could have a bearing on the students' lives (high involvement), and the otherswere informed that they were participating a study on decision making styles, andthat they would first be given a practice task to become acquainted with the exercise(low involvement). However, both groups received the same stimulus material anddependent measures.SIMILARITY was manipulated by varying the social group to which the sourcebelonged. A similar source consisted of "business students" and a dissimilar sourceconsisted of "non-students" (such as a janitor or a research associate working oncampus). When the arguments supporting the university came from a similar source,the arguments supporting the private company came from a dissimilar source, andvice-versa. Each side of the argument was supported by either a minority of peopleor a majority of people. The positions advocated by the minority-majority werecounterbalanced within each cell.RESPONSE-INVOLVEMENT was manipulated by instructing half the subjects57that their responses to this survey would be strictly confidential and that their namesor identities are not required. The other half of the subjects were told that they wouldhave to publicly state their opinions in front of their classmates.4.2.5 MeasuresThe main dependent variable consisted of a two-item measure of the subjects'preference for the two options available. The statements "I think the University(Private Company) should run the cafeteria" were followed by seven-point scalesanchored strongly agree - strongly disagree).4.2.6 Demand Assessment and DebriefingAfter the measures were taken, in order to assess any demand effects thesubjects were asked to write down in the back of the questionnaire their thoughtsabout the intentions of the study. After completion of this task, all subjects were fullydebriefed. Since the entire data for all cells of the design were collectedsimultaneously, a common debriefing was possible.4.2.7 ResultsThe subjects were unable to correctly identify or guess the purpose of the studyor any of the major hypotheses. This indicated that the cover story as well as thetreatments were well disguised. During the debriefing subjects generally expressedsurprise over the intended purpose of the study.58Since the sample sizes were small (4 to 6 subjects per cell), at this stage theprimary concern was not the statistical significance of the results, but the generaldirection of the means. Since this was an exploratory study, the analysis was limitedto mean comparison through t-tests. For Involvement the prediction was that greaterminority influence would occur under high involvement. The means under high andlow involvement respectively were 1.30 and 0.32 (where higher number indicatesgreater minority influence). The predicted effect for Similarity was that greaterminority influence would occur when the source was similar. The means for similarand dissimilar sources respectively were 1.50 and 0.00 (where higher numberindicates greater minority influence). The third treatment variable, response-involvement, had a relatively weaker effect, with greater minority influence occurringunder public response rather than private response as anticipated (1.1 versus 0.40),but this effect was very weak. None of the effects was significant. None of themean differences reported here was significant. Since each of the t-tests for maineffects had two groups with at least 20 subjects, these results indicated a need forstrengthening the treatments.4.3 Revision of Manipulations and Stimulus Based on these pretest result a careful scrutiny of the manipulations wasundertaken. The involvement manipulation was to be strengthened in two ways. (i)Offer of a financial incentive in the high involvement condition. Subjects in the high59involvement condition were to be informed at the outset that their names would beentered in a lottery with three cash prices: $100, $75 and $50. This was expectedto motivate them to pay greater attention to the issues and interest in the task, ascompared to the low involvement subjects. 2 (ii) The high involvement subjects werealso to be informed that a decision regarding the cafeteria in their university wasimminent (to occur within 6 months) and the low involvement subjects were informedthat the decision would be implemented after four years.The similarity manipulation showed the strongest effect of the threemanipulations, even though the main effect was not significant. Given the generaldirectional support, it was anticipated that with a larger sample size this effect wouldbecome significant. Response-involvement was the weakest manipulation. In thepretest subjects read the group discussion text in which some of the discussion groupparticipants were business students (same faculty as the subjects) and others werenon-business students or non-students. The Response-type manipulation involved thesubjects either having to state their position publicly or not. It was felt that thestrength of this manipulation could be increased by informing subjects that some ofthe group discussion participants were actually their classmates. This might causethem to closely examine the position of similar others (classmates/business students)in the public-response condition. These changes came out of feedback from the2 At the end of the data collection, however, both the high and the lowinvolvement subjects were to be offered the chance to enter their names in a lottery.However, only the high involvement subjects were to be made aware of thisbeforehand.60participants in the pretest as well as with other experts in the field who hadconsiderable experience in involvement manipulations. A pilot study, incorporatingthese changes was carried out. The details of this study are presented next.4.4 Pilot StudyThe pilot study involved the same focal topic - should a new campus cafeteriabe run by the university or a private franchise - and involved the same design as in thepretest described above. This study adopted the same experimental design andprocedure described in section 4.2 of this chapter, except for the revised treatmentsdiscussed in section 4.3. The sample of 120 subjects used in this study was drawnfrom the same subject pool as in the pretest. It was hoped that the larger sample sizeand the modifications made to the treatments would produce stronger results or atleast shed more light on the underlying processes. The following sections discuss theresults and the implications of the results.4.4.1 Measures and their ReliabilityAs prescribed in the marketing literature, multiple items were used to tap eachconstruct (Churchill 1979). As per the standard practice, Cronbach's Alpha wascomputed for each set of measures (see summary in Table 2). The two scalesmeasuring attitude towards the two viewpoints (university vs. private-franchise) hadan alpha of 0.92, thus indicating good internal consistency. The manipulation check61measures for involvement (concentration while reading, whether all alternatives wereconsidered and how much attention they paid to arguments on both sides) had analpha of 0.77, which is considered to be adequate (Nunnally 1978). The similaritymanipulation check measures (whether similar others provide a more useful basis ofcomparison, whether they perceive similar others to hold similar views) had anr2 = 0.50, but an alpha of only 0.67, which is somewhat low (the problem with themeasures is discussed later in this chapter). The six belief measures (which requiredsubjects to agree or disagree with the statements made by the group discussionmembers) also fared reasonably well with an alpha of 0.77. Finally, the measuresused as indicators of source credibility fared poorly with an alpha of 0.52. At thisstage it was conclusive that the scales had to be reworded and redesigned.The data were then submitted to a factor analysis with varimax rotation to getfurther insight into how "clean" the measures were. The 17 items used in the factoranalysis were as follows: two measures of overall attitude, one measure ofconfidence, six belief items, three measures of involvement manipulation check, twomeasures of similarity manipulation check, and three items relating to sourcecredibility and quality of arguments. The analysis yielded a total of four factors thataccounted for only 57% of the variance. A scrutiny of the rotated factor matrixrevealed that a simple factor structure was not obtained (see factor analysis summaryin Table 3). There were instances of the same items having fairly high loadings(greater than 0.4) on more than one factor. This result indicated the need for furtherrevision of the measures.624.4.2 Manipulation CheckISSUE-INVOLVEMENT: A t-test was conducted by the variable obtained bysumming two manipulation check measures (amount of attention paid and how muchthey concentrated on the message when they were reading the message). 3 Theexpectation was that the high involvement group would have a significantly lowerscore, indicating a higher issue-involvement (the low and high ends of the compositescale was 2 and 14 with a mid-point of 8, and the data were coded in such a fashionthat a lower score on the scale meant a higher involvement). The t-test producednonsignificant results [t(117) = 0.10, p > 0.90], with both groups having identicalmeans [High = 5.1, Low = 5.2]. These mean involvement scores indicate a high tomoderate level of issue involvement in the two groups. Another indicator ofinvolvement - degree of confidence in one's opinion - also failed to reveal anydifferences between the groups [t(117) =1.12, p> .10]. The possible reasons for thefailure of the manipulation and the implications are discussed in subsection 5.3.4.SIMILARITY: The items used to infer similarity did not directly refer to eitherof the sources in the group discussion text (commerce students and non-students) andask the subjects which of the two sources they thought were similar to them (it wasfelt that such a direct question may make transparent the objectives of the study).The items instead were designed to get a general feeling for whether the subjectspreferred to seek information from similar others or dissimilar others. Hence a clean3 These measures were used because both degree of attention and concentrationof attention to message cues are considered to be important indicators ofinvolvement.63manipulation check was not possible. However, a t-test conducted using thesimilarity/dissimilarity of the minority as the two groups revealed that when theminority consists of a similar source, subjects were more likely to consider a similarsource as a useful source of information than when the minority consists of adissimilar source [t(117)=4.76, p < 0.001]. While this can be taken as an indicationof support for the similarity manipulation, the need to design more precisemanipulation check measures is discussed in a later section.RESPONSE-INVOLVEMENT: To verify the effectiveness of this manipulation,subjects were merely asked to recall whether or not they were previously instructedto state their opinions publicly. In the private response condition 93% of the subjectssaid NO and in the public response condition 89% of the subjects said YES, whichsuggests that this manipulation had registered in most people's minds.MINORITY-MAJORITY: Finally, subjects were required to state how manypeople supported the university running the cafeteria and how many people supportedthe private franchise. Accurate recall of this would indicate that subjects did payattention to the majority-minority configuration of the group discussion team. Theresults indicated that in both conditions - when the minority was supporting theuniversity and when the minority was supporting the private franchise - subjects haddifficulty recalling accurately the number of people in the majority condition. Whenthe minority of two supported the university: 96% recalled that two people supportedthe university, while the response for the majority ranged from 4-6 with only 30%accurately recalling that six people supported the private franchise. In the other64condition, (minority supporting private franchise), accuracy of minority number recallwas 93%, while accuracy of majority number recall was only 35% (the responsesranged from 4 persons to 8 persons). The data were split by involvement (high andlow issue-involvement) to see if those in the high involvement condition paid lessattention to the number of people in each group as compared to those in the lowinvolvement condition. The results indicated that subjects in both involvement levelshad difficulty recalling the exact number of individual in the majority condition. Theseresults call into question the effectiveness of the minority-majority manipulation.Overall, it seems the manipulations were not very effective. At this point, thediscussion of hypotheses tests is presented.4.4.3 Hypotheses TestsFirst a 2x2x2 ANCOVA with Similarity, Issue-Involvement and Response-Involvement as the three factors was conducted, using political-orientation as thecovariate. The dependent measure was the attitude towards the two positions. Thefirst scale measured attitude towards support for university and the second onemeasured attitude towards the private enterprise. An agreement on the first scalewould mean less preference for private enterprise, and vice-versa. The scales wereappropriately recoded and summed to produce a composite score. The ANCOVA wasconducted to pest the three main effects as well as all the interaction effectshypotheses (see detailed results in Table 4).MAIN EFFECTS: None of the three predicted main effects was significant. It65was anticipated that the a similar minority would be more influential than a dissimilarminority, whereas the test showed an insignificant difference between the groups[F(1,118) = 0.023, p > 0.80]. For Issue-involvement, the prediction of greater minorityinfluence under high involvement was not confirmed [F(1,118) = 0.022, p > 0.80]. ForResponse-Involvement, greater minority influence was expected in the private(confidential) response condition, which was not supported [F(1,118) = .069, p > .70].INTERACTION EFFECTS: None of the interaction effects was significant,however, these effects seemed stronger than the main effects and did reveal someinteresting patterns. The Similarity X Issue-involvement interaction, while notstatistically significant indicated an interesting pattern [F(1,118) = 2.56, p > 0.10].While low involvement subjects did respond better to a similar minority (Mean = 0.14),rather than a dissimilar minority (Mean = -0.06), unexpectedly the high involvementsubjects responded more favourably to the dissimilar minority (Mean =0.19 ), than thesimilar minority (Mean = -0.07).The Similarity X Response interaction was also not significant [F(1,118) = 2.24,p> 0.10]. An examination of the means revealed that in the public response conditiona similar minority (Mean =0.21) was more influential than a dissimilar minority(Mean = 0.00) as expected. However, in the private response condition, it seems, thesubjects felt less social pressure to take the side of a similar source and the directionof means were reversed (Similar minority = -0.15; Dissimilar Minority = 0.13).Finally, The Issue-involvement X Response-involvement interaction was insignificantwith no discernable pattern in means [F(1,118)=1.6, p > 0.20 ]66COVARIATES: The covariates used in the study explained more variance thandid any of the main or interaction effects. Position (advocated by the minority) washighly significant [F(1,118) = 8.69, p < 0.005] and Political-orientation of theindividuals also had some effect, albeit not significant at the 0.05 level[F(1,1 18) = 2.65, p = 0.10]. An examination of the means for the Position covariateindicated that the minority was significantly more influential [t(118) = 3.10, p < 0.005]when it advocated a pro-franchise position (Mean = 0.38) than when it supported theuniversity (Mean = -0.18).4.4.4 DiscussionThree major concerns arose from the results of the pilot study. First, thereliability and validity of the measures were challenged by the reliability and factoranalysis results, indicating that the measures might have to be reconceptualizedand/or reworded. Second, the treatments were apparently not at all effective,indicating serious problems with the experiment and the design. Third, moreimportantly, was the conceptualization of the theory and hypotheses itself flawed.Let us consider the three issues in the reverse order, with the last issue addressedfirst.First, the validity of the basic experimental paradigm as well as foundationaltheory of this research has been previously established (e.g., Clark and Maass 1988a;Maass and Clark 1983; Martin 1988a; Perez and Mugny 1987). A key differencebetween this and past studies (e.g., Clark and Maass 1988a; Martin 1988a) was the67incorporation of issue-involvement into minority-majority influence process. It seemedthat the examination of two competing motives - a motive for protecting one'spersonal consequences (issue-involvement) and a motive for social acceptance orbelonging to a group (caused by the group setting) - would be the next logical stepin furthering our knowledge on conformity versus minority influence effects. Theimportance of this line of enquiry has been previously acknowledged (e.g., Chaikenand Stangor; Maass and Clark 1984). Thus, it seemed that the model of socialinfluence suggested here did have solid theoretical grounding.Second, problems with the manipulations and the design, merits seriousattention. Two specific manipulations need closer scrutiny at this point.(i) Minority-majority manipulation: Subjects' inability to recall accurately the numberof people in the majority and minority group indicated that they were either not payingattention to the numbers or were too highly involved in the issue, which might explainlack of attention to this peripheral cue. The self-reported issue-involvement score inthe two issue-involvement conditions revealed that in both the low involvement(Mean = 5.2) and the high involvement (Mean = 5.1) conditions and the amount ofattention paid to the message was quite high, indicating high involvement with theattitudinal issue (probable causes for this observed effect are explored in the followingparagraphs). The other possibility was that the numerical gap between the majorityand the minority was not sufficiently vivid. Yet another possibility was that theformat in which the information was presented (i.e., in two columns) was differentthat used in past research (which was not in columnar form). In this case subjects68were not told anything about the number of people in each group, as was done inearlier research (e.g., Maass and Clark 1983) and given the high involvement in theissue, the subjects may have been oblivious to the number of people on either side ofthe issue, which diminishes the effectiveness of the minority-majority manipulation.(ii) Issue-involvement manipulation and validity problems: Issue-involvementwas manipulated in the standard manner, using "your university" versus "otheruniversity" format. The issue chosen as the focal topic (whether cafeteria should berun by university or a private franchise) was not one that was intrinsically involving(according to earlier pretests). Why then was the manipulation ineffective? Anexternal confound not taken into consideration at the time of data collection providesthe best possible answer. The data for this study were collected at The University ofBritish Columbia in March-April 1992 when there was a campus-wide strike of allunion employees (which included all non-teaching staff). This strike resulted in theclosure of all campus cafeterias for several weeks, including the university operatedfood services at campus residences. Thus, at the time the study was conducted,students were clearly dissatisfied with the university operated food services. As aresult, the issue of campus food services had probably become a highly involving topicfor everyone and the experimental manipulation intended to vary the involvement leveldid not succeed. Also, among the business students who constituted the samplethere is a usual predilection for private enterprise, which could have been further69heightened by these uncontrollable external events. 4Just prior to the data collection it was learned that in the classes where thedata collection was scheduled (an undergraduate marketing class with multiplesections), the instructor had discussed the campus strike in detail and had asked thestudents how it affected their behaviour as consumers. Thus a few days before thedata collection, the students had been sensitized to this subject. At this stage sincethere was no alternative but to go ahead with the data collection, and so somemeasure was necessary to prevent possible impact of these external/historical events.To accomplish this, as the subjects were introduced to the study and briefed aboutthe "purpose" of the study (which was varied for high and low issue-involvementgroups), they were informed that the two private franchises under consideration bythe university (as alternatives to university run food services) both had unionizedemployees, and thus the private franchise was also equally susceptible to be shutdown in case of any campus strike (see Appendix 1). By adding this comment, atthat time, it was felt that the students may now not assign any advantage to theprivate enterprise on this particular issue.The results, however, indicate that the effect of the strike had clearly resultedin a very strong preference for the private enterprise which could not be altered by themanipulation. In other words, a clear zeitgeist had evolved regarding this issue, whichwas not the case during the earlier pretest (see section 4.1), and in accordance with4 Unfortunately the data collection scheduled could not be postponed because itwas the last week of classes for that term, and there were no summer classes in theFaculty of Commerce available for data collection.70prior findings the minority was persuasive only when its message was consistent withthe zeitgeist (e.g., Paicheler 1976; Mugny 1979). Thus, at this point, one couldconclude that an internal validity threat caused by "history" (the strike affecting foodservices) was a major factor in weakening the treatment effects (Cook and Campbell1979).(iii) Response-involvement manipulation: The data were collected in a third yearmarketing class where disagreements during class discussions are not uncommon.Thus, the instruction intended to produce public pressure (i.e., to state their opinionpublicly) may not have been strong enough to induce the desired group pressure. Insupport of this proposition, Fitzpatrick and Eagly (1981) found that moderate attitudeswere expressed only when subjects anticipated a discussion with experts, and in thecase of expected discussion with peers, opinions were quite polarized (meaning lessself-presentational concerns). The response-involvement manipulation used in thepretest and the pilot study required public expression of opinions in front of a groupof peers (classmates who were reasonably familiar with each other). This couldaccount for the ineffectiveness of the manipulation. Further discussion regardingresponse-involvement is carried out in the next section.Thirdly, reliability analysis and factor analysis called into question the internalconsistency as well as validity of the measures. For instance one of the questionsread: " I think the group members with backgrounds similar to mine talked aboutissues that I myself would have considered." Discussion with subjects after theexperiment, to obtain feedback, suggested that this statement was open to multiple71interpretations (some felt that a "similar background" meant commerce students in thegroup discussion, and others interpreted it as those who were advocated a positionsimilar to their own - in other words opinion similarity). The literature on socialcomparison with similar and dissimilar others does suggest that there are twopossible types of similarity - related attributes similarity (Goethals and Nelson 1973;Fazio 1979) and opinion similarity (Kruglanski and Mayseless 1987). Thus, it seemsthe statement required better wording to eliminate dual interpretation. Thus, thesignificant result for similarity, reported earlier, is clouded with some doubts aboutthe validity of the measure used. This and other problems of the same nature led toa careful examination of the measures and revision/addition of measures in the nextstage.In conclusion, the failure of this study did serve as a useful learning experienceand did point out to several problem areas that need attention. The next sectionaddresses some of the corrective measures undertaken.4.4.5 Steps UndertakenAt this stage, an assessment of the entire research was undertaken to chart acourse of action. 5 First, it was felt that the extension of the minority influence theoryto marketing would serve a useful purpose and the examination of issue-involvementwould contribute to the minority influence research. Hence, at this stage, a decision5 The action plan described here evolved out of meetings with the SupervisingCommittee of this dissertation.72was made to rectify the limitations with the measures and treatments and continuethe investigation.Second, after closely examining the independent variables, and evaluating theirrespective contribution to this research, it was decided that response-involvementwould be dropped from further investigation for the following reason. Even thoughthe motivational conflict arising as a function of issue-involvement (where personalconsequences are important) and response-involvement (where self-presentation isimportant) is an interesting issue, Leippe and Elkin (1987), who studied the clashbetween these two motives suggest that the relative strengths of the two motivesmay ultimately decide how the conflict is resolved (i.e., in favour of issue or response-involvement). This suggests that in an experimental setting the stronger of the twomanipulations might win. Also, it felt that since issue-involvement had not beenexamined previously in the context of majority-minority influence, this research wouldmake a greater contribution by carefully investigating the role of issue-involvement atthis stage. Further, since response-type had been examined previously in a similarresearch context (e.g. Martin 1988a; Moscovici and Lage 1976; Moscovici andPersonnaz 1980) and the role of issue-involvement had not been studied before, itwas felt that a smaller experiment devoted to understanding the impact of issue-involvement would be a more meaningful first step. Hence, the examination of theclash between the two types of involvement was deferred to future research.Third, given the impact of the external events on the validity of the research,further attention was devoted to selection of a focal topic (for the group discussion),73which would have minimal influence from any external factors occurring during orprior to the data collection. From the topics tested initially, the joint venture topicwas considered suitable since it satisfied the condition of moderate attitudes amongsubjects and no clear zeitgeist.Finally, it was decided that the dependent and manipulation check measureswould be carefully examined, again compared with measures used in past research,and then submitted to a small pretest to ensure no ambiguities in the questions. Withthese decisions following from the results of the pilot study, the next stage involvedreconceptualization of the study and preliminary tests of the joint-venture study.4.5 Pretest: Joint Venture Topic The "joint venture" topic was used in the group discussion (i.e., stimulusmaterial) in this pretest. Initially small groups of about 4-6 individuals were used toexamine if the manipulations were believable and feedback from subjects wasobtained on the understandability of the measures to ensure there were noambiguities. At this point there was sufficiently encouraging feedback to suggest thatthe measures were meaningful (had face validity) and that the manipulations wereconsistent and believable.MANIPULATIONS: For Similarity, after holding a small group discussion withstudents to identify what they consider to be a "similar" or a "dissimilar" source, itwas decided that "students of the same university" would constitute a similar source,74i.e., the opinions of other students would provide a good point of reference in mostcases. Also, feedback from students indicated that the "faculty of the sameuniversity" would constitute a sufficiently dissimilar source. 6The involvement manipulation was basically the same as in the earlier pretestand the pilot study. The lucky draw used to increase interest in the cafeteria studywas dropped after consultation with scholars familiar with involvement manipulation.'The minority-majority manipulation was further strengthened by making the numberof people in both groups more visible. Feedback from a small group of subjects(similar to those who were to participate in the final study) indicated that a four-eightsplit of minority and majority was clearly visible and did convey the impression thatone group was clearly smaller than the other.QUESTIONNAIRE: Many items in the dependent measures questionnaire werereworded after consultation with experts or on the basis of feedback from the pilotstudy. New measures added included: risk-aversion and feelings towards the source.Manipulation check measures for minority-majority as well as the two theoretical6 It has to be borne in mind that the dissimilar source cannot be extremelydissimilar for the student sample to reject the opinions of that source outright. Inother words, when the subjects read the summary of group discussion (the stimulus)they should be more drawn to the opinions of one group (similar source) than theother (dissimilar source), but the opinions of the dissimilar source should still drawsome consideration for the social categorization manipulation to have face validity.The students used in the small group pretests did suggest that they do consider theopinions of faculty in many cases, but identify more with the students in case of anycommon cause.' Thanks is due to Dr. Jong Won Park, who at the time of this research was atthe University of British Columbia, for his invaluable suggestions in this regard and forsharing his knowledge on issue-involvement manipulations.75variables of interest, involvement and similarity, were also reworded (seequestionnaire of main study in Appendix 2).Another pretest of the entire design at this stage could not be carried out dueto time and resource constraints. Hence, only the main effects were tested to verifythe effectiveness of the treatments before undertaking the final data collection effort.Two classes of undergraduate marketing and economics students were used, with thesimilarity manipulation being tested in one class and the involvement manipulation inthe other. 8SUMMARY OF RESULTS: Similarity manipulation produced results in theanticipated direction, with a similar minority (Mean = 1.67) being more influential thana dissimilar minority (Mean = 0.26), with the t-value being significant at the a = 0.10level [t(38) =1.41, p <0.10]. The involvement manipulation also had directionalsupport, although not significant, with the high involvement condition (Mean =1.39)producing more minority influence than the low involvement condition (Mean = 0.60).4.5.1 Pretest Summary and Overview of Final DesignThe verbal discussions held with small groups of subjects were very useful inredesigning the questionnaire and refining the manipulations. The limited datacollection undertaken at this point did produce encouraging results and more8 The subjects for this pretest as well as the main experiment (see results inChapter VI) came from Simon Fraser University, whereas the subjects for earlierpretests and pilot study were from The University of British Columbia. Except for thisdifference, subjects were similar in terms of age, academic year and the fact that theywere taking a similar third year marketing course.76importantly open-ended feedback provided by subjects in both the similarity andinvolvement conditions revealed that the manipulations had face validity (werebelievable) and the dependent measures were clear with no ambiguities.The main experiment involved a 2(high/low minority similarity) X 2 (high/lowinvolvement) design with the minority-majority being a within-subjects design. Themain dependent measure was the preference for the minority versus majority opinion.The joint venture topic was used as the setting for the manipulations and the stimulusmaterial. Further details of the methodology are in Chapter V.77V. METHODOLOGY5.0 OverviewThis chapter outlines the methodology of the main experiment (and thecorresponding results are in Chapter VI). Since several pretests and a pilot study wereconducted before the main experiment, specifics of the methodology for the pretestsand the pilot study were discussed separately in Chapter IV. The purpose of thischapter is to first discuss the methodological issues in configuring the empirical study,then to present an outline of the research methodology used in the main experiment5.1 Methodological Issues 5.1.1 "Real" versus "Nominal" GroupsSince Moscovici's initial studies, a variety of paradigms have been utilized totest minority influence. In many of these studies the groups are composed ofconfederates who act as the majority and/or minority (e.g, Maass and Clark 1983).Asch's (1951) experiments on conformity effect also involved "real" groups withconfederates, where the subject came face to face with other individuals in the group.Other researchers have resorted to the use of "nominal" groups in which subjectswould read a transcript of a group discussion without any face-to-face contact withthe group members (e.g., Mongeau and Garlick 1988; Clark and Maass 1988a, Clark78and Maass 1988b; Martin 1988a).The critical issue here is whether the use of face-to-face groups would produceresults that are different from what would be obtained through a nominal groupapproach. Scholars in the area of group influence research have emphasized that thephysical presence of group members is not necessary to feel the effects of grouppressure. For instance, Rabbie and Horwitz (1988) have argued that face-to-faceinteraction is not necessary for individuals to perceive common fate orinterdependence or group pressure.An examination of the results do indicate that the nominal group approach,while being easier to use, does not in any significant way alter the results. The useof either methodology is prevalent in this area and the results seem to be consistentacross the two methods. Often the same researchers have used both methods indifferent studies. For instance, Clark and Maass (1988a) found that ingroup minoritieswere more influential than outgroup minorities, and this result was confirmed in twodifferent experimental paradigms - one involving face-to-face meetings and directinteraction and the other using written information. Similarly, Clark and Maass(1988b) used a group discussion text in lieu of a face-to-face meeting and reportedresults that were consistent with their past research involving face-to-face meetings.It seems that based on past evidence one may conclude that the twoapproaches will produce more or less consistent results. While one might argue thatthe written text approach is synthetic, a survey of results does not reveal anysignificant differences in results due to the paradigms employed. In this research,79given the lower cost and ease of operationalizing groups, the group discussion textmethod was adopted.5.1.2 Single Source Versus Dual SourcesMany studies on minority influence require the subjects to read a messagepurported to emanate from only one source - either a majority or a minority (e.g.,Clark and Maass 1988a; Mugny and Papastamou 1980). Other studies, such asMaass and Clark (1983), have simultaneously exposed subjects to minority andmajority sources.The present research is conceptualized as a study of multiple source effectsrequiring a simultaneous social influence attempt by both the majority source and theminority source. Hence previously uncommitted subjects were simultaneouslyexposed to a text of a "group discussion" where the majority and the minorityexpressed contrary positions on an issue relevant to marketing. The subject had theopportunity to consider both points of view before expressing an opinion.5.1.3 Attitude Versus Attitude ChangeIt has been a customary practice in minority influence research to assess thedegree of attitude change brought by the source of influence. Typically the differencebetween pre and post-test attitude scores are obtained for this purpose (e.g., Clarkand Maass 1988a; Martin 1988a). Alternatively, the pretest score has also been usedas a covariate to control for any differences due to pre-test score effect (e.g., Clark80and Maass 1988b; Perez and Mugny 1987).The pretest measures are often used to select subjects who do not holdextreme attitudes on the focal issue (e.g., Clark and Maass 1988b; Maass and Clark1983; Mackie 1987). This is done to ensure that attitude change is demonstrablewithin an experimental setting and sometimes to ensure that either message (majorityor minority message) is not strongly pro or counterattitudinal. In most studiesinvolving pre-test measures, the focal issue is often something where strong priorattitudes might exist. For instance, issues like abortion, gay rights, tuition increaseand financial support to students are used. Studies without pre-test attitude measure,while relatively uncommon, do exist (e.g., Mongeau and Garlick 1988; Moscovici andLage 1976).In this research, a judgement had to be made about whether attitude changeor attitude itself should be measured. While using a pretest measure would facilitatecomparison with previous research, the issue of sensitizing subjects through a pretestmeasure had to be weighed carefully (Campbell and Stanley 1963). 1 It was also feltthat the need to filter subjects through a pre-test (by selecting only those withmoderate attitudes) could be overcome by selecting an issue which the subjects maynot have previously considered and hence were unlikely to have prior attitudes.Hence it was decided to adopt a post-test only design, with the subjects beingrandomly assigned to different experimental cells.1 The data for the pre-test as well as the main study were collected during classtime. It was felt that taking both the pre-test and the post-test measures within ashort time interval of about 20 minutes would lead to pre-test sensitization.815.1.4 Stimulus Material and FormatAnother important methodological issue was the format of the group discussiontext. A determination had to be made as to whether the stimulus, i.e., text of a groupdiscussion, should be presented with minority and majority viewpoints interspersedor if all arguments on one side of the issue should be presented first, followed by thearguments of the other side. Previous research in minority influence had employeddifferent presentation formats, with some using a transcript of a group discussion(e.g., Mongeau and Garlick 1988 and others using a brief summary of the majorityand/or minority arguments (e.g., Perez and Mugny 1987)Since no clear guide was available from past research, pretests were used tostudy several formats of information presentation - minority and majority argumentsinterspersed, minority and majority arguments on separate columns of the same page,and all arguments of one side (minority or majority) followed by all arguments of theother side. In the final study, the third format was used as it was found to be easyfor subjects to grasp the information (see Appendix 2). In the pretest as well as thepilot study described this chapter, the information was presented in columns, withthe minority and majority arguments on one side (see Appendix 1).5.2 Main ExperimentThe main experiment involved tests of only the source-similarity and issue-involvement hypotheses. The response-involvement factor was dropped due to82conceptual as well as operational issues which were discussed in Chapter IV. Withresponse-involvement being omitted from the study, Hypotheses 6 through 8 werenot considered in this study. Thus, the objective of this study was then to testHypotheses 1 through 5, relating to main and interaction effects of similarity andinvolvement. The rest of this section describes in detail the research design andmethodology of the main experiment.5.2.1 Experimental DesignA 2 [Similarity] X 2 [Involvement] between-subjects design was adopted, wherethe minority was either similar or dissimilar (when the minority was similar themajority was dissimilar and vice-versa) and issue-involvement was either high or low.The source (minority and majority) was a within-subjects factor since each subjectreceived a message from both sources. The order of presentation for the two sourcesas well as the position advocated by each of the two sources were counterbalancedwithin each cell.5.2.2 SubjectsUndergraduate Business students participated in the study for course credit.the subjects in the pretest as well as the final study came from the same pool ofsubjects, all of whom were third year students taking a required marketing course.A total of 75 subjects participated in the study, and 72 usable responses wereobtained. The age of the subjects ranged from 18-35 years, with the mean age being83approximately 20 years. Of the 72 subjects, 42 were male and 30 were female.5.2.3 ProcedureSubjects were randomly assigned to one of the four cells in the design, with 18subjects per cell. The data collection took place in a large class. The instructorrequested the cooperation of the student in a faculty research project. They were toldthat their refusal to participate in the study, for whatever reason, would not in anyway affect their course grade. Without any further introduction, packages containingthe treatments and the measures were distributed in random order. Each package hadtwo booklets, the first one contained the treatments and stimulus material and thesecond one contained the dependent measures. The second booklet was identical forall subjects, where as the first one varied (see Appendix 2 for Booklets One and Two).One half of the subjects received the Booklet One which had a cover title"Administration Policy Survey" and the other half of the subjects received the BookletOne which was titled "Decision Making study." The former contained the highinvolvement manipulation and the latter the low involvement manipulation. Themanipulation is described in detail below.After all subjects had received the package containing the two booklets, theexperimenter instructed the subjects to go through Booklet One first before openingBooklet Two. Booklet One presented the decision problem to the subjects (theinvolvement manipulation being embedded in the problem presentation) and thenprovided them an opportunity to see the views of "others" in the form of a group84discussion summary (the similarity manipulation was embedded here). After thesubjects had read Booklet One, they were asked not to reopen Booklet One again andproceed to Booklet Two. Each subject was allowed to go through each booklet athis/her own pace. After all subjects had responded to all measures, the subjects werefully debriefed. Since the entire data collection (all cells of the design) was done inone sitting in a large class room, it was felt that there was no possibility of anycontamination of the data due to communication between subjects. The entireexperiment took about 15-20 minutes to complete.5.2.4 ManipulationsISSUE-INVOLVEMENT: Issue-involvement was increased or decreased bymaking subjects believe that a certain decision would have personal implications intheir university (or city) or another university (or city) (Madsen 1978; Petty andCacioppo 1979; Petty, Cacioppo and Schumann 1983).Low-involvement: Half the subjects were led to believe that they were takingpart in a "Decision Making Study" and were presented with the a problem faced byMcMaster University (in Ontario). The problem had to do with whether that universityshould enter into a joint-venture agreement with a private firm to manufacture andmarket an invention made by that university's faculty members. In the cover story,where this problem was presented, it was indicated that the intention of the study isto understand the differences in decision making process across people, and that theopinions expressed by them (subjects) will have no actual bearing on the situation at85McMaster University.High-involvement: The other half of the sample assigned to the highinvolvement condition was told that their university (i.e., Simon Fraser University inGreater Vancouver, British Columbia) is faced with the decision of whether to enterinto a joint-venture agreement with a private firm or not, and the subjects were alsoalerted to the bearing such an agreement would have on their lives as students of theuniversity. These subjects were informed that they are participating in a surveyconducted by the university administration, which aims to obtain input from studentson this matter before drawing a policy (hence the title, "Administration Policy Survey).To further increase involvement subjects were informed that their opinions could reallyshape the university's policy on this matter.SIMILARITY: After the subjects had read the introduction to the problem (attheir university or at the other university), at end of the same page they wereinformed that a group discussion was conducted on the same topic, involving bothstudents and faculty members of their university to obtain different views on thesubject, and that a summary of this discussion would be made available to them now.The group discussion was presented in a summarized form with the opinions favouringeither side being presented together. When the majority was composed of facultymembers (outgroup or dissimilar), the minority was always composed of students(ingroup or similar), and vice-versa. Thus, similarity was manipulated through socialcategorization.MAJORITY-MINORITY: In the summary provided to the subjects, each position86(i.e., "the university should" or "should not" enter into the joint venture agreement)was supported by either the majority or the minority. When the majority supportedthe joint venture, the minority opposed it, and vice-versa. The minority and majorityadvocacies were counterbalanced across all cells of the design. The majority wascomposed of 8 individuals and the minority consisted of four individuals. In previousresearch, sometimes in small groups of six, two people have represented the minorityand the rest the majority (e.g., Nemeth and Wachtler 1983). The numerical splitbetween the minority and the majority was, thus, fairly consistent with past research.5. 2. 5 StimulusThe treatment, as discussed earlier, consisted of a summary of a purportedgroup discussion. The text contained arguments in favour of or against the jointventure, emanating from either a minority or a majority. The arguments on both sidesof the issue were pretested using a sample of 12 subjects (drawn from the samesubject pool) to ensure that they did not differ in quality. A t-test conducted to verifythe differences in the sample's ratings of pro and con joint venture arguments wasnonsignificant [t(11) =1.0, p > .20]. The text also contained an equal number ofarguments by both the majority and the minority, with each group presenting sixarguments. The lengths of the minority and majority arguments were identical, eachoccupying 12 single-spaced lines.875. 2. 6 MeasuresAfter the subjects had gone through Booklet One (the treatments and thediscussion text), they proceeded to Booklet Two, which contained the manipulationcheck, measure of attitude towards the issue, as well as measures of covariates.MANIPULATION CHECKS: (i) Minority-Majority: A recall measure of thenumber of people supporting each position in the discussion and whether or not theywere students or faculty was used to verify if the minority-majority manipulation (awithin subjects factor) had registered in the subjects' minds.(ii) Similarity: A three item scale measuring the extent to which the subjects identifiedwith the faculty or other students was used. Subjects responded to statementsanchored "strongly agree - strongly disagree." (iii) Issue-involvement: Level ofinvolvement in the decision task was assessed using three items which tried to tapthe personal relevance of the joint-venture issue to the subjects. A measure of theirconfidence in their own opinion was also used to further verify if confidence in thedecision varied by involvement level.DEPENDENT MEASURES: The main dependent variable was the subjects'attitude towards the issue of joint venture. A two-item seven-point scale was used.Other dependent measures included: source credibility, feelings towards the source.COVARIATES: Relevant covariates were partialed out using analysis ofcovariance (ANCOVA). Risk-aversion, a predisposition known to affect decisions(Fagley and Miller 1990) was one of the covariates. Since the persuasion literaturesuggests that women are more susceptible to persuasive communication than men88(Eagly and Chrvala 1986; Moschis and Churchill 1977), it was felt that the degree ofminority influence or conformity may have a gender bias. Hence, gender was alsoproposed as a covariate. Both covariates were entered into the analysis together.5.2.7 AnalysisFirst, t-tests were performed to verify the effectiveness of the involvementmanipulation. The three manipulation check measures were summed up to form asingle measure for this purpose. The manipulation check for the similarity treatmentwas done through a chi-square test. This test facilitated the examination of whethermore subjects identified with the similar source (other students) or the dissimilarsource (faculty members). The Minority-Majority manipulation was checked bycomputing a simple percentage of the correct responses for the question asking themto identify the number of students/faculty supporting each position.The major hypotheses were tested using an Analysis of Covariance (ANCOVA).Similarity and Involvement served as the two factors along with Risk-aversion (anindividual personality measure) and gender as the covariates. This analysis wasessential to test the main and interaction effects hypotheses for Similarity andInvolvement factors. Additionally, t-tests (with Bonferroni adjustment) were used forplanned multiple comparisons. To examine the roles of source-credibility and source-feelings in the proposed model of social influence, for each variable a t-test wasconducted by splitting the sample along the mean and examining if the degree ofminority influence (or alternatively, conformity) varied between high and low scorers.89To obtain further insight into the mediating effects of source-credibility and source-feelings, a path analysis was undertaken. Multiple regressions using OLS estimationprocedure were used to conduct the path analysis.5.2. 8 SummaryThe main experiment involved a test of hypotheses relating to similarity andissue-involvement effects (Hypotheses 1 through 5). A 2x2 design with Similarity andInvolvement serving as the two between-subjects factors was adopted. Minority-Majority was a within-subjects factor. The position advocated by each of thesesources as well the order of presentation were counterbalanced in each cell. Standardinferential statistical methods were used in the data analysis, in addition to a causalmodelling analysis. The next chapter presents in detail the preliminary testsconducted, which shaped the research methodology discussed here.90VI. RESULTS6.0 OverviewThe results of the Main Experiment are presented in the following order. First,the scales used in the dependent measures were subjected to a reliability test toensure that the scales were internally consistent. In addition to examining theCronbach's alpha levels of the scales, a confirmatory factor analysis was alsoperformed. A five factor confirmatory factor model was suggested by the theory asthere were five major categories of dependent variables - scales measuringinvolvement, subjects' perception of source similarity, perceived source credibility,feelings towards the source and finally attitude towards the issue in the scenario.After discussion of these preliminary results, details of manipulation checks, as wellas the hypotheses tests are presented. Then, path analysis conducted to verify thepresence of mediating variables is discussed, followed by a brief discussion of otherpost-hoc tests.6.1 Reliability Assessment 6.1.1 Cronbach's AlphaIn accordance with previous guidelines provided by many scholars in marketing(e.g., Churchill 1979), multiple measures were used to tap each construct. The91coefficient alpha for involvement items, similarity items, source credibility items, risk-averseness, feelings, and attitudes were respectively 0.84, 0.94, 0.68, 0.84, 0.83and 0.82. Except the source credibility items the rest are consistently high. Even inthis case the alpha is quite close to 0.70, which is generally used as a benchmark foracceptable internal consistency (cf. Nunnally 1978).6.1.2 Confirmatory Factor AnalysisCausal modelling can be used to represent and test the reliability ofmeasurements (Bagozzi 1980). A confirmatory factor model including the five majorcategories of variables was tested through LISREL to further gauge the reliability ofthe measurements. This analysis also provides some indication regarding the validityof the hypothesized constructs. The five constructs included in this analysis wereinvolvement (2 items), similarity (3 items), source credibility (2 items), feelingstowards source (2 items), and attitude toward issue (2 items).First, in accordance with the procedure described by Bagozzi (1980), a GeneralReliability Model was tested using LISREL VI (Joreskog and Sorbom 1984). TheGeneral Reliability Model posits that the five true-scores (corresponding to the fivelatent variables) explain the entire pattern of relationship among all the observedmeasurements (Bagozzi 1980, p180.). The model is presented in Figure 4, and theparameters estimates are presented in Table 5. The results revealed that the modelprovided a good fit to the data. A X2 = 47.91 with df = 44 and p = 0.317 wasobtained.92Next an Equal Units of Measurement Model was estimated (see Figure 5 andparameter estimates in Table 5). This model posits that all observations have equalunits of measurement (Bagozzi 1980). This hypothesis was tested by assigning avalue of 1.0 to all As in the measurement model. The estimation of this modelresulted in a x2 = 59.95 with df = 51 and p = 0.18, indicating that there was areasonable fit.The Parallel Forms Model, which in addition to restricting the As to equal 1.0makes a further restrictive assumption which requires the error variances of indicatorsof common true scores are be equal (see Figure 6 and parameter estimates in Table5). The results indicated a further deterioration in fit (X2 = 73.64, df = 53, p = 0.08).Even though Model Two (Equal Unit of Measurement) produced a reasonablefit, the model was rejected because the difference in the goodness-of-fit testsbetween the first and second reliability models produces a X2 = 12.04 with df = 6 andp =0.05. Similarly the difference in goodness-of-fit tests between the Parallel Formand Equal Units Models produces a X2 = 13.69 with df =2 and p <0.005. Hence theparallel form model was also rejected. Thus there was adequate confirmation that themeasurements used in this study were internally consistent and acceptable.6.2 Validity AssessmentBoth convergent and discriminant validity can be assessed using the LISRELmodel by comparing the fit of rival confirmatory factor models. Convergent validityexamines if there's a great deal of commonality between measures of different93constructs, in other words are the hypothesized constructs really different from eachother. Discriminant validity is the extent to which measures of different constructsdiffer (Nunnally 1978). In accordance with the procedure described in Dillon (1986)the validity of the constructs was evaluated.6.2.1 Convergent ValidityFirst, as suggested by Dillon (1986) a Single-Factor Model was tested (seeFigure 7). A Confirmatory Factor Model with a single factor was tested to see ifmeasures of different constructs converged. The convergent validity of the single-factor model was rejected because the model provided a poor fit to the data [x 2 =423.23; df = 54; p < 0.001]. As against this, the multi-factor model (which is thesame as the General Reliability Model with true-scores) discussed earlier, provided avery good fit.6.2.2 Discriminant ValidityDiscriminant validity of the constructs was evaluated by examining the commonvariance between measures of different constructs (Dillon 1986). A high covariationwould lead to doubts about the uniqueness of the measures and/or constructs. Byexamining the Psi matrix (P), which is the factor correlation matrix, one could drawconclusions about the uniqueness of the measures/constructs. In order to determinethe discriminant validity, the Psi matrix from the multi-factor model (General ReliabilityModel) was examined. According to Dillon (1986), the correlations should be94"reasonably" smaller than 1.0 in order to conclude that there is discriminant validity.In this case all inter-factor correlations, except one, were below 0.50, yet weresufficiently larger than zero.The high inter-factor correlation (0.76) was between the factors ATTITUDE(i.e., preference for the issues) and FEELING (i.e., feelings toward the two sourcesexpressing two different viewpoints). Conceptually, however, the distinction betweenthese two constructs seems clear. It should be borne in mind that there is asignificant causal structure that exists among the factors (the structural equation -causal modelling - is explained in the next chapter), hence one would expect strong,but not too high, correlations between factors. Hence, given the conceptual clarityof the two constructs and the fact that the factors are causally related one may arguethat the discriminant validity of the measures is not violated.Using the method described in Dillon (1986), two contrasting models weretested to see if ATTITUDE (towards the issue) and FEELINGS (towards the source) aredifferent constructs or not. First a One-Factor Model was constructed, with all fourmeasures as indicators of a single construct. This model did not fit the data well [x 2= 5.84, df = 2, p = 0.05]. Next, a Two-Factor Model was tested by hypothesizingFEELINGS and ATTITUDE as two separate constructs. This model provided anexcellent fit to the data fx2 = 1.53, df =1, p = 0.217]. These results suggest that inspite of the high correlation between the latent variables FEELINGS and ATTITUDE,these are indeed distinct and separate constructs.956.3 MANIPULATION CHECKS 6.3.1 Involvement ManipulationFor the purpose of verifying the effectiveness of the involvement manipulation,the three involvement measures (concerned whether the joint venture is started ornot, whether joint venture is relevant to their life as a student of that university, andwhether they think the outcome of such a joint venture will have importantconsequences for the students of that university) were summed to form a singleinvolvement score. The high involvement condition (where the joint venture was tooccur in "their" university) was expected to get a higher overall involvement score,as compared to the low involvement condition (where the joint venture was to occurat "another" university). A t-test conducted to test the difference between the twogroups indicated that the involvement manipulation was indeed very effective[t(df = 70) = 4.71, p <0.001], with mean involvement being greater in the highinvolvement group (Mean = 17.55) as compared to the low involvement group(Mean = 13.19). 1In addition to this manipulation check, another t-test was conducted to confirmthe effectiveness of involvement manipulation. Studies have previously shown thatinvolvement contributes to greater confidence in one's attitudes (cf. Berger andMitchell 1989). Since level of confidence in expressed preference was one of thedependent variables, it provided an opportunity to test this effect. A t-test wasconducted to see if the high and low involvement subjects differed in their confidence1 The scales were recoded so that a higher score indicated greater involvement.96levels. The subjects were split into to two approximately equal groups (34 and 38subjects) at the mean value of the total involvement score and the differences in theirconfidence levels were examined through a t-test. While direction of the meansindicated that high involvement subjects were relatively more confident (Mean = 5.3)than the low involvement subjects (Mean = 4.7), the result was not significant[t(df = 70) = 1.84, p = 0.071. 26.3.2 Similarity ManipulationThe manipulation check measures were intended to capture the subjects'identification with the similar or dissimilar source not in relation to the persuasivecommunication they received from that source, but they were intended to capture theextent of identification with the source even outside the limited laboratory setting (seeAppendix 2 for measures). Such a measure allowed for verifying that the experimentused groups with real "histories" outside the experimental setting. Since each subjectwas exposed to a similar as well as a dissimilar source (one of whom was a minorityand the other a majority), the manipulation check measures were designed to seewhich of the two sources (students/similar or faculty/dissimilar) were perceived by thesubjects as being "similar" to them. For example, subjects were asked to indicatetheir agreement to the statement:2 The scale measuring confidence was recoded so that a higher score indicatedgreater confidence. The difference in confidence levels remained almost the samewhen the t-test was conducted using the two involvement conditions as the groups.97"My views on many issues are likely to be similar to that of the students,rather than the faculty who participated in the group discussion."(7-point scale; strongly agree - strongly disagree).Since the subjects' perception of the two sources based on a direct comparisonwas sought, a t-test is not an appropriate statistic to test the similarity manipulation.To conclude that the manipulation was successful, more people should agree with theabove statement. Two seven-point scales were combined and a chi-square test wasconducted to see if there were more people on the agreement side of the scale asopposed to the disagreement side.The null hypothesis would be that "there are an equal number of people in theagreement side as in the disagreement side of the scale". If this is rejected, it wouldindicate the effectiveness of the similarity manipulation. The numbers on either sideof the mid-point were collapsed and two categories were formed. 3 A chi-square testwas conducted to verify the hypothesis. The obtained chi-square [x 2(df = 61) = 10.09]was highly significant at p = 0.005. This result called for a rejection of the nullhypothesis. Since the chi-square could have been large because there were morepeople in one of the two categories (i.e., agree or disagree) than the other, therejection of the null hypothesis by itself does not indicate if the subjects showedgreater identification with a similar, rather than a dissimilar source.The frequencies in the two categories were further examined to see what ledto the rejection of the null hypothesis. It was found that only 30% of the subjects3 Eight subjects who were in the mid-point of the scale were dropped and only thesubjects who had indicated an agreement or a disagreement were used in this test.98were below the mid-point of the scale (meaning they did not identify with other fellowstudents more than they did with the faculty), while 70% of the subjects were at orabove the mid-point of the scale (meaning they were on the agreement end of thescale). This result indicates that the similarity manipulation was indeed successful.6.3.3 Minority-Majority Manipulation CheckOne of the last items on the dependent measures questionnaire was a questionasking the subjects to identify the exact number of students and faculty (in and outgroups) who supported and/or opposed the joint venture decision. The responsecategories provided were: Students Supporting (Opposing) Joint Venture and FacultyOpposing (Supporting) Joint Venture. After deleting the cases with missing values,the results indicated that 88% of the subjects had correctly identified the number ofin and outgroup members who supported and/or opposed the joint venture. Thus, itseems the minority-majority manipulation correctly registered in most subjects' minds.6.4 Hypothesis Tests: Similarity and InvolvementTo test for the main and interaction effects of Similarity and Involvement a two-way Analysis of Covariance (ANCOVA) with risk-aversion and gender as thecovariates. Since the minority advocated either a pro-joint venture position or an anti-joint venture position, it was necessary to determine if the position advocated by theminority in any way influenced the persuasiveness of the minority. The dependent99variable was the preference for the two opposing positions. A summation of twopreference measures was used as the dependent variable in the analysis. In theANCOVA both covariates were insignificant, with risk-aversion [F(1,66) = 3.09,p > 0.05] being stronger than gender [F(1,66) =1.54, p > 0.20].6.4.1 Main EffectsThe results indicate strong support for the similarity main effect [F(1,66) = 9.56,p < 0.005]. Thus, H1 is tenable, meaning that a similar minority is more persuasivethan a dissimilar minority. H4, the involvement main effect, was not supported[F(1,66) = 0.083, p > 0.75], meaning that the predicted effect of greater minorityinfluence under high involvement does not hold.6.4.2 Interaction EffectInvolvement does play a critical role in the minority influence process, asindicated by the strong interaction between involvement and similarity [F(1,66) = 4.16,p < 0.05]. Since a significant interaction by itself does not indicate if the nature ofthe interaction is in accordance with the ELM based hypotheses, an examination ofthe cell means was necessary.6.4.3 Comparison of MeansAs planned a comparison of means was undertaken. T-tests, using the Dunn(1961) multiple comparison method, were conducted. With Bonferroni adjustment an100a level of 0.01 was used for the mean comparisons. The t-tests compared the meansfor similar and dissimilar minority-sources at each of the two levels of involvement.The results indicated that the means for minority influence (for similar anddissimilar minority) were not significantly different under high involvement[t(34) = 0.73, p > 0.4], but were significantly different under low involvement[t(34) ---- 3.98, p < 0.001]. This is consistent with the interaction effect hypothesisbased on the Elaboration Likelihood Model (see Figure 8).6.5 Hypothesis Tests: Credibility and FeelingsAdditional analyses were conducted to determine the role of some of theproposed intervening variables. Hypotheses 2 and 3 were concerned with therelationship between source similarity on the one hand, and source credibility andsource feelings on the other hand. While these two hypotheses are not central toestablishing the minority influence effect, they do allow us to understand the minorityinfluence process better.6.5.1 Source CredibilityAccording to Hypothesis 2, a similar minority was supposed to be perceived asmore credible than a dissimilar minority. Even though minority status in a group, byitself leads to perception of diminished credibility (cf. Moscovici 1980), it wassuggested that a similar minority will be held in higher esteem due to the shared101"social identity" between the recipient and the source of the message. A test wasconducted with the two groups being similar and dissimilar minority and thedependent variable was source credibility. This measure was obtained by summingtwo measures of this construct (believe and credible). 4 The obtained result indicatedsupport for the hypothesis [t(70) = 2.13, p <0.05], with the similar minority(Mean = 7.6) being perceived as more credible than the dissimilar minority(Mean =8.6). 5The total sample was split into two groups along the mean credibility score,yielding two groups of approximately equal size. Using these two groups a t-test wasconducted to see if higher perceived credibility of the minority source was associatedwith higher minority influence. The result was affirmative [t(70) = 2.04, p <0.05].Thus credibility of the source did seem to have an effect on the extent of thatsource's influence. This finding concurs with Martin (Clark and Maass 1988b), whoalso found that higher perceived credibility was associated with higher minoritypersuasion.6.5.2 Source-related FeelingsSimilar to the credibility hypothesis, Hypothesis 3 suggested that a similarsource would generate more favourable feelings than a dissimilar source, even if the4 A lower score on the summed measure indicates higher source credibility.5 A lower score on the composite credibility measure indicates a higher level ofcredibility.102source had a minority status. Again, a t-test using similar and dissimilar minority asthe two groups was conducted, with source-related feelings serving as the dependentmeasure. This measure was a composite of two scales. 6 The results revealed strongsupport for this hypothesis [t(70) = 3.05, p < .005], with a similar minority(Mean =1.33) leading to more positive feelings than a dissimilar minority (Mean = -0.39).Similar to the analysis carried out with source credibility, the sample was splitinto two groups along the mean score on the source-related feelings, and a t-test wasconducted to see if those with favourable or unfavourable feelings toward the minoritydiffered in their acceptance of the minority's position. Again, the results stronglyconfirmed the existence of such an effect [t(70) = 6.19, p < 0.001].6.6 Mediating Effects It is not sufficient to set out the conditions leading to conformity effect or aminority influence effect, but it is also essential to understand the process underlyingsuch an effect. With the exception of a few studies which have examinedattributional variables like credibility (e.g., Clark and Maass 1988b; Martin 1988a), orthe subjects' cognitive responses (e.g., Maass and Clark 1983; Trost et al. 1992),6 Since each scale measured the attitude toward one source (either the majorityor the minority), a high score on one scale meant a low score on the other. Hence,before obtaining a composite score appropriate recoding was done. A high (positive)score indicates more favourable feelings towards the source, and a low (negative)score indicates the reverse.103there has not been a lot of effort towards understanding the role of mediatingvariables in the minority-majority influence process. In this research an importantattributional measure, i.e., source credibility, was measured. Furthermore, since theidentification with the source (i.e., through similarity) was considered essential inproducing the minority influence effect, subjects' feelings towards the source wasalso measured. As explained in Chapters 1 and 3, both source credibility and sourcefeelings were conceptualized as mediating variables. The rest of this section detailsthe path analysis undertaken to verify if these two variables played a mediating role.6.6.1 MethodologyIn testing mediating effects, marketing applications (e.g., Homer 1990;MacKenzie, Lutz and Belch 1986) have quite frequently used Joreskog's LISREL model(Joreskog and Sorbom 1984). Causal modelling or structural equations modelling hassome distinctive advantages: it considers the modest reliability of most observedmeasures and allows for incorporating latent or unobserved variables, and it alsoallows for incorporation of multiple measures used in tapping each construct (Dillonand Goldstein 1984). However, the LISREL model uses the Maximum LikelihoodEstimation procedure, and it is generally recommended that this technique be usedonly with large samples (Dillon and Goldstein 1984). Given the limited sample sizeof seventy-two subjects, it was felt that the causal modelling method would not beappropriate.In the marketing literature there are several instances where a multiple104regression procedure has been used to examine mediating effects (e.g., Batra and Ray1986). This procedure essentially involves path analyses without considering themeasurement errors. Keeping this weakness in mind, a sequential regression analysiswas undertaken to examine the existence of mediating effects, if any.The objective of this exercise was not to test the effects of the two treatmentvariables, which was earlier accomplished through the analysis of covariance, but toexamine the inter-relationships amongst the dependent measures. Hence, thetreatment variables Similarity and Involvement were excluded from the analysis.Instead, felt similarity and felt involvement (denoted as SIMILARITY-F andINVOLVEMENT-F respectively henceforth) were used. Other variables entered into theregression equations were source credibility, source feelings and attitude toward theissue of joint venture (all these three were self-reported measures). Since there weremultiple measures for each construct (see earlier discussion in section 6.1), anaverage score of the measures was obtained for each construct. These compositemeasures were used in the regression estimation.The regression equations were estimated using the OLS procedure in threephases. First, the effects of SIMILARITY-F and INVOLVEMENT-F on each of theproposed mediating variables (i.e., CREDIBILITY and FEELINGS) was examined to seeif the two predictor variables significantly affected these criterion variables. If theseresults were not significant, further use of CREDIBILITY and FEELINGS as mediatingvariables could not be justified. Second, the relationship between the mediatingvariables was examined. Here CREDIBILITY was used as one of the predictors and105FEELINGS was the criterion variable. The reason behind this was the expectation thatFEELINGS (towards the source) will not only be influenced by the extent ofidentification with that source, but also the extent of perceived credibility of thesource (the variable CREDIBILITY) will have an effect on FEELINGS, with a credibleand believable source producing more favourable feelings.' Third and last, a seriesof regression models were examined to see if CREDIBILITY and FEELINGS played amediating role in determining the extent of minority-majority influence (represented bythe variable ATTITUDE).6.6.2 CREDIBILITY and FEELINGS as Criterion VariablesThe effects of SIMILARITY-F and INVOLVEMENT-F on CREDIBILITY were testedfirst (Model 1 in Table 7). In accordance with the expectation, both predictorvariables had significant betas, with SIMILARITY-F being the stronger (beta =0.499)than INVOLVEMENT-F (beta = 0.202). An R 2 of 0.328 was obtained, indicating anacceptable fit.Model 2 (Table 7) used SIMILARITY-F and INVOLVEMENT-F as predictors ofFEELINGS. As anticipated, SIMILARITY-F had a significant beta (0.441), however,INVOLVEMENT-F was not significant (beta = 0.036). But the model still provided arespectable R 2 of 0.2027 Such a hypothesis has support from past research which indicates that theperceptions regarding the source's experience or expertise (which is a indicator ofsource credibility) is liked to the extent of liking expressed for that source, such thatthe greater the perceived expertise or credibility, the more favourable the liking orfeeling (Feick and Higie 1992).106Finally, the possibility of CREDIBILITY of the source having an effect on theFEELINGS towards the source was also examined (Model 3 in Table 7). This modelused SIMILARITY-F, INVOLVEMENT-F and CREDIBILITY as the predictors. In thismodel only CREDIBILITY had a significant beta (0.52), while the predictive power ofSIMILARITY-F declined (beta =0.17, t-value ns), meaning that CREDIBILITY andSIMILARITY-F strongly covaried. With an R2 =0.366, the model offered a significantfit.Models 2 and 3 were compared to see if the addition of CREDIBILITYsignificantly added to the explanatory power. The gain in R 2 from the addition ofCREDIBILITY was 81 percent. 8 Using the procedure described in Dillon and Goldstein(1984; p231-34), the null hypothesis that CREDIBILITY does not add significantly tothe explained variance was tested. 9 The null hypothesis was rejected, meaning thatCREDIBILITY of the source has an important role in predicting the source-relatedFEELINGS [F(1,69) = 17.84, p <0.001].Thus, there is confirmation that the hypothesized mediating variables8 Gain in R 2 = (difference in R 2 between Models 1 and 2)/(R2 of Model 2)9 The F-statistic for the model comparison is computed as follows:(R2 p-R 2)F- (Mg) (1-R2 )(n-p)where F has df of (p-1),(n-p); p and q are the numberof parameters in the two models.107(CREDIBILITY and FEELINGS) do have strong relationships with SIMILARITY-F and/orINVOLVEMENT-F. Having established that these two main predictor variables explaina good proportion of the variance in the two mediating variables, and one of thehypothesized mediators (CREDIBILITY) is an important predictor of the other mediator(FEELINGS), the next stage involved testing of models involving ATTITUDE.6.6.3 ATTITUDE as the Criterion VariableThe next set of models reported in Table 8, used ATTITUDE as the criterionvariable. A sequential approach was used, and first only the two primary predictorvariables, SIMILARITY-F and INVOLVEMENT-F, Model 4 was estimated. The resultsindicated that SIMILARITY-F was a significant predictor of ATTITUDE (beta =0.44),whereas INVOLVEMENT-F was not (beta=-0.13). Overall, the model provided asignificant fit (R 2 = 0.218).Model 5 was tested by adding CREDIBILITY, a hypothesized mediator, into theequation. This model produced an R2 of 0.274, giving a 25.7% increase in theexplained variance. The F-statistic (see footnote 7) was significant [F(1,69) = 5.32,P < 0.051. Again, SIMILARITY-F was the only statistically significant predictor. Eventhough CREDIBILITY did not have a significant beta (beta = 0.255), it seemed to addsignificantly to the variance explained, and hence warranted retention. 1°The final tested (Model 6) involved the addition of FEELINGS to the equation.10 In a similar case, Batra and Ray (1986) have argued that testing for themagnitude of gain in the R 2 could provide a better insight regarding the significanceof the variable(s) added to the equation.108This model showed a very substantial improvement in the variance explained withR2 = 0.518, where only FEELINGS was a significant predictor (beta = 0.492). Againthe shared variance between the predictor variables accounted for the insignificanceof variables that were previously significant. The percentage increase in R 2 betweenModels 5 and 6 was 89%. The test for model comparison (in footnote 7) yielded anF(1,68) = 34.42 (p <0.001). The substantial increase in R2 between Models 5 and 6is very likely attributable to multicollinearity between the independent variablesCREDIBILITY and FEELINGS."From Models 4, 5 and 6, it is evident that CREDIBILITY and FEELINGS do playan important mediating role in determining the ATTITUDE. The R 2 obtained by usingonly the two primary predictors, SIMILARITY-F and INVOLVEMENT-F, improvessubstantially by incorporating the mediating variables. Thus, the hypotheses that thetwo source-related variables (CREDIBILITY and FEELINGS) will play an importantmediating role in determining the extent of minority-majority influence are supported.Further discussion of the regression results is deferred to Chapter VII.6.6.4 Test of SIMILARITY-F and INVOLVEMENT-F InteractionIn the regression analyses described in the last two sections, the interactionterm between SIMILARITY-F and INVOLVEMENT-F was omitted. However, the" Batra and Ray (1986), who encountered a similar problem, have suggested thatin cases the significance of the increase in R2 between two successive models is morereliable than the individual betas. When multicollinearity exists the values of betastend to unstable (cf. Howell 1982), and hence the contribution of each independentvariable is not emphasized here.109existence of such an interaction effect was verified. The ANCOVA reported earlierin this chapter did indicate a significant interaction effect between the treatmentvariables similarity and involvement. The regression model using SIMILARITY-F andINVOLVEMENT-F, along with a product term (representing the interaction) wastesting, with ATTITUDE as the criterion variable. The interaction term was notsignificant (p > 0.15). When CREDIBILITY and FEELINGS were used as the criterionvariable, results along the same lines were obtained.Several factors need to be borne in mind here. Firstly, the variablesSIMILARITY-F and INVOLVEMENT-F represent "felt or perceived similarity" and "feltinvolvement" respectively. Hence they are not the same as the treatment variablesused in the ANCOVA. Secondly, the tests for main and interaction effects of thetreatment variables was conducted with two covariates, which were not used herein the regression models. These two points may account for the lack of significanceof the interaction effect. Since the objective of this modelling exercise was touncover mediation effects, if any, and not to test the interaction effect, this issue isnot pursued further.6.7 Additional AnalysesFirst, using a demographic variable, i.e., ethnicity, an investigation was carriedput to see whether there was greater (or lesser) minority influence in certaindemographic categories than in others. This variable was not treated as covariate in110the main ANCOVA because there was no a priori reason for doing so. There were 6categories provided for ethnicity based on knowledge of the demographic profile ofthe sample. The six categories were: Anglo-Canadian, Canadian of Eastern EuropeanOrigin, French-Canadian, Chinese-Canadian, lndo-Canadian, and Other. The responsewas made optional because: (i) in earlier pretests many subjects had difficultyaccurately naming their ethnicity, and (ii) some subjects felt uneasy about thisquestion. As a result, a total of 22 subjects did not respond to this question. Withthe remaining subjects, a one-way ANOVA was conducted and the result indicatedno significant effect.Next, subjects' perceptions of the quality of the arguments were examined.First, there was no significant difference in the perception regarding consistency andnovelty of the arguments proposed by those advocating a pro or an anti joint-ventureposition. 12 When the same measure of argument quality was used to see if the highand low involvement subjects perceived a difference in argument quality, there wasa indication that the perception of argument quality was slightly higher under highinvolvement (Mean = 7.3) as compared to low involvement (Mean = 8.02), but theresult was not significant [t(70) =1.31, p = .195, two-tailed test]. It is plausible thatthe under high involvement subjects paid more attention to the message and hencewere more aware of the quality of the arguments.A final test carried out to see to what extent the subjects were aware of the12 This analysis was carried out after appropriately recoding the two scalesmeasuring novelty and consistency of arguments, and then summing them into asingle measure.111specific technology (i.e., fuzzy logic) used in the experimental setting. Subjects wereasked to indicate if they had previously heard of this technology. 77.8% said theyhad not heard of the technology, and the remaining subjects indicating someawareness of this technology.6.8 Summary First all the manipulation checks were significant in accordance with earlierpredictions indicating that the manipulations were effective. The Similarity XInvolvement ANCOVA (with Minority Position and Risk-aversion) produced significantresults as anticipated, except for the main effect of the involvement manipulation.But the results confirm that involvement plays a crucial role in determining whenconformity or minority influence will occur. Further, two source-related variables,CREDIBILITY (an attributional type variable) and feeling (emotional reaction to thesource) played a significant mediating role in determining the extent of minorityinfluence. Finally, other extraneous variables such as demographic characteristics(age and ethnicity) were insignificant, thereby enhancing the confidence in the validityof the experimental design and the observed results.112VII. CONCLUSION7.0 Overview This chapter will first summarize the key issues this research set out to addressalong with the results obtained. This is followed by a thorough discussion of theresults and the implications of the research. Then, the potential contributions of thisthesis to the marketing and social psychology literature are addressed, followed by adiscussion of the limitations of this research. Finally, several ideas for future researchare expounded.7.1 Summary of Major Issues A primary objective of this research was to examine the conditions under whichconformity effect (or majority influence) versus minority influence effect occurred.Given that issue-involvement is a major factor in persuasive settings (cf. Johnson andEagly 1989), this research was particularly interested in understanding the role ofissue-involvement in the minority-majority influence process. Since similarity of theminority (operationalized through varying social identity of the minority) is known tolead to enhanced minority influence (e.g., Clark and Maass 1988a; Martin 1988a), thisresearch set out to examine if the role of similarity, a peripheral cue in the ELMparlance, was dependent on the level of the subject's involvement.113The main experiment, described in Chapter IV, predicted that source-similaritywill have a main effect, meaning that a similar minority would be more persuasivethan a dissimilar minority. In other words, greater conformity effect will occur onlywhen the minority consists of dissimilar individuals. Also, a main effect forinvolvement was predicted, where higher involvement will diminish the importance ofperipheral cues (such as status of source or number of people supporting eachargument) and hence facilitate greater minority influence. Finally, an interaction effectbetween similarity and involvement was hypothesized, indicating that source-similarityof the minority will enhance persuasion only under low involvement.7.2 Summary of Major Results 7. 2. 1 MeasurementThe five measured variables were: felt similarity (SIMILARITY-F), feltinvolvement (INVOLVEMENT-F), credibility of the source (CREDIBILITY),feelings toward the source (FEELINGS) and attitude toward the position advocated bythe minority or the degree of minority influence (ATTITUDE). Each construct wasmeasured with multiple-item scales. The scales were internally consistent andexhibited good discriminant and convergent validity as well.1147.2.2 Effectiveness of ManipulationsSimilarity was manipulated by varying the social identity of the source (studentsversus faculty), and as anticipated an overwhelming majority of the subjectsidentifying with a similar source (students) to a dissimilar source (Faculty). Issue-involvement was manipulated by varying the personal consequences of the problemto the subject by having the problem occur in the subjects' own university or atanother university. In line with prior expectations, subjects under high involvementreported greater personal relevance towards the issue. Finally, a large number ofsubjects were able to correctly identify the number of individuals in the minority andmajority groups.7.2.3 HypothesesThe three main hypotheses were the main effects for both source similarity andissue-involvement, and the interaction effect between these two variables. Two ofthe three hypotheses were supported. A similar minority was more persuasive thana dissimilar minority. This main effect, however, was moderated by the significantinteraction between involvement and similarity, with a similar minority being morepersuasive than a dissimilar minority only in low involvement conditions. Under highinvolvement there was no significant difference between the similar and dissimilarminority sources. The hypothesis relating to issue-involvement, which predictedgreater minority influence under high involvement, was not tenable. In this entireanalysis, the effects of risk-aversion and gender were statistically controlled.115It was expected that a similar minority would be perceived as more credible andgenerate more favourable feelings than a dissimilar minority. Both these expectationswere met. T-tests conducted to examine the effect of credibility on minority influencerevealed that a minority perceived as more credible was capable of greater persuasion.Along the same lines, higher favourability of the feelings towards the minority sourceresulted in higher persuasion.Regression analyses conducted to examine the effects of the two hypothesizedmediating variables, credibility and feelings, indicated that these two variables did playa significant mediating role. The overall predictive power increased substantially wheneach of these variables was entered into the equation along with the two primarypredictor variables, felt similarity and felt involvement. Thus the perceived credibilityof the source and feelings (favourable/unfavourable) towards the source are importantfactors in determining the extent of its influence in a minority-majority context.7.3 Discussion Based on Moscovici's Theory of Conversion Behavior (1980), many researchershad found a minority to be more persuasive under a private, rather than a publicresponse condition. In this research, the response-condition (or response-involvement)was held constant by having only private responses. Also, a simultaneous socialinfluence paradigm (e.g., Clark and Maass 1988a) was used, where each individualreceived conflicting persuasive messages from two different sources, with each116message having a high or a low social support. Let us consider the major findings ingreater detail.7.3.1 SimilarityAt the very outset it was contended that in consumer settings, individuals willoften not only know which opinions have more support, but will also be aware of thecharacteristics of the sources expressing these opinions. Hence, a source-characteristic variable, similarity, was manipulated. The result was consistent withthe social comparison literature which indicates that information from a similar "other"(who shared similar values and tastes) is more relevant than information from adissimilar source (cf. Goethals and Darley 1977; Moschis 1976). The manipulationcheck for similarity clearly supported that the contention that the subjects (who wereuniversity students) felt they shared more common beliefs, tastes and values withother students (similar source) than the faculty members (dissimilar source).Researchers reviewing the literature in the field of minority-majority influencehave pointed out using "minimal" groups with no history or no future, is unlikely tocontribute much to the understanding of complexities involved in conformity effectsversus minority influence (see Chaiken and Stangor 1987). In some of the pastresearch the social categories of the subjects have questionable relation to thepersuasive message being examined (e.g., Martin 1988a) and other studies have failedto report any manipulation checks (e.g., Perez and Mugny 1987) or have reportedmanipulation checks that do not capture the extent of subject's identification with the117similar or dissimilar source (e.g., Clark and Maass 1988a). The last mentioned studydid a manipulation check to see if a minority was perceived as more consistent thana majority, but did not report any manipulation checks to confirm that there wasgreater identification with an ingroup or similar minority (heterosexual) rather thanwith the outgroup or dissimilar individual (homosexual).In this research, as mentioned earlier, only private responses were obtained.However, some of the past research (e.g., Perez and Mugny 1987) has used indirectattitude measures (i.e., attitude on an issue related to the main issue discussed) as asubstitute for private measures.' Others have examined the similarity/dissimilarityof the minority when the minority is advocating a pro or anti-zeitgeist position (e.g.,Clark and Maass 1988a). Furthermore, only a few studies examining the effects ofsimilarity have used a simultaneous social influence paradigm (or a multiple sourcesetting) as in this research (e.g., Perez and Mugny 1987), while others (e.g., Martin1988a) have compared similar and dissimilar minorities without any reference to amajority source, in other words the subjects received a persuasive message only froma minority source.The limitations in past research (e.g., failure to provide appropriate manipulationchecks) as well as these differences between present and past research limit theextent of comparison possible. One study which also used a simultaneous social1 As Maass and Clark (1984) pointed out, different types of private acceptancemeasures have been used previously. Some studies use private responses (whichprovide confidentiality to the subject) similar to the one used in this research, and yetother have used indirect attitude measures to capture private acceptance. It is feltthat these measures are distinct.118influence paradigm, found that the a similar minority was more influential than adissimilar majority (Perez and Mugny 1987). Given the consistency of the resultsobtained in the present research with the theoretical predictions, the similarity effectcan be accepted with confidence.7.3.2 InvolvementSince personal consequences become a primary concern under highinvolvement, it was felt that under high involvement the motive to make thesepersonal consequences favourable to oneself would clash with the self-presentationaland other motives induced under conformity settings. In such a clash of motives itmight be reasonable to expect that the best alternative available will be chosen,regardless of whether it comes from a minority or a majority. This is in accordancewith the Elaboration Likelihood Model (Petty and Cacioppo 1986), which suggeststhat attention to source-characteristics will diminish under high involvement. Underlow involvement, due to lower interest in the issue, the status of the source (i.e.,majority) might serve as an important peripheral cue shaping the attitude. Theseexpectations were not supported by the data, and there was essentially no differencebetween subjects in the high and low involvement conditions. At this stage one couldspeculate as to why such a result might have occurred. In addition to using the sizeof the source (minority or majority) as a peripheral cue, another very visible peripheralcue used was the degree of source similarity. Under low involvement, it is possiblethat the most visible or most vivid peripheral cue may have been used by the subjects.119In this case, source similarity may have been a more vivid (or perhaps even important)peripheral cue than minority or majority status.Having said this, it should also be noted that while the low involvementsubjects were swayed by the minority's similarity (in reference to the interactioneffect), the high involvement subjects not only did not seem to pay much attentionto source-similarity, but were also relatively impartial between the minority and themajority (as indicated by their a low minority influence score of 0.30, which is closeto a neutral point between the minority and the majority). This suggests that perhapsthe high involvement subjects, in accordance with the ELM, may have ignored allperipheral cues. Cognitive response data may help to further untangle this result.Trost et al. (1992), who had collected cognitive response data, reported that aminority's position evoked resistance under high involvement, and this effect wasreversed under low involvement. 2 The findings of this research suggest that whenindividuals were provided a basis for strong social identification, under highinvolvement a moderate minority influence occurs, while under low involvement a veryhigh level of minority influence can occur, if the minority possesses similarcharacteristics. The differing objectives and differing methodologies limit furthercomparisons between these two studies at this point. However, a continued2 A methodological difference between Trost et al. (1992) and the presentresearch should be noted. Trost et al. (1992) had exposed subjects only to a majorityor a minority, rather than both. Also, they had the minority and the majority advocateonly a counter-attitudinal position. In this research, a simultaneous social influencemodel was adopted with the subjects being exposed to both the majority and theminority opinions. Further both the minority and the majority presented argumentssupporting or opposing the attitudinal issue.120exploration of the role of issue-involvement in producing conformity versus minorityinfluence is likely to be fruitful.7.3.3 Mediating EffectsInitially t-tests revealed that greater minority influence occurred under higherperceived credibility of the minority and also when the feelings towards the minoritywas more favourable. Path analyses conducted to examine the mediating effects didindicate that these two variables played an important mediating role in the minority-majority influence process.The results of the path analyses suggest that SIMILARITY-F by itself had asignificant impact on reported FEELINGS, with a similar source generating morefavourable feelings. This is consistent with past research (e.g., Feick and Higie1992). Felt involvement did not affect FEELINGS directly. Both SIMILARITY-F andINVOLVEMENT-F significantly predicted CREDIBILITY. Thus, the two primarypredictor variables, SIMILARITY-F and INVOLVEMENT-F, had significant paths to thehypothesized mediating variables. Further, the second set of regression models testedwith ATTITUDE as the criterion variable indicated that the addition of bothCREDIBILITY and FEELINGS in incremental steps, significantly increased the R 2 eachtime. When all the four variables - SIMILARITY-F, INVOLVEMENT-F, CREDIBILITY andFEELINGS - were entered into the regression equation with ATTITUDE as the criterionvariable, only FEELINGS was significant, which meant any effect of CREDIBILITY onATTITUDE was further mediated through FEELINGS (see Figure 9).121While the role of source credibility has been identified previously (Clark andMaass 1988b), FEELINGS towards the source has not been examined in a similarcontext. The results of this research indicate that in addition to the credibility of thesource, the feelings towards the source, which is a related but distinct construct,plays a significant role in determining the minority-majority influence.7.3.4 Support for TheoryOverall, the results provide confirmation for the simultaneous social influenceparadigm. Previous findings indicating that an ingroup or similar minority is morepersuasive than an outgroup or dissimilar minority under private measures (Perez andMugny 1987) is supported here. However, this research indicates that the role ofsimilarity depends on the level of involvement, with similarity having an impact onlyunder low involvement. Regarding the role of involvement, the initial expectation wasthat under high involvement the group pressure might be diminished because ofincreased awareness about the personal consequences of the issue, and this in turnmight facilitate greater acceptance of the minority position. The lack of influence ofeither peripheral cue (similarity and minority-majority) on the high involvement subjectsindicates that they may have engaged strictly in a central process. This does suggestthat issue-involvement may diminish the conformity effect. Clearly, the role ofinvolvement in a group influence context needs further investigation.1227.4 Limitations of Research7.4.1 Generalizability and External ValidityIn this research the attitudinal issues examined, in the pretests as well as in thefinal study, were of a policy nature. In contexts input from consumer groups andother interest groups normally do shape the policies. Since this research wasconducted within such a limited context, at this point it may not be judicious toextrapolate these findings to all consumer group contexts. Along the lines of a recentconsumer behaviour paper on conformity effects (Rose, Bearden and Teel 1992),extending the study of minority influence and the simultaneous social influenceparadigm (or the multiple sources model) to other consumption situations is essential.The generalizability is also limited by the specific nature of the involvement andsimilarity manipulations as well.Does the group influence process created in the experiment match closely theprocess that one might observe outside the laboratory? While face-to-face groupsprovide greater realism, this research has used a group discussion summary in lieu ofa real group. Many scholars in this area have done this in the past (e.g., Clark andMaass 1988a; Martin 1988a; Perez and Mugny 1987; Trost, Maass and Kenrick1992) and others have argued that face-to-face groups are not necessary to createinterdependent relationships or social influence (Rabbie and Horwitz 1988). Thus,there is strong support for the methodology used. However, lack of personalknowledge about the sources as well as limited exposure to group influence do limit123the external validity. There is a need for extending this type of research to morenatural settings.Another issue that might have some bearing on external validity is the natureof the sample used. Will the results obtained from a student sample extend to othergroups? The study did use a sample consisting of both genders and different ethnicgroup, which does increase the external validity.On the whole, given that this research had the extension and testing of a theoryas its objective, the concerns regarding internal validity weighed more heavily. Sucha position has support from marketing scholars (Calder, Philips and Tybout 1983).7.4.2 Design and Experimental ExecutionThe experimental design involved comparisons between a similar minority anda dissimilar majority or vice-versa. The use of a control group or an experimentalcondition where both the majority and minority share the same social identity (eithersimilar or dissimilar to the subject), as done by Perez and Mugny (1987), wouldprovide greater insight. For instance, the role of issue-involvement could be clarifiedfurther if this factor was explored in the presence of only one source-related cue (i.e.,majority-minority sources) as opposed to the two source-related cues used in thisresearch (see Trost et al. 1992 for a such a design).In terms of the execution of the experiment, the data were collected in a largeclassroom with over 70 students. Data could not be collected in small groups due tofinancial and time considerations, as well as the difficulty of getting a sufficiently large124sample within a specific time. However, the collection of the entire data in onesetting did ensure that the "environmental" factors occurring during the experimentwere identical for all subjects, and this does help strengthen the internal validity.Extensive instructions were provided to all students who consented to participate inthe study to ensure there was no communication between subjects. At this stagethere is no reason to suspect that collection of data in a large class setting may haveaffected the result.7.4.3 MeasuresWhile the measures, in general, showed good internal consistency as well asdiscriminant and convergent validity, cognitive response data were not examined, ashas been done in some recent research (Trost, Maass and Kenrick 1992). Thecognitive response data may shed more light on the effect of involvement. Eventhough, the generally consistent results obtained in this research do lend support tothe Elaboration Likelihood Model, a thorough understanding of the subjects' thoughtprocess could be very useful.7.5 Implications of the Research7.5.1 Theoretical ImplicationsThis research complements the recent efforts in marketing aimed at125understanding the social influence phenomena (e.g., Bearden and Rose 1990; Rose;Bearden and Teel 1992). While these recent efforts have involved the examinationof conformity effect as well the personal characteristics which enhance or diminishsuch an effect, this research offers a simultaneous social influence model thatincorporates influence attempts by both the minority and the majority sources. Asempirically demonstrated, such a model not only allows for conformity effects tooccur, but will also allows deviant or minority opinions to be persuasive. Given thatconsumers are very likely to be exposed to multiple opinions with varying levels ofsupport, such a model is likely to further advance our understanding of consumerbehaviour.The social psychology literature has incorporated the social identity or similarityof the source as an important variable in such a social influence context. However,the role of issue-involvement, known to be an important variable in persuasion, indetermining the extent of minority versus majority influence had not been studiedpreviously. Using the Elaboration Likelihood Model as the framework, this researchhas made an important advance in bridging these two streams of research. Eventhough the results did not support the prediction of enhanced minority influence underhigh involvement, involvement did seem to affect the nature and degree of socialinfluence through its interaction with similarity. In terms of similarity effects onminority influence, the interaction effect obtained in this research also contributes toa greater understanding of the social identification effect that has been known toincrease minority influence. This research has made an important, but small step, and126further research on the role of involvement is essential.7.5.2 Managerial ImplicationsThis research has important implications for managers as well. As mentionedat the outset, early adoption of many innovations is mostly confined to a minority.Within social groups, one may find that there are one or two individuals who arewilling to stray away from the group norm. Understanding the process through whichsuch individuals influence other groups members can be of great significance tomarketers.In this research, the minority was not portrayed as an "expert" or "opinionleader". Yet the minority was persuasive under the right conditions when thesimilarity was high and involvement low. This may mean that adoption of newproducts that are not very highly involving, such as many continuous innovations(Assael 1992), could be enhanced by increasing source-recipient similarity in thecommunication efforts. This is consistent with the homophily effect in diffusionresearch (Rogers 1983). Further, since the results indicate a low level of socialinfluence (both in terms of minority-majority influence and in terms of the effect ofsource similarity) for high involvement subjects, increasing the involvement in aninnovative product or idea may facilitate an independent decision with minimalconsideration to social influence factors. At this stage one has to be cautious aboutsuggesting specific marketing applications based on the results obtained. Furtherresearch using specific products and more realistic settings is needed.1277.6 Future Research It is essential to further study the role of issue-involvement in the socialinfluence process. It may be worthwhile examining the effects of involvement in theabsence of similarity (as in Trost et al. 1992), while retaining the two sources as awithin-subjects factor. To fully understand the contribution of issue-involvement, theuse of cognitive response measures may be essential.The work of some social psychologists suggests further avenues of research,which may be very productive for marketers (Nemeth and Wachtler 1983; Tanfordand Penrod 1984). Nemeth (1985) has argued that operationalizing social influencein terms of movement toward the position advocated by one or the other source isrestrictive, and has instead examined how the existence of a minority opinion in agroup can diffuse the tension (conformity pressure) in the group leading to morecreative decisions (Nemeth 1985; Nemeth and Kwan 1985; Nemeth and Wachtler1983). This line of enquiry suggests that in groups with dominant majorities,creativity is stifled and consequently the involvement of individuals is diminished,whereas when a group has dissenting viewpoints individuals may feel less pressurefrom the majority and may actively search for a solution. Nemeth's work hasimportant implications for consumer behaviour. While there is evidence suggestingconflict in consumer groups (e.g., Belch, Belch and Sciglimpaglia 1980), it seemsconflict created by minority may lead to consideration of new alternatives andsolutions leading to better consumer decisions (cf. Folkes and Kiesler 1991).Currently the work on how the presence of minority leads to creative decisions is128based mostly on perceptual stimuli. It seems that this theory can be readily tested inconsumer behaviour settings.In this research an individual's reaction to a majority as well as a minority wasconsidered. Another possibility is to examine the majority's reaction to a minority,and vice-versa. Such a conflict between the majority and the minority can be set-upby using confederates (e.g., Moscovici and Lage 1976). The issue here is whethera conformist (who has accepted the majority's opinion) will adopt a less popularposition, and if so, when might this happen? This issue has relevance to marketersbecause social changes (including adoption of new products) have started as aminority view and then have been accepted by the majority, who may have beensceptical or disdainful in the beginning. Hence understanding the role of minority inmodifying the attitudes and behaviours of conformists is important.This study examined the conflict in motives occurring due to personalinvolvement versus group pressure. The response-involvement variable was omittedafter the pilot test. In future an examination of the clash between issue and response--involvement may be undertaken. But a thorough understanding of the role of issue-involvement may be necessary before venturing in that direction.As mentioned earlier, future research should attempt to incorporate morecommon consumer situations involving specific products or services. It may beinteresting to see what kind of product choices (e.g., public versus private products)are influenced by the majority and the minority. For instance, will the public productsbe more influenced by the majority (similar to greater majority influence in public129measures)? Will the minority be more influential for privately consumed products?If such effects are observed then marketers will gain more insight into promotingpublic versus private products.Finally, some of the methodological limitations identified earlier will have to beaddressed in future research. The use of control groups as well as face-to-facegroups will have to be considered. Additional measures including cognitive responsemeasures as well as other mediating variables (for instance attributional variables)should be incorporated to enhance our understanding of minority influence. Overall,this research has opened several interesting avenues for future research.130REFERENCESAsch, Solomon E. (1951), "Effects of Group Pressure upon Modification andDistortion of Judgements," in H. Guetzkow (ed.), Groups, Leadership and Men,Pittsburgh, PA: Carnegie Press, 177-190.Assael, Henry (1987), Consumer Behavior and Marketing Action, 3rd Edition, Boston,Mass: Kent Publishing Co. p372.Bagozzi, Richard P. (1980), "Causal Models in Marketing." John Wiley and Sons,New York, NY.Bagozzi, Richard P., and Youjae Yi (1989), "On the Use of Structural Equation Modelsin Experimental Designs," Journal of Marketing Research, 26 (3), 271-284.Batra, Rajeev and Michael L. Ray (1986), "Affective Responses Mediating theAcceptance of Advertising," Journal of Consumer Research,  13 (September),234-249.Bearden, William 0., and Randall L. Rose (1990), "Attention to Social ComparisonInformation: An Individual Difference Factor Affecting Consumer Conformity,"Journal of Consumer Research, 16 (4), 461-471.Bearden, William 0. Richard G. Netemeyer and Jesse E. Teel (1989), "Measurementof Consumer Susceptibility to Interpersonal Influence," Journal of ConsumerResearch, 15 (4), 473-481.Belch, Michael A., George E. Belch and Donald Sciglimpaglia (1980), "Conflict andFamily Decision Making: An Exploratory Investigation," in J. C. Olson (ed.),Advances in Consumer Research, 7, Ann Arbor, MI: Association for ConsumerResearch, 476.Berger, Ida E.. and Andrew A. Mitchell (1989), "The Effect of Advertising on AttitudeAccessibility, Attitude Confidence, and the Attitude-Behaviour Relationship,"Journal of Consumer Research, 16 (3), 269-279.Berscheid, Ellen (1966), "Opinion Change and Communicator-Communicatee Similarityand Dissimilarity," Journal of Personality and Social Psychology,  4, 670-680.Bray, Robert M., Dennis Johnson and John T. Chilstrom Jr. (1982), "Social Influenceby Group Members with Minority Opinions: A Comparison of Hollander andMoscovici," Journal of Personality and Social Psychology, 43 (1), 78-88.131Brehm, Jack W. (1966), Theory of Psychological Reactance, New York: AcademicPress.Brown, Jacqueline and Peter H. Reingen (1987), "Social Ties and Word-of-MouthReferral Behavior," Journal of Consumer Research, 14, 350-362.Calder, Bobby J., Lynn W. Philips and Alice M. Tybout (1983), "Beyond ExternalValidity," Journal of Consumer Research,  10 (1), 112-114.Campbell, Donald T., and Julian C. Stanley (1963), "Experimental and Quasi-Experimental Designs for Research," Houghton Mifflin and Company, Dallas,TX.Campbell, Jennifer D. and Patricia J. Fairey (1989), "Informational and NormativeRoutes to Conformity: The Effect of Faction Size as a Function of NormExtremity and Attention to the Stimulus," Journal of Personality and Social Psychology, 57 (3), 457-468.Chaiken, Shelly and Charles Stangor (1987), "Attitude and Attitude Change," AnnualReview of Psychology, 38, 575-630.Churchill, Gilbert (1979), "A Paradigm for Developing Better Measures of MarketingConstructs," Journal of Marketing Research, 16 (1), 64-73.Cialdini, R. B., Levy, A., Herman, C. P., Kozlowski, L. T., Petty, R. L. (1976),"Elastic Shifts of Opinion: Determinants of Direction and Durability," Journal ofPersonality and Social Psychology,  34, 663-672.Clark, Richard D., and Anne Maass (1988a), "Social Categorization in MinorityInfluence: The Case of Homosexuality," European Journal of Social Psychology,18 (5), 347-364.^, and Anne Maass (1988b), "The Role of Social Categorization and PerceivedSource Credibility in Minority Influence," European Journal of Social Psychology, 18 (5), 381-394.Cocanougher, A. Benton and Grady D. Bruce (1971), "Socially Distant ReferenceGroups and Consumer Aspirations, "Journal of Marketing Research,  8, 371-381.Coleman, James S., Elihu Katz and Herbert Menzel (1966), "Medical Innovation: ADiffusion Study," Indianapolis, IN: Bobbs-Merrill.Davis, Leslie L. (1984), "Judgement Ambiguity, Self-consciousness, and Conformityin Judgement of Fashionability," Psychological Reports, 54 (2), 671-675.132Deutsch, M and H. B. Gerard (1955), "A Study of Normative and Informational SocialInfluence upon Individual Judgement, "Journal of Abnormal and Social Psychology, 51, 629-636.Dholakia, Ruby and Brian Sternthal (1977), "Highly Credible Sources: PersuasiveFacilitators or Persuasive Liabilities," Journal of Consumer Research, 3 (March),223-232.Dillon, William R. (1986), "Building Consumer Behaviour Models with LISREL: Issuesin Applications," in Perspectives on Methodology in Consumer Research, eds.David Brinberg and Richard J. Lutz. Springer-Verlag, New York, 107-154.Dillon, William R., and Matthew Goldstein (1984), "Multivariate Analysis: Methodsand Applications," John Wiley and Sons, New York, 430-489.Dunn, 0. J. (1961), "Multiple Comparisons among Means," Journal of AmericanStatistical Association, 56, 52-64.Fagley, N. S., and Paul M. Miller (1990), "Effect of Framing on Choice: Interactionswith Risk-taking Propensity, Cognitive Style and Sex," Personality and Social Psychology Bulletin, 16 (3), 496-510.Fazio, Russel H. (1979), "Motives for Social Comparison: The Construction-ValidationDistinction," Journal of Personality and Social Psychology, 37 (10), 1683-1698.Feick, Lawrence and Robin A. Higie (1992), "The Effects of Preference Heterogenietyand Source Characteristics on Ad Processing and Judgements aboutEndorsers," Journal of Advertising, 21 (2), 9-24.Fitzpatrick A. R., and Alice H. Eagly (1981), "Anticipatory Belief Polarization as aFunction of Expertise of a Discussion Partner," Personality and Social Psychology Bulletin, 7, 636-642.Eagly, Alice and Carole Chrvala (1986), "Sex Differences in Conformity: Status andGender Role Interpretations," Psychology of Women Quarterly, 10 (3), 203-220.Festinger, Leon (1954), "A Theory of Social Comparison Processes," HumanRelations, 7, 117-140.Folkes, Valerie and Tina Kiesler (1991), "Social Cognition: Consumers' Inferencesabout the Self and Others," in T. R. Robertson and H. H. Kassarjian (eds.),Handbook of Consumer Behavior, Englewood Cliffs, NJ: Prentice-Hall.133Gardner, Meryl P., Andrew A. Mitchell and J. Edward Russo (1985), "LowInvolvement Strategies for Processing Advertisements," Journal of Advertising,14(2), 4-12.Goethals, George. R. and John M. Darley (1977), "Social Comparison Theory: AnAttributional Approach," In J.M.Suls and R.L. Miller (eds.), Social ComparisonProcesses: Theoretical and Empirical Perspectives, New York: Wiley, 259-278.^, and R. Nelson (1973), "Similarity in the Influence Process: The Belief-ValueDistinction," Journal of Personality and Social Psychology, 25, 117-122.Grubb, Edward L. and Bruce L. Stern (1971), "Self-Concept and Significant Others,"Journal of Marketing Research, 8, 382-385.Halifax Chronicle-Herald (1990), "Polls Shows Sunday Shopping Unpopular in theEast." A5.Hamilton, Mark A., John E. Hunter and Michael Burgoon (1990), "An Empirical Testof an Axiomatic Model of the Relationship between Language Intensity andPersuasion," Journal of Languages and Social Psychology, 9 (4), 235-255.Hansen, Flemming (1969) "Primary Group Influence and Consumer Conformity," inP. R. McDonald (ed.,), Proceedings of the American Marketing Association's Educator's Conference. Chicago: American Marketing Association, 1970, 300-305.Harkins, Stephen G. and Richard E. Petty (1987), "Information Utility and the MultipleSource Effect," Journal of Personality and Social Psychology, Vol.52(4),260-268.Hartley, R. E. (1968), "Norm Compatibility, Norm Preference, and the Acceptance ofNew Reference Groups," in H. H. Hyman and E. Singer (eds.), Readings in Reference Group Theory and Research, New York: Free Press, 247-256.Heaven, Peter C. (1986), "Correlates of Conformity in Three Cultures,"Personality and Individual Differences, 7 (6), 883-887.Hollander, Edwin Paul (1958), "Conformity, Status, and Idiosyncrasy Credit,"Psychological Review, 65, 117-127.^ (1961), "Some Effects of Perceived Status on Responses to InnovativeBehaviour," Journal of Abnormal and Social Psychology, 63, 247-250.(1964), "Leaders, Groups, and Influence." Oxford University Press,134New York.Hovland and Weiss (1951), "The Effect of Source Credibility on CommunicationEffectiveness," Public Opinion Quarterly, 15, 635-650.Howell, David C., (1982), "Statistical Methods for Psychology." Duxbury Press,Boston, Massachusetts, p.441.Johnson, Blair T., and Alice H. Eagly (1989), "The Effects of Involvement onPersuasion: A Meta-Analysis," 106 (2), 290-314.Joreskog, Karl G., and Dag Sorbom (1984), "LISREL VI: Analysis of Linear StructuralRelationships by Maximum Likelihood." Scientific Software, Mooresville, IN.Kardes, Frank (1988), "Spontaneous Inference Processes in Advertising: The Effectsof Conclusion omission and Involvement on Persuasion," Journal of ConsumerResearch, 15 (2), 225-233.Kelly, Harold (1972), "Causal Schemata and the Attribution Process." GeneralLearning Press, Morristown, NJ.Kelman, Herbert C. (1961), "Processes of Opinion Change," Public Opinion Quarterly,25, 57-78.^ , and Carl I. Hovland (1953), "Reinstatement of the Communicator in DelayedMeasurement of Opinion," Journal of Abnormal Social Psychology, 48 (July),327-335.Kruglanski Arie W., and Ofra Maylseless (1987), "Motivational Effects in the SocialComparison of Opinions," Journal of Personality and Social Psychology, 53 (5),834-842.Krugman, Herbert E. (1965), "The Impact of Television Advertising: Learning WithoutInvolvement," Public Opinion Quarterly, 29, 349-356.Lalancette, Marie-France and Lionel G. Standing, (1990), "Asch Fails Again," SocialBehavior and Personality, 18 (1), 7-12.Larsen, Knud S. (1990), "The Asch Conformity Experiment: Replication andTranshistorical Comparisons," Journal of Social Behavior and Personality,(Special Issue), 5 (4), 163-168.Latane, Bibb (1981), "Psychology of Social Impact," American Sociologist, 36 (4),343-356.135Latane, Bibb and Sharon Wolf (1981), "Social Impact of Majorities and Minorities,"Psychological Review, 88, 438-453.Leippe, Michael R. and Roger A. Elkin (1987), "When Motives Clash: IssueInvolvement and Response Involvement as Determinants of Persuasion," Journalof Personality and Social Psychology, Vol.52(2), 269-278.Maass, Anne and Russell D. Clark III (1983), "Internalization Versus Compliance:Differential Processes Underlying Minority Influence and Conformity," EuropeanJournal of Social Psychology, 13, 197-215.^ and Russell D. Clark III (1984), "Hidden Impact of Minorities: Fifteen Years ofMinority Influence Research," Psychological Bulletin, 95 (3), p428-50.Mackie, Diane M. (1987), "Systematic and Nonsystematic Processing of Majority andMinority Persuasive Communications," Journal of Personality and Social Psychology, 53 (1), 41-52.Martin, Robin (1988a), "Ingroup and Outgroup Minorities: Differential Impact uponPublic and Private Responses," European Journal of Social Psychology, Vol.18,39-52^, (1988b), "Minority Influence and Social Categorization: A Replication,"European Journal of Social Psychology, Vol.18, 369-373.McGill, Andrew R. (1990), "The Influences of Complexity and Conformity inOrganizational Performance: An Exploratory Field Study," UnpublishedDissertation.McGinnies (1973), "Initial Attitude, Source Credibility and Involvement as Factors inPersuasion," Journal of Experimental and Social Psychology, 9, 285-276.Mongeau, Paul A. and Rick Garlick (1988), Social Comparisons and PersuasiveArguments as Determinants of Group Polarization," Communication ResearchReports, 5 (2), 120-125.Moore, David J. and Richard Reardon (1987), "Source Magnification: The Role ofMultiple Sources in the Processing of Advertising Appeals," Journal ofMarketing Research, 24(4), 412-417.Moschis, Goerge P. (1976), "Social Comparison and Informal Group Influence,"Journal of Marketing Research, 13, 237-244.136^, and Gilbert A. Churchill Jr., (1977), "Mass-Media and Interpersonal Influenceson Adolescent Consumer Learning," in B. Greenberg and D. Bellenger (ed.),Educators' Proceedings, Chicago: American Marketing Association, 68-71.Moscovici, Serge (1976), "Social Influence and Social Change,"^EuropeanMonographs in Social Psychology 10 (Series Editor: Henri Tajfel), AcademicPress, London.^ (1980), "Toward a Theory of Conversion Behavior," in L. Berkowitz (ed.),Advances in Experimental Social Psychology, 13, 209-239.^ and C. Faucheux (1972), "Social Influence, Conformity Bias, and the Studyof Active Minorities," in L. Berkowitz (ed.), Advances in Experimental SocialPsychology, 6, New York: Academic Press, 149-202.^, Gabriel Mugny and Eddy Van Avermaet (eds.), "Perspectives on MinorityInfluence," Cambridge and New York: Cambridge University Press, 1985.^ and E. Lage (1976), "Studies in Social Influence III: Majority versus MinorityInfluence in a Group," European Journal of Social Psychology, 6 (2), 149-174.^, E. Lage, and M. Naffrechoux (1969), "Influence of a Consistent Minority onthe Responses of a Majority in a Color Perception Task," Sociometry, 32, 365-379.^ and Charlan Nemeth. (1974), "Social Influence: II. Minority Influence," in C.Nemeth (ed.), Social Psychology: Classic and Contemporary Integrations,Chicago; Rand McNally, 217-249.^ and Bernard Personnaz (1980), "Studies in Social Influence: V. MinorityInfluence and Conversion Behavior in a Perceptual Task," Journal ofExperimental Social Psychology, 16 (3), 270-282.Mugny, Gabriel (1975), "Negotiations, Image of the Other and the Process of MinorityInfluence," European Journal of Social Psychology, 5 (2), 209-228.^ (1982), "The Power of Minorities,"^European Monographs in SocialPsychology, 31 (Series Editor: Henri Tajfel), Academic Press, London.^ and Stamos Papastamou (1980), "When Rigidity Does Not Fail:Individualization and Psychologization as Resistance to the Diffusion of MinorityInnovations," European Journal of Social Psychology, 10, 43-61.Nemeth, Charlan J. (1985), "Influence Processes, Problem Solving and Creativity,"137in Social Influence: The Ontario Symposium,^ and Julianne L. Kwan (1985), "Originality of Word Associations as a Functionof Majority vs. Minority Influence," Social Psychology Quarterly, 48 (3),277-282.^, M. Swedlund and B. Kanki (1976), "Patterning of Minority Responses andtheir Influence on the Majority," European Journal of Social Psychology,4, 53-64.^ and Joel Wachtler (1973), "Consistency and Modification of Judgement,"Journal of Experimental Social Psychology, 9, 65-79.^ and Joel Watchler (1983), "Creative Problem Solving as a Result of Majorityand Minority Influence," European Journal of Social Psychology, 13, 45-55.^, Joel Watchler and J. Endicott (1977), "Increasing the Size of Minority:Some Gains and Some Losses," European Journal of Social Psychology, 1,11-23.Nunnally, Jum C. (1978), "Psychometric Theory." McGraw-Hill, New York, 225-255.O'Brien, Marc S. and William B. Bankston (1984), "The Moral Career of the ReformedCompulsive Eater: A Study of Conversion to Charismatic Conformity," DeviantBehavior, 5 (1-4), 141-150.O'Guinn, Thomas C. and Ronald J. Faber (1985), "New Perspectives on Acculturation:The Relationship of General and Role Specific Acculturation with Hispanics'Consumer Attitudes," in E. C. Hirschman and M. B. Holbrook (ed.,), Advancesin Consumer Research, Provo, UT: Association for Consumer Research,113-117.Paicheler, G (1976), "Norms and Attitude Change:1. Polarization and Styles ofBehavior," European Journal of Social Psychology, 6, 405-427.Park, C. Whan and V. Parker Lessig (1977), "Students and Housewives: Differencesin Susceptibility to Reference Group Influences," Journal of ConsumerResearch, 4, 102-110.Perez, Juan A., and Gabriel Mugny (1987), "Paradoxical Effects of Categorization inMinority Influence: When Being an Outgroup is an Advantage," EuropeanJournal of Social Psychology, 17, 157-169.Petty, Richard E., and John T. Cacioppo (1979), "Issue-Involvement Can Increase or138Decrease Persuasion by Enhancing Message-Relevant Cognitive Responses,"Journal of Personality and Social Psychology, 37 (10), 1915-1926.^, and John T. Cacioppo (1986), "The Elaboration Likelihood Model ofPersuasion," In L. Berkowitz (ed.), Advances in Experimental and Social Psychology, 19, Orlando, FL: Academic Press, 122-205.^, John T. Cacioppo and David Schumann (1983), "Central and PeripheralRoutes to Advertising Effectiveness: The Moderating Role of Involvement,"Journal of Consumer Research, 10, 135-146.Rabbie, Jacob M., and Murray Horwitz (1988), "Categories versus Groups asExplanatory Concepts in Intergroup Relations." European Journal of Psychology,Vol. 18, 117-123.Rogers, Everett M. (1983), "Diffusion of Innovations." Free Press, New York.Rose, Randall L., William 0. Bearden and Jesse E. Teel (1992), "An AttributionalAnalysis of Resistence to Group Pressure Regarding Illicit Drug and AlcoholConsumption," Journal of Consumer Research, 19, 1-13.Roskos-Ewoldsen, David R., and Russel H. Fazio (1992), "On Orienting Value ofAttitudes: Attitude Accessibility as a Determinant of an Object's Attraction ofVisual Attention," Journal of Personality and Social Psychology, 63 (2), 198-211.Ross, Lee, Gunter Bierbrauer and Susan Hoffman (1976), "The Role of AttributionProcesses in Conformity and Dissent," American Psychologist, 31, 148-157.Savoni, Anuradha (1989), "A Comparative Study of Alcoholics and NonalcoholicsAlong the Parameters of Conformity and Self-confidence," Indian Journal ofApplied Psychology, 26(2), 23-29.Schiffman, Leon H. and Leslie L. Kanuk (1978), Consumer Behavior, Englewood Cliffs,NJ: Prentice-Hall.Sherif, Muzafer (1953), "The concept of Reference Groups in Human Relations," inM. Sherif and M. 0. Wilson (eds.), Group Relations at the Crossroads, NewYork: Harper & Row.^ and Carl I. Hovland (1961) Social Judgement: Assimilation and ContrastEffects in Communication and Attitude Change, New Haven, CT: YaleUniversity Press.139Stafford, James (1966), "Effects of Group Influence on Consumer Brand ChoicePreference," Journal of Marketing Research, 3, 68-75.Stang, David J. and Lawrence S. Wrightsman (1981), "Dictionary of Social Behaviorand Social Research Methods," Monterey, CA: Brooks/Cole Pub. Co.Tajfel, Henri (1972), "Le Categorization Sociale," in Introduction a la PsychologieSociale, ed. S. Moscovici, Larousee, Paris.^ (1982), "Instrumentality, Identity and Social Comparison." In H. Tajfel (ed.),Social Identity and Intergroup Relations. Cambridge University Press,New York.^ and John C. Turner (1979), "An Integrative Theory of Intergroup Conflict,"in The Social Psychology of Intergroup Relations, eds. H. Wiegman andS. Worchel. Brooks/Cole, Moneterey, CA.Tanford, Sarah and Steven Penrod (1984), "Social Influence Model: A FormalIntegration of Research on Majority and Minority Influence Processes,"Psychological Bulletin, 95 (2), 189-225.Troldahl, V. C. and F. A. Powell (1965), "A Short-form Dogmatism Scale for Use inField Studies," Social Forces, 44, 211-215.Turner, John C. (1982), "Towards a Cognitive Redefinition of the Social Group," inSocial Identity and Intergroup Relations, ed. H. Tajfel, Cambridge UniversityPress.^ (1987), "Comments on Diose's lndiridual and Social Identities in IntergroupRelations," European Journal of Social Psychology, 18, 113-116.Venkatesan, M (1966), "Experimental Study of Consumer Behaviour Conformity andIndependence," Journal of Marketing Research, 3, 384-387.Wallendorf, Melanie and Michael D. Reilly (1983), "Ethnic Migration, Assimilation, andConsumption," Journal of Consumer Research, 10 (3), 292-302.Wilson Glenn D. and Anathony E. Reading (1989), "Pelvic Shape, Gender RoleConformity and Sexual Satisfaction," Personality and Individual Differences,10(5) 577-579.Witt, Robert E. (1969), "Informal Social Group Influence on Consumer Brand Choice,"Journal of Marketing Research, 6, 473-477.140^ and Grady D. Bruce (1970), "Purchase Decisions and Group Influence,"Journal of Marketing Research, 7, 533-535.141APPENDIX 1PILOT STUDY: TREATMENTS, STIMULUS MATERIAL AND MEASURES142Key to Appendix 1 1. This Appendix contains the treatments, stimulus materialand dependent measures used in the pilot study reported inChapter V. The treatments and measures were given to subjectsin separate booklets - labelled Booklet One and Two.2. To facilitate easier understanding, the treatments used for low andhigh issue-involvement as well as low and high response-involvementare presented in successive pages in this appendix. For the reader'sbenefit these pages are labelled appropriately.143(COMMON TITLE SHEET FOR ALL TREATMENTS)Faculty of Commerce and Business AdministrationUniversity of British ColumbiaBOOKLET ONEMARCH 1992144(HIGH ISSUE-INVOLVEMENT TREATMENT)* The University Administration is faced with an important managerialdecision regarding a specific student service.* In order to make an effective decision, the University is seeking input fromstudents in a few randomly selected classes. Your class is among the onesselected.* Initially a group discussion on this same managerial problem was held inFebruary 1992. You will read a summary of this group discussion. Then,you will be asked to CHOOSE between TWO alternative decisions.* All opinions expressed by participants will be summarized and presented to theUniversity Administration. Thus, you will be providing useful input for adecision which will affect a specific student service on campus.* To show our appreciation for your effort and time, we will automatically enteryour name in a lottery. There are three cash prices in this lottery:$100, $75 and $50.Important Note: Please DO NOT communicate with anyone until this study iscompleted. Please follow ALL instructions carefully.PLEASE PROCEED TO THE NEXT PAGE.145(HIGH ISSUE-INVOLVEMENT TREATMENT)Introduction:* A study commissioned by the Administration to evaluate campus services hasindicated that given the substantial increase in student population in recentyears a new cafeteria will have to be established by the end of 1992 (DavidLam Research Centre is a possible location).* At this stage the University Administration has to decide whether thisproposed cafeteria should be run by the University or by aprofessional food service company.* To aid the University in making this decision a random sample of studentsparticipated in a group discussion.^This group discussion, held inFebruary 1992, focused on the following:"Should a new student cafeteria to be established on Campus (by Dec 1992) be run by theUniversity or by a professional food service company?"Please Note: There are two private companies interested in opening a cafeteria oncampus. Both companies have unionized employees. Thus, in the event ofa strike (by any of the unions on campus), the operations of the privately runcafeteria will also be affected.PLEASE PROCEED TO THE NEXT PAGE.146(LOW ISSUE-INVOLVEMENT TREATMENT)* We would like you to participate in a study that attempts to understanddecision making styles of individuals.* Our research shows that it is possible to acquire more reliable data ondecision making styles if people are given a PRACTICE TASK first.* Hence, before you actually do the main task, we would like you to gothrough a simple PRACTICE TASK , to familiarize you with the "rules of thegame." Booklets One and Two contain the practice task.Important Note: Please DO NOT communicate with anyone until both studies arecompleted. Please follow ALL instructions carefully.PLEASE PROCEED TO THE NEXT PAGE.147(LOW ISSUE-INVOLVEMENT TREATMENT)Practice TaskIntroduction:* From the projections of increase in student enrolments at UBC, it is anticipatedthat a new student cafeteria (in addition to the existing ones) will be requiredby the year 1996-97.* The University Administration has to decide whether this proposed cafeteria shouldbe run by the University or by a professional food service company.• To aid the University in making this decision a random sample of studentsparticipated in a group discussion.^This group discussion, held inFebruary 1992, focused on the following:"Should a new student cafeteria to be established on Campus (in 1996-97) berun by the University or by a professional food service company?"Please Note: There are two private companies interested in opening a cafeteria oncampus. Both companies have unionized employees. Thus, in the event of a strike(by any of the unions on campus), the operations of the privately run cafeteria willalso be affected.PLEASE PROCEED TO THE NEXT PAGE.148(HIGH RESPONSE-INVOLVEMENT TREATMENT)Your Task:* First, you will read a summary of the group discussion, which is provided in thenext page. You will also find a brief background of each group member. Wehave used only the initials of the participants in order to protect their identitySome of the participants were Commerce students from various sections ofCommerce 396 (some of them could be your classmates).* After reading the summary of the group discussion, you will express youropinions in a questionnaire. Later, we will ask you to PUBLICLY state youropinion in front of your classmates.PLEASE PROCEED TO THE NEXT PAGE.149(LOW RESPONSE-INVOLVEMENT TREATMENT)Your Task:* First, you will read a summary of the group discussion, which is provided in thenext page. You will also find a brief background of each group member. Wehave used only the initials of the participants in order to protect their identity.Some of the participants were Commerce students from various sections ofCommerce 396 (some of them could be your classmates).* After reading the summary of the group discussion, you will express youropinions in a questionnaire. Your responses will be STRICTLYCONFIDENTIAL and will not be revealed to anyone. Your name oridentity is not required. Please feel free to express your views.PLEASE PROCEED TO THE NEXT PAGE.150(STIMULUS - order and position advocated were randomized; regular font size was used, and not as presented here)Summary of Group DiscussionEach participant in the group discussion provided arguments favouring one of the two alternatives. After three rounds ofdiscussion, a consensus was not reached. The fmal arguments made by each participant are presented below:Supporting the University^ Supporting the Professional Co.AS (COMMERCE undergrad at UBC):An outside food service company will notre-invest its profits in the University tothe benefit of its target consumers(i.e., students).LR (COMM/ECON undergrad at UBC):The University has run the UBC Food Servicesfor many years. It has a lot of experiencein operating cafeterias.PR (COMMERCE undergrad at UBC):Since the University's mandate is not to makea profit, prices will be lower as compared toa privately run service.KG (COMMERCE undergrad at UBC):I agree that the University should run the newcafeteria.RJ (COMMERCE undergrad at UBC):If the University manages the new cafeteria,students' suggestions will get prompter responses.EC (COMM/ECON undergrad at UBC):A University run cafeteria will provide morecampus job opportunities to students.JK (exchange student, Europe, no major):A professional company can offer more variety.They are likely to be more responsive to consumerneeds. I think they will provide a more courteousservice.PA (Research Associate in Engineering):A professional food service company will be ableto lower the prices due to higher volume of business.They are likely to make the best use of resources andminimize waste.PLEASE CLOSE BOOKLET ONE AND PROCEED TO BOOKLET TWO.PLEASE DO NOT RETURN TO BOOKLET ONE.151(COMMON COVER PAGE FOR ALL TREATMENTS)Faculty of Commerce and Business AdministrationUniversity of British ColumbiaBOOKLET TWOMARCH 19921521. Please Circle the Appropriate Number:I think the University should run the new cafeteria.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think the professional company should run the new cafeteria.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeHow confident are you that you have made the right judgement?Very confident 1 2 3 4 5 6 7 Not at all confidentPLEASE GO TO THE NEXT PAGE. DO NOT RETURN TO THIS PAGE1532. I think the professional company should run the cafeteria: YES ^ NO ^3. Please Circle the Appropriate Number:If the cafeteria is run by the University, students' suggestions will get prompterresponses.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeA professional company will be able to offer greater variety of food items.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeA professional company will not re-invest its profits in the University for the benefit of its targetconsumers (i.e., students).Strongly agree 1 2 3 4 5 6 7 Strongly disagreeAs compared to the University, a professional company will make better use ofresources.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeA University-run cafeteria will not be profit-oriented, hence prices will be lower.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeA professional company can lower the prices due to higher volume of business.Strongly agree 1 2 3 4 5 6 7 Strongly disagreePLEASE PROCEED TO THE NEXT PAGE. DO NOT RETURN TO THIS PAGE.1544. Please rate the university and the professional company on the following factors, usinga five-point scale (5 = very good; 4 = good; 3 = average; 2 = poor; 1 = very poor):UNIVERSITY^PROFESSIONALPriceFood VarietyEfficient use ofResourcesReceptive toConsumer needsOther (specify)5. Please state the number of people in the group discussion who supported eachalternative:University ^ Professional Company^6. Please Circle the Appropriate Number:In the group discussion, the persons supporting the University were more similar to me ascompared to those supporting the professional company.Strongly agree 1 2 3 4 5 6 7 Strongly disagreePLEASE PROCEED TO THE NEXT PAGE. DO NOT RETURN TO THIS PAGE.1557. Please Circle the Appropriate Number:While reading the group discussion summary, I was:Paying a lot of attention 1 2 3 4 5 6 7 Paying very little attentionConcentrating very hard 1 2 3 4 5 6 7 Concentrating very littleI carefully considered the opinions on both alternatives before forming my ownopinions.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeIn the group discussion, the persons supporting the professional company were morebelievable than those supporting the University.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeIn the group discussion, the persons supporting the University gave more usefulinformation than those supporting the professional company.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeIn the group discussion, the persons supporting the professional company expressed morenovel (or original) opinions than those supporting the University.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think the group members with backgrounds similar to mine talked about issues thatI myself would have considered.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think group members with backgrounds similar to mine provided me a better basis forevaluating my opinions on this topic, as compared to those with dissimilar backgrounds.Strongly agree 1 2 3 4 5 6 7 Strongly disagreePLEASE PROCEED TO THE NEXT PAGE. DO NOT RETURN TO THIS PAGE.156Please provide the following personal information. This information will be used for academicresearch purposes only. This information is strictly confidential.1. AGE:2. SEX:^MALE^FEMALE3. ETHNICITY (define it the way you want):4. How would you define your political views:left-oriented/liberalneither left nor right-oriented ^ right-oriented/conservativenot sureWere you instructed earlier (in Booklet One) that you will be asked to state your opinionsPUBLICLY in front of your classmates :YES  ^NO ^ NOT SURE ^cut hereIn order to show our appreciation for your effort and time, we would like to enter your namein a lottery, which consists of three cash prices: $100, $75 and $50. Please provide your nameand contact phone number. Only winners will be contacted on or before April 30, 1992.NAME ^STUDENT#^PHONE# (valid until April 30, 1992) THANK YOU FOR YOUR TIME AND COOPERATION.157APPENDIX 2MAIN EXPERIMENT: TREATMENTS, STIMULUS AND MEASURES158Key for Appendix 2:1. This appendix contains the treatments, stimulus material andmeasures for the main experiment described in Chapter IV.The treatments and measures were presented separately intwo different Booklets labelled One and Two.2. The Booklets One and Two which contain the title "AdministrationPolicy Survey" were given to high involvement subjects, andthe Booklets with the title "Decision Making Study" weregiven to the low involvement subjects. The questionnairescontained in Booklet two were identical for both involvementgroups, except for the title of the questionniare booklet.3.^In the following pages, the different treatments are labelledfor easy identification.159(HIGH INVOLVEMENT COVER PAGE)SIMON FRASER UNIVERSITYADMINISTRATION POLICY SURVEYBOOKLET ONE160(LOW INVOLVEMENT COVER PAGE)SIMON FRASER UNIVERSITYDECSION MAKING STUDYBOOKLET ONE161(HIGH INVOLVEMENT MANIPULATION)INTRODUCTIONWe would like you to participate in a survey that will provide the SFU Administration usefulinput on an important decision. This decision will directly or indirectly affect every studentin SFU. Hence your input is important.BackgroundRecently, a team of SFU scientists (to be called "inventors" henceforth) made a significantscientific breakthrough that has commercial potential. The new technology is based on thefuzzy logic theory. It has potential applications in many fields such as process controlequipment, robotics and other areas of industrial automation.How can the University best benefit from this invention? At SFU there is no policy thataddresses this question. In light of this situation, the inventors have made an offer to theUniversity. This offer requires SFU to participate in a joint venture with a private firm.The University is required to invest $2 million dollars which will enable the inventors topurchase additional equipment needed to carry out ongoing research in this area. The privatehi-tech firm will invest $8 million and will undertake the manufacturing and marketing ofthe technology. The University can remain a silent partner and will not directly participatein the commercial venture, and thus will not violate the University's Charter. The inventors,who will provide the technology for this venture, the University and the private firm willeach hold shares in the new venture, and will share any profits or losses. Since thetechnology is in a ready-to-market stage, the commercial success or failure of this venturewill become apparent within a year.In order to make a sound decision, SFU hired a management consultant to access the marketpotential, competition etc. The consultant's report provided useful information in accessingthe feasibility of the offer made by the inventors. In addition, the University will also beusing input from students and faculty in the decision process. The final decision will bemade shortly.SFU Administration Policy SurveyAs a first step towards getting input from students and faculty, the University recently askeda group of students and faculty to discuss the merits and demerits of the joint ventureproposal. We would like you to read a summary of this group discussion, and then expressyour views on the joint venture. Your input can shape the University's decision in thismatter.PLEASE PROCEED TO THE NEXT PAGE.162(LOW INVOLVEMENT MANIPULATION)INTRODUCTIONWe would like you to participate in a study on decision making. Our focus is onunderstanding differences in decision making process across people. To facilitate this task,we will now present you with an actual decision problem faced by the McMaster Universityadministration.Problem at McMasterRecently, a team of McMaster University scientists (to be called "inventors" henceforth)made a significant scientific breakthrough that has commercial potential. The newtechnology is based on the fuzzy logic theory. It has potential applications in many fieldssuch as process control equipment, robotics and other areas of industrial automation.How can the University best benefit from this invention? At McMaster University there isno policy that addresses this question. In light of this situation, the inventors have made anoffer to the University. This offer requires McMaster University to participate in a jointventure with a private firm. McMaster University is required to invest $2 million dollarswhich will enable the inventors to purchase additional equipment needed to carry out ongoingresearch in this area. The private hi-tech firm will invest $8 million and will undertake themanufacturing and marketing of the technology. The University can remain a silent partnerand will not directly participate in the commercial venture, and thus will not violate theUniversity's Charter. The inventors, who will provide the technology for this venture, theUniversity and the private firm will each hold shares in the new venture, and will share anyprofits or losses. Since the technology is in a ready-to-market stage, the commercial successor failure of this venture will become apparent within a year.In order to make a sound decision, McMaster University hired a management consultant toaccess the market potential, competition etc. The consultant's report provided usefulinformation in accessing the feasibility of the offer made by the inventors. The Universitywill also be using input from students and faculty in the decision process. A final decisionwill be made shortly.Decision Making StudyIn the first phase of our ongoing study on decision making, we presented the McMasterdecision problem along with a summary of the consultant's report (obtained from McMasterUniversity) to a group of SFU students and faculty. We asked them to engage in a groupdiscussion, focusing on the merits and demerits of the joint venture proposal at McMaster.We would like you to read a brief summary of this group discussion by SFU students andSFU faculty. Naturally, your opinions on this problem will not affect the final decision atMcMaster University.PLEASE PROCEED TO THE NEXT PAGE.163(MINORITY - SIMILAR; MAJORITY - DISSIMILAR)Group DiscussionThe group discussion was held in June 1992 to discuss obtain preliminary feedback fromstudents and faculty. In the group discussion, both the student and faculty participants wererandomly selected to ensure that their opinions represented their respective groups. Thediscussion group was made up of 12 participants - 8 were faculty members and 4 werestudents. The average age of the faculty members was 43.5 years and the average age ofthe students was 22 years. Each group discussion participant was aware of the findings ofthe management consultant. The group session lasted one hour. The opinions of thestudents as well as the faculty carried equal weight in the discussion. We would like youto read a brief summary of this group discussion.PLEASE PROCEED TO THE NEXT PAGE.164(MINORITY - DISSIMILAR; MAJORITY - SIMILAR)Group DiscussionThe group discussion was held in June 1992 to discuss obtain preliminary feedback fromstudents and faculty. In the group discussion, both the student and faculty participants wererandomly selected to ensure that their opinions represented their respective groups. Thediscussion group was made up of 12 participants - 8 were students and 4 were facultymembers. The average age of the faculty members was 43.5 years and the average age ofthe students was 22 years. Each group discussion participant was aware of the findings ofthe independent consultant. The group session lasted one hour. The opinions of the studentsas well as the faculty carried equal weight in the discussion. We would like you to read abrief summary of this group discussion.PLEASE PROCEED TO THE NEXT PAGE.165(ORDER and POSITION ADVOCATED by each source was randomized)Discussion Summary The hour long group discussion ended without a consensus as to whether the Universityshould enter this joint venture or not. Interestingly, all four of the students seemed veryenthusiastic about this venture, while all the eight faculty members opposed this idea. Sincethe interests of students and faculty members seemed to differ and both sides maintainedtheir positions consistently, we have summarized the different opinions separately for yourconvenience.Supporters of the new venture: Four students. Summary of their arguments:The University has not previously benefited from the inventions of its faculty. This jointventure gives the University an additional source of income, and profits from this venturecan be used in upgrading many campus services such as library and computer labs. Thisproject can create additional campus jobs for the students, for instance the University's shareof the profits can be used to create summer jobs and Research Assistantships. This jointventure will bring the University closer to industry, which is important for making Canadianindustry globally competitive. Involvement in this hi-tech venture will increase the prestigeof the University as a major player in a state-of-the-art technology. In the long run, thisventure is likely to trigger more industry support for the research activities on campus andwill generate more research grants. According to the independent consultant, this technologyis significantly superior to what available in the market currently, which is a distinctivecompetitive advantage. On the whole, there are more advantages than disadvantages to theUniversity's involvement in this venture.Opposers of the new venture: Eight Faculty members. Summary of their arguments:Given reduced government funding to Universities and the economic downturn, how can theUniversity raise $2 million? The University may have to cut back on other services toprovide capital for this venture. The consultant's report mentions that several major hi-techcompanies are doing their own in-house R&D in this area, which means heavy competitionis likely. Since the total investment in this project is not very large, the joint venturepartners cannot maintain a technological lead over big competitors. If the venture incurslosses, the University will have to recoup the money through other sources such as tuitionhikes, which will affect the students. If this joint venture is established, in future theresearch activities on campus will be dictated by the industry, and the University will notonly lose its independence but will also fail in its mission to carry out basic research. Theprofessors engaged in this venture will spend more time in this venture and less time withthe students, hence quality of education will be affected. While the University should benefitfrom this invention, a joint venture is not the answer.PLEASE CLOSE BOOKLET ONE. DO NOT RETURN TO BOOKLET ONEPLEASE PROCEED TO BOOKLET TWO.166(HIGH INVOLVEMENT CONDITION - QUESTIONNAIRE COVER PAGE)SIMON FRASER UNIVERSITYADMINISTRATION POLICY SURVEYBOOKLET TWO167(LOW INVOLVEMENT QUESTIONNAIRE COVER PAGE)DECISION MAKING STUDYBOOKLET ONE168PLEASE READ THE FOLLOWING INSTRUCTIONS:1. This booklet contains a series of questions. Please answer ALL questions.2. If there are specific instructions at the beginning of a question or at the endof a page, please follow them carefully.3. We are interested in your first impressions/reactions to the questions. Pleasetry to record your actual feelings/thoughts.4. All responses will be kept strictly confidential. Please DO NOT identifyyourself by name or any other means.PLEASE PROCEED TO THE NEXT PAGE.1691. Please Circle the Number that Closely Corresponds to Your Feelings/Opinions: I support the idea of the University being involved in this joint venture.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think the University should not get involved in this joint venture.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think the proposed joint venture is too risky.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think the benefits of the proposed joint venture far outweigh the risks.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeHow confident are you that your judgement on this issue is right?Very confident 1 2 3 4 5 6 7 Not at all confidentAfter reading the discussion summary, I have sympathy for the side with which I disagree.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI carefully considered the opinions of all group members before forming my ownopinion.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeAs an SFU student, I am concerned about whether the joint venture is started or not.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeThe opposers of joint venture presented more believable arguments than the supporters.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI think in the group discussion, the supporters of the venture were less credible than theopposers.170Strongly agree 1 2 3 4 5 6 7 Strongly disagreeWhether the joint venture is actually started or not, is not relevant to my life as a student atSFU.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeMy feelings towards the persons supporting the venture:Very favourable 1 2 3 4 5 6 7 Very unfavourableMy feelings towards the persons opposing the venture:Very favourable 1 2 3 4 5 6 7 Very unfavourableThe arguments of the supporters of the venture were more consistent than the arguments ofthe opposers.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeThe opposers of the venture provided more novel and original arguments than the supporters.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeOn most issues, I think I am more likely to agree with other SFU students than SFU faculty.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeIt is likely that the students in the group discussion and I have similar tastes and preferences.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeIt is likely that I share more common beliefs with the SFU students, rather than the faculty,who participated in the group discussion.Strongly agree 1 2 3 4 5 6 7 Strongly disagreePLEASE PROCEED TO THE NEXT PAGE. DO NOT RETURN TO THIS PAGE.171Please Circle the Number that Closely Corresponds to Your Feelings/Opinions: If the University participates in the joint venture, its success or failure will have importantconsequences for SFU students.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeThe students in the group discussion provided arguments which I had not previouslyconsidered.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeIn my life, I am usually quite conservative and do not take risks.Strongly agree 1 2 3 4 5 6 7 Strongly disagreeI believe in taking risks.Strongly agree 1 2 3 4 5 6 7 Strongly disagree2. Are you familiar with the technology discussed in this study (Circle One): YES ?CoIf YES, tell us where you heard about it and what you know about it:3. In the group discussion summary you read, how many SFU students:Supported the venture:  ^Opposed the venture: ^In the group discussion summary you read, how many SFU faculty members:Supported the venture:  ^Opposed the venture:^PLEASE PROCEED TO THE NEXT PAGE. DO NOT RETURN TO THIS PAGE.172Please provide the following personal information. This will be used for researchpurposes only.Age:  ^Sex: ^Ethnicity/Nationality (optional): ^Major Field of Study:  ^Year of Study: ^Courses enrolled this term (specify course numbers): ^THANK YOU FOR YOUR TIME AND COOPERATION.173TABLE 1: Test of Main Effects for Pretest 1 TREATMENT MEANS T-VALUE DF SIG'SIMILARITYHigh (n =22) 1.50 1.64 40 0.054Low (n =20) 0.00ISSUE-INVOLVEMENTHigh (n =20) 1.30 1.05 40 0.150Low (n = 22) 0.32RESPONSE-INVOLVEMENTPrivate/Low(n =22)1.13 0.79 40 0.22Public/High(n = 20)0.401 One-tailed test.174TABLE 2: Pilot Study - Reliability of MeasuresCONSTRUCT^NO.OF ITEMS^CRONBACH'S ALPHAATTITUDE^2^ 0.92INVOLVEMENT 3 0.77(Manipulation Check)SIMILARITY^ 2^ 0.67(Manipulation Check)BELIEFS^ 6^ 0.77SOURCE CREDIBILITY^3 0.52175TABLE 3:^Pilot Study - Rotated Factor MatrixVARIABLE FACTOR 1 FACTOR 2^FACTOR 3^FACTOR 4URUN .53477 .60657 .30312PRUN .54475 .59562CONFID -.74446SUGRESP .78332VARIETY .70244REINVES .38735 .36621RESORCE .66411 .30729PROFIT .64875VOLUME .74566SIMILMC -.33296 -.62947ATTENT .89100CONCTR .89501CONALT .64394BELIEVE .67080 .34080USEFUL .63980NOVEL .61715SIMTALK .40255 -.30277 .30388 .33644PercentageVarianceExplained 29% 13.4% 8.9% 6.6%NOTE: All Factor Loadings above 0.30 are shown.176TABLE 4: Pilot Study - Similarity X Involvement X Response ANCOVADependent Variable = ATTITUDESum of^Mean^ SigSource of Variation^Squares^DF^Square^F^of FCOVARIATES 1Position 8.546 1 8.546 8.696 .004Political-View 2.604 1 2.604 2.649 .106MAIN EFFECTSSimilar .023 1 .023 .023 .880Involvement .022 1 .022 .022 .882Response .068 1 .068 .069 .793INTERACTION EFFECTSSim X Inv 2.509 1 2.509 2.553 .113Sim X Res 2.201 1 2.201 2.239 .137Inv X Res 1.575 1 1.575 1.602 .208Sim X Inv X Res .002 1 .002 .002 .961EXPLAINED 18.572 9 2.064 2.100 .035RESIDUAL 107.126 109 .983TOTAL 125.697 118 1.0651 Position refers to the position advocated by the source, which could favour the university or theprivate franchise. Political-view refers to the individual's political orientation - ranging from liberal toconservative.177TABLE 5: Main Study - Reliability of MeasuresCONSTRUCT^NUMBER OF ITEMS^CRONBACH'S ALPHASIMILARITY-FELT^3^ 0.94INVOLVEMENT-FELT^3 0.84SOURCE CREDIBILITY' 2 0.69FEELINGS^2^ 0.83ATTITUDE 2 0.82RISK AVERSION^2^ 0.841 It should be noted that the measure used for credibility was differentfrom the measures used previously in the literature (e.g., Feick and Higgie 1992).In this research, the measures of credibility required the subjects to compareto different sources and then state which source was more credible.178TABLE 6: Parameter Estimates of Reliability ModelsGeneral Reliability Model Equal Msmt Units Model Parallel Forms ModelA l 0 A2 0 A3 041.00* 0.33 1.00 0.34 1.00 0.351.02 0.30 1.00 0.32 1.00 0.350.93 0.42 1.00 0.40 1.00 0.351.00* 0.52 1.00 0.48 1.00 0.401.27 0.22 1.00 0.31 1.00 0.401.08 0.43 1.00 0.40 1.00 0.401.00* 0.25 1.00 0.44 1.00 0.490.67 0.66 1.00 0.56 1.00 0.491.00* 0.15 1.00 0.19 1.00 0.320.80 0.45 1.00 0.43 1.00 0.321.00* 0.18 1.00 0.21 1.00 0.300.85 0.41 1.00 0.39 1.00 0.300 5 05 c050.67 0.65 0.650.48 0.62 0.600.75 0.51 0.500.85 0.73 0.690.82 0.72 0.70GOODNESS-OF-FIT INDEX:X2 = 47.91 X2 = 59.95 X2 = 73df = 44 df = 51 df = 58p = 0.317 p = 0.18 p = 0.08GOODNESS-OF-FIT DIFFERENCE:Model 1 and Model 2X2 = 12.04df = 6p = 0.05Model 2 and Model 3X2 = 13.69df = 2p < 0.005* = one variable in each column constrained to be equal 1.0 (Bagozzi 1980, p180).1 = the first three items are measure of felt similarity, the next three indicate felt involvement, and of the lastsix measures, two each represent source credibility, source feelings and attitudes respectively in that order.2 = all As constrained to equal one (see Bagozzi 1980, p181)3 = all As constrained to equal one (see Bagozzi 1980, p181)4 = error variances of indicators of common true-scores constrained to be equal(i.e., 811  = 0 22 =833; °44 = °55 = 966; ^ 911,11 =912,12)5 = the (13. matrix was a symmetric matrix; only the diagonal elements are shown here.179Table 7: Summary of Manipulation Checks for Main StudyTreatment^Test Used Statistic^df^SigSIMILARITY'^Chi-square 10.09^61^0.005INVOLVEMENT^T-test^4.71^70^0.001(Highmean= 17.55)(Lowmean = 13.19)1 Further examination of the data showed that the significant chi-square wasbecause a greater proportion of the subjects (70%) expressed felt similarityor identification with other students, and a much smaller proporation ofthe sample expressed felt similarity with the faculty members (30%).180TABLE 8: Main Study - Similarity X Involvement ANCOVADependent Variable = ATTITUDESum of^Mean^ SigSource of Variation^Squares^DF^Square^F^of FCOVARIATESRisk-aversion 27.386 1 27.386 3.085 .084Gender 1.396 1 1.396 .157 .693MAIN EFFECTSSimilarity 84.735 1 84.735 9.545 .003Involvement .740 1 .740 .083 .774INTERACTION EFFECTSSim X Inv 36.927 1 36.927 4.160 .043EXPLAINED 149.957 5 29.991 3.378 .009RESIDUAL 585.918 66 8.878TOTAL 735.875 71 10.364181Table 9:^Path Analysis - CREDIBILITY and FEELINGS as Criterion VariablesModelCriterionPredictors' R2SIM-F INV-F CRED1 CREDIBILITY 0.499 0.202 0.328(0.001) (0.050)2 FEELINGS 0.441 0.037 0.202(0.001) (0.777)3 FEELINGS 0.173 -0.132 0.529 0.366(0.219) (0.300) (0.001)1 SIM-F stands for SIMILARITY-F; INV-F stands for INVOLVEMENT-F; and CREDstands for CREDIBILITY. Numbers reported under the predictor variablesare betas, with the significance level within parantheses.182Table 10: Path Analysis - ATTITUDE as the Criterion VariablePredictors'^ R2ModelCriterionSIM-F INV-F CRED FEEL74-1834 ATTITUDE 0.443 -0.132 0.218(0.005) (0.379)5 ATTITUDE 0.344 -0.139 0.256 0.274(0.050) (0.340) (0.110)6 ATTITUDE 0.101 -0.201 0.204 0.492 0.518(0.600) (0.334) (0.334) (0.05)1 SIM-F stands for SIMILARITY-F; INV-F stands for INVOLVEMENT-F; CREDstands for CREDIBILITY; and FEEL stands for FEELINGS. Numbers reportedunder the predictor variables are betas, with the significance level withinparantheses.183FIGURE 1: Conformity Influence ModelMinorityorIndividuals MinorityorIndividualsMinorityorIndividuals•INFLUENCEMinorityorIndividuals•MinorityorIndividualsFIGURE 2: Dual Influence ModelVAJORITY(LARGER GROUP)INFLUENCEMinorityorIndividualsFIGURE 3 Simultaneous Social Influence Model MAJORITY(LARGER GROUP)MINORITY(SMALLER GROUP)INFLUENCEINDIVIDUALFIGURE 4 General Reliability Model (_True-Scores) FIXED PARAMETERS^FREE PARAMETERSX-X 4 -A 7 -X g -A 11- 1,0^X2 X3 X 5 X ^X ^X10 X 121 - 61202 1 05,LATENT VARIABLES= SIMILARITY= INVOLVEMENT2 = SOURCE CREDIBILITY3, = SOURCE FEELINGSO = ATTITUDE12101 ^x 2 X^ , 4 ^x 5x 6CI 'b 3^\21326 12S 41S5IS 8IS oI IX11S,^62^63I I I6^67 81 1X4X5X6X9X10X12XiX2X3X7X8FIGURE 5: Equal Units Reliability Model FIXED PARAMETERS^FREE PARAMETERS^LATENT VARIABLESAll X s - 1,0 = SIMILARITY, = INVOLVEMENT, = SOURCE CREDIBILITY= SOURCE FEELINGS5 = ATTITUDE1X11,  x,^\x, x 5/x6A 3 ^\12f:/° „ ^ 5 4403204210XSe 57^ 8^ 9^101X10X11S ,2X7X8X9X2X3X4X5X6X12FIGURE 6: Parallel Form Reliability Model PARAMETERSAll X s •All OsAll Error Variances of IndicatorsConstrained to be Equal6 1 6 2 = (5 3 ;^6, = 6 6 and so on.LATENT VARIABLES, SIMILARITY= INVOLVEMENT= SOURCE CREDIBILITY= SOURCE FEELINGS5 = ATTITUDEFIGURE 7: Single Factor Confirmatory Factor Modelp = 0 00012  1a-2X = 423,23df = 546 ...... X1 1, 1x12,1FIGURE 8: Main Hvnotheses and ResultsMINORITY^ MINORITY^MINORITYINFLUENCE INFLUENCE INFLUENCEPredictedActual High High HighSimilar/(1.83)(0.72)+ve(1,28)+ye0-yeLow(0.31)Low0-ye(0.11)(-0.86) -0,11)Low (-1.61)dissimilarLow^HighISSUEINVOLVEMENTLow^HighSIMILARITYOF MINORITY High^LowISSUEINVOLVEMENTCO^ 1 = Figure not drawn to scale; cells means are in paranthesisCREDIBILITYINVOLVEMENT-FFEELINGSFIGURE 9: Mediating Effects of Credibility and Feelings ATTITUDENOTE:^Only significant paths are shown in this figure.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items