the group polarization phenomenon - david myers€¦ · the group polarization phenomenon david g....

Psychological Bulletin1976, Vol. 83, No. 4, 602 627

The Group Polarization Phenomenon

David G. MyersHope College

Helmut LammUniversitat Mannheim, West Germany

Experiments exploring the effects of group discussion on attitudes, jury deci-sions, ethical decisions, judgments, person perceptions, negotiations, and risktaking (other than the choice-dilemmas task) are generally consistent with a"group polarization" hypothesis, derived from the risky-shift literature. Recentattempts to explain the phenomenon fall mostly into one of three theoreticalapproaches: (a) group decision rules, especially majority rule (which is con-tradicted by available data), (b) interpersonal comparisons (for which thereis mixed support), and (c) informational influence (for which there is strongsupport). A conceptual scheme is presented which integrates the latter twoviewpoints and suggests how attitudes develop in a social context.

In 1961, James Stoner discovered a reliablebut nonobvious effect of group discussion. Toexamine the popular notion that groups aremore cautious and less daring than individ-uals, Stoner had six persons at a time respondas individuals to a series of story problemscalled "choice-dilemmas" items (developed byKogan & Wallach, 1964). The subject's taskwas to advise the fictional character of eachitem as to how much risk he should take infacing a given decision dilemma. A sampleproblem illustrates the task.

George, a competent chess player, is participatingin a national chess tournament. In an early matchhe draws the top-favored player in the tournamentas his opponent. George has been given a relativelylow ranking in view of his performance in previoustournaments. During the course of his play with thetop-favored man, George notes the possibility of adeceptive though risky maneuver which might bringhim a quick victory. At the same time, if the at-

This paper was written while the first author wasin residence at the Sonderforschungsbertich fiirSozial-und Wirtschaftspsychologische Entscheidungs-forschung, Universitat Mannheim, during the summerof 1974. Financial support for this collaboration wasprovided by a Deutsche Forschungsgemeinschaftgrant and National Science Foundation grant GS2891A#1.

We are greatly indebted to Robert S. Baron,James H. Davis, Willem Doise, Norbert Kerr, SergeMoscovici, Dean Pruitt, Amiram Vinokur, andMaryla Zaleska for their helpful comments on earlierdrafts of this paper. However, we alone bear respon-sibility for its ultimate content.

Requests for reprints should be sent to David G.Myers, Department of Psychology, Hope College,Holland, Michigan, 49423.

tempted maneuver should fail George would be leftin an exposed position and defeat would almostcertainly follow.

Imagine that you are advising George. Pleasecheck the lowest probability that you would consideracceptable for the risky play in question to beattempted.

George should attempt to play if the chances areat least

_1 in 10 that the play would succeed_2 in 10 that the play would succeed_3 in 10 that the play would succeed_4 in 10 that the play would succeed_5 in 10 that the play would succeed_6 in 10 that the play would succeed_7 in 10 that the play would succeed_8 in 10 that the play would succeed_9 in 10 that the play would succeed-George should attempt the play only if it iscertain (i.e., 10 in 10) that the play wouldsucceed.

After individually marking their advice onall the items, the participants then assembledas a group and discussed each item until theyagreed. The finding that groups were on thewhole more risky than their average individ-ual member (generally by about one scaleunit or less) was immediately dubbed therisky-shift phenomenon and was followed bya wave of investigations of group risk taking.These studies have taken place in a dozendifferent nations, indicating that Stoner's re-sults are not peculiar to his subject popula-tion. They also reveal that group decisionmaking is not an essential component of theprocedure; a brief period of discussionfollowed by individual responses will alsoproduce a shift in the group average.

602

GROUP POLARIZATION 603

This mass of risky-shift studies conductedduring the 1960s, and thought by some (e.g.,Smith, 1972) to have been a fruitless fad, isnow appearing to have had more heuristicvalue than previously realized. This papersummarizes recent experiments examining thegenerality of the group shift effect uncoveredin the risky-shift experiments, and it reviewsrecent efforts to construct a theoretical homefor the group shift phenomenon.

The risky-shift literature, which is founda-tional to this review, has been well sum-marized elsewhere (Cartwright, 1971; Clark,1971; Dion, Baron, & Miller, 1970; Pruitt,1971a, 1971b; Vinokur, 1971b). Hence, wedo not again summarize data pertaining tothe conditions under which group shift occurson choice-dilemmas items or to theoretical ap-proaches which have been discounted and areno longer actively pursued.

ORIGIN AND DEFINITION OF THE GROUPPOLARIZATION CONCEPT

It is now widely recognized that the desig-nation risky shijt was a misnomer that un-fortunately induced many investigators toperceive the phenomenon from the perspectiveof the dependent variable, risk taking, ratherthan to think in broader theoretical termsabout the effects of intragroup communicationon attitudes and behavior. The risky-shiftlabel continued to guide experimentation longafter it was well established that shift togreater caution could be reliably demonstratedon certain choice-dilemmas items. In fact,one powerful and important finding from thechoice-dilemmas studies is that the individualitems differ from one another in (a) meaninitial response and (b) mean shift and thata and b are very highly correlated (generally.70 to .90 [e.g., Teger & Pruitt, 1967]). Itis important to note that this correlation isoccurring at the level of the item, with dataaveraged over groups. Items which elicit rela-tively risky initial tendencies generally elicitfurther shift toward the risky extreme afterdiscussion. Items with relatively cautious ini-tial means are more likely to elicit furthershift in the cautious direction.

Restating this finding in more general termswe have the group polarization hypothesis:

The average postgroup response will tend tobe more extreme in the same direction as theaverage of the pregroup responses. The termgroup polarization originates from the writingsof Serge Moscovici and his colleagues (e.g.,Moscovici & Zavalloni, 1969). Note thatpolarization refers to an increase in the ex-tremity of the average response of the subjectpopulation. (This use of polarization in therelevant literature is somewhat specialized, asin general usage the term may also refer toa split within a group of people.)

This polarization concept should also bedistinguished from a related concept, extremi-zation. Whereas polarization refers to shiftstoward the already preferred pole, extremiza-tion has been used to refer to movement awayfrom neutrality, regardless of direction. Sinceall instances of group polarization are in-stances of extremization, but not vice versa,extremization may be easier to demonstratethan polarization.

It should also be noted that conclusionsabout group polarization need not apply toindividuals. As Fraser, Gouge, and Billig(1971) illustrated, a three-person group withscores of +3, +1, and —2 on some +3 to —3attitude scale might all put down +1 afterdiscussion. At the group level, a slight moveto extremity (which in this case is also polari-zation) has occurred in the shift from a meanof +.67 to +1. But none of the individualsis more extreme than prior to the discus-sion. In fact, since group members generallyconverge, this pattern of responses is faiilytypical.

Finally, it should be noted that defininggroup polarization as an enhancement of thetendency initially dominant in the wholesample requires the definition of a subjectiveneutral point on a bipolar scale. Very oftenthe definition of a psychological midpoint(where there is no dominant tendency) is nosimple matter. On the choice-dilemmas items,for examine, it probably is not 5 chances in10, or even S.S in 10 (the mathematical mid-point of the scale). For example, post hocregression analyses indicate that items withan initial mean between 6 and 7 in 10 willgenerally not yield much shift, while thosewith initial means below 6 will generally yield

604 DAVID G. MYERS AND HELMUT LAMM

risky shift and those with initial means above7 will often elicit shift toward the higherprobability pole (e.g., Myers & Aronson,1972). Post hoc regression functions are, ofcourse, a circular way of defining neutralpoints to then use in testing the group polari-zation hypothesis. More direct data collectedby Miller (Note 1) indicate that with itemson which subjects' positions are located nearthe mathematical midpoint of the scale (5.5),subjects nonetheless tend to verbalize a pref-erence for the risky alternative when askedto advise whether the protagonist should takethe risk.

We now examine group discussion experi-ments using measures other than the choicedilemmas to see whether the group polariza-tion hypothesis derived from the choice-dilemmas literature has external validity withother response dimensions. The generalizabil-ity of the phenomenon has theoretical as wellas empirical and practical relevance. If grouppolarization can be demonstrated with a va-riety of measures and be given a satisfactorygeneral explanation, it will make unnecessarythose explanations of the risky-shift phenome-non which are risk specific, as well as thosewhich attribute the shift effect to peculiaritiesin the instructions or the metric of the choice-dilemmas items.

GENERALITY OF GROUP-INDUCEDPOLARIZATION

For convenience we have organized thesestudies into seven categories: attitudes, jurydecisions, ethical decisions, judgments, personperceptions, negotiation behavior, and riskmeasures other than the choice dilemmas.This categorization is admittedly somewhatarbitrary. Within each category we brieflypresent two or three illustrative studies andthen summarize findings from the remainingrelevant studies.

Attitude Studies

Varying the stimulus materials. Two differ-ent research paradigms have yielded datarelevant to the group polarization hypothesis.The first strategy has been to engage subjectsin discussion of stimulus materials which haveelicited some dominant predisposition and then

to determine whether talking in groups tendsto enhance the dominant leaning (just aschoice-dilemmas items have elicited risky orcautious initial tendencies and correspondingshift tendencies).

Most studies of group discussion in socialpsychology and in speech-communicationswere designed for purposes other than exam-ining a strengthening of the initial opinionaverage. Hence they report no prediscussionopinion average or they involved an initialopinion average near the neutral point(e.g., McCauley, 1972; Utterback, 1950). Inother studies, however, it is possible to inferthe direction of initial preferences. Robinson(1941) conducted lengthy discussions of twoattitudes. On attitude toward war, where stu-dents were initially quite pacifistic, there wasa nonsignificant shift to even more pacifismfollowing discussion. On attitude toward capi-tal punishment, to which students were ini-tially opposed, there was a significant shift toeven stronger opposition. Most other studiespredating the contemporary group shift litera-ture also evidence some degree of polariza-tion effect (Geier, Forston, & Larson, 1970;Hopkins, 1964, p. 131; Utterback, 1954),but not Miller and Biggs (1958).

More contemporary attempts to extendgroup shift findings to the attitude realmhave met with consistent success. Moscoviciand Zavalloni (1969) observed that Frenchstudents' initially positive attitudes towardDeGaulle and negative attitudes toward Amer-icans were strengthened through discussion.Doise (1969b) reported that discussion en-hanced the negative attitudes which Frencharchitectural students had toward their school.Judging from the number of citations thesetwo papers have received, they appear to havebeen a stimulus for much of the researchwhich followed. Subsequent studies investi-gating attitudes regarding social issues (Gouge& Fraser, 1972; Paicheler & Bouchet, 1973)and life situation dilemmas (Myers & Bishop,1971) have yielded similar polarization effects.

Group composition. Another set of attitudeexperiments has explored group polarizationusing a strategy dissimilar to that of therisky-shift paradigm. Groups were composedof subjects sharing common inclinations rele-


vant to the discussion materials and theiraverage shift was contrasted with that ofgroups of subjects sharing the opposite tend-ency. The group polarization prediction wasthat discussion with similar others will in-crease the attitude gap between separatedhomogeneous groups.

Myers and Bishop (1970) composed homo-geneous groups of relatively high-, medium-,or low-prejudice subjects, using a racialattitude inventory. Group members then re-sponded to eight new racial attitude itemsbefore and after discussion. The discussionwith others having similar racial attitudessignificantly increased the gap between thehigh- and low-prejudice groups. Mitnickand McGinnies (1958) observed a similarphenomenon using racial attitude materials.

Since there is ample evidence that peopledo in fact prefer being with others who havesimilar attitudes and values (Byrne, 1971),the question of whether separation on thebasis of shared values generally produces in-creased intergroup polarization is of more thanmere academic interest. Attempts to replicatethe intergroup polarization effect have metwith uneven success. Myers and Bach (1974)did not observe increased polarization be-tween separated groups of pacifistic and mili-taristic subjects after dis.cussion of items rele-vant to their value differences. Rather, allgroups significantly increased in pacifism. How-ever, Myers (1975) did observe that groupswhich were liberal in attitudes regarding therole of women polarized further from groupswhich were traditional after discussing rele-vant items. And Cvetkovich and Baumgard-ner (1973) found that groups of subjectswho were initially nonpunitive in their atti-tudes toward civil disobedience became evenmore tolerant after discussion, while morepunitive groups did not. In these studies, andsome others to be considered later, the generaltrend of shift averaged over all conditionsseems consistent with the prevailing externalnorm.

Field observations of social polarization ap-pear consistent with the laboratory studies ofintergroup polarization. Coleman (1957) con-cluded from his analyses of opinion polariza-tion during community conflict that

group discussion . . . is such an important phenome-non in community controversies that in the casestudies examined most descriptions of behaviorduring the intense part of the controversy were de-scriptions of discussion and of attempts to persuadeor reinforce opinion, (p. 18)

Homogeneous grouping was a source of com-munity polarization and the occurrence ofsocial conflict further heightened

the proliferation of associations among those whofeel one way, and the attenuation of associationbetween those who feel differently. One's statementsmeet more and more with a positive response; oneis more and more free to express the full intensityof his feeling, (p. 14)

Research on student change during col-lege, summarized by Feldman and Newcomb(1969), reveals even clearer evidence of inter-group polarization.

Initial differences among students in different collegesand in different curricula are accentuated or ampli-fied as students progress through college. Instancesof this same phenomenon also occur with respect toinitial differences among students entering differenttypes of residences, (p. 209)

For example, the tendency for fraternitymembers to be more conservative and preju-diced than independents tended to be smallestat the freshman and sophomore levels andlargest at the senior level. Feldman and New-comb surmised that this may occur partlybecause "the reciprocal influences of memberson one another reinforce and strengthen extantorientations (p. 223)." The likelihood of thistype of communication effect is one of thereasons Pettigrew (1969) has argued sostrongly that increased racial separation isantithetical to racial reconciliation.

To summarize, the weight of the evidencefrom attitude studies strongly confirms thegroup polarization phenomenon when stimulusmaterials are used which elicit a dominant ini-tial tendency in the sample population, andthe evidence is generally indicative of inter-group polarization effects when the groupcomposition paradigm is used.

Jury Decisions

Juries are small groups which are given thetask of arriving at a decision following dis-cussion. Do decisions following jury delibera-


tiott differ in any predictable way from theaverage of the predeliberation opinions of in-dividual jury members? Three small sets ofstudies, the first two involving experimentalsimulations, are summarized.

Varying the stimulus materials. Myers andKaplan (1976) engaged their subjects in dis-cussion of stimulus materials which elicited adominant predisposition of guilty or notguilty. After discussing traffic cases in whichthe defendants were made to appear as lowin guilt, the Subjects Were even more definitein their judgments of innocence and morelenient in recommended punishment. Afterdiscussing "high-guilt" cases, the subjects po-larized toward harsher judgments of guilt andpunishment.

Izzett and Leginski (1974) observed, how-ever, that an initial tendency for unattractivedefendants to receive harsher sentences thandid attractive defendants was reduced ratherthan exacerbated by discussion. Foss and Foss(Note 2) and Rumsey and Castore (Note 3)also observed greater leniency after discus-sion, although Heimbach (Note 4) observedgreater harshness by groups. In all four ofthese experiments, however, it is difficult todescribe the subjects as initially tendingtoward leniency or harshness, since no psycho-logical neutral point can be defined.

Group composition studies. Vidmar (Note S)composed groups of jurors high or low indogmatism. The high-dogmatism juries shiftedtoward harsher sentences following discussion,and the low-dogmatism groups shifted towardmore lenient sentences, despite the factthat the high- and low-dogmatism juries didnot differ in their predeliberation judgments.Laughlin and Izzett (Note 6) observed thatgroups composed of subjects which were atti-tudinally similar to the defendant shiftedtoward greater leniency following group dis-cussion, whereas subjects who were attitudi-nally dissimilar did not shift.

Actual court decisions. These findings, fromexperimental simulations of the jury process,do not lend themselves to clear-cut generaliza-tions. They do, however, suggest the possibil-ity that response-enhancement processes mayoperate in judicial group settings. Walker andMain (1973) provided more direct evidence.They compared civil liberties decisions by in-

dividual federal district court judges to civilliberties decisions by three-judge panels. Thegroup condition produced substantially morelibertarian decisions than did the single-judgecondition (65% versus 30%). A subset ofthese decisions also involved rulings of theconstitutionality of statutes. Main and Walker(1973) observed that these constitutionalitydecisions were also more libertarian in thegroup condition (65% versus 45%). Althougha minority of the single-judge decisions wereprolibertarian, Walker and Main surmisedthat the preexisting private values of thejudges were actually prolibertarian and thattheir decisions made alone were compromisedin the face of antilibertarian public pressure.Their private values were then supposedly re-leased and reinforced in the professionalgroup context,

Kalven and Zeisel (1966) presented evi-dence from 225 trials that is consistent withthe group polarization hypothesis (the initialmajority predicts the consensus outcome 90%of the time), and they concluded that

the deliberation process might well be likened towhat the developer does for an exposed film: itbrings out the picture, but the outcome is pre-determined. . . . From what we have been able toperceive thus far, the process is an interesting com-bination of rational persuasion, sheer social pressure,and the psychological mechanism by which individualperceptions undergo change when exposed to groupdiscussion, (p. 489)

Ethical Decisions

Our third category of experiments dealswith ethical and moral judgments. Home andLong (1972) engaged their subjects in discus-sion of situations involving conflicts betweenuniversalistic and particularistic moral obliga-tions. The subjects shifted toward increaseduniversalism following discussion, especiallyon items on which universalism was initiallyvalued. Alker and Kogan (1968) asked theirsubjects to predict their own behavior in anethical conflict situation. They observed aninitial leaning away from the universalistic(ethical) pole and an enhancement of thistendency after discussion. Myers, Schreiber,and Viel (1974) used some simple ethical-legal dilemmas (e.g., a scenario in which amature-looking but underage teenager is de-ciding whether to order a drink), The subjects,


male felons residing at a correctional institu-tion, initially favored the illegal behavior andfavored it even more following discussion.

Two other studies concern altruistic behav-ior. Schroeder (1973) had his subjects adviseprotagonists as to how much time or moneythey should sacrifice for a particular need.The altruistic direction, initially socially de-sired, was enhanced by group interaction.Baron, Roper, and Baron (1974) asked theirsubjects how much of their own money theywould pledge to Bengali relief and how muchthey would recommend their student bodycontribute. Although generosity was presumedto be socially desirable (because the subjectsperceived themselves as more generous thanothers), the pledges were actually more stingyafter group discussion. (It might be noted,however, that actual prediscussion pledges—less than $2.00 per person—were also on thestingy side.) The idea of contributing fromstudent body funds elicited greater generosityinitially and a trend toward further increasesafter discussion. Both of these studies are ofambiguous relevance to the group polarizationhypothesis, however, as it is impossible todefine a 0 point on the altruism-selfishnesscontinuum.

Finally, a series of studies by Rettig andhis associates (Chapko, 1972; Rettig, 1966,1972; Rettig & Turoff, 1967; Rettig, Note 7)has examined the effects of discussion onpredictions of ethical risk taking. Rettig hassuggested that his findings of increased pre-dicted risk taking following group discussionrun contrary to theorizing about the riskyshift, since ethical risks are socially deviant.But close scrutiny of his data reveals that thediscussions enhanced predicted ethical risktendencies when and only when risk was ofhigh reinforcement value (money stolen or"borrowed" was needed for a crucial medicaloperation). In this situation, the individualswere more likely to justify the illegal behav-ior, and social interaction further increased itsperceived desirability.

Judgment

This set of studies centers more on judg-ments of fact than on social evaluations. Assuch, it borders on the group problem-solvingliterature, which we do not review. Kogan and

Wallach (1966) compared the average abso-lute extremity of individual pregroup judg-ments (e.g., "What are the chances that anAmerican family owns its own house?") withthe extremity of consensus judgments. Theaverage consensus was not more extreme thanthe mean absolute extremity of initial judg-ments. Doise (1971) reanalyzed these data insearch of group polarization that may none-theless have been occurring (enhancement ofthe initial algebraic mean of responses to anitem). He found that "when the initial polari-zation is strong enough, the consensus signifi-cantly moves toward the pole that alreadyattracted the individual responses (p. 515)."Moscovici and Neve (1973) observed asimilar result with social judgments.

Johnson and Andrews (1971) had theirsubjects judge their probable satisfaction withnew consumer products that were preselectedfor high or low desirability. The desirableproducts were judged as even more satisfac-tory after discussion; the opposite was foundfor the undesirable products. However, Seminand Glendon (1973) intensively studied onereal business committee and found that theirjudgments of the relative importance of dif-ferent job criteria were unaffected by discus-sion. Their procedures made the pregroupjudgments highly salient, however.

In judgments of the actual probability ofsuccess on choice-dilemmas items, Madarasand Bern (1968) and Lamm, Trommsdorff,and Kogan (1970) observed greater pessimismafter group discussion. It is difficult to fit thegroup polarization hypothesis to their results,since the subjects were generally near themiddle of the pessimism-optimism scale priorto discussion. Lamm and Trommsdorff (1974)failed to replicate the pessimistic shift withprobability judgments concerning social andpolitical change. In this study, too, the sub-jects were generally near the middle of thescale prior to discussion. Finally, Vidmar(1974) reported broader category-width judg-ments by groups than by individuals. Acrossall items, individuals already tended toward"broadness" in total score. But contrary to thepolarization hypothesis, the shift to broadercategories was not greatest on the individ-ual items, on which initial judgments werebroadest. Thus, while there is some evidence


of group polarization in studies of factualjudgment, the phenomenon is not as reliablyconfirmed as in studies requiring a greaterdegree of social evaluation.

Person Perception

Another set of experiments has exploredthe effects of group interaction on evaluationsof hypothetical or real stimulus persons.Johnson and his associates engaged studentsin an evaluation of faculty. In one experiment(Andrews & Johnson, 1971), the subjectswere provided with favorable or unfavorablecues describing a hypothetical faculty. Withfavorable cues, groups were more positivethan individuals and with unfavorable cues,they were more negative. However, Shrews-berry and Johnson (Note 8) failed to repli-cate this effect. Subsequent studies (Johnson,Note 9; Krapf, Note 10) engaged studentsindividually and then as groups in judgmentsof their actual professor. Groups tended to bemore negative than individuals when theaverage individual ratings were already some-what negative and more positive when indi-vidual ratings were positive. Myers (1975)also used a faculty evaluation task. The sub-jects responded to 200 word descriptions of"good" or "bad" faculty with a scale judg-ment and by distributing a pay increasebudget among the hypothetical faculty. Aspredicted by the group polarization hypothe-sis, good faculty were rated and paid evenmore favorably after the group interaction,and contrariwise for the bad faculty.

Muscovici, Zavalloni, and Louis-Guerin(1972) and Moscovici, Zavalloni, and Wein-berger (1972) reported increased extremiza-tion following group discussion of impressionformation materials (descriptive adjectivesand photographs). However, they presentedno data enabling one to determine whetherthis was also a polarization effect. Doise(1970) used paragraphs portraying a stimu-lus person as relatively extroverted or intro-verted and did observe greater polarity ingroup impressions of extroversion-introver-sion. Thus, in general, the work on personperception supports the group polarizationhypothesis, especially when the stimulus ma-terials are more complex than just a singleadjective.

Negotiation and Conflict

Might group interaction effects also extendto bargaining and negotiation situations? Theavailable data are minimal but intriguing. Ina labor-management simulation, Rabbie andVisser (1972) instructed union bargainingteams to set expectation levels individuallyand then as groups. Individuals set higheraspiration levels when the issue was importantand when the bargaining position was strongand lower aspiration levels when the issuewas less important and when the bargainingposition was weak. These tendencies werestrengthened in the group condition.

Lamm and Sauer (1974) observed a similarpolarization in a bargaining experiment whichinvolved real consequences. When asked todistribute 18 profit units between themselvesand another player, individuals initially pro-posed giving 64°/o to themselves. After dis-cussing their positions with other individuals(whom they were not competing against),they increased their demand to almost 707c.Control subjects retested after private studyshowed no change.

Myers and Bach (1976) compared theconflict behavior of individuals and groups,using an expanded prisoner's dilemma matrixcast in the language of a gas war. There wasno difference in their conflict behavior (bothindividuals and groups were highly non-cooperative). But on postexperimental scalesassessing the subjects' evaluations of them-S2lves and their opponents, individuals tendedto justify their own behavior, and groupswere even more inclined toward self-justifica-tion. This demonstration of group polariza-tion supports Janis's (1972) contention thatin situations of intergroup conflict, groupmembers are likely to develop a strengthenedbelief in the inherent morality of their actions.

Risk Taking

The general finding of risky shift on choice-dilemmas items stimulated numerous experi-ments in which group-induced risk takingwas sought on other risk measures. Since itwas risky shift and not group polarizationthat was being sought, many of these experi-ments used betting materials in which theexpected value of all outcomes was near zero


and the initial mean did not suggest a domi-nant prediscussion trend toward either risk orcaution. Not surprisingly, a risky shift wasoften not realized in these studies, causinginvestigators to question the generality ofrisky shift. Burnstein (1969) and Doise(1969a) reviewed research on skill and chancerisk measures and observed, consistent withthe group polarization hypothesis, that inthose experiments which did produce a riskyshift, initial responses tended to be risky.

More recent research can also be seen tofit the polarization process. Manipulating theexpected value of bets such that the takingof increased risk would increase the expectedvalue of one's outcome has substantial effectson group-induced shift (Davis, Kerr, Sussman,& Rissman, 1974; Marquis & Reitz, 1969).When the expected value of the bet ispositive, groups will take more risk thanwill individuals. When the expected value isnegative, groups will take less risk than willindividuals.

When the expected value of the bet isheld at zero, then the direction of the shiftmay be seen to vary with other parameters inaccord with the affect of these parameters onpregroup betting tendencies. Using a proce-dure adapted from Pruitt and Teger (1969),Zaleska (1974, 1976) and Lamm and Ochs-mann (1972) we observed that group risk tak-ing exceeded that by individuals only when thestake (potential loss) was small. Zaleskafound that it is in precisely this conditionthat individual pregroup inclinations alsotend to be quite risky. When the stake waslarge (and the potential gain also high),individuals were more cautious and groupsexceeded them in caution.

A reanalysis of the Lamm-Ochsmann data(Lamm, Myers, & Ochsmann, in press) re-veals the same pattern of results. Eight betswith varying stakes were used, with initialrisk taking higher on low-stake bets. Acrossthe eight bets the correlation between themean of initial risk to a bet and mean riskyshift elicited by that bet was .80. Recall thatit was a comparable correlation obtainedon choice-dilemmas items which inspired thegroup polarization hypothesis.

Various other risk measures have also beenused. With an investment simulation task,

Deets and Hoyt (1970) observed that aver-age initial preferences were for securitiestoward the high-risk end of the continuumand that this tendency was strengthened bydiscussion. Runyan (1974) observed that sub-jects who advised another on risk taking ini-tially elected more risk than did subjects whomade binding decisions for another. Thisdifference was slightly magnified by groupinteractions (p < .10). In a field experiment,McCauley, Stitt, Woods, and Lipton (1974)observed that groups were more cautious inrace track betting than were individuals. Theinitial inclination of most bettors was alsoto prefer the favored horses and to avoid thelong shots.

Finally, there are three small studies(Clement & Sullivan, 1970; Cohen & Ruis,1974; Colligan & Giambra, Note 11) in whichthe investigators sought risky shift in class-room contexts in which students were asked,before and after discussion, to recommend oneof several grading procedures (from quiterisky—e.g., heavy weight on one final exam—to more conservative). In each of the threestudies a shift to greater caution followeddiscussion, causing the investigators to ques-tion the reliability of risky shift or to labelit an artifact of the laboratory. However, intwo of these studies the investigators pro-vided both initial means and shift means, andthe findings are perfectly compatible with thegroup polarization hypothesis. For example,Clement and Sullivan observed that "theprivate choices before discussion were all fromthe conservative half of the eight choices pre-sented (p. 244)." In the third study no pre-discussion data are presented.

All in all, it appears that recent demonstra-tions of risky and cautious shifts in risk be-havior are quite compatible with the general-ized group polarization hypothesis. In studiesin which risky shift has not been observed,the results may be seen to confirm rather thanrefute the basic group locomotion phenome-non observed on choice-dilemmas items.

Conclusion

A review of recent literature on the effectsof group discussion on various types of re-sponses indicates that the evidence is gener-ally consistent with the group polarization


hypothesis derived from research with choice-dilemmas items. Group enhancement of ini-itial mean tendencies is not always observed,and when it does occur it is generally ofsmall absolute magnitude. A further difficultywhen trying to document the generality ofa phenomenon is the likelihood that negativeresults have more often gone unreportedthan have rejections of the null hypothesis(Greenwald, 1975). But the trends are con-sistent enough across a variety of task situa-tions to conclude that some generality ofgroup polarization phenomenon has been es-tablished. It appears that the conclusion of-fered by Roseborough (1953) regarding thestate of knowledge on group problem solvingmore than 20 years ago is equally appropriateto the research literature on group polariza-tion 20 years later:

We need not be further persuaded that group dis-cussion processes have an effect on individual per-formance even though there is a selective processoccurring in the reporting of studies. This proof hasonly opened up new and troublesome problems con-cerning the mechanisms by which this influence isachieved and the conditions under which such anempirical observation holds, (p. 279)

THEORETICAL EXPLANATION OF GROUPPOLARIZATION

Having ascertained the empirical generalityof group polarization, we now consider pos-sible explanations of it. What are the mecha-nisms by which this group influence isachieved and the precise conditions underwhich the empirical observation holds? Ourgoal is to understand the phenomenon in away that will account for the known condi-tions under which group polarization occurs,will predict the effects of other conditions notyet studied, and will suggest when an en-hancement of the initial mean tendency is notto be expected. The initial mean tendency isnot causing the shift, it is only a crude indexto some actual group dynamics which do pro-duce the effect. Thus, if we can identify thegeneral mechanism responsible for the en-hancement of initial mean tendencies, perhapswe can then use this understanding to betterpredict the expected direction and extent ofgroup shift in a particular situation.

The findings we review in the remainderof the paper are based mostly on choice-

dilemmas research. This operational consist-ency has aided in the development of a well-established set of empirical parameters. Theevidence reviewed earlier indicating that riskyshift is merely one subclass of a generalpolarization phenomenon gives grounds foralso extending theory-testing research withchoice dilemmas to other situations in whichgroup polarization is observed. It is possible,however, that the causal group dynamics willvary somewhat across tasks. A good theorywill predict any situational differences.

The earlier reviews have already docu-mented that group shifts do not appear to bedue, to any large extent, to the diffusion ofresponsibility via the development of affec-tive bonds among group members, or to thereduction of uncertainty resulting from mereincreased exposure to the stimulus materials,or to a relationship between leadership abilityand propensity to take risks.1 We thereforeconcentrate our analysis on three theories thathave recently received considerable attention.Group decision rule explanations (e.g., major-ity rule) predict shift by using statisticalschemes for combining individual preferencedistributions into an expected group product.According to one version of these statisticalexplanations, no actual psychological changesin preference need be postulated. Interper-sonal comparison theories explain shift interms of social motivation. Subjects desire toperceive and present themselves favorably, soexposure to others' positions may stimulatethe subject to readjust his response in orderto maintain his image of social desirability.2

1 A potentially more viable notion than the gen-eral leadership theory is that shift is due to dispro-portionate influence by extreme group members. Theextremity of a subject's response to a specific itemmay be related to the degree of his commitment,awareness of arguments, and/or confidence. Avail-able data do not strongly support this explanation(see Davis, Kerr, Meek, & Rissman, in press; Myers& Murdoch, 1972), although it is possible to incor-porate extremity influence affects within the theoriesdiscussed below,

2 Some have also called this value theory, a labelwe avoid because of its ambiguity. All that theinterpersonal comparisons approach needs to presumeis that subjects enter into social interaction withsome shared initial suppositions as to what attitudeor course of action is desirable. It does not seemnecessary to debate whether, in a particular instance,these suppositions also point to cultural values.


Informational influence explanations suggestthat discussion generates arguments predomi-nantly favoring the initially preferred alter-native, including some persuasive argumentsthat the typical subject has not previouslyconsidered. Thus, a "message effect" evokesresponse change resulting from new cognitivelearning.

Group Decision Rules

Several investigators have recently sug-gested that schemes for combining individualpreferences might successfully predict thegroup product. Davis (1973) has enumeratedseveral of these possible social decisionschemes. The most popular has been the sug-gestion of majority rule (e.g., Burnstein,1969; Cartwright, 1971; Lambert, 1969,1971;Lamm, Trommsdorff, & Rost-Schaude, 1973).None of these authors contended that major-ity rule is the sole determinant of polariza-tion, but they did convincingly point out theplausibility of this idea. A social decisionscheme of majority rule predicts shift towardthe dominant pole when the majority favorsthat direction and when there is skewness inthe distribution of initial choices (which therewill often be when the response mean departsfrom the midpoint of the scale). Perhapswhen such a majority-minority constellationexists, deviant minority persons in the tail ofa skewed distribution are moving towardagreement with the majority opinion, therebycreating a shift in the group mean. A statis-tical decision scheme such as this could poten-tially account for group polarization withoutrecourse to postulating any real changes inindividual preferences. (A decision schemeapproach is not necessarily this indifferent toindividual psychological change. Although onemight consider shift a statistical artifact ofa majority rule decision procedure, anotherview is to assume that majority influenceproduces genuine change in the minority.) Agroup decision scheme approach requires someinitial diversity in choices and implies thatthe variance among choices will then generallybe reduced after the discussion; and thereis ample evidence that these conditions gen-erally hold when shift is obtained (e.g.,McCauley, 1972). Since the decision schemeapproach is so intuitively compelling, it may

be a disappointment that numerous otheravailable data so sharply contradict it, atleast as far as majority rule is concerned. Wenow enumerate in a roughly ascending orderof significance those known facts which con-tradict the suggestion that group polarizationis a statistical artifact of the application ofsocial decision schemes. (Not all of these factscontradict a social influence interpretation ofdecision schemes.)

1. Group-induced shift appears to be inter"nalized, not a temporary group product. Theshift effect is about equally robust regardlessof whether a group decision is required(Kogan & Wallach, 1967a; Lamm, 1967;Marquis, 1962; Wallach & Kogan, 1965;Wallach, Kogan, & Burt, 1968). Of course,when overt consensus is not required, animplicit consensus might nonetheless emerge.But it is furthermore true that nonpartici-pating observers of the group discussion alsoshift (e.g., Lamm, 1967) and that the group-induced changes in participant responses willpersist for several weeks afterward (Johnston,1968; Wallach, Kogan, & Bern, 1962—although this might be an after-effect of theconsensus commitment).

2. When the opportunity for combiningpretest decisions according to a decision ruleis eliminated, shift still occurs. Shift effectshave recently been obtained in numerous ex-periments using between-groups designs with-out pretest (e.g., Baron, Baron, & Roper,1974; Davis et al., 1974; Gaskell, Thomas,& Farr, 1973; Myers & Bach, 1974; Myers,Bach, & Schreiber, 1974). Still, it might becontended that individual preferences arenonetheless expressed early in the discussionand then combined according to some implicitscheme. But when subjects discuss the sub-stance of the items without an awareness ofany response scale, shift still obtains (Myers,Bach, & Schreiber, 1974). This is also thecase when subjects discuss a related responsedimension (utilities of outcomes) and thenrespond on the probability scale they have notdiscussed (e.g., Burnstein, Miller, Vinokur,Katz, & Crowley, 1971). These are clearindications that genuine preference changesdo occur, producing response change even inthe absence of an opportunity for decisionschemes to operate on the response dimension.


It is still possible, though, that a groupmight adjudicate their opinions (if not theirresponses) via some decision rule.

3. Skewness cannot account for group po-larization. This is particularly relevant to themajority rule scheme, which depends on askewed distribution of initial choices. Onchoice dilemmas, positively skewed distribu-tions (i.e., with a risky majority) shouldproduce risky shift, and negatively skeweddistributions should yield a conservative shift.Several findings refute this prediction.

Vinokur (1969) reanalyzed raw data fromearlier studies that had yielded risky shift andnoted that negative skewness was equally asprevalent as positive skewness. Furthermore,both the positively and negatively skewedgroup distributions on the three most riskyitems produced equivalent shift.

Shifts in the group median, althoughslightly attenuated, are not significantlysmaller than shifts in the group mean (Fraser,1971; Myers & Aronson, 1972; Lamm et al.,in press; Abend & Kogan, Note 12). This iscontrary to any model that depends on skew-ness effects. For example, if the majority isreally unaffected by the group discussion,then the median group member (who is oneof the majority) should also be unaffected.

Group shift has also been shown to occur indyads (although somewhat reduced), whereobviously there can be no skewness in theinitial responses (Baron, Baron, & Roper,1974; Bateson, 1966; Lamm, Schaude, &Trommsdorff, 1971; Myers & Aronson, 1972;Swap & Miller, 1969). Most of the availablegroup decision schemes would not predictpolarization effects in dyads.

The majority rule model not only predictsa correlation between the initial mean of re-sponses to an item and mean shift but, morespecifically, also implies the same correlationat the level of specific group decisions (acrossgroups within items). In other words, givena single item, the riskier a group's initialmean on that item, the more it should shifttoward the dominant pole (because of in-creased likelihood that there will be a major-ity favoring that direction). Another groupthat on the same item initially tends in theopposite direction should then shift in that

opposite direction. But in fact the correlationbetween initial mean and mean shift acrossgroups within items is near zero (e.g., Teger& Pruitt, 1967). (Actually it is somewhat inthe reverse direction, although ceiling andregression effects might contribute to this—agroup initially near the extreme simply can-not move much further in that direction.) Wereturn to this finding later.

Although group discussion generally resultsin decreased intragroup variance, as groupdecision schemes would predict, this may bea concomitant rather than an inherent fea-ture of the shift phenomenon. By eliminatingboth pretesting and discussion of the responsescale per se, Myers, Bach, and Schreiber(1974) demonstrated polarization withoutgroup convergence (as compared to variancein a control condition).

Finally, attempts to directly apply amajority rule model to discussion-producedshifts have produced only mixed results (e.g.,Cvetkovich & Baumgardner, 1973; Moscovici& Zavalloni, 1969; Zaleska, 1976).

In summary, it is evident that while groupdecision models may be useful in other situa-tions in which discussion is minimal or absentand the task is to reach agreement (e.g.,Lambert, 1969), the models (or at least themajority rule model stressed in this analysis)are not a sufficient explanation of the grouppolarization findings we are seeking to ex-plain. There are still a variety of other deci-sion schemes that can be explored and withother specific tasks. But clearly, group-induced shift on choice dilemmas is somethingmore than a statistical artifact.

We conclude with two general observationsabout prediction based on group decisionmodels. First, the most stringent test of anymodel is not to predict the general trend ofshift (since the general trend might occur forreasons irrelevant to the model), but ratherto predict specific group outcomes. In specificgroup situations a given model might be quitean inaccurate predictor, but across groups thealgebraic sum of its deviations may nonethe-less be close to zero (because positive andnegative errors cancel out, making the trendprediction accurate). In one study (Abend &


Kogan, Note 12) this in fact occurred. Theirmodel was successful in predicting the generaldistribution of group products, but failed topass the acid test of minimizing the sum ofabsolute deviations from actual choices. Thus,it is important that investigators who usethis approach give their model the most rigor-ous possible test. (See, for example, Cart-wright's, 1973, and Davis et al.'s, in press,analyses of percentage of "hits" in predictingspecific group outcomes.)

Second, we suggest that even if statisticalprediction by a model should prove to bequite accurate, this does not necessarily con-stitute a psychological explanation of thedynamics actually producing change in indi-vidual subjects. For example, it is conceivable(as Lamm, Trommsdorff, & Rost-Schaude,1972, have pointed out) that a majority rulemodel might be successful for reasons impliedby other available theories (e.g., Utterback,1962, observed majorities to be more influen-tial because in fact they possessed morecogent arguments). Put more positively, somesocial psychological explanations might betranslatable into precise social decision schemepredictions (see Burns, 1967; Davis et al.,in press). Thus, decision schemes need not beconsidered merely a statistical artifact ap-proach to explaining group effects; they can,in fact, be a rigorous operational definition ofcertain psychological explanations.

Interpersonal Comparisons

A second class of explanations attributesgroup polarization to a type of normativesocial influence (connoting social-emotionalprocesses and concern for favorable self-perception and self-presentation). More spe-cifically, several versions of an interpersonalcomparison approach suggest that mere expo-sure to the preferences of others is the neces-sary and sufficient condition for shift. Pruitt(1971a, 1971b) itemized the variations on thistheme. In general, they suggest that a subjectchanges when he discovers that others sharehis inclinations more than he would have sup-posed, either because the group norm is dis-covered to be more in the preferred directionthan previously imagined or because the sub-ject is released to more strongly act out his

preference after observing someone else whomodels it more extremely than himself. Thistheory, taken by itself, suggests that relevantnew information which emerges during thediscussion is of no consequence. Group polari-zation is a source effect, not a message effect.

A confusing array of reliable findings rele-vant to interpersonal comparison dynamicsnow exists. These diverse findings constitutea challenging intellectual puzzle to anyone whocares to search for a simple principle that willbring order to the mixture.

Differences between self, presumed other,and ideal scores. One well-known and widelysubstantiated assumption of the interpersonalcomparisons approach is the observation fromchoice-dilemmas research that if, after re-sponding, the subjects go back over the itemsand guess how their average peer would re-spond and then go back over the items athird time and indicate what response theywould actually admire most, they tend to esti-mate the group norm as more neutral thantheir own initial response and their ideal asmore extreme (see Myers, 1973, for a listingof more than 24 studies). Lamm et al. (1972)have also shown that not only do subjectsindicate their ideal as more extreme than theiractual response, but they also suspect that thesame is true of their peers. The tendency ofpeople to perceive themselves as more in whatthey consider to be the socially desirabledirection than their average peer extendsbeyond the choice dilemmas (see Codol,Note 13). For example, most businessmenbelieve themselves to be more ethical than theaverage businessman (Baumhart, 1968), andthere is evidence that people perceive theirown views as less prejudiced than the normof their community (Lenihan, Note 14).

On the choice-dilemmas task there are somepuzzling order effects, however. The tendencyto perceive others as "behind" oneself existsonly when the self response is made priorto estimating the group norm (McCauley,Kogan, & Teger, 1971; Myers, 1974). Evi-dently it is after one has decided for himselfthat there is then a tendency to consider one'saction as relatively admirable (by perceivingthe average person as less admirable thanoneself).


One version of interpersonal comparisontheory suggests that in the group context, thesubject typically discovers that the groupnorm is more supportive of his position thanhe had supposed (he is therefore outshiningthe other group members less than he hadpresumed). Consistent with this line of think-ing, it has been shown that after discussion,subjects will indeed revise their estimates ofthe group norm more in line with reality, evenwhen mention of initial responses has beensuppressed (Clark et al., 1971; Ferguson &Vidmar, 1971; Myers, Bach, & Schreiber,1974; Myers, Wong, & Murdoch, 1971).

Finally, it has been reliably demonstratedthat subjects perceive other persons who haveresponded more extremely than themselves(in the direction of their ideal) as more so-cially desirable than persons who have not(Baron, Monson, & Baron, 1973; Jellison &Davis, 1973; Jellison & Riskind, 1970, 1971;Madaras & Bern, 1968). A parallel findingexists in the attitude literature (Eisinger &Mills, 1968): An extreme communicator onone's side of an issue tends to be perceived asmore sincere and competent than a moderate.

While these findings are consistent withthe interpersonal comparisons approach, theirtheoretical significance is somewhat ambigu-ous. For example, Burnstein, Vinokur, andPichevin (1974) took an informational influ-ence viewpoint and showed that people whoadopt extreme choices are presumed to pos-sess cogent arguments and are then presum-ably admired for their ability. They alsodemonstrated that subjects have much lessconfidence in others' choices than in theirown, suggesting that the tendency to perceiveothers as more neutral than oneself simplyreflects ignorance about others' choices. Giventhe ambiguous significance of these self-other-ideal difference scores, the importantquestion becomes whether these differencescores predict response shifts, as interpersonalcomparison theories would imply.

Predicting shift with self, presumed other,and ideal difference scores. This has been at-tempted in two different ways, correlationallyand experimentally. First, we may inquirewhether the perceived difference between one-self and others does in fact correlate with an

individual's shift on a specific item. When asubject perceives himself as way ahead of hisaverage peer, he should be more likely tosuffer disconfirmation of his perceived rela-tive position and so should be more stimulatedto shift when informed of the actual groupnorm. But to the contrary, the self-otherdifference score on a particular item does notpredict a subject's change following discussionof that item, even if his own initial responseis partialed out (Lamm, et al., 1971; Myerset al., 1971; Pruitt & Teger, 1967). Thus, adirect examination of interpersonal compari-son theory at the level of particular decisionsby individuals does not confirm the theory.

The experimental strategy is to composegroups on the basis of self versus perceivedother difference or self versus ideal differenceand observe whether this manipulation affectsshift magnitude. Here the data are somewhatmore comforting to interpersonal comparisontheorists. Composing groups on the basis ofself versus perceived other difference hasyielded mixed results. Clark et al. (1971)observed increased risky shift when their sub-jects strongly underestimated peer risk ac-ceptance, but Lamm et al. (1972) observedno effect of a similar manipulation. However,composing groups on the basis of the extentto which their ideal exceeds their actual re-sponse does affect shift in accord with inter-personal comparison theory (Lamm et al.,1971).3 The latter two studies by Lamm andhis associates—combined with the observationthat self-ideal difference scores are less af-fected by order of measurement than selfversus perceived other differences (Myers,1974)—suggest that the self-ideal discrep-ancy may be the more crucial element of aviable interpersonal comparisons approach.This conclusion is further reinforced by severalindependent observations of group-induced

3 One other set of group composition experimentshas grouped subjects high or low in social depen-dence (field dependence, need for approval, internal-external control, Machiavellianism). Subjects high insocial dependence have not evidenced greater shift(Kogan & Wallach, 1967b; Lamm & Myers, in press;Lamm et al., 1971; Minton & Miller, 1970; Myers,Murdoch, & Smith, 1970; Wallach, Kogan, & Burt,1967).


shift toward rather than away from the per-ceived norm (Baron, Roper, & Baron, 1974;Myers, Schreiber, & Viel, 1974; Vidmar,1974). In two of the three studies it seemsreasonable to presume that the subject's idealmay in fact have been in the same directionas the perceived norm but that an additionalsocial constraint, surveillance by an adult whorepresented a conflicting norm, compromisedthe initial choices. The group interactionmight then have liberated the subjects fromthe conflicting external norm.

Given these mixed results from attemptsto predict shift with self-other-ideal differ-ence scores, we now seek clarification of therole of interpersonal comparison dynamics inthe final set of relevant studies. These at-tempt to provide the presumed necessary andsufficient conditions for shift—mere exposureto others' responses.

Choice shift following exposure to others'responses. One set of studies has manipulatedthe information about others' responses byproviding fake norms. More than a dozenseparate studies all show that subjects willmove toward the manipulated norm (seeMyers, 1973). This simply demonstrates thatconformity effects can be demonstrated onchoice-dilemmas items, as with numerous othermeasures. It is of interest, though, that themagnitude of the conformity effect is notgenerally increased much by making the fakenorms consistent with the direction of one'sideals. Baron, Monson, and Baron (1973)recently demonstrated greater influence whenthe group deviated in the valued directionrather than the nonvalued direction. It is notcertain whether this was due to the manipula-tion of confederate positions or to a con-comitant increased potency of confederateinformation when arguing in the valueddirection.

Other studies expose subjects to informa-tion about the actual initial choices of othersubjects without any discussion or exchangeof arguments. In general, the results from aseries of such studies (see Myers, 1973, forbibliography) reveals either reduced but sig-nificant shift or no shift at all. Burnstein andVinokur (197S) suggested that where an ef-fect of mere exposure to others' choices was

obtained, it may have been due to the stimu-lation this provided to think of argumentsthat others might have had for their choices.Consistent with this idea, they observed thatexposure to others' choices produced shiftonly when subjects then wrote arguments onthe item. If knowledge of others' choices wasdenied or if an opportunity to rethink theitem was denied, no shift occurred.

On the other hand, it may be reasoned thatin each of the studies producing minimal ornonexistent shift after exposure to others'attitudes, the subjects were first induced tobind themselves publicly to a pretest choiceand then simply exposed to others' choices.It takes only a quick recall of some classicconformity studies (e.g., Asch, 1956) torealize that this was an excellent procedurefor inhibiting response change. Perhaps dis-cussion only serves the face-saving functionof rationalizing response tendencies that areimplicit after learning of others' preferences.

A recent experiment (Myers, Bach, &Schreiber, 1974) removed this problem byhaving the subjects respond to three riskychoice dilemmas without a pretest commit-ment after merely being informed of the dis-tribution of responses by 40 subjects in acontrol condition. While this provided a largerdose of interpersonal comparison than did thetypical small group treatment, it maximizedthe possibility of a comparison effect uncon-taminated by other variables. A surprisinglyrobust effect (increased risk taking) was evi-denced when comparing these posttreatmentresponses with the control responses the sub-jects observed. This finding is counterintuitivein that it is the opposite of conformity—thesubjects were fairly accurate in their guessof the average of the responses they hadobserved, and yet this exposure elicited adifferentiation (polarization) from the ob-served norm.1

As we noted at the outset, the studies oninterpersonal comparison dynamics do notlend themselves to a simple summary. How-ever, enough evidence has been uncovered in

4 LeMaine (1974) has recently discussed the con-tribution to one's identity of differentiating oneselffrom others.


support of certain assumptions of the theorythat we later attempt to integrate theseassumptions into a theory of group-inducedattitude change.

Informational Influence

The third major theoretical approach at-tributes the observed response changes tocognitive learning resulting from exposure toarguments during discussion. Discussion gen-erates arguments predominantly favoring theinitially preferred alternative, and for anygiven subject, some of these are likely to bepersuasive new arguments. Thus, cognitivelearning occurs mostly in the dominant direc-tion and responses are modified accordingly.Each subject is processing and weighing in-formation by a process akin to Anderson's(1971) information integration scheme, ratherthan comparing himself with the actual posi-tions taken by others. The evidence relatingto informational influence is largely of twotypes: experimental manipulations of theavailability of arguments and content analy-ses of arguments generated by individualsand groups.

Several experiments have sought to stimu-late argument exchange while eliminating in-formation about others' choices. Normal shifteffects result even when the subjects are pro-hibited from mentioning their initial choices(Clark et al., 1971; Myers et al., 1971; St.Jean, 1970) and even when the subjects ex-change arguments without pretest and withoutawareness of the probability scale to whichthey will subsequently be exposed (Myers,Bach, & Schreiber, 1974). Similarly, discuss-ing utilities of choice-dilemmas alternativesproduces subsequent change in the probabil-ities chosen, even when probability choicesare not directly discussed (and vice versa,discussing probabilities affects utility judg-ments—Burnstein et al., 1971; Vinokur,197la). While it is therefore clear that argu-ments have a persuasive impact, it is less clearwhether this results from their teaching thesubject something he did not know (informa-tional influence) or from their telling thesubject something about the general opinionpositions of the persons who spoke them(interpersonal comparison). An ingenious ex-periment by Burnstein and Vinokur (1973),

which provided arguments without clues asto the speakers' real attitudes (by keepingthe subjects unaware of whether other groupmembers were advocating their own initialchoice or role playing support for a contraryone), indicates that informational influenceis involved. In a recent attitude change study,Eagly (1974) has also experimentally demon-strated an effect of information comprehen-sion. Studies of political attitudes in naturalsituations further evidence that increasedinformation tends to polarize opinion (seeSears, 1969).

The second type of relevant evidence looksinside the "black box" of discussion to ex-amine the actual distribution and nature ofarguments expressed. It is now well docu-mented that the direction of discussion argu-ments elicited by an item is an excellent pre-dictor of the direction of shift generally ob-tained on that item (correlations in the areaof .9 if a variety of items are used—e.g., Bishop& Myers, 1974; Ebbesen & Bowers, 1974;Morgan & Aram, 1975; Vinokur & Burnstein,1974). These studies indicate that the initialmean of responses to an item predicts verywell the trend of prediscussion and discussionarguments which in turn predict very wellthe mean shift. The predictive power of theinitial response mean is thus captured almostentirely by the content of the subsequent dis-cussion, suggesting that it is the nature of theexpressed arguments which mediates the rela-tionship between initial mean and mean shift.

Vinokur and Burnstein (1974, Note 15),Vinokur, Trope, and Burnstein (in press),and Bishop and Myers (1974) have formu-lated mathematical models of the presumedinformational influence mechanisms. Thesemodels assume that the amount of group shiftwill be determined by three factors: thedirection of each argument (which alternativeit favors), the persuasiveness of each argu-ment, and the originality of each argument(the extent to which it is not already knownby the group members before discussion). Indiscussion, the potency of an argument willbe zero if either the rated persuasiveness iszero (it is trivial or irrelevant) or if all groupmembers considered the argument before dis-cussion (in which case no cognitive learningwill occur when the argument is heard). Since


the simple direction of arguments is such anexcellent predictor of shift (without consider-ing persuasiveness and originality), it is noteasy to demonstrate the superiority of themodels over a simple analysis of argumentdirection as undertaken by Ebbesen andBowers (1974). Nonetheless, exhaustive analy-ses by Vinokur and Burnstein support theirinformational influence model by demon-strating that arguments do center on utilitiesof outcomes, that arguments consistent withthe dominant tendency are rated as more per-suasive than arguments opposing polarization,that the relevant arguments are indeed onlypartially shared prior to discussion, and thatnovel arguments have more impact thando unoriginal arguments. Furthermore, theirmodel is even successful in predicting varia-tion among group shifts within items. Recallthat the prediscussion response means forparticular groups do not account for variationamong shifts within items. Earlier we notedthat this latter finding was contrary to themajority rule model. It can, however, be ac-counted for by the informational influenceassumption that there exists a pool of per-suasive arguments for each item. A group thatis fairly polarized on a particular item beforediscussion is presumably already in generalpossession of those arguments which polarizea group. A less extreme group has more togain from the expression of partially sharedpersuasive arguments.

Two other sets of evidence are relevant toinformational influence theory. First, experi-mentally manipulating the direction of thearguments to which subjects are exposed affectsshift in accord with informational influencepredictions (Baron, Dion, Baron, & Miller,1971; Ebbesen & Bowers, 1974; Roberts &Castore, 1972; Silverthorne, 1971; Russo,Note 16). However, most of these studieswere probably simultaneously manipulatinginformation about others' positions, so theymight also be interpreted as demonstratinginterpersonal comparison or simple conform-ity effects.

While the above collection of findings con-stitutes compelling and consistent support forthe causal role of cognitive learning in pro-ducing group polarization, the final set offindings suggests that cognitive learning is

not a complete explanation by itself. Passivereceipt of arguments outside an interactivediscussion context generally produces reducedshift (e.g., Bishop & Myers, 1974; Burnstein& Vinokur, 1973; St. Jean, 1970; St. Jean &Percival, 1974). Likewise, listening to agroup discussion generally elicits less shiftthan actual participation (e.g., Bell & Jamie-son, 1970; Lamm, 1967; St. Jean, 1970).These findings—that active discussion gener-ates more change than the passive receipt ofinformation—probably should not have sur-prised us in light of the classic work byLewin (1947) on the relative effects of dis-cussion and lecture, educational research com-paring the relative efficacy of informationpresentation versus participative discussion(McKeachie, 1968), and the observed ineffec-tiveness of information communications infield settings (Crawford, 1974). Studies byMoscovici, Doise, and Dulong (1972) andMoscovici and Lecuyer (1972) further indi-cate that instructions and seating patternswhich interfere with the natural group inter-action process reduce the shift effect.

A hint as to how the informational influ-ence approach might be refined to accom-modate these findings comes from theory andresearch on the role of cognitive learning andrehearsal in attitude change. McGuire (1972)pointed out that attention to and comprehen-sion of arguments (cognitive learning) mustbe followed by conditions which also produceyielding in order for attitude change to beevidenced. Mere awareness of information isnot a sufficient condition for attitude change.Consistent with McGuire's analysis, Green-wald (1968) observed that cognitive learningin a passive context was not sufficient toproduce attitude change. Cognitive rehearsalof self-generated cognitive responses was alsonecessary for attitude change to occur. Thesubject must actively reformulate the infor-mation he has received in order for it tostimulate an internalization of attitude change.As John Dewey (1900) once wrote, "Athought is not a thought unless it is one's own(p. 66)." It seems quite reasonable to pre-sume that the social confrontation inherent indebate and discussion would motivate anactive rehearsal process, even in those who arequietly contemplating their next remark.


Closely related to the rehearsal notion isthe possibility of internalization resultingfrom the public verbal commitments that thesubject makes during discussion. His verbalexpressions are both one form of activerehearsal and a possible source of dissonance.Thus, attitude change may be motivated bya need for consistency with the verbal behav-ior elicited in discussion. Verbal commitmentcould produce the increased sense of involve-ment and certainty that Moscovici and Zavol-loni (1969) believe to be inherent in grouppolarization. It should be noted that Ebbesenand Bowers (1974) did not observe thatforced private generation of risky or cautiousarguments affected shift. However, their pro-cedure minimized both perceived choice andpublic exposure of the commitment, factorsknown to be important determinants of com-mitment effects (Kiesler, 1971). A moresubtle procedure for manipulating the direc-tion of rehearsed arguments did affect responseshift (Burnstein & Vinokur, 1973).

In summary, the evidence for informationalinfluence processes is compelling. However, italso appears that group polarization is notfully explained by a passive process of cogni-tive learning. More dynamic processes ofcognitive rehearsal and verbal commitmentare also likely contributors to the effects oftalking in groups.

Conceptual Integration

We have seen that the isolation of impor-tant variables (i.e., arguments, interpersonalcomparisons) demonstrates that each can haveeffects when other variables are held constant.But in reality the variables interact with andfeed one another rather than functioning asseparate, additive components of group influ-ence. In life outside the laboratory, the ex-pression of arguments and of one's positionare often a part of the same communicationsequence. Arguments convey implicit informa-tion about one's opinion. On the other hand,social motivations may influence the emissionof arguments. Groups composed of subjectswhose reported ideals exceed their actual re-sponses are likely to select arguments whichfavor and reinforce their aspirations.

What we need is a conceptual integration

which captures the interaction among theseelements of social influence. Our proposal fora more holistic view of the group influenceprocess (schematically represented in Fig-ure 1) is inspired by the literature on socialinfluence, which amply demonstrates thatman is indeed both a social and a rationalbeing, and by Kelman's (1974) analysis ofthe dynamics of attitude change. The schemeis simply an integrated summary of conceptsestablished in past research on group shift.It is proposed as a net which can catch thediverse set of known facts about group shiftand also place the group polarization conceptwithin the scope of attitude theory—where itmay profit from concepts developed in atti-tude research and in turn contribute to ourunderstanding of how attitudes develop intheir social context.

We assume that man's social motivationand rational faculties are dynamically inter-acting in social contexts. Figure 1 attemptsto visualize this interaction and may beverbally summarized as follows. Social moti-vation produces a small direct impact bymotivating the person to verbalize argumentsthat are socially desirable and in accord withhis ideals. By offering arguments that tendtoward the outer limits of his range of accept-ability, the individual tests his ideals and alsopresents himself favorably to the group since,as we noted earlier, extremity in the directionof the ideal connotes knowledgeability andcompetence. (The expression of arguments is,of course, also determined by the individual'sexisting cognitions.) The expression of thesearguments may have some direct impact onthe relevant attitude, through a dissonancereduction or self-attribution process, and italso constitutes both a form of cognitiverehearsal for the speaker and of informationto be received and responded to by the othergroup members. The resultant cognitive learn-ing and rehearsal contributes significantly tosubsequent attitude change. The strength ofthe evidence for informational influence ef-ects is symbolized by the double arrows fromthe "cognitive foundation." Indeed, the evi-dence indicates that some polarization effectcan result from a pooling of prediscussionarguments, apart from any added effects ofsocial motivation.


Action Commitment

(Comparison with others'attitudes—motivation to

(Arguments verbalized—tend to be biased toward

outer limits of one'slatitude of acceptance)

ATTITUDECHANGE

rehearsed—persuasivearguments related to

FIGURE 1. Summary conceptual scheme.

To summarize, social motivation (a desireto perceive and present oneself favorablyrelative to others) may change an attitudethrough the interpersonal comparison process,but it also motivates the person to expresssocially desirable arguments. This verbalcommitment may enhance the attitude, and italso serves a cognitive rehearsal function forthe speaker and a cognitive learning functionfor the listeners. Thus, the effects of socialmotivation may be partly mediated by thelearning and rehearsal that accompanies thehearing and speaking of discussion arguments.

There is one key assumption of this con-ceptual integration for which evidence has notyet been provided, namely, that discussionarguments are indeed influenced by the socialcontext. Recent research provides such evi-dence. Arguments spoken in discussion moredecisively favor the dominant alternative thando written arguments.3 (This is indicated byevidence from Bishop & Myers [1974] andEbbesen & Bowers [1974] and by a com-parison of the data from Myers & Bishop[1971] and Silverthorne [1971] with the datafrom Burnstein & Vinokur [1973], Myers& Bach [1974], Stokes [1971], and Vinokur& Burstein [1974].) In writing an advis-

ory brief or list of arguments, the subject ismore impartial than when discussing the casewith his peers. This might partly be due toa demand for balance in the instructions ofthe written brief conditions, but the findingis reliable enough to suggest the hypothesisthat people are generally more polar in con-versation than in writing. Perhaps this isbecause in conversation people are respondingto other people, whereas when working alonethe individual is responding only to the ma-terials. A field demonstration observation ofthe effects of social context on communica-tion content is Crawford's (1974) observationof Catholic priests adapting their sermonarguments to fit the perceived attitudes oftheir parish. Laboratory research (e.g., Manis,Cornell, & Moore, 1974) also confirms that"the average person will often 'censor' theinformation that he (or she) relays to others,so as to reduce the dissonance between the

0 The tendency for d'scussion arguments to beone-sided is probably not equal for all phases of agiven discussion. Studies in speech-communications(see Fisher, 1974) suggest that one-sided discussionis especially likely after a choice direction has im-plicitly emerged and group members mutually rein-force their shared inclination.


listener's apparent views and the views ex-pressed in the transmitted message (p. 92)."

The strength of the various vectors isexpected to vary across situations. In morefact-oriented judgment tasks (group problem-solving tasks being the extreme case), thecognitive determinants will likely be para-mount, although people will still be moti-vated to demonstrate their abilities. On mat-ters of social preference, in which the socialdesirability of actions is more evident, thedirect and indirect attitudinal effects of socialmotivation are likely to appear. The directimpact will occur in situations in which theindividual has ideals that may be compro-mised by presumed norms but in which expo-sure to others' positions informs him that hisideals are shared more strongly or widely thanhe would have supposed. These situations—inwhich expressed ideals are a step ahead ofprior responses—will also tend to elicit discus-sion content that is biased toward the ideals.

Informational influence mechanisms implythat shift will occur when discussion generatescognitive learning and rehearsal predomi-nantly in one direction. Usually the mean ofinitial preferences is an index to the prevail-ing direction of cognitive learning, hence thegroup polarization phenomenon. But the meaninitial preference is likely to be only a crudeindex of the predominant direction of cogni-tive change. For example, in the Myers andBach (1974) study of separated pacifistic andmilitaristic groups, arguments generated bythe two groups were similar, despite theirinitial differences in attitudes. In some otherstudies (e.g., some of the judgment and simu-lated jury experiments) the manipulation ofthe initial mean tendency may have been sounambiguous (e.g., by simple descriptive ad-jectives) as to preclude any persuasive newarguments relevant to the manipulation. It isalso possible to conceive of situations in whicha very familiar alternative, A, is initiallyfavored over an unknown alternative, B, butshift occurs toward B because the potencyof new information learned and rehearsed insupport of B is greater than it is for A. Per-haps this is why Walker and Main (1973)observed bold new libertarian decisions injudicial groups even though individual judges

did not make predominantly libertarian deci-sions when deciding alone.

SIGNIFICANCE OF RESEARCH ON GROUPPOLARIZATION

A legitimate question raised by some whohave assessed the risky-shift literature (Cart-wright, 1973; Smith, 1972) is whether all theresources and time expended will in the endprove to have been worth it. With the addi-tion of recent empirical and theoretical studieson group polarization, we again raise thequestion, What is being gained by the totalresearch effort? The payoffs appear to be oftwo types: (a) knowledge directly gainedabout the effects of group discussion and (b)indirect heuristic payoffs in the form of newideas about social influence and group process.

Knowledge Gained About Group Effects

The group polarization phenomenon is ap-parently a general outcome of group discus-sions of various sorts. The significance of thisphenomenon may be evaluated in terms ofthe importance of small group communica-tion for attitude change and in terms of pos-sible pragmatic implications, as well as interms of the actual experimental data sum-marized earlier.

What is the extent of small group influenceon attitudes? McGuire (1969) noted, "It isclear that any impact that the mass mediahave on opinion is less than that producedby informal face-to-face communication of theperson with his primary groups, his family,friends, co-workers, and neighbors (p. 231,)."Possible reasons for the greater effect of face-to-face communication are summarized inFigure 1. In social interaction the target per-son is motivated to present himself favorably,and he is engaged in active cognitive rehearsaland verbal commitment. Thus, it is not sur-prising that in Western culture, group discus-sion seems increasingly integral to our socialand organizational existence. Reliance on indi-vidual psychotherapy is being discarded infavor of mutual assistance in small groups(see Tyler, 1973), and group-centered orga-nizational methods are displacing authoritar-ian management hierarchies (see Vroom,1970, for a discussion of the impact of par-


ticipative methods on employee morale andattitudes). Pilot studies by Eraser (Note 17)suggest care in generalizing group polariza-tion studies to real decision-making groups,however.

The group polarization principle may helpto explain what some observers feel are nega-tive effects of group interaction (e.g., radicalmovements emerging from homogeneous sub-cultures and the difficulty of rehabilitation inclosed environments such as penal and mentalinstitutions), as well as effects that are oftenconsidered to be beneficial (e.g., in grouptherapy situations and in strengthened ethnicor religious identity subsequent to a groupingwith similar others). Thus, we speculate thatif the social and informational forces likely tobe evoked by social interaction are in the de-sired direction, then discussion will likely be auseful strategy for producing more "effective"attitudes. If not, then the leader should eithercontrol the communication content so as toelicit desired arguments (as Lewin, 1947, didin some of his classic experiments on groupdiscussion) or simply introduce the action asfait accompli (see Allport, 1954).

Some of the findings we have noted are ex-perimental demonstrations of "groupthink"processes, which Janis (1972) has proposed tohelp explain decision fiascoes such as the failureto anticipate the Pearl Harbor attack, theinvasions of North Korea and the Bay ofPigs, and the escalation of the Vietnam war.For example, his suggestion that group mem-bers "show interest in facts and opinions thatsupport their initially preferred policy andtake up time in their meetings to discussthem, but they tend to ignore facts andopinions that do not support their initiallypreferred policy" (p. 10) seems to be con-firmed in the finding that discussion argu-ments are more in line with the dominant polethan are written arguments. Janis attributedthis to a "suppression of deviant thoughts"resulting from "concurrence-seeking as a formof striving for mutual support" in order to"maintain self-esteem" (pp. 201-203).

This literature not only provides experimen-tal confirmation for some of the groupthinkdynamics proposed by Janis, it also suggestssome additional ways to prevent them. The

observations of comparison-induced polariza-tion suggests that a group leader might profit-ably suppress mention of initial preferenceswhile eliciting relevant arguments. The find-ing of greater polarity in discussion argumentsthan in written briefs suggests the usefulnessof generating arguments by having partici-pants individually write pro and con consider-ations for later sharing with the group. (Theindividual production of ideas has also beenshown to generate a greater quantity ofideas than does group brainstorming—see thereview by Lamm and Trommsdorff, 1973.)

One practical suggestion for doing this isthe "nominal group" technique suggested byVan de Ven and Delbecq (1971):

Imagine a meeting room in which seven to tenindividuals are sitting around a table in full view ofeach other. However, they are not speaking to eachother. Instead, each individual is writing on a padof paper in front of him. At the end of 10 to 20minutes, a very structured sharing of ideas takesplace. Each individual, in round-robin fashion, pio-vides one idea from his private list which is writtenon a flip-chart by a recorder in full view of othermembers. There is still no discussion, only therecording of privately generated ideas. This round-robin listing continues until each member indicatesthat he has no further ideas to share. . . . Gener-ally, a spontaneous discussion then follows for aperiod (in the same fashion as an interactive groupmeeting) before nominal voting. Nominal votingsimply means that the selection of priorities . . . isdone by each individual privately, and the groupdecision is the pooled outcome of the individualvotes, (p. 204)

Heuristic Payoffs

Pruitt (Note 18) regarded research ongroup shift primarily "as a heuristic proce-dure for generating a new theoretical para-digm in the general area of group locomotion(p. 3)." He suggested that the elucidation ofa convincing explanation of the shift effectmay enliven the field of group dynamics "aspeople begin to develop and test the implica-tions of the theoretical assumptions underlyingthis explanation (p. 3)."

Very recently investigators in this fieldhave begun to cease thinking of the groupchange effect as risky shift and are insteadusing the paradigm to develop basic knowl-edge about group process and about the dy-namics of attitude change in social situations


(as contrasted with research on passive recip-ients receiving programmed communications).Hopefully, theoretical ideas developed to ex-plain group polarization will be useful beyondthis paradigm.

For example, interpersonal comparison the-ory stimulates questions about the attitudinaleffect of exposure to others' attitudes. Mightthere be conditions under which social con-tagion effects result from feedback concerningothers' attitudes? Brown (1974) stated thegeneralized hypothesis clearly:

The suggestion is that just as there are manysituations in which individuals will be motivated toconform, to huddle up close to the central tendency(as in perceptual judgments), there is another classof situations in which people will be motivated tofall on one or the other side of the central tendencybecause they seek not to be average but better thanaverage, or virtuous. To be virtuous, in any of anindefinite number of dimensions, is to be differentfrom the mean—in the right direction and to theright degree, (p. 469)

For example, if an opinion poll reveals thatothers support your inclinations to an extentthat surprises you, does this tend to reinforceyour opinion on the matter? There has beena substantial amount of research on the effectsof misinforming a subject concerning others'attitudes and behavior, but almost no researchin social psychology of public opinion usinga paradigm that more closely approximatessocial reality by giving accurate informationconcerning others' attitudes. There is knownto be no general "bandwagon effect" resultingfrom the publication of political polls. Butinterpersonal comparison theory provides amore precise prediction: Attitudes will changeif and when there is a discrepancy betweenone's social perception of the opinions of hisreference group and his revised perceptionafter exposure to the actual distribution ofopinions.

Of course, we are observing others' posi-tions in many ways other than opinion polls.Alker (Note 19) suggested that recent in-creases in political attitude polarization notedin national polls "may have been facilitatedby repeated exposure in the media of personsholding extreme views." This suggestion iscompatible with Pruitt's (1971 a, 1971b) "re-lease theory" of interpersonal comparison ef-fects. The media present models who embody

their viewpoint with great clarity, consist-ency, and courage. We have noted earlier theadmiration which is accorded extreme repre-sentatives of one's viewpoint. This may there-fore release a sympathizer to then align him-self with the more extreme forms of his ownopinion.

In summary, research on group polarizationhas provided new knowledge about the out-comes of group discussion, and the need toexplain the phenomenon has stimulated newconcepts which may enrich our understandingof attitude change in social situations. Iffurther research on these concepts shouldprove to be fruitful, then we may consoleourselves that the effort represented by morethan 300 manuscripts since Stoner's (1961)discovery of the risky shift has indeed beenof value.

REFERENCE NOTES

1. Miller, H. Group decision making: The riskyshift phenomenon explained in terms of confi-dence. Unpublished manuscript, University ofMichigan, 1968.

2. Foss, R. D., & Foss, N. E. Effects of irrelevantvariables and group discussion on simulated jurydecisions. Paper presented at the meeting of theWestern Psychological Association, April 1973.

3. Rumsey, M. G., & Castore, C. H. The effect ofdefendant's character and group discussion on in-dividual sentencing judgments. Paper presented atthe meeting of the Midwestern PsychologicalAssociation, May 1974.

4. Heimbach, J. T. Social psychology in the juryroom: The effects of evidence, confession, andgroup interaction in sentencing. Paper presentedat the meeting of the Midwestern PsychologicalAssociation, May 1970.

5. Vidmar, N. Group-induced shifts in simulatedjury decisions. Paper presented at the meeting ofthe Midwestern Psychological Association, May1972.

6. Laughlin, E. R., & Izzett, R. R. Juror-defendantattitude similarity and choice shift in the jurytrial. Paper presented at the meeting of the Mid-western Psychological Association, May 1973.

7. Rettig, S. Risk shift and information flow inrelation to predicted ethical risk taking. Unpub-lished manuscript, Hunter College, 1969.

8. Shrewsberry, R. D., & Johnson, D. L. A com-parison of group and individual responses on animpression formation task. Paper presented at themeeting of the Southeastern Psychological Asso-ciation, April 1972.

9. Johnson, D. L. Small group polarization ofratings in student evaluation of faculty. Paper


presented at the meeting of the SoutheasternPsychological Association, April 1972.

10. Krapf, G. Polarization of responses in studentevaluation of teachers as a function of groupdiscussion. Unpublished master's thesis, RadfordCollege, 1972.

11. Colligan, M. J., & Giambra, L. M. A failure tofind the risky shift in an undergraduate statisticsclass. Paper presented at the meeting of theMidwestern Psychological Association, May 1970.

12. Abend, R., & Kogan, N. A mathematical scale-of-measurement model for the risky-shift effect.Paper presented at the meeting of the EasternPsychological Association, May 1973.

13. Codol, J. P. Sur le comportement dit de "con-formite superieur de soi": Vingt researches experi-mentales. Unpublished manuscript, Universite deProvence, 1974.

14. Lenihan, K. J. Perceived climates as a barrierto housing desegregation. Unpublished manu-script, Bureau of Applied Social Research, Co-lumbia University, 1965.

15. Vinokur, A., & Burnstein, E. Novel argumenta-tion and attitude change: The case of group-induced shifts on choice. Unpublished manuscript,University of Michigan, 1974.

16. Russo, J. E. Information control and the shiftto risk. Paper presented at the meeting of theEastern Psychological Association, April 1970.

17. Eraser, C, Determinants of individual and groupdecisions involving risk. England: University ofBristol, 1974.

18. Pruitt, D. G. Comment on Cartwright's articleon The Choice Shift. Unpublished manuscript,State University of New York at Buffalo, 1973.

19. Alker, H. A. Social psychological explanationsof political attitude polarization. Paper presentedat the meeting of the American Political ScienceAssociation, Chicago, 1971.

REFERENCES

Alker, H. A., & Kogan, N. Effects of norm-orientedgroup discussion on individual verbal risk takingand conservatism. Hitman Relations, 1968, 21, 393-405.

Allport, G. VV. The nature of prejudice. Cambridge,Mass.: Addison-Wesley, 1954.

Anderson, N. H. Integration theory and attitudechange. Psychological Review, 1971, 78, 171-206.

Andrews, I. R., & Johnson, D. L. Small grouppolarization of judgments. Psychonomic Science,1971, 24, 191-192.

Asch, S. E. Studies of independence and conformity:A minority of one against a unanimous majority.Psychological Monographs, 1956, 7(9, Whole No.416).

Baron, P. H., Baron, R. S., & Roper, G. Externalvalidity and the risky shift: Empirical limits andtheoretical implications. Journal of Personality andSocial Psychology, 1974, 30, 95-103.

Baron, R. S., Dion, K, L., Baron, P. H., & Miller, N.

Group consensus and cultural values as determi-nants of risk-taking. Journal of Personality andSocial Psychology, 1971, 20, 446-455.

Baron, R. S., Monson, T. C., & Baron, P. H. Con-formity pressure as a determinant of risk taking:Replication and extension. Journal of Personalityand Social Psychology, 1973, 28, 406-413.

Baron, R. S., Roper, G., & Baron, P. H. Group dis-cussion and the stingy shift. Journal of Personalityand Social Psychology, 1974, 30, 538-545.

Bateson, N. Familiarization, group discussion, andrisk-taking. Journal of Experimental Social Psy-chology, 1966, 2, 119-129.

Baumhart, R. An honest profit. New York: Holt,Rinehart & Winston, 1968.

Bell, P. R., & Jamieson, B. D. Publicity of initialdecisions and the risky shift phenomenon. Journalof Experimental Social Psychology, 1970, 6, 329-345.

Bishop, G. D., & Myers, D. G. Informational influ-ence in group discussion. Organizational Behaviorand Human Performance, 1974, 12, 92-104.

Brown, R. Further comment on the risky shift.American Psychologist, 1974, 29, 468-470.

Burns, J. F. An extremity-variance analysis of groupdecisions involving risk. Unpublished doctoral dis-sertation, Massachusetts Institute of Technology,1967. (Cited in D. G. Marquis, Individual andgroup decisions involving risk. Industrial Manage-ment Review, 1968, 9, 69-75.)

Burnstein, E. An analysis of group decisions in-volving risk ("The risky shift")' Human Relations,1969,22, 381-395.

Burnstein, E., Miller, H., Vinokur, A., Katz, S., &Crowley, J. Risky shift is eminently rational. Jour-nal of Personality and Social Psychology, 1971,20, 462-471.

Burnstein, E., & Vinokur, A. Testing two classes oftheories about group-induced shifts in individualchoice. Journal of Experimental Social Psychology,1973, 9, 123-137.

Burnstein, E., & Vinokur, A. What a person thinksupon learning he has chosen differently fromothers: Nice evidence for the persuasive-argumentsexplanation of choice shifts. Journal of Experi-mental Social Psychology, 1975, 11, 412-426.

Burnstein, E., Vinokur, A., & Pichevin, M. F. Whatdo differences between own, admired, and attrib-uted choices have to do with group induced shiftsin choice? Journal of Experimental Social Psy-chology, 1974, 10, 428-443.

Burnstein, E., Vinokur, A., & Trope, Y. Interpersonalcomparison versus persuasive argumentation: Amore direct test of alternative explanations forgroup-induced shifts in individual choice. Journalof Experimental Social Psychology, 1973, 9, 236-24S.

Byrne, D. The attraction paradigm. New York:Academic Press, 1971.

Cartwright, D. Risk taking by individuals andgroups: An assessment of research employingchoice dilemmas. Journal of Personality andPsychology, 1971, 20, 361-378.


Cartwright, D. Determinants of scientific progress:The case of the risky shift. American Psychologist,1973, 28, 222-231.

Chapko, M. K. The source of increased emphasis ofgain in predicted ethical risk taking in groups.Proceedings of the 80th Annual Convention of theAmerican Psychological Association, 1972, 7, 219-220.

Clark, R. D. III. Group-induced shift toward risk:A critical appraisal. Psychological Bulletin, 1971,76, 251-270.

Clark, R. D. Ill, Crockett, W. H., & Archer, R. L.Risk-as-value hypothesis: The relation betweenperception of self, others, and the risky shift.Journal of Personality and Social Psychology,1971, 20, 42S-429.

Clement, D. E., & Sullivan, D. W. No risky shifteffect with real groups and real risks. PsychonomicScience, 1970, 18, 243-245.

Cohen, S. L., & Ruis, C. B. Wary shift or riskyshift? Bulletin of the Psychonomic Society, 1974,3, 214-216.

Coleman, J. S. Community conflict. New York: FreePress, 1957.

Crawford, T. J. Sermons on racial tolerance and theparish neighborhood context. Journal of AppliedSocial Psychology, 1974, 4, 1-23.

Cvetkovich, G., & Baumgardner, S. R. Attitude po-larization: The relative influence of discussiongroup structure and reference group norms. Jour-nal of Personality and Social Psychology, 1973,26, 159-165.

Davis, J. H. Gioup decision and social interaction:A theory of social decision schemes. PsychologicalReview, 1973, SO, 97-125.

Davis, J. H., Kerr, N. L., Meek, D., & Rissman,A. K. The group position as a function of mem-ber attitudes—Choice shift effects from the per-spective of social decision scheme theory. Journalof Personality and Social Psychology, in press.

Davis, J. H., Kerr, N, L., Sussman, M., & R;ssman,A. K. Social decision schemes under risk. Journalof Personality and Social Psychology, 1974, JO,248-271.

Deets, M. K., & Hoyt, G. C. Variance preferencesand variance shifts in group investment decisions.Organizational Behavior and Human Performance,1970, 5, 378-386.

Dewey, J. School and society. Chicago: Universityof Chicago Press, 1900.

Dion, K. L., Baron, R. S., & Miller, N. Why dogroups make riskier decisions than individuals? InL. Berkowitz (Ed.), Advances in experimental so-cial psychology (Vol. 5). New York: AcademicPress, 1970.

Doise, W. Jugement collectif et prise de risque despetits groupes. Psychologie Franfaise, 1969, 14,87-95. (a)

Doise, W. Intergroup relations and polarization ofindividual and collective judgments. Journal ofPersonality and Social Psychology, 1969, 12, 136-143. (b)

Doise, W. L'importance d'une dimension principaledans les jugements collectifs. L'Annee Psycholo-gique, 1970, 70, 151-159.

Doise, W. An apparent exception to the extremiza-tion of collective judgments. European Journal ofSocial Psychology, 1971, 1, 511-518.

Eagly, A. H. Comprehensibility of persuasive argu-ments as a determinant of opinion change. Journalof Personality and Social Psychology, 1974, 29,758-773.

Ebbesen, E. B., & Bowers, R. J. Proportion of riskyto conservative arguments in a group discussionand choice shift. Journal of Personality and SocialPsychology, 1974, 29, 316-327.

Eisinger, R., & Mills, J. Perception of the sincerityand competence of a communicator as a functionof the extremity of his position. Journal of Ex-perimental Social Psychology, 1968, 4, 224-232.

Feldman, K. A., & Newcomb, T. The impact ofcollege on students. San Francisco: Jossey-Bass,1969.

Ferguson, D. A., & Vidmar, N. Effects of group dis-cussion on estimates of culturally appropriate risklevels. Journal of Personality and Social Psychol-ogy, 1971, 20, 436-445.

Fisher, B. A. Small group decision making. NewYork: McGraw-Hill, 1974.

Fraser, C. Group risk-taking and group polarization.European Journal of Social Psychology, 1971, 1,493-510.

Fraser, C., Gouge, C., & Billig, M. Risky shifts, cau-tious shifts, and group polarization. EuropeanJournal of Social Psychology, 1971, 1, 7-30.

Gaskell, G. D., Thomas, E. A. C., & Farr, R. M.Effects of pretesting on measures of individual riskpreferences. Journal of Personality and SocialPsychology, 1973, 25, 192-198.

Geier, J. G., Forston, R. F., & Larson, C. U. Smallgroup discuss:on versus the lecture method: Astudy in individual decision making. WesternSpeech, 1970, 34, 38-45.

Gouge, C., & Fraser, C. A further demonstration ofgroup polarization. European Journal of SocialPsychology, 1972, 2, 95-97.

Greenwald, A. G. Cognitive learning, cognitive re-sponse to persuasion, and attitude change. In A. G.GrcenwaM, T. C. Brock, & T. M. Ostrom (Eds.),Psychological foundations of attitudes. New York:Academic Press, 1968.

Greenwald, A. G. Consequences of prejudice againstthe null hypothesis. Psychological Bulletin, 1975,82, 1-20.

Hopkins, T. K. The exercise of influence in smallgroups. Totawa, N.J.: Bedminster Press, 1964.

Home, W. C., & Long, G. Effect of group discussionon universalistic-particularistic orientation. Journalof Experimental and Social Psychology, 1972, 8,236-246.

Izz-.tt, R. R., & Leginski, W. Group discussion andthe influence of defendant characteristics in asimulated jury setting. Journal of Social Psychol-ogy, 1974, 93, 271-280.


Janis, I. L. Victims oj gro-upthink. Boston: Hough-ton Mifflin, 1972.

Jellison, J. M., & Davis, D. Relationships betweenperceived ability and attitude extremity. Journalof Personality and Social Psychology, 1973, 27,430-436.

Jellison, J. M., & Riskind, J. A social comparisonof abilities interpretation of risk taking behavior.Journal of Personality and Social Psychology,1970, 15, 375-390.

Jellison, J. M., & Riskind, J. Attribution of risk toothers as a function of their ability. Journal ojPersonality and Social Psychology, 1971, 20, 413-415.

Johnson, D. L., & Andrews, I. R. The risky-shifthypothesis tested with consumer products as stim-uli. Journal of Personality and Social Psychology,1971, 20, 382-385.

Johnston, L., Jr. The immediate and short-termeffects oj group discussion and individual study onrisk-levels. Unpublished doctoral dissertation, Flor-ida State University, 1968.

Kalven, H. G., Jr., & Zeisel, H. The American jury.Boston: Little, Brown, 1966.

Kelman, H. C. Attitudes are alive and well and gain-fully employed in the sphere of action. AmericanPsychologist, 1974, 29, 310-324.

Kiesler, C. The psychology oj commitment. NewYork: Academic Press, 1971.

Kogan, N., & Wallach, M. A. Risk taking: A studyin cognition and personality. New York: Holt,Rinehart & Winston, 1964.

Kogan, N., & Wallach, M. A. Modification of ajudgmental style through group interaction. Jour-nal of Personality and Social Psychology, 1966, 4,165-174.

Kogan, N., & Wallach, M. A. Effects of physicalseparation of group members upon group risktaking. Human Relations, 1967, 20, 41-48. (a)

Kogan, N., & Wallach, M. A. Group risk taking as afunction of members' anxiety and defensivenesslevels. Journal of Personality, 1967, 35, 50-63. (b)

Lambert, R. Extremisation du comportement de prisede risque en groupe et modele majoritaire. Psy-chologie Franc,aise, 1969, 14, 113-125.

Lambert, R. Extremisation du risque en groupes.Journal de la Sode.lt de Statisque de Paris, 1971,112, 11-22.

Lamm, H. Will an observer advise higher risk takingafter hearing a discussion of the decision problem?Journal oj Personality and Social Psychology,1967, 6, 467-471.

Lamm, H., Myers, D. G., & Ochsmann, R. On pre-dicting group-induced shift toward risk or caution:A second look at some experiments. PsychologischeBeitrdge, in press.

Lamm, H., & Myers, D. G. Machiavellianism, dis-cussion time, and group shift. Social Behavior andPersonality, in press.

Lamm, H., & Ochsmann, R. Factors limiting the gen-erality of the risky-shift phenomenon. EuropeanJournal of Social Psychology, 1972, 2, 99-102.

Lamm, H., & Sauer, C. Discussion-induced shifttoward higher demands in negotiation. EuropeanJournal of Social Psychology, 1974, 4, 85-88.

Lamm, H., Schaude, E., & Trommsdorff, G. Riskyshift as a function of group members' value ofrisk and need for approval. Journal of Personalityand Social Psychology, 1971, 20, 430-435.

Lamm, H., & Trommsdorff, G. Group versus indi-vidual performance on tasks requiring ideationalproficiency. A review. European Journal of SocialPsychology, 1973, 3, 361-388.

Lamm, H., & Trommsdorff, G. Group influences onprobability judgments concerning social and polit-ical change. Psychological Reports, 1974, 35, 987-996.

Lamm, H., Trommsdorff, G., & Kogan, N. Pessi-mism-optimism and risk taking in individual andgroup contexts. Journal of Personality and SocialPsychology, 1970, IS, 366-374.

Lamm, H., Trommsdorff, G., & Rost-Schaudc, E.Self-image, perception of peers' risk acceptance andrisky shift. European Journal of Social Psychology,1972, 2, 255-272.

Lamm, H., Trommsdorff, G., & Rost-Schaude, E.Group-induced extremization: Review of evidenceand a minority-change explanation. PsychologicalReports, 1973, 33, 471-484.

LeMaine, G. Social differentiation and social origi-nality. European Journal of Social Psychology,1974, 4, 17-52.

Lewin, K. Group decision and social change. InT. M. Newcomb & E. L. Hartley (Eds.), Read-ings in social psychology. New York: Holt, 1947.

Madaras, G. R., & Bern, D. J. Risk and conservatismin group decision-making. Journal of ExperimentalSocial Psychology, 1968, 4, 350-365.

Main, E. G., & Walker, T. G. Choice shifts andextreme behavior: Judicial review of the federalcourts. Journal oj Social Psychology, 1973, PI,215-222.

Manis, M., Cornell, S. D., & Moore, J. C. Trans-mission of attitude-relevant information througha communication chain. Journal of Personality andSocial Psychology, 1974, 30, 81-94.

Marquis, D. G. Individual responsibility and groupdecisions involving risk. Industrial ManagementReview, 1962, 3, 8-23.

Marquis, D. G., & Reitz, H. J. Effect of uncertaintyon risk taking in individual group decisions. Be-havioral Science, 1969, 14, 281-288.

McCauley, C. R. Extremity shifts, risky shifts andattitude shifts after group discussion. EuropeanJournal of Social Psychology, 1972, 2, 417-436.

McCauley, C., Kogan, N., & Teger, A. I. Ordereffects in answering risk dilemmas for self andothers. Journal of Personality and Social Psychol-ogy, 1971, 20, 423-424.

McCauley, C., Stitt, C. L., Woods, K., & Lipton, D.Group shift to caution at the race track. Journalof Experimental Social Psychology, 1973, 9, 80-86.

McGuire, W. J. The nature of attitudes and attitudechange. In G. Lindzey & E. Aronson (Eds.),


Handbook of social psychology (Vol. 3). Reading,Mass.: Addison-Wesley, 1969.

McGuire, W. J. Attitude change: The informationprocessing paradigm. In C. G. McClintock (Ed.),Experimental social psychology. New York: Holt,Rinehart & Winston, 1972.

McKeachie, W. J. Psychology at age 75: The psy-chology teacher comes into his own. AmericanPsychologist, 1968, 23, SS1-S57.

Miller, K. M., & Biggs, J. B. Attitude changethrough undirected group discussion. Journal ofEducational Psychology, 1958, 4°, 224-228.

Minton, H. L., & Miller, A. G. Group risk takingand internal-external control of group members.Psychological Reports, 1970, 26, 431-436.

Mitnick, L., & McGinnies, E. Influencing ethno-centrism in small discussion groups through a filmcommunication. Journal of Abnormal and SocialPsychology, 1958, 56, 82-90.

Morgan, C. P., & Aram, J. D. The preponderanceof arguments in the risky shift phenomenon.Journal of Experimental Social Psychology, 1975,11, 25-34.

Moscovici, S., Doise, W., & Dulong, R. Studies ingroup decision II: Differences of positions, differ-ences of opinion and group polarization. EuropeanJournal of Social Psychology, 1972, 2, 385-400.

Moscovici, S., & Lecuyer, R. Studies in group deci-sion I: Social space, patterns of communicationand group consensus. European Journal of SocialPsychology, 1972, 2, 221-244.

Moscovici, S., & Neve, P. Studies on polarizationof judgments: III. Majorities, minorities and socialjudgments. European Journal of Social Psychology,1973, 3, 479-484.

Moscovici, S., & Zavalloni, M. The group as apolarizer of attitudes. Journal of Personality andSocial Psychology, 1969, 12, 125-135.

Moscovici, S., Zavalloni, M., & Louis-Guerin, C.Studies on polarization of judgments: I. Groupeffects on person perception. European Journal ofSocial Psychology, 1972, 2, 87-91.

Moscovici, S., /avalloni, M., & Weinberger, M.Studies on polarization of judgments: II. Personperception, ego involvement and group interaction.European Journal of Social Psychology, 1972, 2,92-94.

Myers, D. G. Summary and bibliography of experi-ments on group-induced response shift. Catalog ofSelected Documents in Psychology, 1973, 3, 123.

Myers, D. G. Social comparison processes in choicedilemma responding. Journal of Psychology, 1974,86, 287-292.

Myers, D. G. Discussion-induced attitude polariza-tion. Human Relations, 1975, 28, 699-714.

Myers, D. G., & Aronson, S. J. Enhancement ofdominant risk tendencies in group discussion. Psy-chological Reports, 1972, 30, 615-623.

Myers, D. G., & Bach, P. J. Discussion effects onmilitarism-pacifism: A test of the group polariza-tion hypothesis. Journal of Personality and SocialPsychology, 1974, 30, 741-747.

Myers, D. G., & Bach, P. J, Group discussion effectson conflict behavior and self-justification. Psycho-logical Reports, 1976, 38, 135-140.

Myers, D. G., Bach, P. J., & Schreiber, F. B. Nor-mative and informational effects of group inter-action. Sociometry, 1974, 37, 275-286.

Myers, D. G., & Bishop, G. D. Discussion effects onracial attitudes. Science, 1970, 169, 778-789.

Myers, D. G., & Bishop, G. D. Enhancement ofdominant attitudes in group discussion. Journal ofPersonality and Social Psychology, 1971, 20, 386-391.

Myers, D. G., & Kaplan, M. F. Group-inducedpolarization in simulated juries. Personality andSocial Psychology Bulletin, 1976, 2, 63-66.

Myers, D. G., & Murdoch, P. H. Is risky shift dueto disproportionate influence by extreme groupmembers? British Journal of Social and ClinicalPsychology, 1972, 11, 109-114.

Myers, D. G., Murdoch, P. H., & Smith, G. F.Responsibility diffusion and drive enhancement ef-fects on risky shift. Journal of Personality, 1970,38, 418-425.

Myers, D. G., Schreiber, F. B., & Viel, D. Effects ofdiscussion on opinions concerning illegal behavior.Journal of Social Psychology, 1974, 92, 77-84.

Myers, D. G., Wong, D. W., & Murdoch, P. H.Discussion arguments, information about others'responses and risky shift. Psychonomic Science,1971, 24, 81-83.

Paicheler, G., & Bouchet, J. Attitude polarization,familiarization, and group process. European Jour-nal of Social Psychology, 1973, 3, 83-90.

Pettigrew, T. F. Racially separate or together?Journal of Social Issues, 1969, 25, 43-70.

Pruitt, D. G. Choice shifts in groups discussion: Anintroductory review. Journal of Personality andSocial Psychology, 1971, 20, 339-360. (a)

Pruitt, D. G. Conclusions: Toward an understandingof choice shifts in group discussion. Journal ofPersonality and Social Psychology, 1971, 20, 495-510. (b)

Pruitt, D. G., & Teger, A. I. Is there a shift towardrisk in group discussion? If so, is it a group phe-nomenon? If so, what causes it? In J. R. Braun(Chair), The Risky Shift Phenomenon: CurrentStatus. Symposium presented at the meeting of theAmerican Psychological Association, 1967.

Pruitt, D. G., & Teger, A. I. The risky shift ingroup betting. Journal of Experimental SocialPsychology, 1969, 5, 115-126.

Rabbie, J. M., & Visser, L. Bargaining strength andgroup polarization in inter-group polarization.European Journal of Social Psychology, 1972, 2,401-416.

Rettig, S. Group discussion and predicted ethical risktaking. Journal of Personality and Social Psychol-ogy, 1966, 3, 629-633.

Rettig, S. Trust and locus of control in predictivejudgments of unethical behavior. Proceedings ofthe 80th Annual Convention of the American Psy-chological Association, 1972, 7, 221-222.


Rettig, S., & Turoff, S. F. Exposure to group dis-cussion and predicted ethical risk taking. Journalof Personality and Social Psychology, 1967, 7,177-180.

Roberts, J. C., & Castore, C. H. The effects of con-formity, information, and confidence upon sub-jects' willingness to take risk following a gioupdiscussion. Organizational Behavior and HumanPerformance, 1972, 8, 384-394.

Robinson, K. F. An experimental study of the effectsof group discussion upon the social attitudes ofcollege students. Speech Monographs, 1941, 8, 34-57.

Roseborough, M. E. Experimental studies of smallgroups. Psychological Bulletin, 1953, 50, 275-303.

Runyan, D. L. The group risky-shift effect as afunction of emotional bonds, actual consequences,and extent of responsibility. Journal of Personal-ity and Social Psychology, 1974, 29, 670-676.

Schroeder, H. E. The risky shift as a general choiceshift. Journal of Personality and Social Psychol-ogy, 1973, 27, 297-300.

Sears, D. 0. Political behavior. In G. Lindzey &E. Aronson (Eds.), Handbook of social psychology(Vol. 5). Reading, Mass.: Addison-Wesley, 1969.

Semin, G. R., & Glendon, A. I. Polarization and theestablished group. British Journal of Social andClinical Psychology, 1973, 12, 113-121.

Silverthorne, C. F. Information input and the groupshift phenomenon in risk taking. Journal of Per-sonality and Social Psychology, 1971, 20, 456-461.

Smith, M. B. Is experimental social psychology ad-vancing? Journal of Experimental Social Psychol-ogy, 1972, 8, 86-96.

St. Jean, R. A reformulation of the value hypothesisin group risk taking. Proceedings of the 78thAnnual Convention of the American PsychologicalAssociation, 1970, S, 339-340.

St. Jean, R., & Percival, E. The role of argumenta-tion and comparison processes in choice shifts:Another assessment. Canadian Journal of Behav-ioural Science, 1974, 6, 297-308.

Stokes, J. P. Effects of familiarization and knowl-edge of others' odd choices on shifts to risk andcaution. Journal of Personality and Social Psy-chology, 1971, 20, 407-412.

Stoner, J. A. F. A comparison of individual andgroup decisions involving risk. (Unpublishedmaster's thesis, Massachusetts Institute of Tech-nology, 1961.) Cited in D. G. Marquis, Individualresponsibility and group decisions involving risk.Industrial Management Review, 1962, 3, 8-23.

Swap, W. C., & Miller, H. Risky shift and socialexchange in dyads. Psychonomic Science, 1969, 17,249-250.

Teger, A. I., & Pruitt, D. G. Components of grouprisk taking. Journal of Experimental Social Psy-chology, 1967, 3, 189-205.

Tyler, L. E. Design for a hopeful psychology.American Psychologist, 1973, 28, 1021-1029.

Utterback, W. E. The influence of conference onopinion. Quarterly Journal of Speech, 1950, 35,365-370.

Utterback, W. E. Independent variables in the con-ference situation. Quarterly Journal of Speech,1954, 40, 381-387.

Utterback, W. E. Majority influence and cogency ofargument in discussion. Quarterly Journal ofSpeech, 1962, 48, 412-414.

Van de Ven, A. H., & Delbecq, A. L. Nominalversus interacting group processes. Academy ofManagement Journal, 1971, 14, 203-211.

Vidmar, N. Effects of group discussion on categorywidth judgments. Journal of Personality and So-cial Psychology, 1974, 29, 187-195.

Vinokur, A. Distribution of initial risk levels andgroup decisions involving risk. Journal of Person-ality and Social Psychology, 1969, 13, 207-214.

Vinokur, A. Cognitive and affective processes influ-encing risk-taking in groups: An expected utilityapproach. Journal of Personality and Social Psy-chology, 1971, 20, 472-486. (a)

Vinokur, A. A review and theoretical analysis of theeffects of group processes upon individual andgroup decisions involving risk. Psychological Bul-letin, 1971, 76, 231-250. (b)

Vinokur, A., & Burnstein, E. The effects of partiallyshared persuasive arguments on group inducedshifts: A group problem solving approach. Journalof Personality and Social Psychology, 1974, 29,305-315.

Vinokur, A., Trope, Y., & Burnstein, E. A decisionmaking analysis of persuasive argumentation andthe choice-shift effect. Journal of ExperimentalSocial Psychology, in press.

Vroom, V. E. Industrial social psychology. In G.Lindzey & E. Aronson (Eds.), Handbook of socialpsychology (Vol. 5). Reading, Mass.: Addison-Wesley, 1970.

Walker, T. G., & Main, E. C. Choice-shifts in po-litical decision making: Federal judges and civilliberties cases. Journal of Applied Social Psychol-ogy, 1973, 2, 39-48.

Wallach, M. A., & Kogan, N. The roles of informa-tion, discussion and consensus in group risk taking.Journal of Experimental Social Psychology, 1965,1, 1-19.

Wallach, M. A., Kogan, N., & Bern, D. G. Groupinfluence on individual risk taking. Journal ofAbnormal and Social Psychology, 1962, 65, 75-86.

Wallach, M. A., Kogan, N., & Burt, R. B. Grouprisk taking and field dependence-independence ofgroup members. Sociometry, 1967, 30, 323-338.

Wallach, M. A., Kogan, N., & Burt, R. B. Are risktakers more persuasive than conservatives in groupdiscussion? Journal of Experimental Social Psy-chology, 1968, 4, 76-88.

Zaleska, M. The effects of discussion on group andindividual choices among bets. European Journalof Social Psychology, 1974, 4, 229-250.

Zaleska, M. Majority influence on group choicesamong bets. Journal of Personality and SocialPsychology, 1976, 33, 8-17.

(Received February 28, 1975)