Effects of Evidence on Attitudes: Is Polarization the Norm?

Effects of Evidence on Attitudes: Is Polarization the Norm? Author(s): Deanna Kuhn and Joseph Lao Source: Psychological Science, Vol. 7, No. 2 (Mar., 1996), pp. 115-120 Published by: Sage Publications, Inc. on behalf of the Association for Psychological Science Stable URL: http://www.jstor.org/stable/40062922 . Accessed: 15/06/2014 23:00 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Sage Publications, Inc. and Association for Psychological Science are collaborating with JSTOR to digitize, preserve and extend access to Psychological Science. http://www.jstor.org This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/action/showPublisher?publisherCode=sage http://www.jstor.org/action/showPublisher?publisherCode=assocpsychsci http://www.jstor.org/stable/40062922?origin=JSTOR-pdf http://www.jstor.org/page/info/about/policies/terms.jsp http://www.jstor.org/page/info/about/policies/terms.jsp EFFECTS OF EVIDENCE ON ATTITUDES: Is Polarization the Norm? Deanna Kuhn and Joseph Lao PSYCHOLOGICAL SCIENCE Research Report Columbia University Abstract- A 1979 study by Lord, Ross, and Lepper has been widely cited as showing that examination of mixed evidence on a topic leads to polarization of attitudes. The polarization phe- nomenon, we suggest, in fact encompasses two distinct change patterns- a shift from an initially moderate to a more extreme position (regarded here as genuine polarization) and a shift from an initially neutral to a moderate position (which might better be termed (< articulating a position"). The findings re- ported here indicate that genuine polarization is a real but in- frequent outcome of exposure to mixed evidence. In addition, we offer data in support of the view that the effective compo- nent of such interventions is simply cognitive engagement, rather than exposure to new evidence. A 1979 study by Lord, Ross, and Lepper in Journal of Per- sonality and Social Psychology has had remarkable impact for a single piece of research, especially one that has major meth- odological limitations acknowledged by its authors and, until recently, has been neither replicated nor refined. Descriptions of the ' 'biased assimilation" of mixed evidence and subsequent 4 'attitude polarization" reported by Lord et al. now commonly appear in introductory textbooks (Myers, 1990; Worchel, Coo- per, & Goethals, 1991). The uncritical acceptance of this study is even more striking because the findings contradict an as- sumption basic to much educational thought and prevalent in our culture more broadly - the assumption that engaging people in thinking about an issue will lead them to think better about the issue. The study by Lord et al. challenges this assumption by showing that examination of mixed evidence regarding the efficacy of capital punishment led college students to biased evaluation of the evidence (treating evidence favoring their own opinions less critically than contrary evidence), followed by increased extremity of attitudes. In other words, rather than moderating their views after contemplating mixed evidence, subjects became more extreme in their original opinions - the polarization phenomenon. In discussing these findings, Lord et al. took the position that biased assimilation is not necessarily nonnormative. Selective use of evidence to bolster one's orig- inal opinion (the polarization phenomenon), they claimed, how- ever, is a cause for alarm and suggests that exposure to evidence regarding complex, controversial issues may have effects that are more detrimental than beneficial- a conclusion having sweeping practical implications if taken seriously. The major methodological weakness in the Lord et al. study is its confinement to self-reported rather than directly assessed attitude change. Despite the fact that the two constructs are quite different, this feature is rarely noted in citations of the study. Lord et al. explained in a footnote that they were pre- cluded from using the more straightforward direct assessment of pre- and posttest attitude that they had intended to use be- cause many subjects' initial attitudes were at the extremes of the scale (precluding the possibility of observing polarization in a pretest-posttest design). The researchers therefore substituted a measure in which subjects reported whether they believed their attitudes had changed following examination of the evi- dence. It was for this self-report measure that Lord et al. ob- tained evidence of polarization. Noting this distinction, Miller, McHoskey, Bane, and Dowd (1993) conducted a study that included both self-reported and di- rect assessment of attitude change. Although they used differ- ent materials, their design was similar to that of Lord et al. (1979). For capital punishment (one of two topics they exam- ined), Miller et al. found significant self-reported polarization, replicating the findings of Lord et al. When directly assessed, however, in a pretest-posttest design, overall attitude change was nonsignificant. Still, this result does not resolve the question of whether polarization ever occurs as the consequence of exposure to mixed evidence. Some subgroup of subjects may genuinely ex- hibit polarization, even if they do not produce an overall sig- nificant effect because they are few in number or are offset by another subgroup who show centration (movement toward the midpoint of the scale). Such a result would be of theoretical and practical significance and is investigated here, using the original stimulus materials of Lord et al. and including an examination of individual patterns of change. In addition to probing the existence and frequency of the polarization phenomenon, we address the conditions for its oc- currence. If polarization is not in fact a consequence of expo- sure to mixed evidence, this finding must be reconciled with a large and diverse body of literature that reports polarization as a consequence simply of thought about a topic (Chaiken & Yates, 1985; Downing, Judd, & Brauer, 1992; Liberman & Chaiken, 1991; Millar & Tesser, 1986; Tesser, 1978; Tesser & Leone, 1977; Tesser, Martin, & Mendolia, in press). This liter- ature, which has rarely been connected to the study by Lord et al., raises the possibility that the effective component of their treatment is simply cognitive engagement with the topic, rather than exposure to the specific evidence presented (or any evi- dence at all). To investigate this possibility, we compared an evidence evaluation condition with several comparison condi- tions in which subjects were engaged with the topic but did not examine evidence. Finally, if polarization is a significant consequence of exam- ination of evidence on a topic, it should be enduring. To inves- tigate this question, we included a delayed posttest in which subjects' attitudes were reassessed 2 weeks following their ini- tial participation. Address correspondence to Deanna Kuhn, Department of Psychol- ogy, Teachers College, Columbia University, New York, NY 10027; e-mail: [email protected]. VOL. 7. NO. 2, MARCH 1996 Copyright © 1996 American Psychological Society 115 This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/page/info/about/policies/terms.jsp PSYCHOLOGICAL SCIENCE Attitude Polarization METHOD Subjects Subjects were ethnically diverse students at an urban com- munity college where enrollees are somewhat older and less academically prepared than typical undergraduates. Subjects were enrolled in introductory courses in psychology or critical thinking skills and participated as a course requirement. The group of subjects included 93 in the evidence condition, 48 in a dialogue condition, 38 in an essay condition, and 49 in a low- engagement condition. Procedure Evidence condition The evidence that subjects were asked to evaluate was iden- tical to that used in the original Lord et al. study.1 However, both direct and self-reported assessments of attitude change were included. To minimize the problem of extreme pretest scores encountered by Lord et al. (and leading them to abandon direct assessment), our 13-point opinion scale included elon- gated ends. For example, the 3 most extreme points at the positive end of the scale ( + 4, +5, +6) all included the state- ment "I am totally in favor of capital punishment," but the + 5 and + 6 positions included additional statements: "I will never change my mind, no matter what new information or arguments I hear" ( + 5 and +6 positions) and "I can't imagine how any- one else could believe differently" ( + 6 position).2 This tech- nique was effective in minimizing use of the extremes of the scale ( + 6 and - 6) at the initial assessment. The midpoint (0) position on the scale was labeled "I have mixed or undecided feelings about capital punishment," and the adjacent points (+1, - 1) also included an indication of uncertainty (e.g., for + 1, "I am somewhat in favor of capital punishment but I'm not sure"). A 21 -point self-report scale was included at posttests, with subjects asked to indicate how their views about capital pun- ishment (CP) had changed as a result of participation in the study. The midpoint of the scale indicated no change, and the remaining points bore labels varying from "slightly more in favor" to "much more in favor" along the positive side, with corresponding labels on the negative side. At each assessment, subjects were also asked to write a brief essay that "ex- plains and justifies and supports your view about capital pun- ishment." Following Lord et al., we counterbalanced the specific con- tent of the pro and con evidence (whether it involved a com- parison of states with and without CP or a comparison of crime rates before and after institution of CP) across the two positions (pro and con) and counterbalanced order of presentation (pro or con first) across subjects. Initial assessment was followed im- mediately by exposure to pro and con evidence, a question designed to enhance processing of the evidence ("Please sum- marize your reactions to the studies you've just read"), and then posttest assessment. The posttest was repeated 2 weeks later. Following Lord et al., pre- and posttest assessments in- cluded both an opinion measure (already described) and a belief measure ("How certain are you that CP reduces crime?"). The latter was also a 13-point scale, with points ranging from "to- tally certain CP reduces crime" to "totally certain CP does not reduce crime." Both posttests included the self-report scale. Dialogue condition Subjects in this condition began with the same assessment as subjects in the evidence condition. Each subject was then paired with a partner and assigned the task of engaging in a phone conversation that evening with the partner on the CP topic. The partners were asked to try to resolve any differences and to prepare a one- to two-page joint position paper due in class 5 days later, at which point a posttest assessment occurred. Essay and low-engagement conditions Subjects in the essay condition received the same assess- ments at the same intervals as subjects in the dialogue condi- tion, with no intervening treatment. Subjects in the low- engagement condition received only the direct (opinion and be- lief) and self-reported scale assessments (essay justifications were omitted) at the same pre- and posttest intervals, the pur- pose of this condition being to reduce to a minimum subjects' cognitive engagement with the topic. RESULTS Evidence Condition Consistent with the findings of Lord et al. (1979) and Miller et al. (1993), following exposure to mixed evidence, a sizable number of subjects (30%) reported that their opinions had be- come more extreme. For change scores based on direct assess- ment, however, our results (like those of Miller et al.) do not show the polarization phenomenon reported by Lord et al.: Change scores overall did not differ significantly from zero for either opinion or belief, M = 0.086, t(92) = -0.47, p > .64, for opinion and M = -0.420, f(92) = 0.42, p > .67, for belief. Side (pro vs. con), strength of opinion (moderate vs. strong), and their interaction also failed to yield significant effects. (Nor did effects appear for sex, course enrolled in, or order of evidence presentation, for either direct or reported change scores.) For reported change, strength of opinion was nonsignificant, but the effect of side (pro, con, or neutral) was significant, F(2, 90) = 10.71, p < .001. Magnitude of reported polarization was 1. We thank Lord et al. for making these materials available to us. 2. Extending the attitude scale to points beyond "I am totally in favor of (opposed to) capital punishment" (which we considered essen- tial to avoiding the ceiling problems encountered by Lord et al., as explained in the text) necessitated incorporating a certainty component into the scale, because one cannot get much more extreme than "totally in favor of or "totally opposed to" as a pure measurement of the attitude itself (and we wanted to attach verbal labels to the scale points to maximize the task's meaningfulness to subjects). A study by Pomer- antz, Chaiken, and Tordesillas (in press) provides justification for in- corporating certainty into an extremity scale for the present purposes of examining polarization. Factor analytic methods showed extremity and certainty to define a common factor that differed from other aspects of attitude and predicted subsequent processing. 116 VOL. 7, NO. 2, MARCH 1996 This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/page/info/about/policies/terms.jsp PSYCHOLOGICAL SCIENCE Deanna Kuhn and Joseph Lao comparable to that reported by Miller et al. - an average of 1.98 points in the pro direction for subjects with initial pro positions and 1.75 points in the con direction for subjects with initial con positions, with neutral subjects reporting a mean change of 0.38 points in a con direction. Unlike Lord et al. or Miller et al., we included subjects with initial neutral opinions in these analyses, although the results remained comparable when neutral sub- jects were excluded. Our rationale for retaining neutral subjects is that the polar- ization construct potentially encompasses two distinct phenom- ena: (a) initially neutral or near-neutral opinions that become moderately (or strongly) pro or con and (b) initially moderate (or strong) positions that become stronger. Despite the overall nonsignificant change when attitudes are assessed directly, did any subjects show either of these patterns? Analysis of individ- ual patterns of performance showed that attitudes of 6 of the 93 subjects became more extreme, while about the same num- ber - 8 subjects - showed less extreme posttest attitudes (cen- tration). In these analyses, changes of 2 or more scale points were classified as genuine change. Changes of 1 scale point were considered nonsignificant deviations from initial position and were classified as no change. (Nine subjects changed 1 point in a polarization direction, and 8 changed 1 point in a centration direction.) Of the 6 subjects whose attitudes increased in extremity, 4 changed from a neutral or near-neutral ( - 1 , 0, or + 1) to a mod- erate (-2 to -4 or +2 to +4) position. The other 2 changed from a moderate to a more extreme position. In our view, only the latter pattern warrants the label of polarization. The former might more appropriately be termed "articulating a position." Of the 4 subjects in the neutral-to-moderate group, 2 reported change in the direction consistent with their directly assessed change scores; 2 reported that there had been no change in their views. Of the 2 genuine polarizers, both reported change in a direction consistent with their directly assessed change scores. Thus, although polarizers appear to be aware that their opinions have become more extreme, subjects who are articulating a position may not be aware of change. When genuine polarization does occur, is it enduring? The answer here is affirmative for the 2 relevant subjects. In one case, the subject's second-posttest position remained at its ex- treme (+6) first-posttest level, and the self-report measure sim- ilarly remained at the most extreme scale point. The other sub- ject's opinion at the second posttest also remained at its first- posttest level; however, at this time, the subject reported "no change" in opinion as a result of participation in the study. Thus, awareness of having changed may disappear once the individual has become accustomed to or integrated the new position. Qualitative data from the essay justifications corroborate these two cases as genuine instances of attitude polarization. For example, the first subject's initial justification of a moder- ately pro position read in part as follows: I believe CP can be a deterrent to crime, even though some people are sound critics of it. . . . People will think otherwise before they commit crimes. . . . If the law was that if someone kills intentionally has to be sentenced to death, then in fact people will be intimidated. Therefore the rate of crime committed everyday would decrease drastically. After then reading and being asked to summarize the mixed evidence, the subject stated simply, "The study shows that CP deters." Consistent with the "biased assimilation" reported by Lord et al., the subject made no mention of the contrasting evidence. In the immediately following justification of his opin- ion, which had polarized to the extreme pro position, the sub- ject incorporated the attitude-consistent evidence into his jus- tification: The study shows that the states with CP have a lower level of crimes. . . . Though some people argue that CP does not deter, I be- lieve it does. Due to lack of CP in some states, mostly we see the rate of crime skyrocketing. If criminals had to be punished accordingly, they would think otherwise before they commit crimes. By the time of the second posttest, the subject's scale position remained at the extreme position and the reasoning underlying it had solidified and simplified (e.g., there was no longer any mention that other people may think differently), with the new evidence thoroughly incorporated and no longer differentiated from the belief: If every state had CP, crime rates would decrease. Criminals would be afraid because they know they will be punished accordingly if they commit crimes. Comparison Conditions Results for the remaining conditions are compared to those for the evidence condition in Tables 1, 2, and 3. The results presented are for the opinion scale only; findings were compa- rable for the belief scale. Table 1 presents mean amount of change by condition, irrespective of its direction, for both di- rectly assessed and self-reported change. Note that mean amounts of direct and self-reported change in Table 1 cannot be Table 1. Mean absolute amount of directly assessed or reported change by condition Condition Change Evidence Dialogue Essay Low-engagement measure (n = 93) (n = 48) (n = 38) (n = 49) Direct assessment 0.82 1.02 0.92 0.73 Self-report 1.88 1.42 0.89 0.86 VOL. 7, NO. 2, MARCH 1996 117 This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/page/info/about/policies/terms.jsp PSYCHOLOGICAL SCIENCE Attitude Polarization Table 2. Numbers and percentages of subjects showing various patterns of directly assessed change Condition Change Evidence Dialogue Essay Low-engagement patterna (n = 93) (n = 48) (n = 38) (n = 49) Polarization 2 (4%) 5 (17%) 3 (16%) 1 (4%) Articulation 4 (14%) 2 (12%) 3 (20%) 1 (6%) Centration 8(11%) 4(11%) 4(14%) 2(5%) Side change 2 (3%) 0 (0%) 1 (3%) 2 (5%) No change 77(83%) 37(77%) 27(71%) 43(88%) Note. Percentages do not add to 100% as they are based on the number of subjects for whom each pattern was possible (see text). aPolarization = movement from an initially moderate to a more extreme position (with one exception, as noted in the text); articulation = movement from a neutral or near-neutral to a moderate position; centration = movement from a more to a less extreme position; side change = movement from a pro to a con position, or vice versa; no change = movement of 1 or no points in any direction. compared directly because different scales are involved. Di- rectly assessed change does not differ significantly across con- ditions, F(3, 224) = 0.34, p > .79. Similarly, reported change (Row 2, Table 1) does not differ overall across conditions, F(3, 224) = 1.96, p > .12; however, a contrast of the evidence and dialogue conditions against the essay and low-engagement con- ditions yields a marginally significant difference, t(222) = 1.99, p = .048. Table 2 (for directly assessed change) and Table 3 (for self- reported change) provide a picture of the numbers of subjects showing change and the patterns of change exhibited. In Table 2, percentages do not add to 100% because each is based on the number of subjects for whom a given pattern was possible; in the cases of centration and side change, for example, this num- ber excludes subjects whose initial scores were neutral. Except for 1 subject in the essay condition who changed from a near- neutral ( + 1) to an extreme ( + 5) position and was classified as showing the polarization pattern, all subjects classified as show- ing this pattern changed from an initially moderate to a more extreme position. All subjects showing the articulation pattern changed from a neutral or near-neutral to a moderate position. There are two main points we wish to make regarding the directly assessed change summarized in Tables 1 and 2. First, directly assessed change is infrequent across all conditions. Second, and this is our most important point, the various forms of change portrayed in Table 2 occur at least as commonly in other kinds of engagement conditions as they do in the evidence condition. An implication is that it may not be the evaluation of evidence per se that is the effective component responsible for the change that occurs in this condition. Consistent with this contention are data for the low-engagement condition. It is of course not possible to completely eliminate engagement be- cause assessment itself is a form of engagement (Downing et al., 1992), but when engagement is minimized, as it was in this Table 3. Percentages of subjects showing various patterns of reported change Condition Change pattern Evidence Dialogue Essay Low-engagement Report of change in same direction as initial opinion (articulation or polarization) :a Consistent with directly assessed change scoreb 5% 4% 3% 2% Inconsistent with directly assessed change score 24% 13% 11% 10% Report of change in direction opposite to initial opinion (centration or side change): Consistent with directly assessed change scoreb 1% 0% 0% 4% Inconsistent with directly assessed change score 0% 15% 5% 2% Report of no change 70% 69% 82% 82% aSubjects having initial neutral opinions and reporting change are included in this group. bReported change was classified as consistent even if directly assessed change was in the reported direction but of a magnitude of only 1 point (classified as no change in the analysis of directly assessed change scores). 118 VOL. 7, NO. 2, MARCH 1996 This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/page/info/about/policies/terms.jsp PSYCHOLOGICAL SCIENCE Deanna Kuhn and Joseph Lao condition, incidence of the various change patterns overall di- minishes even further.3 Reported change, as reflected in Tables 1 and 3, is slightly more frequent in the evidence and dialogue conditions than it is in the conditions in which subjects' opinions were merely as- sessed (essay and low-engagement conditions). It is possible, therefore, that subjects asked to process new information or listen to other people's opinions feel under some pressure to change their opinions. Also notable in Table 3, and consistent with the preceding conjecture, is the fact that in only a small proportion of cases is reported change also reflected in directly assessed change. DISCUSSION The major conclusions we believe can be drawn from these findings are the following: 1. Polarization, which should be distinguished from a pattern of change from neutral or near-neutral to moderate opinion, is not an inevitable, nor even a common, consequence of ex- posure to mixed evidence. A small percentage of cases do fit this pattern, however. In these few cases, polarization ap- pears genuine (detectable by direct change measures), dura- ble (in repeated assessment), and phenomenologically acces- sible (in self-report measures). 2. Because other forms of cognitive engagement with a topic are as likely to lead to polarization as is the evaluation of mixed evidence, simple engagement is the most probable effective component of such interventions (and thus the most parsimonious explanation for the change observed). Accordingly, the evaluation of mixed evidence is best re- garded as a subtype of such engagement, rather than as an experience uniquely associated with attitude polarization. 3. Although by no means universal, self-reported polarization is much more common as a consequence of exposure to mixed evidence than is directly assessed change. It is not clear how self-reported polarization should be interpreted, beyond being differentiated clearly from directly assessed attitude polarization. Because self-reported polarization is more likely to occur when subjects are exposed to new sources of information (in contrast to other forms of engage- ment) and often occurs in the absence of directly assessed change, its interpretation as a demand characteristic of such interventions is perhaps the initial hypothesis warranting in- vestigation. If these conclusions are accepted, an immediate question of both theoretical and practical importance is whether there is any way to identify the minority of individuals whose engage- ment with a topic is likely to result in polarization. Researchers who have examined polarization as a consequence of engage- ment (Chaiken & Yates, 1985; Liberman & Chaiken, 1991; Millar & Tesser, 1986) have suggested consistency of initial attitude as a condition for polarization. Individuals whose atti- tudes consist of unintegrated, possibly inconsistent fragments, it is proposed, will not show polarization, whereas those who have consistent, integrated positions may become more ex- treme in these positions following contemplation. In the present case, an index of consistency is derivable by examining the relation between a subject's positions on the be- lief and opinion scales. Lord et al. reported similar results for the two scales and treated them as equivalent measures for assessing the impact of evidence. Conceptually, however, they are quite different. We would expect examination of empirical evidence regarding the effects of CP to have direct influence on a subject's beliefs about such effects. This influence might or might not lead to change in the subject's opinion regarding the desirability of CP. Our data are consistent with this interpretation. On the belief scale, compared with the opinion scale, subjects in the evidence condition were somewhat more likely to change, with belief becoming either stronger (10 subjects, vs. 6 on the opinion scale) or less strong (13 subjects, vs. 8 on the opinion scale). Data on consistency scores (the difference between a subject's position on the opinion scale and position on the belief scale) are also revealing in this respect. Mean difference at the initial assessment was 2.00 points for the evidence group. If we divide subjects into high-consistency (difference ̂ 2.00; n = 53) and low-consistency (difference > 2.00; n = 40) groups, the high- consistency group shows significantly less belief change than the low-consistency group - means of 1.06 versus 2.20 scale points, r(91) = 3.23, p < .002. This difference does not appear for opinion change scores. Thus, the evidence has relatively less influence on beliefs of high-consistency subjects and rela- tively greater influence on beliefs of low-consistency subjects. If high-consistency subjects do show change, however, they may be the most likely candidates for polarization of their con- sistent beliefs and opinions. (Of the 2 subjects showing opinion polarization, both had high consistency scores.) An additional finding that merits mention is an increase in consistency with time. In the evidence condition, by the time of the first posttest, the initial mean difference of 2.00 between opinion and belief scores had decreased to 1.41, and by the second posttest it had decreased further to 1.06. This finding is significant in suggesting that engagement has the positive effect of increasing consistency of attitudes. In the specific case of the evidence condition and our opinion-belief measure, engagement appears to have had the effect of bringing a subject's opinion regarding the desirability of CP into greater alignment with the subject's belief regarding the efficacy of CP. More broadly, and in summary, the small proportion of in- stances in which engagement leads to genuine polarization does not appear to warrant renouncement of the traditional wisdom that cognitive engagement can be expected to have largely pos- itive consequences, especially in view of findings (such as the one just noted) that offer evidence of positive effects. More comprehensive evidence, we believe, requires qualitative ex- 3. A further supportive finding is the fact that in the dialogue con- dition, all partner-subject combinations (of pretest opinions) occurred among polarizers. Polarization is hence not predictable from partner mismatch, suggesting that simple cognitive engagement (rather than any particular influence of ideas introduced by the partner) is the effective component of such interventions. VOL. 7, NO. 2, MARCH 1996 119 This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/page/info/about/policies/terms.jsp PSYCHOLOGICAL SCIENCE Attitude Polarization animation of the thinking that underlies attitudes (Kuhn, 1991). Although space does not allow us to discuss the reasoning that subjects in each of our engagement conditions offered in justi- fication of their opinions, we believe that qualitative analysis of such data is essential to fully answering the question of how engagement affects thinking and beliefs. REFERENCES Chaiken, S., & Yates, S. (1985). Affective-cognitive consistency and thought- induced attitude polarization. Journal of Personality and Social Psychol- ogy, 49, 1470-1481. Downing, J.W., Judd, CM., & Brauer, M. (1992). Effects of repeated expressions on attitude extremity. Journal of Personality and Social Psychology, 63, 17-29. Kuhn, D. (1991). The skills of argument. New York: Cambridge University Press. Liberman, A., & Chaiken, S. (1991). Value conflict and thought-induced attitude change. Journal of Experimental Social Psychology, 27, 203-216. Lord, C.G., Ross, L., & Lepper, M.R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evi- dence. Journal of Personality and Social Psychology, 37, 2098-2109. Millar, M.G., & Tesser, A. (1986). Thought-induced attitude change: The effects of schema structure and commitment. Journal of Personality and Social Psychology, 51, 259-269. Miller, A.G., McHoskey, J.W., Bane, CM., & Dowd, T.G. (1993). The attitude polarization phenomenon: Role of response measure, attitude extremity, and behavioral consequences of reported attitude change. Journal of Per- sonality and Social Psychology, 64, 561-574. Myers, D.G. (1990). Social psychology (3rd ed.). New York: McGraw-Hill. Pomerantz, E., Chaiken, S., & Tordesillas, R. (in press). Attitude strength and resistance processes. Journal of Personality and Social Psychology. Tesser, A. (1978). Self-generated attitude change. In L. Berkowitz (Ed.), Ad- vances in experimental social psychology (Vol. 1 1 , pp. 289-338). New York: Academic Press. Tesser, A., & Leone, C. (1977). Cognitive schemas and thought as determinants of attitude change. Journal of Experimental Social Psychology , 13, 340-356. Tesser, A., Martin, L., & Mendolia, M. (in press). The impact of thought on attitude extremity and attitude-behavior consistency. In R.E. Petty & J.A. Krosnick (Eds.), Attitude strength: Antecedents and consequences. Hills- dale, NJ: Erlbaum. Worchel, S., Cooper, J., & Goethals, G. (1991). The psychology of attitude change and social influence. New York: McGraw-Hill. (Received 10/10/94; Revision accepted 6/8/95) ^fflP £SiFGICAL Want the Pest candidates for your /fiSliiNW ^tpF I academic, applied, or research /ftMlifQlP^/ (A&&&\ \ Advertise in the jflMH \§nS^eW«*o ^^ \ American Psychological Society's fS^pBBe/ \B^0^iJAPS \ ^^ " Observer Employment Bulletin s§«5a \ and1^^^^^- " (ISSN: 1050-4672) Three easy ways to place your Here's why you should . . . aU Itl the bulletin ... ^ The odse/ver has the most competitive ad rates. * Fax'\ ^ ,M nnnn Readers include nearly 16,000 academic, applied, and 202-783-2083 ^ ,M nnnn research psychologists in all subdisciplines. //Jterj^R@Aps Wasnjn ton DC us Ads are searchable on Internet before they are received in the mail. ai APS OBSERVER * There are no typesetting charges for display ads. 10? 0 V6th RlntSA n'S' NWt(S^t ̂ jo?8"* * There is a shortf three"week lead time t0 Publ|sh your ad. Washington, DC 20005-4907 Readers easily find ads with the one-of-a-kind job subject (202) 783-2077 (ext. 3022) index in each issue. I Employment line-ads are $6 per line (34 characters fit on a line). Display-ad rates and a publication calendar are available on request. Include PO # with order. APS 1010 Vermont Ave., NW, Suite 1100 Washington, DC 20005-4907 Tel. 202-783-2077 Fax 202-783-2083 120 VOL. 7, NO. 2, MARCH 1996 This content downloaded from 195.78.109.54 on Sun, 15 Jun 2014 23:00:07 PM All use subject to JSTOR Terms and Conditions http://www.jstor.org/page/info/about/policies/terms.jsp Article Contents p. 115 p. 116 p. 117 p. 118 p. 119 p. 120 Issue Table of Contents Psychological Science, Vol. 7, No. 2 (Mar., 1996), pp. 63-124 Front Matter Feature Review Review: The Verdict on Freud [pp. 63-68] The Anatomy of Impact: What Makes an Article Influential? [pp. 69-75] Research Articles The Role of Task and Context in Preference Measurement [pp. 76-82] Cognitive Suppression during Saccadic Eye Movements [pp. 83-88] Cerebral Asymmetry and Cognitive Development in Children: A Magnetic Resonance Imaging Study [pp. 89-95] Research Reports Word Frequency and Word Difficulty: A Comparison of Counts in Four Corpora [pp. 96-99] Repetition Blindness Interacts with Syntactic Grouping in Rapidly Presented Sentences [pp. 100-104] Averageness, Exaggeration, and Facial Attractiveness [pp. 105-110] Infant Attachment Security and Affective-Cognitive Information Processing at Age 3 [pp. 111-114] Effects of Evidence on Attitudes: Is Polarization the Norm? [pp. 115-120] Commentaries The Difficulty of Measuring Musical Quality (And Quantity): Commentary on Weisberg [pp. 121-122] Causality, Quality, and Creativity: A Reply to Repp [pp. 123-124] Books Received since October 1, 1995 [p. 124-124] Back Matter