Direct and Indirect Effects of Essay Writing and ... - APA PsycNET

1 downloads 0 Views 777KB Size Report
Robert S. Wyer, Jr.2. University of Illinois at Urbana-Champaign. Sets of three propositions (A, B, and C) were selected. In each set, B was more likely to be true if ...
Journal of Personality and Social Psychology 1975, Vol. 31, No. 1, 55-63

Direct and Indirect Effects of Essay Writing and Information About Other Persons' Opinions Upon Beliefs in Logically Related Propositions1 Robert S. Wyer, Jr.2 University of Illinois at Urbana-Champaign Sets of three propositions (A, B, and C) were selected. In each set, B was more likely to be true if A was true than if A was false, and C was more apt to be true if B was true than if B was false. After receiving information that other persons either agreed or disagreed with B, subjects wrote an essay either supporting or opposing this proposition. Both essay writing and others' opinions affected beliefs that B was true (Pjj) in the direction of the position advocated. Effects of these variables upon beliefs that C was true (Po) were in the same direction but less in magnitude. Essay writing affected beliefs that A was true (PA) only when subjects were told that other persons agreed with B. Changes in PA and Po did not depend upon whether A and C were mentioned in the essays about B. Changes in beliefs were similar to those predicted by Wyer's subjective probability model of cognitive organization and were attributed to a tendency for subjects to regain cognitive consistency as denned in terms of this model.

Two general methods have often been used to induce a subject to change his beliefs. In one procedure, the subject is induced to prepare and deliver a communication advocating the position that one wishes him to adopt personally (cf. Brehm & Cohen, 1962; Collins & Hoyt, 1972; Rosenberg, 1965; Scott, 1957; Sherman, 1970; Wallace, 1966). In the second, most often used in studies of conformity, the subject is informed that another person or group (0) advocates this position (cf. Harvey, Kelley, & Shapiro, 1957; Hollander, 1958; Jackson & Saltzstein, 1958; Sampson & Insko, 1964; Tannenbaum, 1967; Wyer, 1966, 1967). At times this information is accompanied by a verbal communication from O containing arguments in support of the position being advocated. In other cases, only O's position is conveyed. Belief change has been successfully induced through each procedure. However, the combined effects of these procedures are unclear, 1 This research was supported by National Science Foundation Grant GS-29241. Appreciation is extended to Marti Bazell and Sara LaBelle for their assistance in preparing materials and running the study. 2 Requests for reprints should be sent to Robert S. Wyer, Jr., Department of Psychology, University of Illinois, Champaign, Illinois 61820.

particularly under conditions in which their implications differ (e.g., when a subject is asked to defend a proposition with which 0 disagrees). Equally unclear is the generalization of these effects to beliefs other than the one to which the procedures directly pertain. The study reported here was concerned with these issues. A technique that changes a belief in Proposition B may also affect beliefs both in propositions that if true, imply B (premises) and in propositions that are likely to be true if B is true (conclusions). McGuire (1960) has hypothesized that when beliefs in logically related propositions are made salient to a subject by presenting them in a questionnaire, cognitive reorganization, involving change in one or more of these beliefs, occurs in order to eliminate inconsistencies among them. Empirical support for this hypothesis has been obtained when consistency is denned by subjective probability models of cognitive organization (Rosen & Wyer, 1972; Wyer & Goldberg, 1970). It follows that if a set of related beliefs is made salient in an initial questionnaire, and if one of these beliefs is then modified using either of the influence techniques described above, corresponding changes in the other beliefs SS

56

ROBERT S. WVER, JR.

should occur in a manner that maintains their internal consistency. Research based upon other theories of cognitive consistency (e.g., Festinger, 1957; Heider, 1958; Osgood & Tannenbaum, 1955) also indicates that changes in one cognition are accompanied by changes in other, related cognitions. To give but a few examples, Tannenbaum (1966) found that information bearing upon the favorableness of a concept X affected not only evaluations of a person whose attitude toward X was known, but also evaluations of a second concept, Y, which this person favored or opposed. Gerard (1965) found that information about the similarity of a subject's judgment of a stimulus to that of another person affected the subject's attitude toward this person. Finally, research designed to test implications of cognitive dissonance theory (for recent summaries, see Collins & Hoyt, 1972; Zajonc, 1968) suggests that behavior-related cognitions (e.g., one's belief that he has decided to engage in a certain activity) may often affect other cognitions (beliefs that the chosen activity is desirable or that alternative behaviors are undesirable). While the tendency to regain cognitive consistency may be one factor underlying the effect of social influence procedures upon beliefs to which they do not directly pertain, there are other possible reasons for these effects. For example, when a subject writes an essay pertaining to B, he is likely to make statements that, if true, support the position he is advocating. In the course of presenting these arguments as persuasively as possible, the subject may become personally more convinced of their validity, and this may be reflected in his subsequent responses to questionnaire items that pertain to these arguments. One implication of this hypothesis is of course that a subject's beliefs in propositions related to B will be affected by essay writing only when these propositions are used in preparing his essay. Thus, if A implies that B is true, it should be used more often in pro-B essays than in anti-B essays, and beliefs in A should change more when the first type of essay is written. Moreover, beliefs in A are less likely to change when

a conformity paradigm is used to affect beliefs in B; if a subject is simply told that 0 advocates B without being given any basis for O's belief, there is no reason to expect that his beliefs in other propositions will change except through a process of cognitive reorganization. The effects of cognitive reorganization are often difficult to isolate because simultaneous changes in several different beliefs may be involved. The relative magnitudes of these changes are hard to predict a priori. For example, a subject may believe that B is true because A is true and A implies B. If he subsequently is induced to believe that B is not true, this could lead to a change in his belief that A is true. However, if it does not, one would not know whether the subject has failed to reorganize his cognitions according to the laws of logic or whether he has simply changed other relevant beliefs (e.g., his belief that A implies B). To understand the dynamics of cognitive reorganization when many beliefs are potentially involved, a precise description of the manner in which these beliefs are theoretically interrelated is required. The subjective probability model proposed by Wyer and Goldberg (1970) may serve this purpose. According to this formulation, the beliefs associated with two propositions, A and B, are related according to the equation: PB =

+ (1 - PA)PB/A;

[1]

where PA and Pn are the subjective probabilities (beliefs) that A and B are true, respectively, and PJI/A and Pn/A- are the beliefs that B is true if A is and is not true, respectively. Similarly, beliefs associated with B and a third proposition are related according to the equation

Pa =

+ (1 - Ptt)P0/B;

[2]

where PC is the belief that C is true and Pc/n and PO/B' are the beliefs that C is true if B is and is not true. Several studies (Rosen & Wyer, 1972; Wyer, 1970, 1972, 1973; Wyer & Goldberg, 1970) suggest that equations similar in form to those above provide accurate descriptions of the manner in which beliefs are related and of the effects of one

BELIEFS IN LOGICALLY RELATED PROPOSITIONS belief upon others. If Equations 1 and 2 are valid, changes in PB as a result of the influence techniques to be considered in this article may affect not only Pn and PC but also the "conditional beliefs" relating A, B, and C (Ps/A, Pc/n, etc.). However, the similarity between obtained and predicted changes in PB (based upon Equation 1), and between obtained and predicted changes in PC (based upon Equation 2), should indicate the extent to which changes in beliefs other than PB are attributable to cognitive reorganization of the sort implied by these equations. An additional concern of the present study was to determine the combined effects of role playing and conformity techniques. It Deemed possible that the effects of the two procedures would be additive. On the other hand, if the belief that 0 disagrees with a proposition is dissonant with the cognition that one is publicly advocating it (Festinger, 1957) and if a change in one's belief that the proposition is true results from an attempt to reduce this dissonance, then essay writing should have less effect when subjects are told that O supports the position advocated in the essay than when they are told that 0 opposes this position. Alternatively, information that O disagrees with the position to be advocated may affect the effort expended in role playing and therefore may affect the change in beliefs produced by it for this reason. These factors would produce interactive effects of the two influence procedures.

METHOD Overview One hundred twenty-eight introductory psychology students participated in the experiment to fulfill a laboratory requirement. Subjects were initially administered a belief questionnaire containing seven items associated with each of four issues. The items associated with each issue concerned three propositions (A, B, and C) which were related in the manner described by Equations 1 and 2. After completing the questionnaire, subjects received information that either most other college students or very few other students agreed with Proposition B pertaining to one of the four issues and then wrote an essay either supporting or opposing B. The original belief questionnaire was then readministered. Eight subjects participated under each of the 16 combinations of "test" issue, others' position (pro-B versus anti-B) and the position advocated in the essay (pro-B versus anti-B).

57

Construction of Belief Items Four sets of three statements (A, B, and C), each pertaining to a different contemporary issue, were selected: Issuel. A. Cigarette sales will be severely restricted by 1978. B. Cigarette smoking will decrease greatly over the next 10 years. C. Lung cancer and respiratory ailments will become less common during the next 10 years. Issue 2. A. People unconsciously imitate the behavior of persons they see on television. B. Watching television causes people to behave more violently. C. Television programs showing unnecessary violence should be banned. Issue 3. A. The number of students per faculty member at UICC [University of Illinois at Chicago Circle] will increase over the next 5 years. B. The personal contact between students and faculty at UIC'C will decrease during ' the next 5 years. C. The quality of undergraduate education at UICC will decrease during the next 5 years. Issue 4. A. Marijuana is physically and psychologically harmless. B. The use of marijuana will be legalized by 1978. C. Marijuana will become more widely used than alcoholic beverages within the next 10 years. In each set of statements, it was expected that B would be believed more likely to be true if A was true than if A was false and that C would be believed more likely to be true if B was true than if B was false. Seven items were constructed for each set of propositions, beliefs which corresponded to the subjective probabilities PA, PS, Pa, PS/A, PB/A>, PO/B, and PO/B' as denned in Equations 1 and 2. (For example, the belief items corresponding to PA, PB, and Pa for the first issue were identical to the three propositions, A, B, and C listed above; the items corresponding to PB/A and PB/A> were "If cigarette sales are severely restricted by 1978, cigarette smoking will decrease greatly during the next 10 years" and "Even if cigarette sales are not severely restricted by 1978, cigarette smoking will decrease greatly during the next 10 years"; similarly, the items corresponding to PO/B and PC/B- were "If cigarette smoking decreases greatly over the next 10 years, lung cancer and respiratory ailments will become less common during the next 10 years" and "Even if cigarette smoking does not decrease greatly

ROBERT S. WYER, JR.

58

TABLE 1 MEAN CONDITIONAL PROBABILITIES ASSOCIATED WITH EACH ISSUE Issue

PlilA

PBM- Difference

Pc/a

PCIU-

Difference

t

.603 .452 .697 .622

.420 .294 .391 .429

.598 .609 .574 .586

.383 .406 .381 .417

.215 .203

2 3 4

.183

.158 .306 .193

.193 .169

over the next 10 years, lung cancer and respiratory ailments will become less common during the next 10 years.") The resulting 28 items were ordered randomly and presented in a questionnaire; each item was followed by an 11-point scale from 0 (not at all likely) to 10 (extremely likely) for subjects to use in recording their belief that it was true. These responses were subsequently divided by 10 to convert them to units of subjective probability.

Procedure Subjects, run in groups of 8 to 10, were told that the experiment was concerned with students' attitudes and beliefs about various contemporary issues. They initially completed the belief questionnaire described above with instructions to estimate the likelihood that each statement was true along the scale provided. Then they were told that a second part of the study was to determine how well persons could prepare persuasive communications on the spur of the moment, without much opportunity to think about the issues before developing their arguments. Each subject was then given a sheet of paper, at the top of which was written the statement labeled B pertaining to one of the four issues described above, and was told to prepare a communication that would argue as persuasively as possible that the statement was either true or false, as indicated on the sheet. Before beginning, subjects were told to note the proportion of introductory psychology students who had expressed agreement with the statement in an earlier experiment, since this information might be helpful to them in deciding what sorts of arguments would be most effective. This proportion, which was typed on the answer sheet below the proposition each subject was asked to discuss, varied slightly over issues but was either between 89% and 91% (under pro-B conditions) or between 9% and 11% (under anti-B conditions). Subjects were given IS minutes to prepare the essay. Upon its completion, they were readministered the questionnaire they had filled out at the beginning of the experiment under the pretense that there was "some evidence that a person reports his beliefs more accurately after he has had a chance to think about them and has become familiar with the type of questionnaire." To decrease commitment effects still further, they were told to consider each answer carefully, since their responses to this questionnaire rather than the preceding one would be

used to represent their true beliefs about the issue described.

RESULTS Preliminary Analyses Analyses of conditional probabilities. To validate the assumption that the propositions pertaining to each issue were related, the relative magnitudes of PB/A and PB/A; and of PO/B and PC/B-, were determined for each test issue. These values, collapsed over influence conditions and the two administrations of the questionnaire, are shown in Table 1. While the difference between conditional probabilities varied over issues, in all cases PR/A > PR/A' and PG/n > PC/H- as expected (F > 125.0, dj = 1/112, p < .001). Analyses of essay content. Each communication was scored for whether the arguments contained in it pertained to premises (propositions which if true implied B) or conclusions (propositions which would be true if B were true) and also for whether any of the arguments pertained to the specific propositions about which beliefs were assessed (A or C). One subject considered both premises and conclusions in writing his essay, and two others wrote essays that were not classifiable. As might be expected, arguments pertained to premises a greater proportion of times (.84) than to conclusions (.16), regardless of whether the essay supported or opposed B. Of those cases in which essays pertained to premises, Proposition A was mentioned a greater proportion of times in pro-B essays (.59) than in anti-B essays (.39; x2 = 4.21, p< .05), and a nonsignificantly lower proportion of times when other persons were ostensibly pro-B (.44) than when they were anti-B (.55; x2 = 1-15, p > .10). Finally, the proportion of times that arguments pertained to A rather than to other premises depended greatly upon the test issue involved (^ 2 = 28.85, p < .01). The implications of these data are considered presently. Effects of Influence Procedures To determine the effect of influence procedures upon the belief in B (Pu), the "relative change" in this belief was calculated for each subject separately by subtracting the

BELIEFS IN LOGICALLY RELATED PROPOSITIONS mean change in Pj> for each of the three "control" issues to which questionnaire items pertained from the change in PB for the test issue. (Pooled over subjects in each influence condition, each issue appeared the same proportion of times as a test issue and as a control issue.) Relative changes in PA and Pa were calculated similarly. An analysis of variance was then performed on these change scores as a function of belief (PA, PB, or PC), others' position (pro or anti), essay position (pro or anti), and test issue. Averaged over beliefs, both influence procedures had the expected effects (in each case, F > 6.31, df = 1/112, p < .02). However, the triple interaction of belief with the two influence variables (F = 6.n, df = 2/224, p< .01) indicated that these effects depended upon the belief involved. Data relevant to this interaction, which was not contingent upon the test issue to which influence procedures pertained (F = 1.S4, dj — 6/224, p > .10) are shown in Table 2. To help clarify the nature of this interaction, relative changes in each belief were analyzed separately. Analyses of change in PB yielded main effects of both others' position (F — 5.59, rf/=l/112, p < .05) and essay position (F = 4.10, df = 1/112, p < .05). Although the interaction of these variables only approached significance (F = 3.42, df = 1/112, p < .10), data in Table 2 indicate that PB decreased only when subjects wrote anti-B essays after being told that others were anti-B; the change under this condition (M = —.127) differed substantially from the change occurring in each of the other three combinations of essay position and others' position (in each case, F > 7.50, df = 1/112, p.lO). However, a post hoc analysis of data pertaining to PB and Pa in combination indicated that the main effects of others' position and essay position were

TABLE 2 MEAN RELATIVE CHANGE IN HEI.IEFS AS A FUNCTION OF INFLUENCE VARIABLES Others' position Belief

Pro-B

PA

Anti-B

M

.167 -.051 .058

.005 .005 .005

.086 -.023 .032

.076 .068 .072

.052 -.127 -.038

.064 -.030 .017

Pro-B essay Anti-B essay

.057 .018 .038

.018 -.036 -.009

.038 -.009 .015

Pro-B essay Anti-B essay

.086 .020 .053

.035 -.036 -.001

.061 -.009 .026

Pro-B essay Anti-B essay

-.005 -.064 -.035

.100 -.042 .029

.048 -.053 -.003

Pro-B essay Anti-B essay

M

PB

Pro-B essay Anti-B essay

M

PC

M

M \(Pc/B + PCIB*) M

both significant (F > 4.34, df = 1/112, p < .05) but did not depend upon the belief involved (F < 2.58, df = 1/112, p > .10). In other words, the effects of influence variables upon PC, although not themselves significant, did not significantly differ from the effects of these variables upon beliefs to which they directly pertained. Analysis of the relative change in PA yielded an interaction of others' position and essay position (F = 6.97, df = 1/112, p< .01). This interaction indicated that writing essays about B affected PA only when other persons ostensibly agreed with B; when others disagreed with B, essay writing had no effect whatsoever upon PA. Effects of type of argument. It was speculated that subjects who write essays advocating a position on B call forth specific arguments in defense of this position and become more personally convinced of the validity of these arguments as a result of their attempt to present them persuasively. However, two comparisons suggest that this cannot account for the effects of influence variables upon PA. First, Table 2 shows clearly that the effect of essay writing affected PA only when others agreed with B, However, the proportion of essays which pertained to A was slightly

60

ROBERT S. WYER, JR.

lower when others agreed with B (.38) than when others disagreed with B (.44). Second, the effect of essay writing upon PA was not significantly contingent upon the issue about which essays were written despite the fact that essays about Issue 1 mentioned A in only 3 of 32 cases, while those about Issue 2 mentioned A in 20 of 32 cases. More specifically, A was mentioned in .125, .625, .812, and .438 of the pro-B essays pertaining to Issues 1-4, respectively, but the relative changes in PA were +.026, -.071, +.062, and +.267. Similarly, the proportion of times A was mentioned in anti-B essays was .062, .625, .187, and .375 when Issues 1-4 were involved, yet the changes in PA. were —.068, -.011, -.029, and +.017, respectively. The proportion of times A was mentioned and the mean change in PA in the direction implied by the essay were correlated .091 over the experimental conditions (n = 8, p > .25). While these data are too limited to be conclusive, there is no evidence of a relationship between the likelihood that a proposition is mentioned in the essay pertaining to B and change in beliefs that this proposition is valid. Effects of cognitive reorganization. If changes in PA and PG are the results of the cognitive reorganization required to maintain consistency among beliefs and if Equations 1 and 2 are valid descriptions of the relations among beliefs, then the values of Pn and PC predicted by these equations should approximate the obtained values of these beliefs, and in addition, obtained and predicted values of these beliefs should be related similarly to the two influence variables. Predicted values of Pn both before and after influence were therefore calculated for each subject separately on the basis of his questionnaire responses to items pertaining to the test issue and also items pertaining to each of the three control issues. The relative change in predicted values was determined for each subject separately in a manner analogous to that used to estimate the relative change in obtained values. An analysis of variance was then performed on the relative change in PB as a function of others' position, essay position, test issue, and "dummy" variable, (predicted versus obtained). If Equa-

tion 1 is valid and if the changes in beliefs related to Pu are the consequence of cognitive reorganization, all main effects and interactions involving this dummy variable should be nonsignificant. This was in fact the case (in each instance, F < 2.02, dj = 1/112, p> .10). A similar analysis was performed on obtained and predicted values of the relative change in Pc based upon Equation 2. Unlike the analysis of obtained values alone, the main effect of others' position was significant (F = 6.40, df = 1/112, p < .02). However, no main effects or interactions involving predicted versus obtained were reliable (F < 2.16, d}= 1/112, p> .10). The preceding analyses indicate that Equations 1 and 2 provide accurate descriptions of the relationships among the beliefs to which they pertain and give indirect support to the hypothesis that beliefs other than PB changed primarily as a result of cognitive reorganization in a manner implied by these equations. Some additional implications of this hypothesis should be noted. Table 2 shows that when other persons disagreed with B, essay writing changed PB substantially but had no effect upon PA.. Moreover, when subjects wrote anti-B essays after being told that other persons supported B, PB increased while PA. decreased. If Equation 1 is valid, these differences could occur only if, under these experimental conditions, PB/A and PH/A' both changed in the same direction as Pn (i.e., if subjects changed their beliefs that B was valid for reasons unrelated to their beliefs in A). Data in Table 2 indicate that the relative change in the mean of PB/A and PB/A- was positive under each of the three conditions in which PB increased and was negative under the condition in which PB decreased. Whether the magnitudes of these changes are sufficient to account for the observed changes in PB is unclear; 3 however, the fact that obtained changes and predicted changes did not sig3 Comparisons of magnitude are somewhat misleading, since according to Equation 1 the relations among various beliefs are multiplicative. The mean change in the product of two beliefs (FtPa/A) cannot be estimated from the mean change in each belief separately.

61

BELIEFS IN LOGICALLY RELATED PROPOSITIONS nificantly differ supports this possibility. It might be argued that the similarity between change in PD and change in the two conditional beliefs is simply a result of the subjects' failure to understand the difference between conditional and unconditional probabilities. However, if this were so, the mean change in PC/B and Pc/n' should also be similar to the change in P0. Data shown in Table 2 indicate that this was not the case. Supplementary Analyses Subjects may be less influenced by the opinions of others when they are initially inclined to disagree with these opinions than when they are predisposed to agree. On the other hand, defending a position with which one personally disagrees may be dissonance arousing (Festinger, 1957) and therefore may produce greater change than arguing a position with which one already agrees. To explore these possibilities, 12 initially pro-B subjects (those whose initial estimates of PB were greater than .5) and 12 initially anti-B subjects (those whose initial estimates of PB were less than .5) were selected under each combination of influence variables, controlling insofar as possible for the proportion of times that each test issue was represented. An analysis of variance was then performed on the relative change in beliefs as a function of initial attitude, belief (PA, PB, or P0), and the two influence variables. While the significant effects obtained in the main analyses were again reliable, only the interaction of belief and others' position was contingent upon initial attitude (F — 3.S3, df = 2/224, p > .01). Data pertaining to this interaction, shown in Table 3, are difficult to interpret psychologically. However, supplementary analyses revealed that the mean initial (pretest) belief was greater than .5 (the scale midpoint) in each cell of Table 3 in which a decrease in beliefs occurred but was less than .5 in each cell in which an increase occurred. Thus, the contingencies described in this table may well be an artifact of regression of beliefs toward the scale midpoint, and not an indication that the effects of experimental variables per se were contingent upon initial opinions.

TABU'', 3 REI.ATIVK BELIEJJ-S AS A [''UNCTION oa' INITIAJ, ATTITUDE AND OTHERS' POSITION Change in Initial opinion

Others' position PA.

Pro-B Anti-B

Pro-B Anti-B Pro-B Anti-B

-.007

.051 .160 -.024

Pa

PC

-.031 -.154 .096 .048

-.011 -.046 .044 .048

DISCUSSION The results of this study can be summarized briefly. Subjects' beliefs that Proposition B was true (PB) were affected both by other persons' beliefs in B and by writing a persuasive communication about B. The combined effect of these variables was positive when subjects either wrote a pro-B essay or received information that other persons agreed with B, a negative change in PB occurring only when subjects wrote anti-B essays under the illusion that others also disagreed with B. The effects of influence variables upon beliefs in a proposition implied by B (P0) were similar in direction to their effects upon PB but were less in magnitude. Writing an essay about B affected beliefs in a premise which if true implied B (PA) only when subjects were told that other persons agreed with B. Changes in PA. under various influence conditions were unrelated to the proportion of times that A was mentioned in the essays subjects had written. Obtained changes in Ps and PC under these conditions did not significantly differ in magnitude from changes predicted by the subjective probability model of cognitive organization proposed by Wyer (see Equations 1 and 2). The latter two findings, considered in combination, suggest that the changes in beliefs other than PB were attributable to a tendency for subjects to reorganize sets of related beliefs in a manner that maintains consistency among them as defined by the probability model. Although changes in Pc under different instructional conditions did not differ significantly from predicted values based upon Equation 2, analyses of changes in P0 alone yielded no significant results. However, as Me-

62

ROBERT S. WYER, JR.

Guire (1960) points out, changes in the conclusion of a syllogism should theoretically be less than changes in beliefs in the premises. (The reason for this is apparent from Equation 2; that is, a change in P0 would be equal in magnitude and direction to a change in PD only if PO/B = 1 and PO/B' = 0. This was clearly not the case in this study; see Table 1.) Perhaps if the direct effects of influence variables upon PB had been greater, the changes in PC would have been sufficiently large to be detected statistically. In other studies, where greater changes in beliefs analogous to PB have been induced, significant changes in beliefs analogous to Pc have typically been detected (cf. Wyer, 1970, 1972). Two findings in this study are difficult to interpret. First, PB increased both when subjects wrote pro-B essays after receiving information that others opposed B and when subjects wrote anti-B essays after learning that others supported B. This suggests that each influence procedure was more effective in gaining support for a proposition than in gaining opposition to it, regardless of the nature of this proposition. Although this interpretation would seem reasonable if subjects were predisposed to agree with B before they were exposed to influence, the fact that the effectiveness of influence procedures was not contingent upon the subjects' initial attitudes argues against this interpretation. It is conceivable that subjects have an "agreement" bias which leads them to accept propositions as true more readily than to reject them as false, all other things being equal. To this extent a given procedure should be more effective in increasing beliefs in the proposition "marijuana will not be legalized by 1978" than in decreasing beliefs in the proposition "marijuana will be legalized by 1978." Although this interpretation is not very convincing, no alternatives seem any more compelling. The second curious aspect of the data reported here concerns the different effects of influence variables upon Pn and PA- When subjects were told that other persons disagreed with B, the changes in PB produced by essay writing were not accompanied by changes in beliefs in A, but rather by changes

in beliefs that B was true for reasons unrelated to their beliefs in A. (This does not mean that subjects changed their beliefs that A was relevant to B under these conditions; the difference between PB/.4 and PB/A' was not significantly affected.) The reason for this, which is apparently not attributable to differences in essay content, can only be speculated. One possibility is suggested by self-perception theory (Bern, 1967) which assumes that subjects infer their beliefs from observations of their own behavior. Thus, in the present study, a person who writes a pro-B (anti-B) essay may infer that he personally agrees (disagrees) with B both if A is true and if A is false but would not necessarily change his belief in A. This explanation is not wholly satisfactory, however, since it does not account for the fact that PA was affected by essay writing when others ostensibly agreed with B. Difficulties are also encountered in applying cognitive dissonance theory (Festinger, 19S7) to these results. It is conceivable that changes in PA occur in order to increase consonance with one's decision to advocate a position regarding B. However, why this should occur only when other persons agree with B is unclear. Moreover, in this study subjects were not told that they could refuse to participate in the essay-writing task. Since evidence for dissonance reduction is typically not obtained in counterattitudinal advocacy situations unless the freedom to refuse is made explicit (cf. Collins & Hoyt, 1972), it may not be a valid explanation of behavior in the present experiment. Thus, while the conclusion seems justified that procedures intended to influence beliefs in B affect beliefs in related propositions through the process of cognitive reorganization, an understanding of which cognitions are affected in any given instance remains elusive. In this regard, some caution should be taken in generalizing the results of this experiment. As noted previously, numerous situational variables affect the magnitude of change induced by the two influence procedures in beliefs to which they directly pertain. Similar variables conceivably affect the generalization of this influence to peripherally related beliefs. The formulation of cognitive organiza-

BELIEFS IN LOGICALLY RELATED PROPOSITIONS tion considered in this article may help to identify the nature of these effects. REFERENCES Bern, D. J. Self-perception: An alternative interpretation of cognitive dissonance phenomena. Psychological Review, 1967, 74, 183-200. Brehm, J. W., & Cohen, A. R. Explorations in cognitive dissonance. New York: Wiley, 1962. Collins, B. E., & Hoyt, M. F. Personal responsibilityfor-consequences: An integration and extension of the "forced compliance" literature. Journal of Experimental Social Psychology, 1972, 8, 5S8-S93. Festinger, L. A theory of cognitive dissonance. Stanford, Calif.: Stanford University Press, 19S7. Gerard, H. B. Deviation, conformity, and commitment. In I. D. Steiner & M. Fishbein (Eds.), Current studies in social psychology. New York: Holt, Rinehart & Winston, 1965. Harvey, O. J., Kelley, H. H., & Shapiro, M. M. Reactions to unfavorable evaluations of self made by other persons. Journal of Personality, 1957, 25, 393-411. Heider, F. The psychology of interpersonal relations. New York: Wiley, 1958. Hollander, E. P. Conformity, status, and idiosyncracy credit. Psychological Review, 1958, 65, 117127. Jackson, J. M., & Saltzstein, H. D. The effect of person-group relationship on conformity processes. Journal of Abnormal and Social Psychology, 19S8, 57, 17-24. McGuire, W. J. A syllogistic analysis of cognitive relationships. In C. I. Hovland, & M. J. Rosenberg (Eds.), Attitude organization and change. New Haven, Conn.: Yale University Press, 1960. Osgood, C. E., & Tannenbaum, P. H. The principle of congruity in the prediction of attitude change. Psychological Review, 1955, 62, 42-55. Rosen, N. A., & Wyer, R. S., Jr. Some further evidence for the "Socratic effect" using a subjective probability model of cognitive organization. Journal of Personality and Social Psychology, 1972, 24, 420-424. Rosenberg, M. J. When dissonance fails: On eliminating evaluation apprehension from attitude measurement. Journal of Personality and Social Psychology, 1965, 1, 28-42. Sampson, E. E., & Insko, C. A. Cognitive consistency

63

and performance in the autokinetic situation. Journal of Abnormal and Social Psychology, 1964, 68, 184-192. Scott, W. A. Attitude change through reward of verbal behavior. Journal of Abnormal and Social Psychology, 1957, 55, 72-75. Sherman, S. J. Attitudinal effects of unforeseen consquences. Journal of Personality and Social Psychology, 1970, 16, 510-520. Tannenbaum, P. H. Mediated generalization of attitude change via the principle of congruity. Journal of Personality and Social Psychology, 1966, 3, 493-499. Tannenbaum, P. H. The congruity principle revisited: Studies in the reduction, induction, and generalization of persuasion. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 3. New York: Academic Press, 1967. Wallace, J. Role reward and dissonance reduction. Journal of Personality and Social Psychology, 1966, 3, 305-312. Wyer, R. S., Jr. Effects of incentive to perform well, group attraction, and group acceptance on conformity in a judgmental task. Journal of Personality and Social Psychology, 1966, 4, 21-26. Wyer, R. S., Jr. Behavioral correlates of academic achievement: Conformity under achievement- and affiliation-incentive conditions. Journal of Personality and Social Psychology, 1967, 6, 255-263. Wyer, R. S., Jr. Quantitative prediction of belief and opinion change: A further test of a subjective probability model. Journal of Personality and Social Psychology, 1970, 16, 559-570. Wyer, R. S., Jr. Test of a subjective probability model of social evaluation processes. Journal of Personality and Social Psychology, 1972, 22, 279286. Wyer, R. S., Jr. Category ratings as "subjective expected values": Implications for attitude formation and change. Psychological Review, 1973, 80, 446467. Wyer, R. S., Jr. & Goldberg, L. A probabilistic analysis of the relationships between belief and attitudes. Psychological Review, 1970, 77, 100-120. Zajonc, R. B. Cognitive theories in social psychology. In G. Lindzey & E. Aronson (Eds.) Handbook of social psychology. Vol. 1. Reading, Mass.: Addison-Wesley, 1968. (Received May 14, 1973)