30, no. 1, transmission of attitude ... - uni-muenster.de€¦ · 1974, vol. 30, no. 1, 81-94...

Journal of Personality and Social Psychology1974, Vol. 30, No. 1, 81-94

TRANSMISSION OF ATTITUDE-RELEVANT INFORMATIONTHROUGH A COMMUNICATION CHAIN 1

MELVIN MANIS,* S. DOUGLAS CORNELL, AND JEFFREY C. MOORE

University of Michigan and U. S. Veterans Administration Hospital, Ann Arbor, Michigan

Female college students summarized the main points of a speech concerningthe legalization of marijuana. These summaries were addressed to discussionpartners who seemingly favored or opposed legalization. Summaries addressedto prolegalization audiences were thought (by naive judges) to derive frommessages that were more favorable to legalization than were summariesaddressed to antilegalization audiences. This transmission bias emerged withincreasing clarity as the retention interval was lengthened. In another experi-ment, subjects rated several different transmitters (summarizers). Although itwas clear that the transmitters did not necessarily endorse the messages thatthey summarized, the ratings that the transmitters received were nonethelessmost favorable if (fortuitously) they had been required to relay informationthat was supportive of the listeners' views.

Communication involves the transfer of in-formation from one person to the next, andthereby enables the individual to learn aboutsituations with which he has not had direct(personal) contact. Social psychologists havelong recognized the importance of informalcommunication networks for disseminatingattitudes and information; systematic com-munication errors are not uncommon, how-ever (Campbell, 1958).

Several investigators have demonstratedthat the respondent's behavior in a communi-cation task may be significantly affected bythe character of the anticipated audience(Grace, 1951; Schramm & Danielson, 1958;Zajonc, 1960; Zimmerman & Bauer, 1956).Most relevant, perhaps, is the study byZimmerman and Bauer, in which an experi-menter came to a number of classrooms;sometimes she presented herself as represent-ing an organization that favored improvedteachers' salaries, and sometimes she seemedto speak for a group of taxpayers who wereconcerned with economy in government ex-

1 All statements are those of the authors and donot necessarily represent the opinions or policies ofthe U.S. Veterans Administration. The authorsexpress appreciation to the Center for Research onGroup Dynamics at the Institute for Social Research,University of Michigan, for making their laboratoryfacilities available.

2 Requests for reprints should be sent to MelvinManis, 4106 Institute for Social Research, Universityof Michigan, P. O. Box 1248, Ann Arbor, Michigan48106.

penditures (and hence opposed pay raises forteachers). Her organization, she explained,wanted to have several students present talkson the subject of teachers' salaries at theirnext meeting and would select these speakerson the basis of some informal speeches thatthey (the students) were to write "nextweek." To help provide ideas for this talk,the experimenter then read a series of state-ments about teachers' salaries that had pre-sumably been condensed from a magazinearticle. In half of the classes, the statementsthat she read supported the need for a teach-ers' pay raise; the other classes heard state-ments that opposed any increase in teachers'salaries. Immediately after the experimenter'spresentation, the students attempted to repro-duce the arguments that they had just heard;there was no sign of an audience effect.When, however, the experimeneter returneda week later and collected a second recallmeasure, the congruence between the state-ments that the students had heard and theapparent views of their anticipated audiencesignificantly affected their performance. Forexample, the statements favoring the payraise were better recalled by those respon-dents who expected that they might laterhave an opportunity to address an audiencethat supported this position, as comparedwith those who expected that they mightaddress a group that did not favor a pay raisefor teachers.

81

82 M. MANIS, S. D. CORNELL, AND J. MOOEE

Schramm and Danielson (1958) replicatedthe Zimmerman-Bauer results in a studywhere the respondents were presented with atwo-sided set of arguments, some favoringand some opposing a reduction in the mini-mum voting age. Subjects were told that thesearguments would be useful to them in pre-paring a speech for a contest that was appar-ently sponsored either by (a) a group thatfavored a reduction in the voting age, or(b) a group that opposed this type of change.When tested for recall a week after expo-sure to the experimental message, the respon-dents were better able to remember the pointsthat supported (rather than opposed) theposition of their anticipated audience.

Despite the replicated audience effect thatwas obtained in both the Zimmerman-Bauerand the Schramm-Danielson studies, this re-sult may have derived in part from a commonset of demand characteristics that wereoperative in the two experiments. In bothcases, the subjects were led to believe thatthe arguments with which they were pre-sentd might subsequently prove useful in a"competition," where their speeches would becompared with those of their classmates.Given the obvious biases of the "sponsoringorganizations," the average subject wouldsurely anticipate that his chances of winningmight be substantially improved if the mainthrust of his remarks were congruent withthe sponsor's views. As a result, he mightplausibly focus his attention on those aspectsof the background arguments that supportedthe sponsor and would be less concerned withremembering the points that conflicted withthe sponsor's position.

Following this line of reasoning, it seemedimportant to see if similar results might beobtained in a somewhat altered experimentalsetting, where the respondents were simplyasked to relay an attitude-relevant messageto a peer who seemingly shared (or did notshare) their views regarding the issue athand. Our prototype study was therefore de-signed to explore the possibility that withina communication chain, the apparent viewsof a given listener might induce a systematicbias in the messages addressed to thatperson by respondents who were attempting

to convey verbally the main points of anattitude-relevant message.

EXPERIMENT 1

The first study in this series was designedto determine if the apparent views of the au-dience would systematically affect the verbaloutput of speakers who were asked unexpect-edly to summarize a speech that they hadheard previously. Following the design of theZimmerman-Bauer study, delays of varyinglengths were interposed between (a) thepresentation of the original message and (b)the respondents' (transmitters') summaries.We anticipated that the difference betweenthe summaries addressed to contrasting audi-ences (pro versus con) would be most strik-ingly manifested as the delay interval waslengthened, since this would presumablyreduce the clarity of the remembered message.

If an audience effect was indeed manifested(at least in the longest of the delay condi-tions), then the identification of those ele-ments that were necessary for its emergenceseemed to be logical "next step." Previousresearch in this area has invariably focusedon respondents who learned about the viewsof their expected audience before receivingthe message that they were subsequently torecall. It therefore seemed possible that thetransmission biases that were exhibited mightderive from a distorted reception (interpreta-tion) of the original information. It is alsoconceivable, on the other hand, that the dis-tortion process mainly resides in the respon-dent's transmission of what he remembers. Ifthe audience effect that is discussed aboveprimarily represents a bias in transmission,then it should be possible to demonstratethis phenomenon among respondents who arenot aware of their listener's views until theyare just about to summarize a body of infor-mation that they had received previously. Anaudience effect that was obtained under thesecircumstances would presumably reflect atransmission bias, rather than a bias in recep-tion. To explore these contrasting possibilities,half of the transmitters who were assignedto the longest of the delay conditions (wherethe occurrence of audience effects seemedmost probable) were not informed of their

TRANSMISSION OF ATTITUDE-RELEVANT INFORMATION 83

partner's opinions until after they had re-ceived the background information and wereabout to pass it on; other respondents inthis long-delay group learned about their part-ner's views before receiving the backgroundinformation.

MethodA group of 96 undergraduate women, all of whom

favored the legalization of marijuana, served as paidsubjects in this experiment.3 These women wererecruited for "an experiment in communication," andthey were led to believe that together with a part-ner, they would have an opportunity to discusssome current social issues. Since the study was tobe concerned with verbal communication, nonverbalinteraction was eliminated by assigning the subjectsto separate booths and requiring them to communi-cate by means of microphones and headsets. Theexperimental arrangements made it possible tocollect data simultaneously from 2 to 4 subjects.

After entering their booths and receiving someintroductory instructions, each subject was informedthat she and her partner were first going to hearone, two, or three tape-recorded messages concerningpresent-day social issues. Through discussion, theywere then to resolve any differences of opinion re-lating to one of the topics, which was to be chosenby the experimenter. Subjects were warned thatthere was no way of knowing beforehand just howmany messages they would hear before starting theirdiscussion; moreover, to make matters somewhatmore difficult, one or both of the partners mightfind that some of these talks had been recordedagainst a noisy background (this last was includedso that later on, the subjects might plausibly beasked to relay a particular message to partners whohad seemingly been unable to make sense of thelittle they had heard).

Early-Information SubjectsFor the respondents who learned of their partner's

views before hearing the material they were later tosummarize, the experimenter began the session byasking the members of each pair to state brieflytheir opinions about the legalization of marijuana.Each subject then heard what appeared to be herpartner (this was actually the tape-recorded voiceof a confederate) state either that she was in favorof (or against) the legalization of marijuana. Thenaive subject then stated her own opinion and heardone or more background talks on different currentissues. For all of the subjects, the first of thesetalks was a relatively neutral S^-minute speech con-

8 We had originally anticipated the recruitment ofan equal number of respondents who opposed legali-zation. However, those opposing legalization consti-tuted a rather small proportion of the populationfrom which we were sampling and hence wereexcluded from the study.

cerning the legalization of marijuana; the speakerdiscussed this controversy rather dispassionately anddid not endorse either of the opposing sides. Tworather similar talks were used for this purpose, eachdealing with several aspects of the marijuana issue,and it was these talks that the respondents weresubsequently to summarize for the benefit of theirpartners (see below). Some subjects (in the no-delaycondition) heard only the marijuana tape beforebeing asked for their summaries. Other subjects wereasked for their summaries after an intervening delayperiod. For these people, the marijuana talk wasfollowed either by (a) a "filler" message lastingabout 6 minutes, concerning the Nixon administra-tion's desegregation policies or the changing pornog-raphy laws in Denmark, or (6) an interval of about12 minutes during which both of these messageswere played.

There were 72 subjects assigned to the early-information groups, and they fulfilled the specifica-tions of a 2 X 2 X 3 factorial design in which themain variables were audience opinion (pro- or anti-marijuana), message (depending on the particularmarijuana speech that had been played), and delayinterval (no delay, brief delay, and long delay). Sixsubjects were assigned to each of the 12 cells inthis design.

Late-Information Subjects

Twenty-four subjects first learned of their part-ners' attitudes (pro or con) just before they wereasked to summarize the marijuana talk. As notedearlier, all of the subjects in this group were in thelong-delay condition, and thus heard two "filler"messages before being asked to summarize the speechabout marijuana. The late-information subjects, com-bined with those early-information respondents whohad (like themselves) been assigned to the long-delaycondition, comprised a 2 X 2 X 2 factorial design, inwhich the independent variables were audience opin-ion, message, and information timing (early versuslate). There were 6 subjects in each of the resultingcells.

The Message Summaries

After the marijuana and "filler" messages (if any)had been played, respondents in all conditions heardthe experimenter announce that the topic for discus-sion was to be the legalization of marijuana; hethen asked "Subject A" to confirm the fact that,for her, message reception had been difficult becauseof the noisy background. Subject A (a confederatewhose comments had previously been tape-recorded)indicated that the interference was so severe that itwas virtually impossible to make sense of the mes-sage. She added that while she herself was in favorof (or opposed to) the legalization of marijuana, shewas really unable to tell much about the speaker'sviews from the little she had heard. To remedy thissituation, the experimenter suggested that Subject B(the "unprogrammed" member of the pair) sum-

84 M. MANIS, S. D. CORNELL, AND J. MOORE

marize as much of the marijuana talk as she couldrecall, so that Subject A would have some back-ground on which to base their upcoming discussion.B's spontaneous summary of the marijuana talk wasthen surreptitiously tape recorded for later analysis(see below).4

After concluding this unexpected summary, eachsubject was asked to complete a short questionnaireconcerning such matters as (a) her present views onthe marijuana issue, (6) her impression of the viewsendorsed by the speaker who provided the back-ground talk on marijuana; and (c) her understand-ing of the purposes of the experiment (to permit theidentification of suspicious respondents).0 The ex-perimental session was then terminated, with apolo-gies that there had not been sufficient time to permitthe discussion that was originally planned. Finally,all of the subjects were interviewed in small groupsof two to four for further evidence of suspiciousnessand were thoroughly debriefed.

Sources of Data

The experiment yielded two primary data sources:each respondent's summary of the marijuana speechand the questionnaires that were completed at theend of the experimental session. Each summary wasconverted into a common quantitative framework bya team of 5-7 judges; fQ%-10Q% of the judges oneach team favored the legalization of marijuana. Thejudges were put in the position of naive listeners;they listened to each of the summaries and then,for each one, used a series of rating scales to indi-cate the viewpoint that they thought had beenexpressed in the original message (the passage thatthe transmitters were summarizing). In performingthis task, the judges were led to believe that no twotransmitters had heard the same original message(this was to encourage an independent assessmentof the various summaries) and were not informedof the views held either by the speaker who hadprovided the summary or by her audience.

The ratings of primary interest focused on theapparent favorability (or unfavorability) of theoriginal speakers' views, as indexed by the judges'ratings. After hearing each message summary, thejudges responded on 9-point scales to each of thefollowing items: (a) Was the original message infavor of or against legalization of marijuana?(6) According to the original message, how detri-mental is marijuana to the individual user? (c) Ac-cording to the original message, how detrimental ismarijuana use to the rest of society?

*A11 of the subjects were subsequently given anopportunity to have these recordings destroyed, ifthey objected to the deceptions that were involved.Only one respondent out of several hundred whohave been tested in this situation asked to have herrecording destroyed.

0 Data from 20 respondents in Experiment 1 wereomitted from further consideration due to excessivesuspiciousness (17) or other irregularities (3).

To arrive at an overall scale value for each of thesummaries, correlation coefficients (r) were calcu-lated between all pairs of judges within a giventeam, item by item. The four judges whose ratingswere most highly intercorrelated were then identifiedfor each item; the transmitters' scores on the indi-vidual rating scales were taken as the mean of theratings assigned by these four "most compatible"judges. The total score for a particular summary(representing the overall favorability toward legaliza-tion that it implied) was then calculated by aver-aging the mean ratings that were obtained on theindividual items.

Results

Figure 1 presents the mean ratings thatwere assigned to summaries generated in thevarious experimental groups. These curvesrepresent pooled results, including summariesfrom both background talks. Although thesummaries that were derived from these"neutral" speeches received significantly dif-ferent ratings, this pattern was unaffected bythe other independent variables in the study(i.e., there were no significant interactionsinvolving the message variable), and hencethe two messages have been combined forsimplicity of presentation. Figure 1A presentsthe results obtained in the early-informationconditions (where the respondents were in-formed of their partners' views before hearingthe speech about marijuana). These resultssuggest that the longer the retention interval,the more promarijuana were the summariesaddressed to listeners who "favored" thelegislation of marijuana, and the more anti-marijuana the summaries addressed to thosewho were "opposed." This pattern is quitecongruent with the results reported in theZimmerman-Bauer and Schramm-Danielsonexperiments. An analysis of variance whichwas performed on these data indicated thatthe difference between the contrasting audi-ence conditions increased linearly with in-creases in the delay variable (F = 5,33, df =1/60, p < .025). The modest reversal of theoverall trend that appears in the no-delaycondition is far from significant.

A second analysis of variance was then per-formed, focusing exclusively on the sum-maries that were composed following thelong-delay interval. (See Figure IB.) Ourprimary concern in this analysis was to com-pare the magnitude of the audience effect for


I A. EARLY INFORMATION GROUPS I B. LONG-DELAY GROUPSEarly vs. late information

7.5 r-

7.0

|

V)

fc 6-5

z

16.0

PRO AUDIENCE

ANTI AUDIENCE

I I I5 10

RETENTION INTERVAL

EARLY LATE

LISTENER'S ATTITUDEREVEALED

FIGURE 1. Inferences made by the early-information groups (Figure 1A) andthe long-delay groups (Figure IB) concerning the "original" message as afunction of (a) the recipient's views, (6) the retention interval, and (c) thepoint at which the transmitter first learned of her partner's views. (Experi-ment 1.)

respondents in the early- versus late-informa-tion groups. If the audience effect is mainlyattributable to a biased reception of the mes-sage to be relayed, then it should be substan-tially reduced in magnitude when the trans-mitter does not learn about her partner'sviews until after she has heard the back-ground talk. Such a pattern of results wouldyield a significant interaction between theviews of the audience (pro- versus antilegali-zation) and the contrasting information groups(early versus late information). This inter-action was, however, far from significant inthe present experiment (p > .20), which sug-gests that the message summaries were notinfluenced by the timing of the listener's an-nouncement concerning her views on legaliza-tion. On the other hand, while there was amain effect attributable to the attitude of theaudience when the early- and late-informationgroups were combined ( />< .06) , the sum-

maries produced by the late-information re-spondents did not show a reliable audienceeffect when analyzed separately. There is, asa result, some uncertainty as to the robust-ness of the audience effect in cases where thetransmitter does not learn about the views ofher audience until after she has received theinformation that is to be passed on.

Inspection of Figure 1A suggests that theobtained audience effects are somewhat"asymmetric," in that with the passage oftime, they appear to develop more intensivelywhen the transmitter addresses her summaryto a prolegalization rather than an anti-legalization discussion partner. Thus, whenconsidered separately, there is a significanttendency for the respondents in the pro-legalization condition to generate messagesummaries that are increasingly favorable tolegalization as the retention interval islengthened. For those assigned to discussion


partners who opposed legalization, by con-trast, there is no significant difference be-tween the three data points shown in Figure1A, despite the general downward drift. Thedifference between these patterns may in partderive from the uniformly prolegalizationviews of the transmitters. Apart from anyother influences, it is conceivable that with in-creases in the retention interval, there is asystematic tendency for the average trans-mitter to recall the original message as beingincreasingly similar to her own prolegalizationviews; this "drift," in turn, may combineadditively with other influences, such as thosethat derive from the listener's apparent atti-tude, to yield the "net" displacement pat-terns that are shown in Figure 1A. Accordingto this analysis, for the respondents who werepaired with prolegalization partners, the twosources of bias (i.e., the transmitter's ownattitude and the apparent views of herlistener) operate in parallel, in that theywould both serve to increase the favorablenessof the resulting summary. In situations wherethe listener seemingly opposes legalization, onthe other hand, the two sources of bias wouldwork in contrasting directions and hencelargely cancel one another, since the trans-mitter's own attitudes presumably induce asummary that exaggerates the favorability ofthe original message, while the apparent viewsof the listener constitute a force in theopposite direction.

Content Analysis

In contrast to the provocative effects thatwere obtained when global ratings were usedto assess the message summaries, a moremolecular content analysis yielded relativelyuninteresting results. When the recorded re-lays were analyzed by counting the numberof pro- and antimarijuana arguments thatwere cited, the contrasting audiences did notdifferentially affect the content of the re-spondents' summaries. It is likely, however,that the global rating procedure provides amore accurate indication of the meaning thatthe summaries would convey in the course ofa normal conversation (it was explicitly de-signed to approximate this type of situation);that is, the rating results may have greater

external validity than the results obtainedfrom the content analysis.

Questionnaire Results

Own views. After relaying what they couldrecall of the original marijuana speech, allof the respondents completed a brief question-naire that included, among other things, arating of their own views about legalization(opposed versus in favor). Our first analysisof these data was restricted to subjects inthe early-information groups. These respon-dents showed an increasing tendency to"migrate" toward the positions that theirpartners had endorsed, as the retention inter-val was increased (the pattern was grosslysimilar to the results shown in Figure 1A).The reliability of this trend was corroboratedby the fact that the interaction between theaudience-attitude and the retention-intervalvariables included a significant linear compo-nent (F = 4.99, dj = 1/60, p < .025). In asubsequent analysis, attention was restrictedto those respondents who had been assignedto the long-delay interval, including subjectsfrom both the early- and late-informationconditions. Here, the influence of the con-trasting audiences emerged as a significantmain effect (p = .OS), while the difference be-tween the early- and late-information groupsdid not significantly affect the obtainedresults.

The striking parallelism between the mes-sage summary results (as depicted in Fig-ure 1) and the own-opinion data suggests theoperation of a self-persuasion process (Janis& King, 1954). On the other hand, an evensimpler explanation is the possibility thatour respondents were directly influenced bythe opinions that their partners had expressed,although this influence was not immediatelymanifested. The absence of consistent within-group correlations between (a) the rated con-tents of the message summaries and (b) thetransmitters' final opinions favors this latterpossibility.

Ratings of the original message. At the endof the experimental session, all of the re-spondents were asked to indicate their judg-ment as to the opinion held by the speakerwho had provided the background informationon marijuana. These ratings were not signifi-


cantly affected either by the views of theanticipated audience or by the retentioninterval.

EXPERIMENT 2

Perhaps the most interesting finding inExperiment 1 was the fact that following afilled retention interval of roughly 12 minutes,the respondents' summaries were systemati-cally displaced toward the views favored bytheir partners. In casting about for a plausibleexplanation for this effect, we were struck bythe possible relevance of the old story con-cerning the Persian messenger who breath-lessly rode up to the king and informed himthat his forces had, unfortunately, lost animportant battle. In return for this service,the king ordered the messenger executed, eventhough the messenger himself clearly was notresponsible for the events that transpired onthe battlefield. Should the messenger havebeen reincarnated following this episode, wewould anticipate that he might have learnedsomething from his traumatic experience. Hemight, for example, retire from his post asmessenger because of the attendant dangers.Or, if this was impossible, his subsequentmessages might well be slanted so as to avoidthe king's displeasure. For example, ratherthan openly announcing the loss of a battle,he would perhaps shade matters a bit, anddescribe such a battle as "difficult," but stillin doubt.

The general notion here is that the innocentbearer of "bad tidings" may well be punishedand may eventually learn that the safestcourse is to systematically bias his messages(within limits) so that they are minimallyoffensive to his listener(s). Recognizing thedangers inherent in such a situation, politicalobservers have often commented on the prob-lems that result when a chief of state unwit-tingly surrounds himself with a staff of "yesmen" who automatically slant their reportsto avoid the chief's displeasure.

One of the most important assumptions inthis account revolves about the actions ofthe message recipient. Our model suggeststhat the receiver of bad news generally re-sponds unfavorably to the messenger; andyet, apart from anecdotal evidence, there doesnot seem to be any systematic evidence on

this particular point. Perhaps the most perti-nent of the relevant studies are those byByrne (1969) and his associates, demonstrat-ing that our liking for others is directly re-lated to the number of attitude areas in whichthey share our preferred views. Note, how-ever, that in the studies by Byrne and hisassociates, the "other person" is liked ordisliked as a consequence of his own personalviews. If, however, as Byrne has suggested,liking for others operates in accordance withan associativelike model, with each area ofagreement functioning as a positive reinforce-ment, then it is possible that any event thatoccurs in the presence of another may havean impact on our liking for that person, eventhough the event itself may be largely for-tuitous and not under the control of the indi-vidual in question. An experiment by Griffittand Veitch (1971) seems to sustain this lineof reasoning.

In the Griffitt and Veitch study, the re-spondents indicated their evaluative reactionsto a stranger who appeared to agree withthem concerning either 6 or 18 of the attitudeissues that were included in a 24-item ques-tionnaire. Some subjects were tested in small,comfortably spaced groups, while others weretested under conditions of high-density crowd-ing. The temperature of the test chamber,which varied between normal (73.4 degreesFahrenheit) and hot (93.S degrees Fahren-heit), constituted an additional independentvariable. Apart from the expected effects ofattitude similarity (greater liking for thosewho shared the respondents' views), Griffittand Veitch reported that the overall responseto these "strangers" was significantly af-fected by environmental factors. Holding at-titude similarity constant, subjects who wereassigned to the hot room responded less posi-tively to the stranger than did those whowere tested under normal temperature con-ditions; similarly, respondents who weretested in a chamber with low-population den-sity rated the "stranger" more positivelythan those assigned to the crowded condition.

These results seem applicable to the mes-sage transmission task if we make the plausi-ble assumption that message receivers nor-mally feel somewhat uncomfortable whenpresented with information that challenges

M. MANIS, S. D. CORNELL, AND J. MOORK

their personal beliefs (Festinger, 1957).Through an associative process, this discom-fort may ultimately produce a negative reac-tion to the communicator who has (inescapa-bly) accompanied his noxious, dissonance-arousing message. Experiment 2 was mainlydesigned to test this hypothesis; we sought toevaluate the proposition that people who con-vey messages that challenge the views of theirlisteners are less favorably evaluated by theselisteners than are those who convey supportivemessages. It is interesting to note the simi-larity between this prediction and the resultsreported by Jones and Harris (1967). Jonesand Harris found that a person who has ex-pressed a particular attitude position is gen-erally assumed to be in favor of that stand,even by listeners who have been informedthat the speaker was merely following in-structions, and was not necessarily expound-ing his own views.

A secondary purpose of Experiment 2 wasto explore the "implicit communication the-ory" of our respondents. If, as seemed intui-tively plausible, our respondents believed thatin the process of summarizing an attitude-relevant passage, most people would selec-tively "shape" the original message contentto yield a summary that conformed moreclosely to their own beliefs, then we mightanticipate that the respondents would "cor-rect" for this bias when attempting to recoverthe main thrust of the original message. Thisreasoning suggests that a message summarywhich was delivered by someone who ap-parently favored the legalization of mari-juana would be decoded differently thanwould that same summary when it appearedto emanate from a messenger (summarizer)who opposed legalization.

MethodOverview

The subjects in this study were recruited for astudy in communication. They listened to a seriesof tape-recorded speakers (college students) whoattempted to summarize the main elements of amessage they had heard concerning the legalizationof marijuana. After hearing each summary, therespondents answered several questions regarding(a) the main thrust of the original message and(6) the personality characteristics of the speaker(transmitter).

In designing this experiment, considerable care wastaken to insure that the respondents (raters) werefully aware that the summaries which were playedto them did not necessarily reflect the opinions ofthe people whose voices they heard. It seemed im-portant to make this point rather clearly, since wewere mainly interested in the reactions evoked by"innocent" message bearers and did not want ourresults contaminated by a mistaken belief that thespeakers (summarizers) were personally sympa-thetic with the views that they transmitted. The dis-tinction between the views of the transmitters andthose conveyed in the messages that they summarizedwas reaffirmed in several ways during the courseof the experiment: (a) The summarizer's purportedviews about marijuana appeared at the top of therating sheet that was to be completed in response toeach summary; (6) to emphasize this "backgroundinformation," in introducing each of the tape-recorded speakers the experimenter read aloud fromthe material that had been provided (e.g., "Thegirl who recorded this summary is a student in theSchool of Education. She is opposed to the legali-zation of marijuana").

After hearing each summary, the respondents werepresented with two types of questions. First, theywere to try to estimate (in a rating format) theviews expressed in the original message on whichthe summary was based. The three rating scales thatwere used here were identical to the items that hadbeen provided to the judges in Experiment 1 (seeabove). The mean of these ratings served as anindex of the respondent's inferences concerning thecontent of the message that each speaker had sum-marized. As in Experiment 1, to encourage a careful,independent response to the various summaries, theraters were informed that many of the passages theywould hear were "based on rather similar inputsthat differed mainly in nuance and detail." Thesummaries were, in fact, derived from four differentoriginal messages: one that was promarijuana, oneantimarijuana, and two that expressed a relativelyneutral point of view.

In responding to each summary, the subjects alsocompleted a second group of rating items thatfocused on the speaker's apparent personality char-acteristics. Here they were instructed to give their"honest, subjective reactions" to each speaker, basedon what little information there was available (in-cluding the speaker's voice, speech pattern, etc.).Each speaker was rated with respect to her likabil-ity, overall intelligence, and emotional maturity.The mean of these ratings provided an overall indexof esteem for the individual speakers.

Experimental Design

The subjects in this study were summer studentsat the University of Michigan (12 men and 12women) who were paid for their efforts; they allfavored the legalization of marijuana. Each subjectresponded to 24 different message summaries. Thesummaries varied with respect to (a) the content of


5i

6.0UlI-K2

5.5

UJ

12 5.0

_, PRO-LEGALIZATION^SUMMARIES

ANTI-LEGALIZATION^SUMMARIES

NEUTRAL>..XNVIMMARIES *-\

V

PRO NEUTRAL UNDECIDED ANTI

PASSER'S ATTITUDEFIGURE 2. Respondents' reactions to the transmitters as a function of (a) the

message content that they conveyed and (b) the transmitters' apparent atti-tudes. (Experiment 2.)

the original messages on which they were based(pro-, neutral-, or antilegalization) and (b) theapparent attitudes of the transmitter (prolegaliza-tion, antilegalization, undecided, or unknown).These two variables were combined factorially toproduce all 12 (3 X 4) of the combinations thatwere possible (e.g., prolegalization summary, unde-cided transmitter; prolegalization, antilegalizationtransmitter; etc.). Each of these 12 combinationswas presented twice to every respondent, once inthe first half of the test session and once again inthe second half; while these two presentations wereconceptually equivalent, they were based on inde-pendently recorded message summaries.

Each of the recorded summaries was presented (todifferent respondents) as if it had emanated fromfour different passers. This was accomplished bypreparing four different response booklets containingdifferent "background information" about the vari-ous transmitters. While respondents who receivedBooklet I might be informed that a certain speakerfavored legalization, those who had received Book-let II (or III or IV) were provided with contrast-Ing background information (e.g., the speaker op-posed legalization or was undecided). Each responsebooklet was completed by three male and threefemale respondents.

ResultsEsteem for the Transmitter

The respondents' reactions to the variousspeakers were quantified by averaging theirratings across a set of three evaluative scales

(likability, intelligence, and emotional ma-turity). Figure 2 shows how these ratingswere affected by the transmitters' own atti-tudes and by the message contents that eachspeaker had been assigned. Note that regard-less of the views that were seemingly favoredby the transmitters themselves, those whohad (fortuitously) been assigned the job ofrelaying the promarijuana speech received themost favorable ratings. Since all 24 of theraters in this study favored the legalizationof marijuana, this result clearly supports ourinitial theorizing. As shown in Figure 2, therespondents reacted negatively not only tothose who had summarized messages thatthey (the listeners) found challenging, butthey reacted in a similarly negative fashion tospeakers who summarized neutral messages.This aspect of the data was somewhat unex-pected and may in part reflect the rater's dis-comfort in decoding summaries that did notendorse any single clearcut position.

An analysis of variance that was applied tothese data confirmed the fact that the re-spondents' ratings were significantly affectedby the message contents that the speakershad summarized (F = 9.9S, dj = 2/32, p <.001). Unfortunately, however, the interpre-


tation of these data is somewhat clouded bythe fact that all of our respondents favoredthe legalization of marijuana. A skepticalreader might be concerned about the possi-bility that the transmitters who deliveredpromarijuana messages might have been re-garded favorably by virtually any respondent,even a respondent who opposed legalization.While this is indeed a logical possibility, itdoes not seem very likely in an intuitivesense, nor does it seem plausibly derivablefrom any coherent theory of attraction. Ourpresent view is that these results support thetheoretical concerns that led to the design ofthis experiment, and testify to the fact thatwe normally respond rather favorably tothose who provide us with supportive informa-tion.

The ratings that the speakers receivedwere also affected by their "own views" (F =7.37, d f = 3/48, p < .001). This result rep-licates Byrne's numerous studies relating lik-ing and attitude similarity; that is, the trans-mitters who seemingly disagreed with therespondents' prolegalization attitudes receivedless favorable ratings than those whose viewsseemed more supportive. Surprisingly, how-ever, the neutral transmitters received ap-proximately the same ratings as those whoappeared to share the respondents' prolegali-zation views.

It is interesting to compare the extent towhich our respondents' liking for the variousspeakers was affected by the attitude infor-mation with which they were provided, ascontrasted with the message contents that thespeakers summarized. When we consider onlythe extremes of these two variables, the over-all differences in liking that are attributableto the content of the original messages (pro-versus antilegalization) are slightly larger inmagnitude than the differences in liking thatderive from the transmitters' own presumedviews (pro- versus antilegalization). In brief,the message contents that the transmitterssummarized seemed at least as important inaffecting their likability as the opinions thatthey themselves appeared to favor.

Decoding the Summaries

After hearing each summary, the respond-ents indicated the attitude position that they

believed to have been expressed in the originalmessage. Not surprisingly, a highly significanteffect was produced by varying the messagecontent that the speakers were summarizing;summaries derived from prolegalizationspeeches were (correctly) thought to haveemanated from relatively promarijuana mes-sages, as contrasted with those derived fromantilegalization speeches. There were, how-ever, no significant effects associated withvariations in the transmitters' apparent views.That is, in decoding the various summaries,our respondents were essentially unaffected bythe information that was provided about thetransmitters' attitudes (the F value for themain effect of the passer's attitude was lessthan 1.00, as was the interaction betweenthis variable and the original message con-tent). These data thus indicate that our re-spondents did not, as a group, implicitly "cor-rect for" biases that might have been intro-duced into the summaries as a consequence ofthe transmitters' preferred views. While it isconceivable that the individual raters mayhave made adjustments of this sort, there wasclearly no consensus on how such adjust-ments should be made, for the overall results(across respondents) were far from significant.

EXPERIMENT 3

Experiment 2 established that the averagelistener did indeed react negatively to thosewho transmitted challenging, attitude-relevantmessages. These results were consistent withthe idea that in relaying attitude-relevant in-formation, a transmitter might plausibly at-tempt to ingratiate herself with her listenerby producing a biased summary, designed todecrease the dissonance between the originalmessage content and the views of the listener.

Starting with these premises, it seems rea-sonable to anticipate that audience-inducedcommunication distortions might be amplifiedif the transmitter was addressing her sum-mary to someone whom she especially liked.The rationale here seems rather straight-forward: If message summaries are slantedbecause of the transmitters' attempts to in-gratiate themselves with their listeners, thenthis biasing effect should be particularlymarked when the listener is someone who isliked, and thus a person whose affective


response to the transmitter would be amatter of some importance. Experiment 3 wasdesigned to test this hypothesis.

MethodExperiment 3 was closely modeled after the meth-

ods that were employed in the long-delay conditionsof Experiment 1. Once again, a group of paid fe-male subjects (A' = 96) was recruited for an experi-ment on "communication" in which, together witha partner, they would each discuss some currentsocial issue.

When the subjects were initially recruited (a fewdays before the actual experiment), they all com-pleted a questionnaire which was presumably to beused in establishing pairs that would work welltogether. Each subject, therefore, indicated the sortof person with whom she could "get along" and"communicate" most effectively, and she also de-scribed herself, using a series of rating scales.

On arriving in their separate experimental cubicles,all of the subjects found a brief note which indi-cated either that (a) the experimenter had succeededin locating a partner whom the respondent wouldprobably find to be a likable person who was easyto get along with; or (6) conversely, that the part-ner was unfortunately the sort of person whom thesubject did not normally like. This method for ma-nipulating liking had previously been successfullyemployed by Berkowitz and Goranson (1964),among others. To enhance the respondents' motiva-tion for eliciting a positive reaction from these part-ners, the introductory notes all ended with anindication that despite the experimenter's knowledgeas to how the subject would probably react to herassigned partner, it was not at all clear how thepartner would react to the subject, and hence,by implication, ingratiating tactics might well benecessary to insure a positive response.

After each subject had finished reading about herassigned "partner," the experimental session pro-ceeded along the lines that had been established forthe early-information conditions of Experiment 1(where the subjects had learned about their part-ner's views before receiving the message that theywere subsequently to summarize). The only note-worthy change was the fact that following receiptof the original marijuana message, the retentioninterval was lengthened to 23 minutes during whichtime each subject heard recorded talks dealing withdesegregation, pornography laws in Denmark, andthe problem of health care in America.

Experimental DesignThrough the procedures that are described above,

a 2 X 2 X 2 factorial design was established, with12 female respondents in each cell. The independentvariables were partner's opinion (pro- versus anti-legalization) , partner's apparent likability (like versusdislike), and message (which of the two "back-ground" messages had this respondent heard?).

ResultsValidity Check

At the end of the experimental session, eachrespondent rated the girl with whom she waspaired in terms of likability, intelligence, andemotional maturity. These ratings were com-bined additively, to provide an overall mea-sure of the respondent's esteem for her as-signed partner, and an analysis of variancewas conducted on the resulting data. Theresults indicated that our attempt to manipu-late the subjects' affective reaction to theirpartners was not fully successful, for the dif-ference between those assigned to the likeversus dislike conditions was only significantat the .10 level (F = 3.48, dj = 1/88, two-tailed). At a descriptive level, the differencebetween the contrasting affectivity conditionsseemed fairly substantial among the respon-dents who were paired with prolegalizationpartners; however, the obtained results weretrivial (although in the anticipated direction),among those whose assigned partners wereseemingly opposed to legalization.

The Summary Ratings

When subjected to analysis of variance pro-cedures, the judges' ratings of the messagesummaries showed only one reliable effectof substantive interest: The summaries ad-dressed to prolegalization listeners werethought to derive from speeches that weremore favorable to marijuana than were thoseintended for listeners who opposed legaliza-tion (F = 10.03, dj = 1/88, p < .005). Theseresults replicated the effects obtained in thelong-delay conditions of Experiment 1, foronce again, following a modest (23-minute)retention interval, the transmitters producedmessage summaries that were displacedtoward the views seemingly favored by theirpartners. By contrast, the liking variable didnot have a reliable impact on the summaries.Most important, while our theoretical analy-sis had suggested that the audience effectmight be especially pronounced when the re-spondents were favorably disposed towardtheir partners, there was essentially no evi-dence of an interaction between the likingvariable and the purported views of theaudience (F < 1.00).


It is, of course, possible that the apparentfailure of the likability hypothesis derivesfrom our relatively limited success in experi-mentally manipulating this variable (see thepreceding Validity Check section). It there-fore seemed reasonable to reanalyze thesedata, this time, however, ignoring the likingconditions into which our subjects had beenassigned, and as a substitute, focusing on theesteem for their partners that the respondentshad expressed in responding to the final ques-tionnaire. A secondary analysis based on therespondents' rated esteem for their partnersdid not, however, yield any reliable evidencefor the ingratiation hypothesis that had moti-vated the design of this experiment.

Questionnaire Results

Own views. The respondents' own viewswere assessed by means of self-ratings at theend of the experimental session. Recall thatin Experiment 1 the subjects assigned tothe long-delay conditions tended to migratetoward their partners' favored positions (pro-or antilegalization). While this directionaltrend was replicated in the present experi-ment, for reasons that are presently unclear,the obtained effect was far from significant(F < 1.00), nor did the respondents' assign-ment to the like (versus dislike) conditionhave a reliable impact on their final opinionsregarding legalization (F < 1.00).

Ratings of the original message. The post-experimental questionnaire included a seriesof rating scales concerning the viewpoint thatthe initial speaker had endorsed in his back-ground speech. The obtained ratings werereasonably consistent with the freely verbal-ized summaries that had previously beencollected, for those respondents who werepaired with prolegalization partners remem-bered the speaker's position as being morefavorable to legalization than did those whosepartners seemingly opposed legalization(F = 3.11, df = 1/88, p < .10).° Note, how-ever, that while the message summaries re-flected the respondents' public recollection ofthe background speech (i.e., these remarkshad been openly addressed to their assigned

6 This finding emerged from an analysis ofvariance based on the respondents' ratings.

partners), the questionnaire results representa relatively private recollection, intended pri-marily for a presumably objective experi-menter, and hence devoid of any conscious(deliberate) bias. The present results thussuggest that the views favored by the audi-ence may affect not only the respondents'public re-creations of messages that they havepreviously received, but they may also havean impact on their private recollection ofthese messages. While this effect was notmanifested in Experiment 1, it is nonethe-less a reasonably reliable finding, for it hasbeen manifested at conventional significancelevels in two additional studies within thisseries (not fully reported here).

DISCUSSION

Taken together, Experiments 1 and 3 pro-vide substantial support for the propositionthat the average person will often "censor"the information that he (or she) relays toothers, so as to reduce the dissonance betweenthe listener's apparent views and the viewsexpressed in the transmitted message. Theseresults seem closely related to recent workby Rosen, Tesser, and their associates (Rosen& Tesser, 1970, 1971; Tesser & Rosen, 1972)which showed that people have a rather gen-eral tendency to remain mum about unpleas-ant messages (the MUM effect). In one ex-periment, for example, the subjects overhearda message intended for someone else (therecipient); the message was a request for thisperson to call home regarding some very good(or bad) news as soon as he arrived at thelaboratory. When the recipient (a confeder-ate) finally arrived, virtually all of the sub-jects mentioned the phone call. Those as-signed to the good news condition, however,were more likely to mention the content(valence) of the message than were those inthe bad news condition. If we make the rea-sonable assumption that information whichchallenges the recipient's views constitutesone form of bad news, the convergence be-tween the present results and those of Rosenand Tesser seems obvious.

Experiment 2 indicated that transmittersare likely to be rejected if they communicateinformation that challenges the views of theirlisteners. This result, in combination with the


audience-induced displacements that are dis-cussed above, suggests that the average per-son may, through bitter experience, learnabout the negative consequences that willoften follow if he transmits excessively chal-lenging information and may, as a result,tailor his verbal output (see Rosen, Johnson,Johnson, & Tesser, 1973, for a somewhatdifferent approach to the relation betweenattraction and communication).

Whether or not this explanation ultimatelyproves viable, the present studies suggest thatbearers of bad (dissonance-producing) newsmay quite regularly elicit negative reactionsfrom their listeners. Apart from the data re-ported in Experiment 2 (see Figure 2), thisrelationship between the agreeableness of thetransmitter's message and her apparent lik-ability was independently confirmed in asubsequent experiment in which the judgeswho rated the contents of several messagesummaries also responded to a series of evalu-ative items regarding such matters as thetransmitters' likability, overall intelligence,warmth and friendliness, and emotional matu-rity. All of the judges favored the legaliza-tion of marijuana, and hence would presum-ably agree with prolegalization summaries;their ratings indicated that they respondedmost favorably to those respondents whosesummaries seemed to derive from relativelyprolegalization speeches. The correlations be-tween the transmitter's likability, as measuredby the mean of the evaluative ratings thatshe elicited, and the agreeableness of her mes-sage (i.e., its favorability toward legalization)were +.372 and +.499 for two independentsamples («i = »2 = 32) ; these correlationswere significant at the .05 and .01 levels, re-spectively. The consistency of this correlationbetween the transmitter's likability and theagreeableness of her message is particularlyinteresting in view of the relatively limitedrange of opinions that were represented in therecorded passages, for all of the summaries inthis study were based on essentially neutralbackground information.

As noted earlier, the fact that our judgesreacted negatively to messengers who broughtthem bad tidings represents something of anunwarranted reaction, for these messengerswere presumably trying, as best they could,

to follow the experimenter's suggestion thatthey relay the main points of a speech thatthey had heard previously, and did not neces-sarily endorse the views expressed in thisbackground talk. Jones and Harris (1967)have reported several studies showing whatis probably a closely related phenomenon.These investigators found that subjects wholistened to a speech in which the speaker hadbeen instructed to take a particular attitudestand (e.g., pro-Castro) and who were fullyaware that the speaker had no choice in theposition that he defended, nevertheless tendedto assume that the speaker's true attitudeswere probably in line with his overt verbalbehavior. Similar results have been reportedby Jones, Worchel, Goethals, and Grumet(1971). Taken together with the present ex-periments, these studies suggest that in inter-preting and evaluating the behavior of others,we are likely to be unduly influenced by theovert acts that we see and hear, and are rela-tively unaffected by the knowledge that theseactions may represent assigned role behaviors,rather than the spontaneous expression ofsome personally felt motive. Jones and Nis-bett (1972) suggest that effects like this mayderive from a general propensity to accountfor the behaviors of others in terms of stablepersonal dispositions, in contrast to the situ-ational variables that seem so potent inguiding our own actions.

Final Remarks

The communication-chain paradigm is wellknown to social psychologists, although it hasrarely (if ever) been subjected to the sortof sophisticated, systematic experimentationthat has characterized such areas as attitudechange or conformity. Given the feasibility ofsuch research and the potential gains thatmay be realized, the study of communicationchains seems to be a promising area for futureinvestigation.

REFERENCES

BERKOWITZ, L., & GORANSON, R. E. Motivationaland judgmental determinants of social judgment.Journal of Abnormal and Social Psychology, 1964,69, 296-302.

BYRNE, D. Attitudes and attraction. In L. Berkowitz(Ed.), Advances in experimental social psychology.Vol. 4. New York: Academic Press, 1969.


CAMPBELL, D. T. Systematic errors on the part ofhuman links in communication systems. Informa-tion and Control, 1958, 1, 334-369.

FESTINGER, L. A theory of cognitive dissonance.Stanford, Calif.: Stanford University Press, 1957.

GRACE, H. A. Effects of different degrees of knowl-edge about an audience on the content of com-munication. Journal oj Social Psychology, 1951,34, 111-124.

GRIFFITT, W., & VEITCH, R. Hot and crowded: In-fluence of population density and temperature oninterpersonal affective behavior. Journal oj Per-sonality and Social Psychology, 1971, 17, 92-98.

JANIS, I. L., & KING, B. T. The influence of roleplaying on opinion change. Journal of Abnormaland Social Psychology, 19S4, 49, 211-218.

JONES, E. E., & HARRIS, V. A. The attribution ofattitudes. Journal of Experimental Social Psychol-ogy, 1967, 3, 1-24.

JONES, E. E., & NISBETT, R. E. The actor and theobserver: Divergent perceptions of the causes ofbehavior. In E. E. Jones et al. (Eds.), Attribu-tion: Perceiving the causes of behavior. Morris-town, N.J.: General Learning Press, 1972.

JONES, E. E., WORCHEL, P., GOETHALS, G. R., &GRUMET, J. Prior expectancy and behavioral ex-tremity as determinants of attitude attribution.Journal of Experimental Social Psychology, 1971,7, 59-80.

ROSEN, S., JOHNSON, R. D., JOHNSON, M. J., &TESSER, A. The interactive effects of news valenceand attractiveness on communicator behavior.Journal of Personality and Social Psychology, 1973,28, 298-300.

ROSEN, S., & TESSER, A. On the reluctance to com-municate undesirable information: The MUM ef-fect. Sociometry, 1970, 33, 253-263.

ROSEN, S., & TESSER, A. Fear of negative evaluationand the reluctance to transmit bad news. Proceed-ings of the 79th Annual Convention of the Amer-ican Psychological Association, 1971, 6, 301-302.(Summary)

SCHRAMM, W., & DANIELSON, W. Anticipated audi-ence as determinants of recall. Journal of Abnor-mal and Social Psychology, 1958, 56, 282-283.

TESSER, A., & ROSEN, S. Similarity of objective fateas a determinant of the reluctance to transmit un-pleasant information: The MUM effect. Journal ofPersonality and Social Psychology, 1972, 23, 46-53.

ZAJONC, R. B. The process of cognitive tuning incommunication. Journal of Abnormal and SocialPsychology, 1960, 61, 159-167.

ZIMMERMAN, C., & BAUER, R. A. The effect of anaudience on what is remembered. Public OpinionQuarterly, 1956, 20, 238-248.

(Received November 8, 1972)

30, no. 1, transmission of attitude ... - uni-muenster.de€¦ · 1974, vol. 30, no. 1, 81-94...

Documents