COMMUNICATION STRATEGIES AFFECTED BYAUDIENCE OPPOSITION, FEEDBACK

AND PERSUASIBILITY

MICHAEL D. HAZEN and SARA E. KIESLER

IT is generally accepted in rhetoricaltheory that speakers systematically

alter their message presentations to suitdiffering audiences. Aristotle argues thatthe speaker needs to take into accountthe views of his audience1 and Burke'sconcept of identification focuses on thespeaker finding common ground with hisaudience.2 Brockriede and Scott said ofStokely Carmichael's speeches to a blackand white audience, "He makes manyof the same points. . . . The two speechesdiffer distinctively in style and persua-sive appeals, however; in each instancestyle and appeals are appropriate to theaudience addressed."3 On the other hand,experimental studies of persuasion havealmost exclusively investigated attitudi-nal, affective and behavioral changes inthe audience. While it is widely assumedthat communicator behaviors vary, nomajor attempt has been made to specifyhow they vary or why. The presentstudies were conducted to examine ex-

Mr. Hazen is Assistant Professor of Speech Com-munication and Theatre Arts at Wake ForestUniversity. Ms Kiesler is Associate Professor ofSpeech Communication and Human Relationsat the University of Kansas. These studies werein part supported by a NICHD contract (NO1-HD-2-2703) with the second author.

1 Lane Cooper, The Rhetoric of Aristotle(New York: Appleton-Century-Crofts, 1932), pp.16-17, 91-181.

2 Kenneth Burke, A Rhetoric of Motives(Berkeley: Univ. of California Press, 1950), pp.19-46.

3 Wayne Brockriede and Robert Scott,"Stokely Carmichael: Two Speeches on BlackPower," in The Rhetoric of Black Power, ed.Robert Scott and Wayne Brockriede (New York:Harper & Row, 1969), p. 115.

SPEECH MONOGRAPHS, Vol. 42, March, 1975

perimentally the effect of audience char-acteristics on communication strategies.

A review of previous empirical re-search and theory sugests that a com-municator, aware that his audience canignore the message, reject its conclu-sions, or derogate the communicatorhimself, will "slant" the message so asto: a) make the position argued seemminimally discrepant from the audience'sand b) make his behavior appear ac-ceptable. Goffman's theoretical work onsocial approval suggests that the com-municator will attempt to satisfy theexpectations of his audience and willemploy various "facework" strategies soas to avoid appearing too discrepantfrom them, or too much in violation oftheir norms.4 Byrne's theory of similar-ity and interpersonal attraction, sup-ported by many empirical demonstra-tions of a positive relationship betweenthe two, predicts that communicatorswill attempt to appear as similar totheir audience—in attitude, as well aspersonality—as possible.5 Moreover,Manis, Cornell and Moore have demon-strated that neutral messages trans-mitted to audiences on each side of anissue are judged by others to be relative-ly congruent with the audiences' opin-ions.6 This research has led us to begin

4 Erving Goffman, The Presentation of Selfin Everyday Life (Garden City, New York:Doubleday, 1959), pp. 1-16.

5 Donn Byrne, The Attraction Paradigm(New York: Academic Press, 1971).

6 Melvin Manis, S. Douglas Cornell, andJeffrey Moore, "Transmission of Attitude-Relevant Information Through a Communica-

COMMUNICATION STRATEGIES 57

with the assumption that the audience'sinitial opposition to the communicator'sposition will be an important factor indetermining his strategy.

There is, however, little empirical evi-dence indicating exactly how messagesare slanted in order to avoid rejectionor derogation. Research on the "MUMeffect" by Abraham Tesser and his col-leagues, showing that people avoid bear-ing bad news to others, suggests thatcommunicators may omit very discrep-ant statements when facing an opposedaudience.7 (In fact, Zimmerman andBauer, and Schramm and Danielsonfound that the communicator tends notto recall those statements most incon-gruent with his anticipated audience'sbeliefs.)8 But many other strategies canbe envisioned. Making the message lessspecific, taking a less extreme positionor using weaker language, and spendingrelatively more time on issues and prob-lems (rather than solutions) are possiblestrategies aimed at reducing the opposedaudience's awareness of the communi-cator's position.

A communicator may also have otherconcerns when he faces an opposed audi-ence. He may believe that greater dis-crepancy results in more counterargu-ing, which may be precluded either bydecreasing explicitness, by "distracting"the audience or by presenting counter-arguments. He may assume that audi-ences who are in major disagreementwill be very difficult to persuade. In that

tion Chain," Journal of Personality and SocialPsychology, 30 (1974), 81-94.

7 For example, Abraham Tesser and Sid-ney Rosen, "Similarity of Objective Fate asa Determinant of the Reluctance to TransmitUnpleasant Information: The MUM Effect."Journal of Personality and Social Psychology, 23(1972), 46-53.

8 Claire Zimmerman, and Raymond Bauer,"The Effect of an Audience on What is Re-membered." The Public Opinion Quarterly, 20(1956), 238-248; Wilbur Schramm and WayneDanielson, "Anticipated Audiences as Determin-ants of Recall." Journal of Abnormal and SocialPsychology, 56 (1958), 282-283.

case the communicator might marshalas many arguments as he can to supporthis position. Or he may concentrate onproblem-centered arguments, hoping toincrease the audience's perception of theneed for change and to prepare them foracceptance of his solutions and conclu-sions.

The first experiment reported hererepresents an attempt to discover whe-ther audience opposition would systema-tically affect the communicator's strat-egy. A procedure was used which allowedcommunicators arguing on one side of anissue to choose more or fewer argumentsvarying in their strength, specificity,and focus. Audience opposition was in-dependently manipulated.

EXPERIMENT I

Method

Eighty volunteers were recruited fromcollege debate and public speaking class-es, and from a high school summer de-bate camp, for a study on the construc-tion of persuasive speeches. Each sub-ject was handed a page of written in-structions, a set of 28 3x5 printed cards,an answer sheet, and a page of questionsto answer later. The instructions stated:"We would like you to develop the mosteffective speech you can with the in-formation we give you. Your job is toconstruct a 5-10 minute speech on thetopic: THE FEDERAL GOVERN-MENT SHOULD CONTROL POPU-LATION GROWTH. You have beengiven a set of cards to use in making upyour speech. Please use the answer sheetto list the cards you will use in yourspeech, and list them in the order theywill be used. You cannot use all thecards because that would make thespeech longer than 10 minutes."

Following these statements was oneof four descriptions of the hypotheticalaudience. Three of these statements

58. SPEECH MONOGRAPHS

varied the opposition of the audience toFederal control of population growth.In the Very Opposed condition, thestatement read: "Assume that your au-dience is a group of women who arevery opposed to Federal governmentalcontrol over population growth. Designyour speech so that it is the most per-suasive speech for this audience." Inother conditions "moderately opposed"or "moderately in favor" was inserted.There was also a No Opposition, com-parison condition. The instructions tothe subjects assigned to this conditionnever mentioned persuasion. The taskwas depicted as one requiring the con-struction of "effective speeches, as in de-bate tournament," where the audiencewould be "a group of experienced debatejudges who will judge your speech on thebasis of its logic, your analysis, andclarity." Random assignment to one ofthe four audience conditions was em-ployed.

Experimental materials and measures.On each 3" x 5" card was printed oneargument which might be used in aspeech favoring Federal involvement inpopulation control. All arguments wereculled from previous public speeches andarticles. Their content concentrated onfive areas: education, legal issues, taxes,birth control and abortion. Eight of thecards were designed to be used if thesubject wanted to arouse concern aboutthe problem of population growth, showhow serious the problem was, or giveevidence that Federal intervention wasnecessary, e .g., "Unwanted children areborn partly because parents have notknown about birth control methods. . . ."These cards differed from another setof IS cards which presented solutionsto the population growth problem bystating how the government might inter-vene, e.g., "Massive programs of mediaeducation on population growth, coupledwith sex education programs in our

schools, and distribution of birth con-trol pamphlets . . . could provide amechanism for informing the public."Of five remaining cards, one stated theconclusion and four could be used ascounterarguments against leaving popu-lation growth uncontrolled, or to the in-fluence of private groups, or to statecontrol.

On the answer sheet were 28 blankspaces. The subject simply listed theidentifying number of each argument hewould use (in order) to construct hisspeech. He was reminded not to useall 28 because of the 10 minute limit.Scoring was accomplished by countingthe total number of arguments listed,the number of problem and solution fo-cused arguments, the number of counter-arguments, and whether conclusions wereemployed. In addition, the strength (ex-tremity) and the specificity of each listof arguments- was scored. First, a sepa-rate class of 40 undergraduates ratedeach argument according to how strongand how specific or general it was sothat each argument could be assigneda strength score and a specificity scorebased upon the average ratings. Thenthe average of the strength scores andspecificity scores for each subject couldbe computed.

When the subject had completed hislist (about 20 minutes), he turned tothe next page and answered three ques-tions. The first asked how effective hisspeech would be; the second, how hepersonally felt about the issue (pre-testing had established that most stu-dents were in favor of Federal inter-vention); and finally there was a checklist of goals he might have had in mindwhen constructing his speech (e.g., tobe persuasive, to offer arguments whichbuilt on previously held attitudes).

Upon completion of the questionnaire,the subject was informed fully of theexperimenter's purpose and procedures.

COMMUNICATION STRATEGIES 59

Results

The subjects apparently responded tothe instructions as intended, using from6 to 21 cards. Approximately 75% ofthe subjects in each condition stated thatone of their goals was to persuade, andmost intended to present a well organ-ized and analyzed speech. There wereno significant differences among condi-tions in perceived effectiveness of theplanned speech and in the subjects' atti-tudes.

One-way analyses of variance wereused to assess the total number of argu-ments, the strength of arguments, thespecificity of arguments, and the use ofconclusions and counterarguments. How-ever, the dependent measure of mostinterest was the use of problem centeredarguments vs. solution centered argu-ments. This measure was tested using a2 x 4 approximate analysis of variancefor profile data with one between factor(audience opposition) and one withinfactor (type of argument). This analysistests for the significance of the differenceof the relative use of each strategy bysubjects.9 The interaction was highly sig-nificant (F — 12.62, df = 3/76, p <.O1).The greater the opposition of the audi-ence, the fewer solution arguments andthe more problem arguments were used.Means are presented in Table 1.

9 Profile analysis is an approximate form ofanalysis o£ variance which is a combinationof the randomized factoral design and thetreatment by subjects design. Such a mixeddesign (or factoral designs with repeated mea-sures) is discussed in Donald F. Morrison, Mul-tivariate Statistical Methods (New York: Mc-Graw-Hill, 1967) pp. 221-258; Seymour Geisserand Samuel Greenhouse, "An Extension ofBox's Results on the Use of the F Distributionin Multivariate Analysis," Annals of Mathe-matical Statistics, 29 (1958), 885-891; SamuelGreenhouse and Seymour Geisser, "On Methodsin the Analysis of Profile Data," Psychometrika,24 (1959-112.

Such a form of analysis involves a betweensubjects factor and a within subject factor.This design leads to three questions: a) arethe population mean profiles similar or paral-lel? b) if they are parallel are they on thesame level? and c) assuming parallelism, are

TABLE 1AVERAGE FREQUENCY OF PROBLEM CENTERED

ARGUMENTS AND SOLUTION CENTERED ARGUMENTSPLANNED FOR USE WITH AUDIENCES VARYING

IN INITIAL OPPOSITION TO THECOMMUNICATOR'S POSITION

Audience

Very OpposedTV — 9 Ai\ £,\J

Mod. OpposedN = 20

Mod. in FavorN = 20

No Opposition

Type ofProblemCentered

(8 possible)

5.50

4.85

4.70

4.80

Argument UsedSolutionCentered

(15 possible)

5.75

6.95

7.15

8.20

Note: The higher the mean, the more fre-quently subjects in that experimental conditionchose to use the type of argument indicated.

There were no other significant (or nearsignificant) differences in strategy amongconditions.

the means different? The first question dealswith the interaction effect, the second questionwith the main effects for factoral treatmentsand the third question with effects for repeatedmeasures. Tests for the second question canfollow traditional analysis of variance withits exact assumptions, but the first and thirdquestions do not meet the assumptions of equalvariance and mutual independance or equalcorrelations. Thus an appropriate F-distribu-tion has been developed for such problems fromthe work of Box (Annals of MathematicalStatistics, 25 (1954), 290-302, 484-498) by Geisserand Greenhouse. This work has also led to aconservative adjustment of degrees of freedomby Greenhouse and Geisser.The summary table for the mixed analysis ofvariance in Experiment 1 is presented below:

SUMMARY OF ANALYSIS OF VARIANCE

Source

Between subjectsAudience

conditionsSubject within

groups

Within subjectsSolution/Problem

ArgumentsAudience x

ArgumentsArguments x

Subjects withinGroups

Total

df79

3

76

80

1

3

76159

ms F P

5.41

5.56

168.10 121.81 .001

17.42 12-62 .001

1.38

197.87

Note: The large main effect for use of argu-ments is trivial because there were al-most twice as many solution centeredcards available.

60 SPEECH MONOGRAPHS

As noted above, most subjects checkedas a goal of their speech that they in-tended to be persuasive, and to presenta well organized speech. There were nodifferences among conditions in thechecking of these goals, or of most ofthe others listed. Two goals, however,were differentially selected, and thesewere consistent with the different strat-egies depicted in Table 1. That is, sub-jects in the more opposed conditionswere more likely to say (in retrospect)that they had intended to select prob-lem centered arguments, and also to offer"arguments appearing to emotion."

Discussion

Given that the subjects could havevaried proposed speeches in a numberof ways, the systematic preference forproblem centered arguments by thoseanticipating more opposed audiences isof interest. But there are several reasonswhy they should have done so. As wediscussed earlier, they may have as-sumed that by avoiding explicit solu-tions and focusing on problems the op-posed audience would be less aware ofthe communicator's actually discrepantposition, and presumably would be lesslikely to reject his message or derogatehim. We call this the "avoidance of re-jection" hypothesis. Or, in accord witha "need to arouse" hypothesis, they mayhave been assuming that audiences op-posed to control of population growthare particularly difficult to persuade.Problem centered arguments may havebeen used for the purpose of arousingconcern with the issue of populationgrowth and thereby preparing the audi-ence to accept new solutions sometimein the future. Finally, the use of problemcentered arguments may have reflecteda lack of motivation among communica-tors asked to persuade opposed audiences(the "giving up" hypothesis). If theyperceived that no real attitude change

could occur, they may have resignedthemselves simply to increasing interestin the issue. The questionnaire data(e.g., equal persuasive goals) make thispossibility doubtful, however.

EXPERIMENT II

The second experiment was designedfor the purpose of replicating the pre-vious results, and distinguishing amongthe possible causes: of differences in strat-egy. In addition, the procedure was al-tered somewhat so as to increase "ex-ternal validity." The design was a 2 x 3factorial, with audience opposition anda new factor, anticipated audience feed-back, varied orthogonally.

Method

As before, subjects were asked to plana persuasive speech favoring Federal in-volvement in population control. Theyexpected that their target audience, an-other person, would actually read theirchosen arguments shortly afterward.The audience was to be either opposedto their position, or somewhat in favorof it. The potential impact of the audi-ence was increased by specifying exactlyhow discrepant the audience's positionwas and by making reference to the diffi-culty (or ease) of changing the audience.It was expected that an audience maineffect but no interaction with feedbackwould support the "need to arouse" hy-pothesis, i.e., that problem centered ar-guments are used with opposed audi-ences to arouse and prepare them forlater acceptance of new solutions.

The feedback manipulation was em-ployed principally to distinguish betweenthe "need to arouse" hypothesis and the"avoiding rejection" hypothesis. It wasassumed that if avoiding rejection oftheir message is of some importance,communicators should be even moreconcerned about rejection when the

COMMUNICATION STRATEGIES 61

audience's assessment of their messagemust be faced. Some previous researchdemonstrates this increased concern whenfeedback is expected. For example,Brown and Garland found that face sav-ing attempts increased when amateursingers expected to interact with theiraudience later.10 In the present study,it was predicted that if problem centeredarguments are employed in order to con-ceal discrepancy, and prevent rejection,they would be chosen more often whenfeedback was anticipated than when itwas not. We would expect either a Feed-back main effect, or an interaction suchthat problem centered arguments arechosen most often when feedback is an-ticipated from a very opposed audience.

Two different feedback manipulationswere employed in order to distinguishbetween the effects of simply anticipat-ing feedback and anticipating that theaudience will personally deliver theirfeedback. Concern with rejection mightbe highest when future interaction withthe audience is expected, as well as theirfeedback.

Subjects. The subjects were collegestudents recruited through an ad in theUniversity of Kansas student newspaperfor a study of "attitudes toward currentissues," which would pay $2.00 for par-ticipation. All of the students who calledduring the specified hours were allowedto participate, but only those who indi-cated on the pretest that they favoredFederal support of population control,a total of 94, were included in the majoranalyses. The subjects were assembledin groups of four to eight in a centralexperimental room, surrounded by in-dividual cubicles.

Procedure. When all subjects had ar-rived, the experimenter asked them to

10 Bert R. Brown and Howard Garland, "TheEffects of Incompetency, Audience Acquaint-anceship, and Anticipated Evaluative Feedbackon Face-Saving Behavior," Journal of Experi-mental Social Psychology, 1 (1971), 490-502.

complete a survey of their attitudes to-ward some current issues. The surveyconsisted of ten statements with whichthe subjects were to indicate their agree-ment or disagreement. One of the state-ments, used as the pretest of attitude,was "The Federal government shouldsupport population control measures."The experimenter scored the surveyswhile the subjects waited. When he fin-ished, he remarked that the variationamong their attitudes was not unusualand went on to explain the major pur-pose of the study: to investigate atti-tudes about population control. Then,he requested that they each enter anindividual cubicle where they were givena written set of instructions.

In both feedback conditions the in-structions said, "We are interested inwhat kinds of arguments people usewhen they are talking about populationcontrol. Half of the participants in thisgroup will be assigned the job of per-suading someone else in the group thatthe Federal government should supportpopulation control measures." The Feed-back subjects were informed that theexperimenter would select those most infavor of Federal support for populationcontrol to be the persuaders. Each ofthose asked to be a "persuader" was toplan a message which would be pre-sented to one of those persons in thegroup who was not in favor of Federalintervention. Actually, all subjects wereassigned the persuader role.

Those in the Feedback conditions werethen further subdivided. Half of the sub-jects expected only that their audiencewould evaluate their arguments in pri-vate and send to them a written evalu-ation. Half expected that in addition toreceiving the written evaluation, theywould meet with their audience so that"the other person could give . . . somepersonal feedback, too."

In the No Feedback condition, the

62 SPEECH MONOGRAPHS

subjects were not told that they wouldhave to persuade a specific person in thegroup, but just that they would haveto construct a persuasive speech infavor of Federal support for populationcontrol.

Audience opposition was manipulatedby varying the instructions page also.In the No Feedback condition the audi-ence was described as either "someonewho is very opposed to population con-trol measures (usually someone who isvery opposed is very hard to persuadebecause people like this tend to haveconsidered all of the alternatives)," or"someone who is moderately in favor. . . (usually very easy to persuade be-cause people like this tend not to haveconsidered all of the arguments)."

Audience opposition in the Feedbackcondition had to be manipulated in aslightly different manner because theaudience was a particular other person.The manipulation was accomplished byshowing the subject his own score onthe attitude survey and the supposedscore of his audience. The latter scorewas bogus—very opposed (5), or mod-erately in favor (25) of population con-trol measures. At the same time, the ex-perimenter made the same remarksabout the audience as in the No Feed-back condition.

The speech construction task was es-sentially the same as in Experiment I.In explaining how to plan their speech,all of the subjects were told that a setof cards with arguments printed on eachwould be used so as to establish "experi-mental control." There were 25 cardsused in this experiment. Ten of the argu-ments were problem centered, ten weresolution centered, three were counterar-guments, and two were conclusions (prob-lem, solution centered). Half of the prob-lem centered and solution centered cardspresented a general argument, and halfadded to the general argument some de-

tailed examples or proposals. One argu-ment in each of the resulting four cate-gories dealt with each of the followingtopics: legal issues, tax incentives, abor-tion, birth control, and education. Thestrength and specificity of each argumentwas rated by two undergraduate classesso as to provide a mean score for each.

The subjects were asked to chooseonly 10 arguments in constructing theirpersuasive messages. (In the Feedbackconditions these would be given to theaudience for evaluation.) They were alsoasked to estimate how long they wouldspend to present their arguments. After30 minutes, the subjects completed apostquestionnaire, to measure the effec-tiveness of the manipulation, attitudechange, and perceptions- of the experi-mental situation. Afterwards, the entireexperiment was explained, and the sub-jects were dismissed.

Results

Effectiveness of the Manipulations.Postquestionnaire items measured theextent to which the feedback and audi-ence opposition manipulations were ef-fective. These were assessed by means of2 x 3 analyses of variance for each item.With respect to the audience's position,the subjects in the Opposed conditionperceived their audience to be muchmore against Federal intervention thandid subjects in the Favorable condi-tion (F = 210.9, # = 1 / 8 8 , p<Wl).The difference, however, was greater inthe two feedback conditions (interac-tion F = 10.23, df = 2/88, p <01). Weattribute this effect to greater salienceof the audience's position being effectedwhen the subject was actually shownhis and the other's pretest score.

Despite the mention of persuasion dif-ficulty in the Opposed condition, sub-jects did not differ in answer to thequestion, "How difficult was it to per-suade your audience?" However, the

COMMUNICATION STRATEGIES 63

subjects in the Opposed conditions feltthat they had been less effective inpersuading their audience than did thesubjects in the Favorable conditions(F = 4.77, df = 1/88, p <.O5).

The question, "How likely is it thatyou will receive feedback about yourarguments?" produced no differencesamong conditions. We suspect that thefailure of this question to discriminatewas due to subjects knowing that allexperiments must include a debriefing.The measure of the subjects' perceptionsthat they would receive personal feed-back did produce differences, such thatthose in the Personal Feedback conditionrated the probability of personal feed-back greater than in the other conditions(F = 12 AS, df-2/88, p<.01). Therewere no differences among conditions forsubjects ratings of their own attemptto persuade or for attitude change.

Strategy differences. To assess the ef-fect of the experimental conditions onpersuasive strategies, both repeated mea-sures and individual-item analyses ofvariance on the various measures wereemployed. Results were very similar, butare difficult to discuss with the repeatedmeasures analyses because there weretwo independent variables and severalmeasures of strategy. Therefore the re-sults of the separate 2 x 3 analyses foreach dependent measure are presentedbelow. As in the first experiment, therewere no significant differences in the pre-rated strength or specificity of the argu-ments used among conditions. Neitherwere there differences among conditionsin planned time to give the completedspeech. Again, the only consistent differ-ence in strategy was in the type of argu-ment chosen. Table 2 indicates the fre-quency of problem centered and solution

TABLE 2

AVERAGE FREQUENCY OF PROBLEM CENTERED ARGUMENTSAND SOLUTION CENTERED ARGUMENTS IN EXPERIMENT II.

iType ofArgument

DetailedProblemCentered

GeneralProblemCentered

2TotalProblemCentered

DetailedSolutionCentered

GeneralSolutionCentered

2TotalSolutionCentered

AudiencePosition

Favorable

Opposed

Favorable

Opposed

Favorable

Opposed

Favorable

Opposed

Favorable

Opposed

Favorable

Opposed

Experimental Conditions

No Feedback

1.83

2-00

1.28

1.00

4.00

4.00

2.61

2.85

2.06

2.15

5.28

5.23

Expected Feedback

ImpersonalFeedback

2.78

2.31

1.28

.69

4.89

4.00

2.28

2.62

1.56

1.62

4.33

4.92

PersonalFeedback

2.11

2.54

1.05

1.54

4.05

4.92

3.00

2.62

1.74

1.62

5.11

4.46

1 Does not include counterarguments for which there were no differences in usage among con-ditions.2 Includes detailed and general arguments, plus statement of conclusion.

64 SPEECH MONOGRAPHS

centered arguments, as well as those con-taining only a general statement andthose including detailed examples or pro-posals.

The data in Table 2 show that sub-jects who were led to expect feedbackused more detailed, problem centered ar-guments than those who did not expectfeedback (F=3.16, df = 2/88, p <.O5).Those in the Feedback conditions alsoused fewer general, solution centered ar-guments (F = 2.90, # = 2/88, p<.07).There was no main effect for audienceopposition, but problem centered argu-ments, overall, were chosen most bythose subjects in the Personal Feedbackcondition who expected an opposed audi-ence. Subjects in the Impersonal Feed-back condition expecting an opposedaudience did not choose problem cen-tered arguments highly (interactionF = 3.56, df = 2/88, p <.O5). While notsignificant, the pattern was reversed forsolution centered arguments.

Discussion

In Experiment II, anticipated audi-ence feedback had somewhat similar ef-fects to audience opposition in the firstexperiment: when feedback was ex-pected, the subjects were more likely todesign persuasive communications thatfocused upon issues and problems thanupon solutions. These data support ourhypothesis that the latter strategy maybe chosen in order to avoid audiencerejection. That the highest use of prob-lem centered arguments was in the Per-sonal Feedback-Audience Opposed con-dition also supports the hypothesis.

Somewhat confusing the interpreta-tion is the relatively low use of problemcentered arguments by those in the Im-personal Feedback condition who ex-pected an opposed audience. These sub-jects used fewest total arguments ofeither type (8.92 vs. an average of 9.25in the other conditions), while at the

same time using more counterarguments.We conclude that these subjects weretrying another strategy which our rela-tively open-ended measurement proced-ure failed to pinpoint. Their pattern isnot entirely inconsistent with the "avoid-ance of rejection" hypothesis, however,since using counterarguments might bean alternative procedure for preventingrejection.

EXPERIMENT III

This experiment was designed for thepurpose of again assessing the effect offeedback on persuasive strategy; the ef-fects of audience opposition and persua-sion difficulty were tested separately.The same procedural paradigm was usedas in the previous experiments (choosingarguments to construct a speech), butthe setting and design was somewhat dif-ferent.

Method

The design might be most succinctlydescribed as two overlapping factorialexperiments. There were three levels ofaudience feedback as in Experiment II(No Feedback, Impersonal Feedback,Personal Feedback) crossed with twolevels of persuasion difficulty (Easy toPersuade vs. Difficult to Persuade). Bothaudiences in the latter condition weredescribed as opposed to the communi-cator's position. Forming a design "leg"were four other conditions where thesubjects also expected no feedback.These were designed to assess the effectof audience tractability when the audi-ence was described as moderately infavor of the communicator's position,and to assess the effect of telling thecommunicator to simply inform, but notpersuade, the audience. Because the totaldesign, depicted in Figure 1, was un-balanced, strategy differences were as-sessed using two types of analyses: an-

COMMUNICATION STRATEGIES 65

alyses of variance of the effect of feed-back, with two levels of audience diffi-culty, and analyses of variance over allNo Feedback conditions of the effect ofdifficulty, and audience opposition andcommunicator intent.

Subjects. Students enrolled in 16 sec-tions of the introductory interpersonalcommunication course at the Universityof Kansas participated in the experi-ment as part of a class exercise. Onemonth prior to the experiment, a pre-test of attitudes toward Federal supportof population control was administeredas part of a general test of student atti-tudes unassociated with the research.Only students whose scores indicated at-titudes in favor of Federal interventionwere included in the analyses of the ex-perimental data. After students withnegative attitudes, students who werenot present at one of the sessions, andthose failing to complete the instrumentswere excluded, 206 subjects were left.The subjects were randomly assigned toexperimental condition within each sec-tion.

Procedure. In each section an experi-menter explained that the class exercisewas also part of a research project onpopulation control. Written instructions(which also included all the manipula-tions), a set of 25 Vf x 5" cards, ananswer sheet, and a postquestionnaire

were distributed to each subject. On thecards were printed the same argumentsfavoring Federal support of populationcontrol as in Experiment II. The speechconstruction task was identical to thatdescribed in Experiment II except thatthe total number of arguments was notrestricted.

The subjects read that part of theresearch program involved preparingtelevision speeches concerning popula-tion control. All but the subjects in theInform conditions were told that thespeeches would persuade the public thatthe Federal government should supportPopulation control. In the Inform condi-tions, all mentions of persuasion weredeleted (the intent of the speeches wasto inform the public about Federal sup-port of population control). The sub-jects in the Persuade conditions wereasked to help construct speeches forpossible television use. To enhance thecredibility of this request the subjectswere also asked to sign a "release form"permitting use of their speech outlinesin TV messages. It was then explainedto all of the subjects that televisionmessages are often directed at specificsub-audiences. The speeches in thepresent case were described as beingconstructed for the subgroup of womenwho were either very opposed to Fed-

EXPERIMENTAL CONDITIONS

CommunicatorInstructedTo:

Persuade

Persuade

Inform

AudiencePosition

Opposed

Favorable

Opposed

Difficultyof ChangingAudience

Easy

Difficult

Easy

Difficult

Easy

Difficult

None

Expected Audience Feedback

Impersonal

XX

XX

XX

XX

Personal

XX

XX

XX

XX

Figure 1. Experimental design o£ Experiment III.

66 SPEECH MONOGRAPHS

eral intervention or moderately in favorof intervention. To manipulate persua-sion (informing) difficulty, the audiencein either case was described, just as inExperiment II, as difficult or easy topersuade (inform).

Feedback was manipulated in the fol-lowing manner. In the No Feedbackcondition the subjects were told abouta second step of the research programin which panels of local women (op-posed or favorable to Federal interven-tion as in the TV audience) would eval-uate the speech outlines they had com-posed. Further they were told, "Sincethese outlines are submitted anony-mously, you will not receive feedback.However, the women will write downtheir reactions to the various outlinesand turn over their results to the re-search staff. On the basis of these re-sults, television messages will then becompleted using the best outlines." Inthe feedback conditions the subjectswere told that the outlines would not beanonymous and that the women wouldwrite their evaluations of the speechoutlines on two forms, one of whichwould be returned to the student. In thePersonal Feedback condition, the sub-jects were informed that 5 of the womenwould visit their class to meet and dis-cuss their speech outlines with them.The subjects were debriefed in the finalweek of classes.

R esults

Effectiveness of the Manipulations.For purpose of brevity, the 3 x 2 an-alyses of feedback and persuasion diffi-culty (for separate measures) will bedescribed as the "Feedback" analysis,whereas the analyses within all No Feed-back conditions will be referred to as"No Feedback" analyses. The Feedback

analyses revealed a significant effect ofthe feedback manipulation upon thesubjects' perception of receiving feed-

back in the direction expected (F = 3.19,# = 2/118, p<.OS), and in addition,in the Personal Feedback condition sub-jects were more likely to expect personalinteraction with a segment of their audi-ence (F = 13.04, df =2/111, p<.01).No such differences were found withinthe No Feedback conditions.

The perceived degree of difficulty ofpersuading the audience was significantlydifferent in the predicted direction also.The Feedback analysis produced a sig-nificant effect at the .01 level (F = 51.00,# = 1 / 1 1 8 ) , as did the No Feedbackanalysis (F = 30.59, # = 1 / 1 1 4 ) . A NoFeedback analysis of perceived audi-ence opposition indicated that the ma-nipulation was successful (F = 49.21,df = 2/114, p <.01). However, there wasalso a tendency, as seen in the Feedbackanalysis for audiences described as diffi-cult to persuade to also be perceived asmore opposed to the communicator'sposition (F = 5.92, # = 1 / 1 1 8 , p <.05).

In contrast to the other manipulations,the persuade/inform manipulation wasnot very successful. Subjects in the in-form conditions were less likely to ex-pect persuasion in their audience, butnot significantly so. There were no dif-ferences in pretest attitudes, attitudechange, or perceived effectiveness ofspeech among conditions.

Effect of Persuasive Intent and Audi-ence Opposition. There were few effectson strategy of the Persuade/Inform ma-nipulation, as would be expected fromthe measured weakness of that manip-ulation. However, in general the sub-jects in the Inform-Opposed conditionsacted much like those in the Persuade-Favorable conditions. That is, the sub-jects in these conditions reacted to audi-ence difficulty in a manner opposite tothe subjects in all other conditions. In-stead of apparently "trying harder"when the audience was depicted as dif-ficult to persuade, they tried less hard,

COMMUNICATION STRATEGIES G7

and vice versa for easy to persuade audi-ences. This statement is reflected inseveral interactions found in No Feed-back analyses: Subjects in the Inform-Opposed condition and in the Persuade-Favorable condition, (as opposed to thePersuade-Opposed condition), estimatedthat with a difficult audience theywould spend less time on their speech(F = 6.36, df = 2/lU, p<.01), wereless likely to use a solution centered con-clusion (F = 320, df = 2/114, p<J05),and chose weaker specific problem cen-tered arguments (F = 2.27, dj — 2/114,p<A2).

Effect of Persuasion Difficulty. Thesubjects who expected a difficult to per-suade audience, except those in the In-form and Audience Favorable conditionsdescribed above, tended to choose astrategy which appears as if they weretrying harder than subjects expectingan easy to persuade audience. Thus,the No Feedback analysis of strengthscores indicated that subjects in thedifficult conditions' used significantlystronger arguments than did thosein the Easy conditions (F = 4.99,# = 1 / 1 1 5 , p<.05). In general, thesame effect occurred across the feedbackconditions, but was mostly reflected inthe strength of solution centered argu-ments (F = 5.ZS, dj =1/121, p<.OS).Feedback analysis of expected time togive the speech also showed a non-sig-nificant tendency for those expecting adifficult to persuade audience to spendmore time. Comparing these data tothose from Experiment II is somewhatdifficult since audience opposition waspurposely confounded with persuasiondifficulty in the experiment, and acci-dently confounded in the present experi-ment. However the subjects did not per-ceive differences in persuasion difficultyin Experiment II, and there were nomain effects for audience opposition,whereas in the present experiment, where

perceived differences in audience diffi-culty were great, there were some effects.Our tentative conclusion is that "tryingharder" results from a perception thatthe audience is difficult to persuade,rather than from perceived opposition,per se.

Effect of Audience Feedback. In con-trast to the previous study, the effectof feedback was quite weak. However,the trends were in precisely the samedirection as in Experiment II. There wasan interaction of feedback with audi-ence difficulty, similar to the feedback-opposition interaction in Experiment II.That is those in the Personal Feedbackcondition expecting to confront a diffi-cult to persuade audience planned lesstime for their speech than those expect-ing an easy to persuade audience, where-as the reverse occurred in the No Feed-back and Impersonal Feedback condi-tions OF = 4.10, # = 2/121, p<.QS).

Discussion

The - experiments described in thispaper differ from most other experimentspublished in the communication litera-ture in that the subjects, instead of beingrestricted to a single range of responses,could have responded in many ways tothe manipulations. For example, in mostattitude change studies, the subjects canchange their measured attitudes but al-ternative ways of reacting, such as der-ogation of the communicator, leavingthe situation, changing the behavior ofthe communicator, or misperceiving thecommunication, are closed. In our at-tempt to measure strategies of persua-sion, however, we made it possible forthe subjects to alter their behavior inmany ways. It is clear that subjects didchoose various ways of constructingspeeches, and it is this variability thatprobably accounts for the difficulty intesting the hypotheses. Upon reflection,it appears that in order to test the

68 SPEECH MONOGRAPHS

"avoidance of rejection," "need toarouse" or "trying" hypotheses morefully, it might be preferable to arbi-trarily choose one message dimension ata time along which subjects might differ.

Still, some effects occurred with regu-larity despite the high variability. Thedata from all three experiments are quiteconsistent with the hypothesis that mes-sage strategies are planned with possibleaudience rejection in mind, particularlyif that rejection may have to be receivedby personal feedback ("avoiding rejec-tion" hypothesis). Avoidance of statingparticular solutions while instead dis-cussing problems and issues was highestunder conditions where an opposed audi-ence was most likely to personally de-liver negative feedback (or when thesubjects role played such a possibility,as in Experiment I) . We would predictthat under conditions where only onestrategic variation is possible, communi-cators facing an audience which may re-ject them will choose the option leastoffensive to that audience.

There were no data to support thehypotheses that problems and issuesare focused on in order to arouse orprepare an opposed audience ("need to

arouse" hypothesis). Neither were theredata consistent with the hypothesis thatcommunicators "give up" when facedwith an opposed audience ("trying"hypothesis). Instead, the third experi-ment produced some data which pointto the opposite conclusion, i.e., that adifficult audience causes the communi-cator to try even harder. These datapose an interesting predictive problem.If communicators trying to persuade anaudience are both afraid of rejection andmotivated to change their audience, howwill they resolve the apparent conflict?The ideal technique might be to usethe strongest possible arguments withinthe audience's "range of acceptance" asdescribed by Sherif and Hovland.11 An-other technique would be to maximizethe communicator's credibility by, forexample, listing his credentials. Wewould predict the latter to occur withgreater frequency, the greater the per-ceived probability of rejection by theaudience. These speculations, of course,are testable in future research.

11 Muzafer Sherif and Carl I. Hovland, SocialJudgment: Assimilation and Contrast Effects inCommunication and Attitude Change (NewHaven: Yale Univ. Press, 1961), pp. 127-145.