Thesis far the Degree of M. A mam; smz UWERSEW WAHER Pi. Cfifi‘éBS 1%8 RRY WWW"!WIN JIJJIIWIII'IHH W 310615 2410 THESIS ‘Wrm‘ hug—1:2: ...:‘ ;: ..' ‘lr—v: v—' w z ‘57 w I F f!" g- 2 a. 'I . “.3? E 1.1} U}, “yrs . _O. 3. *2! 6‘;- :‘ I r... (1.. I A... m u =. .‘v;!)_-‘. " u ‘. - -' "V$ : . . "1‘ .'a \UK ‘v‘d uh." WETLA“" " “QTFHLXJJ .-t,_ ”S; '5" ‘ . A“. I 51:" ‘ 5.'; ‘ ' .‘ . E19 431 .5). .‘y ~‘-".' 3.”. ..;.- , “'11:?- ‘ V 11 . ; nwfimmszfmffih" . .mmy MSU RETURNING MATERIALS: Place in book drop to LIBRARIES remove this checkout from “ your record. FINES wil] be charged if book is returned after the date stamped below. 5;: '\ . "3‘“...I/ ‘ L331! //'Q/"{ ABSTRACT THE EFFECT ON AUDIENCE FEEDBACK ON ENCODING BEHAVIOR OF SPEAKERS by Walter H. Combs Two eXperiments were conducted to study the effects of audience feedback on encoding behavior of public Speakers. The explanation of the findings reported by Amato and Ostermeier (1967) and the procedures they employed formed the basis for the rationale of this thesis. Amato and Ostermeier's procedures included using an audience to administer favorable, neutral, and unfavorable feedback to three Speakers. Audience members were instructed to indicate their responses by holding up appropriate colored cards during each speaker's presentation. At the conclusion of each Speech, the audience evaluated the Speaker on six Speaking characteristics. Results were reported to be supportive of the authors' hypothesis: that unfavorable audience feedback produces a deterioration in Speaker delivery. The purpose of Study I was to demonstrate that there is a more parsimonious explanation of Amato and Ostermeier's findings; namely, that the audience members rated the Speakers consistent with the response- role they were asked to perform. It was hypothesized in Study I that subjects administering positive feedback to a Speaker will rate him more favorably than subjects administering neutral feedback; and conversely, that subjects administering negative feedback to a speaker will rate him more unfavorably than subjects administering neutral feedback. Walter H. Combs Eighty-two students enrolled in classes of public Speaking courses at Michigan State University were randomly assigned to one of three response-role conditions: Favorable, Neutral or Unfavorable. In each class every member of the audience was given a set of IBM cards to indicate their response-role. GREEN cards indicated a favorable response; RED-- an unfavorable response; and WHITE--a neutral one. Approximately one-third of the audience members in each class was assigned to each feedback condition. During the delivery of a 10 minute Speech given by a confederate, the audience held up their cards once eaCh minute for approximately 10 seconds. Immediately following the Speech the audience was asked to rate the speaker along four evaluation dimensions: Content and Analysis, Delivery, Language and Overall Effectiveness. The results of Study I indicated that Amato and Ostermeier's findings lend themselves to a response-role, rather than a Speaking performance interpretation. The data suggested that the response-role instructions influenced the audience's evaluations of the Speaker. Subjects administering positive feedback rated the Speaker Significantly higher than those in the Neutral condition. Subjects administering negative feedback rated the speaker significantly lower than those in the Neutral condition. The purpose of Study II was to replicate the Amato and Ostermeier study using more defensible procedures. Study II differed from Study I in two ways: 1) speakers were used as subjects; and 2) encoding rate and nonfluencies were used as dependent variables. Encoding rate was defined Walter H. Combs as the total number of utterances per minute. Nonfluencies were operationally defined as the total number of repetitions, vocalized pauses, or revisions which interfered with the fluency of utterance. Sixty-one students enrolled in an introductory course in communi- cation at Michigan State University were randomly assigned to one of three feedback conditions: Favorable, Neutral, or Unfavorable. Subjects reported individually to a preparation room where they received in- structions to prepare a 10 minute Speech on: "What factors Should be considered in the choice of a college or university?" At the end of 25 minutes, the experimenter escorted the subject to another room where he delivered his speech to three male and three female confederates. The confederates administered favorable, neutral or unfavorable feedback by holding up GREEN, WHITE, and RED colored cards resPectively. No feedback was administered during the first minute of each speech. At the end of the first minute and every 20 seconds thereafter, the confederates held up their cards for 10 seconds. This feedback procedure was repeated fifteen times, three times a minute for a total of 5 minutes for each subject. Following the Speech, each subject filled out a questionnaire concerning their reactions to the audience and experimental task. In general, the results were supportive of the two major hypotheses of Study II. Subjects receiving favorable feedback exhibited somewhat higher utterance rates and significantly fewer nonfluencies than subjects Walter H . Combs receiving neutral feedback. Conversely, subjects receiving unfavorable feedback exhibited somewhat lower utterance rates and significantly greater nonfluencies than subjects receiving neutral feedback. THE EFFECT OF AUDIENCE FEEDBACK ON ENCODING BEHAVIOR OF SPEAKERS by Walter H . Combs A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1968 Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree. girector of Thesis ACKNOWLEDGMENTS The author wishes to exPress his gratitude to his advisor, Dr. Gerald Miller for his guidance, support and assistance in the preparation of this thesis. The author also wishes to thank Dr. Erwin Bettinghaus and Dr. Larry Sarbaugh for their recommendations and assistance as members of my guidance committee. Several individuals helped in the collection of the data: Mr. Clyde Morris, Mr. Jack Baseheart, Mr. Bill McEwen, Miss Ann Christiansen, Mrs. Betty Blackburn and Mr. Charles Berger. Their cooperation and time was most appreciated. Dr. Brad Lashbrook, Mr. Martin Hunt and Miss Brenda Dervin deserve Special thanks for their assistance in obtaining subjects. The author is indebted to Mrs. Shirley Sherman for her assistance in the typing and final preparation of this thesis. Finally, no words can eXpress the appropriate amount of gratitude to my wife for her nmderstanding, assistance and encouragement through- out my graduate studies. ii TABLE OF CONTENTS CHAPTER Page I STUDY I O O O O O O 0 O O O O O O O O O O O 1 Introduction . . . . . . . . . . . . . . Method . . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . l2 (DWI-4 II STUDY II C O O O O O O O O O O O O O O O 0 1” Introduction . . . . . . . . . . . . . . 1n Method . . . . . . . . . . . . . . . . . 17 Results . . . . . . . . . . . . . . . . . 23 Discussion . . . . . . . . . . . . . . . no BIBLIOGMPHY O O O O O O O O O O O O O O O O O O 0 II"'7 iii Table 10. 11. LIST OF TABLES Pag I Mean ratings and standard deviations for subjects in Favorable, Neutral, and Unfavorable response- mle conditions I O O O O O O O O O O O O O O O O O O O O 8 Smnmary table of simple analysis of variance of Delivery ratings for subjects in Favorable, Neutral, and Unfavorable reSponse-role conditions . . . 9 Summary table of simple analysis of variance of Content and Analysis ratings for subjects in Favorable, Neutral, and Unfavorable response- role conditions . . . . . . . . . . . . . . . . . . . . 9 Summary table of simple analysis of variance of Language ratings for subjects in Favorable, Neutral, and Unfavorable reSponse-role conditions . . . 10 Summary table of simple analysis of variance of Overall effectiveness ratings for subjects in Favorable, Neutral, and Unfavorable response—role conditions.......................10 Summary of four Newman-Keuls Tests for the Significance of differences in ratings of Delivery, Content and Analysis, Language, and Overall Effectiveness for subjects in Favorable, Neutral, and Unfavorable response-role conditions. . . . . . . . . . . . . . . . 11 Total utterance means and standard deviations by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions . . . . . . . . . . 2t} Analysis of variance of total utterance for subjects in the Favorable, Neutral, and Unfavorable feedback cmditims O O O O O O O O O O O O O O I O O O O O O O O 26 Means and standard deviations of nonfluency-fluency ratios by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions . . . . . . . . . . 29 Summary of Newman-Keuls Tests of all possible mean comparisons of nonfluency-fluency ratios for subjects in Favorable, Neutral, and Unfavorable feedback conditions 31 Analysis of variance of nonfluency to fluency ratios for subjects in the Favorable, Neutral, and Unfavorable feed- back conditims O O O O O O O O O O O O O O O O O 0 O 0 32 iv LIST OF TABLES CONTINUED Table Page 12. Means and standard deviations of reSponses of subjects in Favorable, Neutral, and Unfavorable feedback conditions to Questionnaire Items used in Study II . . . . . . . . . 35 13. Simple analysis of variance of reSponses to Questionnaire Item 5: ("What was your personal reaction to the audience?") fer subjects in the Favorable, Neutral, and Unfavorable feedback conditions . . . . . . . . . . . . 37 1a. Pre and post mean ratings of responses of subjects in Favorable, Neutral, and Unfavorable feedback conditions to Questionnaire Item 3: ("How well do you think (feel) you.will do (did) on this Speaking task?") . . . . . . . 39 LIST OF FIGURES Figure Page 1. Arrangement of subject, assistant, and confederates in the testing room . . . . . . . . . 20 2. Mean utterance by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions . . . . . . . . . 27 3. Mean nonfluency-fluency ratios by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions . . . 33 vi LIST OF APPENDICES Appendix Page A Copy of instructions for research participants in Study I O O O O O O O O O O O O O O O O O O O O O O O 0 “9 B Copy of manuscript of confederate's speech used in Study I O O O O O O O O O O O O O O O O O O O O O O 51 C Copy of questionnaire and method of scoring used in Study I O O C C C O O O O O O O O O C O O O O O O O O 57 D Copy of preparation room instructions for research partiCipantS in StUdy II 0 O O O O O O O O O O O O O O 60 E Copy of preparation sheet for research participants in Study II 0 O O O O C O O O O O O O O O O O O O O O 63 F Copy of speech delivery instructions for research participantsinStudyII . . . . . . . . . . . . . . . 65 G Copy of questionnaire and method of scoring used in Study II 0 O O O O O O O O O O O O O O O O O O O O 67 H Copy of sign-out sheet for research participants in Study II 0 O O O O O I O O O O O O O O O O O O O O O O 70 vii CHAPTER I STUDY I Introduction Feedback is an important element in interpersonal communication. Given at least one source and one receiver, feedback may be defined as: "those overt reSponses of a listener that serve to shape and to modify the succeeding communication behavior of a speaker," (Miller, 1966). The effects of feedback have been studied in a variety of settings. Leavitt and Mueller (1951) found that mean accuracy scores were higher when an instructor had orally communicated a series of geometric forms to subjects who were allowed free feedback and lower when subjects were allowed zero feedback. Accuracy scores for intermediate conditions increased as feedback increased. Verplank (1955) studied the effects of positive and negative feedback in situations involving conversation between two persons. In his study one person reacted to personal opinion statements of another by either agreement or disagreement, paraphrasing or remaining silent. Personal opinion statements from subjects in the agreement and paraphrasing conditions were significantly higher than in the conditions of disagreement and silence. Data from another experiment suggest that verbal approval of a prior Speaker produces disruptive effects on certain speech patterns of a second Speaker (Miller, Zavos, Vlandis and Rosenbatm, 1961). Subjects in a condition where the second Speaker received no verbal approval and in a condition where the second Speaker received verbal approval had significantly higher nonfluencies than did subjects in a condition where the prior Speaker had not received verbal approval. Stolz and Tannenbaum (1963) studied the effects of knowledge of performance on encoding behavior in an oral examination setting. Subjects were presented with results of their answers to an item before proceeding to the following item. Positive feedback produced little difference in the number of words per minute, while negative feedback produced a significantly lower mean value. Subjects receiving negative feedback also had significantly higher ratios of unfilled pauses to number of words than did subjects in a control group. Similarly, Miller (1964) demonstrated that Speakers receiving the same kind of reSpcnses as their predecessors had fewer nonfluencies and higher utterance rates than did subjects receiving feedback that was more or less favorable than their predecessors. More recently, Amato and Ostermeier (1967) have reported results concerning the effect of audience feedback on beginning public speakers. The explanation of their findings and the procedures they employed form the basis for the rationale of this thesis. Specifically, the purpose of this thesis is twofold: l) to demonstrate that there is a more parsimonious explanation of the findings reported by Amato and Ostenneier; and 2) to replicate the Amato-Ostermeier study using more defensible procedures. A detailed description of their study will reveal the weakness of central concern. These investigators hypothesized that unfavorable audience feed- back produces a deterioration in Speaker delivery; Specifically eye contact, nervousness, bodily movement and fluency. Twenty-two students enrolled in a fundamentals public speaking course formed the audience. Three other students, unknown to the audience, served as Speakers. Each audience member was given three a" x 6" index cards to use in presenting feedback during the oral presentation of each Speaker. The audience was instructed to assume a predetermined response-role for each speaker. In one condition the audience was instructed to reSpond favorably to the speaker by holding up a.WHITB index card. In the other conditions they were instructed to indicate an unfavorable or neutral reSponse by holding up either a RED or GREEN card respectively. These audience response-role procedures were fellowed for each of the three speakers during the first speaking session. The speakers were given the same instructions concerning the meaning of the colored cards. At the conclusion of each Speech, the audience filled out a questionnaire in which they were asked to evaluate the Speaker on six Speaking characteristics: bodily movement, eye contact, facial ex- pression, fluency, friendliness, and nervousness. A seven interval "good- ba." scale was used to measure each of the Speaking characteristics. A second set of instructions was given to the audience prior to the second round of speeches. In this set only two reSponse-roles were perudtted, either unfavorable or neutral but ng£_favorable. Again the questionnaires were administered following each speech. Results of a three dimensional analysis of variance indicated a significant between treatments effect on four of the six delivery characteristics measured: eye contact, nervousness, bodily movement, and fluency. The authors concluded that negative feedback not only leads to a disruption of encoding behavior but also produces a deterioration inknnm. Amato and Ostermeier‘s results are open to an alternative inter- pretation. Assume that an individual is asked to respond "unfavorably" to a Speaker's presentation. After the presentation, the individual fills out a questionnaire which requires him to evaluate the Speaker on certain delivery characteristics. Obviously, the act of reSponding unfavorably might influence that individual's perception of the Speaker. Also, repeated "unfavorable" responses from an individual while attending to the speech should prediSpose that individual to rate that speaker more unfavorably than if he had been instructed to respond in a "neutral" manner. Moreover, when a subject is instructed by an experimenter to repeatedly reSpond "favorably" to a source's presentation, it could be expected that the subject would evaluate the source's presentation more favorably than if he had been instructed to assume a "neutral" reSponse- rode. Such an expectation is consistent with Orne's (1962) researdh ' concerning subjects' perceptions of the demand characteristics of the psychological exPeriment. These considerations cast doubt on Amato and Ostermeier's inter- pretation of their results. They attribute their obtained differences in audience ratings of the Speakers to the effects of the feedback on the Speaker's delivery. Their differences can just as well be attributed to the nature of the procedures employed; that is, the instructions given to the audience prior to Obtaining their ratings of the speaker. Also, it would seem that the repeated act of assuming a particular response-role would strengthen the sensitizing effect of these in- structions. Therefore, a more parsimonious eXplanation of the Amato and Ostermeier results would be that the response-role instructions in— fluenced the audiences' rating of the speakers: that audience members rated the speakers consistent with the reSponse-role they were asked to perform. If the feedback manipulations had an effect on the Speakers' delivery, this effect was not measured unequivocally because the same individuals who administered feedback also rated the speakers' delivery. Study I To implement the initial purpose of this thesis, an experiment was conducted using a modification of the procedures employed by Amato and Ostermeier. It was hypothesized that subjects administering positive feedback to a Speaker will rate him more favorably than subjects administering neutral feedback; and conversely, that subjects administering negative feedback to a Speaker will rate him more unfavorably than subjects administering neutral feedback. Method Eighty-two students enrolled in four classes of public speaking courses at Michigan State University were randomly assigned to one of three reSponse-role conditions: Favorable, Neutral or Unfavorable. In each class every member of the audience was given a set of ten 3 1/2" x 7 3/8" IBM cards to indicate their particular reSponse-role. .This procedure resulted in a total of twenty-eight subjects being assigned to the Favorable condition and twenty-seven in both the Neutral and Unfavorable conditions. In each class approximately one-third of the audience was assigned to each feedback condition. This means that any differences in ratings had to result from differences in response—role assignments and not from differences in speaking performance. Before the Speaker was introduced, the experimenter furnished the audience with the following instructions: Students who have completed courses in public Speaking are aware of the responses a Speaker receives from his audience. In this particular Speaking situation we are asking you to assume the role of an individual audience member and provide the Speaker with non-verbal feedback according to an assigned role... You have before you a set of 10 cards. These cards will assist you in performing your response-role. Each card represents a different kind of reSponse; GREEN-- a favorable reSponse; RED-- an unfavorable response; and WHITE" a neutral one. These cards are arranged in the order in which you are to present them to the speaker as he is delivering his Speech. The speaker is aware of the kinds of feedback you will be administering . Your particular reSponse role is (favorable, unfavorable, neutral); therefore all the cards you have are (GREEN, RED, WHITE). The speech you will hear is 10 minutes long. At the end of the first minute you will hold up the first card in your deck for about 10 seconds. At the end of the second minute, the second card; at the end of the third minute, the third; and continue with the fourth, fifth, sixth, etc. until the Speech is completed...1 The Speaker, a male confederate of the eXperimenter, was introduced in each class as an undergraduate volunteer taking a "special problems" course in the Department of Communication. He was unknown to the audience prior to the completion of his Speech. The Speech was prepared by the eXperimenter in collaboration with the Speaker and delivered from a typed manuscript.2 During its delivery, the speaker maintained eye contact with the audience, since he had participated in several practice sessions prior to the presentation and was familiar with the content of the speech. Immediately following the Speech the audience was asked to react 'to the Speaker along four evaluation dimensions: Content and Analysis, Delivery, Language, and Overall Effectiveness. Each dimension was accompanied by a brief statement listing some of the criteria to be used in reacting to it. A seven interval rating scale appeared below each dimension.3 Prior research on commonly used Speech rating scales had indicated that the Content and Analysis, Delivery, and 1A copy of the instructions appears in Appendix A. 2A copy of the text appears in Appendix B. QA copy of the questionnaire appears in Appendix C. Language dimensions were somewhat independent (Becker, 1962). The Overall Effectiveness dimension was included to obtain general audience reactions to the total Speech. Results For all statistical tests the .05 level of significance was required. The mean ratings and standard deviations for each feedback condition within the four rating dimensions appear in Table 1. Table 1. Mean ratings and standard deviations for subjects in Favorable, Neutral and Unfavorable reSponse-role conditions; EF28’ 27, 27 reSpectively.* Rat in g Dimens ion Response Role Mean Standard Devi ati on DELIVERY Favorable “.57 1.07 Neutral 3.uu .75 Unfavorable 2 . 81 . 88 CONTENT AND Favorable 5.71 .53 ANALYSIS Neutral 5.33 .83 Unfavorable 4.30 1.23 IJANGUAGE Favorable 5.17 .9” Neutral n.03 .75 Unfavorable 3.55 l. 36 OVERALL Favorable 5 . ll . 31 EFFECTIVENESS Neutral ll . lS . 75 Unfavorable 3.07 l. 36 *7 = most favorable; l = least favorable; therefore, the higher the rating, the more favorable the reaction. Simple analyses of variance were performed on data for the four rated dimensions of speaking effectiveness. Significant F ratios were obtained from these analyses for the Delivery, Content and Analysis, Language , and Overall-Effectiveness dimensions (Tables 2 , 3 , I4, 5) . Table 2. Sunmary table of simple analysis of variance of Delivery ratings for subjects in Favorable, Neutral and Unfavorable response-role conditions . Source df 38 ms F Between Groups 2 #3.68 21.8“ 26.31* Within Groups 79 65.61 .83 Total 81 109.29 Table 3. Sumnary table of simple analysis of variance of Content and Analysis ratings for subjects in Favorable, Neutral and Unfavorable reSponse-role conditions. Source df 85 ms F Between Groups 2 29.n3 1n.71 l7.78* Within Groups 79 65.35 .82 Total 81 9u.78 *Critical value = 3.11, p<.05 10 Table I4. Summary table of simple analysis of variance of Language ratings for subjects in Favorable, Neutral and Unfavorable response-role conditions. Source df ss ms F Between Groups 2 38.36 19.18 17.27* Within Groups 79 87.74 1.11 Total 81 126.10 Table 5. Summary table of simple analysis of variance of Overall Effectiveness ratings for subjects in Favorable, Neutral and Unfavorable reSponse-role conditions. Source df ss ms F Between Groups 2 56.8% 28.112 35.11* Within Groups 79 63.9w .81 Total 81 120.78 —¥ *Critical value = 3.11, p (.05 Table 6. Summary of four New'man-Keuls Tests for the significance of differences in ratings of Delivery, Content and Analysis, Language, and Overall Effectiveness for subjects in Favorable, Neutral and Unfavorable reSponse-role conditions. DE LIVERY Favorab 1e Ne utral Un favorab 1e Favorable 35.00a 52.00a Neutral 17.00a CONTENT 8 ANALYS I S Favorable Ne utral Unfavorab 1e b b Favorable 16.00 un.oo Neutral 28.00b LANGUAGE Favorable Ne utral Un favorab 1e Favorable 36.00c u9.oo° Neutral 13 . 00 OVE RALL E FFECT IVENESS Favorab 1e Ne utral Unfavorab 1e Favorable 31.00d 60.00d Neutral 29.00d aSignificant at < .05. q for the .05 level for two steps is .13. 61+; for three steps apart, 16. 38. b . Significant at < .05. 18 13.35; for three steps apart, 16.01}. q for the .05 level for taro steps cSignificant at < .05. q for the .05 level for two steps is 15.1w; for three steps apart, 18. 59. d Significant at <. 05. q for the .05 level for two steps is J-3~'-F'7; and for three steps apart, 16.18. apart apart apart ap art 12 The Newman-Keuls technique for tests of all ordered pairs of meansl+ was used to investigate the nature of the differences obtained through analysis of variance. This procedure is most useful when attempting to draw inferences from data in which the order and magnitude of differences are being considered. The farther apart two means or treatment totals are on a scale, the larger the difference between them mint be before this difference exceeds its critical value. The obtained 1 values for all possible mean comparisons on each rating dimension appear in Table 6. All Favorable-Unfavorable, Neutral- Favorable and Neutral-Unfavorable mean comparisons were Significantly different on each of the four rating dimensions. These significant 3 values indicate that the direction of the differences obtained was consistent with what was predicted. Favorable mean ratings were significantly higher than neutral mean ratings and unfavorable mean ratings were significantly lower than neutral mean ratings. Discussion The results of Study I demonstrate that Amato and Ostermeier's findings lend themselves to a response-role, rather than a speaking Performance interpretation. The data suggest that the response-role inStruetions influenced the audience's evaluations of the Speaker. Consistently, the Favorable, Neutral and Unfavorable groups differed as Predicted. Subjects administering negative feedback rated the Speaker significantly lower than those in the Neutral condition. Subjects ‘_— u C O In: B.J. Winer, Statistical Principles in Emerimental Design. Ne" York: McGraw-Hill Company, 1962, pp. 80-85. 13 administering positive feedback rated the Speaker significantly higher than those in the Neutral condition. In only one case did the results Show a departure from this consistent rating pattern. Whether or not the feedback had an effect on the delivery of the speakers cannot be determined from the procedures employed by Amato and Ostermeier. A basic design modification would have to be made before the question of the effect of audience feedback on encoding behavior could be studied meaningfully. The manipulation of the independent variable of feedback must be separated from the subsequent dependent variable measures. One way to accomplish this would be to use the Speakers as subjects and actually record their delivery while under the influence of audience feedback. This design modification was incorporated into a second experiment which is reported in the following chapter. CHAPTER I I STUDY II To implement the second purpose of this thesis, Study II was conducted, using the modification in procedures suggested in Chapter I. Study II differed from Study I in mo major ways: Speakers were used as subjects rather than audience members; and measurements of the dependent variable, encoding behavior, were based on the Speakers' actual verbal behavior, not the audience's ratings of Speaker behaviors. Both of these design modifications were employed to maximize the effect of the feedback manipulation and to minimize the effects of extraneous variables on measurements of the dependent variable. These modifications enabled the present investigator to study the effects of audience feedback on encoding behavior while eliminating the contaminating factors of the Amato and Ostermeier study. an .1; I_ntroduction Results of studies concerned with the effects of audience feedback on oral encoding behavior (e.g., Miller, 1964; Stolz and Tannenbaum, 1963; and Miller 9; _a_l_., 1961) indicate that positive feedback facilitates encocling behavior; that is, positive feedback produces increases in the nulliber- of words per minute, decreases in the number of nonfluencies, and decreases in the ratio of unfilled pauses to number of words. Conversely, negative feedback inhibits encoding behavior; that is, it produces increases 1‘4 15 in the number of nonfluencies, increases in the ratio of unfilled pauses to number of words, and decreases in the number of words per minute. The H1: major hypotheses investigated in Study II were: Speakers receiving positive feedback from an audience will exhibit higher encoding rates than speakers receiving neutral feedback; and conversely, Speakers receiving negative feedback will exhibit lower encoding rates than Speakers receiving neutral feedback. Speakers receiving negative feedback from an audience will exhibit a greater number of nonfluencies than Speakers receiving neutral feedback; and conversely, Speakers receiving positive feedback will exhibit fewer nonfluencies than speakers receiving neutral feedback. The following predictions were made about subjects' reactions to the experimental situation as a function of the kind of feedback they received. H 3. Speakers receiving positive feedback from an audience will make Shorter estimates of their actual speaking time than speakers receiving neutral feedback, and Speakers receiving negative feedback will make longer estimates of their actual Speaking time than speakers receiving neutral feedback. Speakers receiving positive feedback from an audience will rate themselves as less nervous while delivering their Speech than Speakers receiving neutral feedback, and Speakers receiving negative feedback will rate themselves as more nervous while delivering their Speech than Speakers receiving neutral feedback. Speakers receiving positive feedback from an audience will rate the assistant recording their presentation more favorable than speakers receiving neutral feedback, and Speakers receiving negative feedback will rate the assistant recording their presentation more unfavorably than speakers receiving neutral feedback. 16 H5: Speakers receiving positive feedback from an audience will rate that audience more favorably than Speakers receiving neutral feedback, and speakers receiving negative feedback will rate that audience more unfavorably than Speakers receiving neutral feedback. H7: Speakers receiving positive feedback from an audience will rate the experimental task higher on enjgyment and satisfaction scales than speakers receiving neutral feedback, and Speakers receiving negative feedback will rate the experimental task lower on enjgflent and satisfaction scales than Speakers receiving neutral feedback. Of secondary importance to Study II was a hypothesis concerning how subjects would respond to a questionnaire item designed to elicit their attitude toward their performance on the experimental task. This task included preparing and delivering a 10 minute Speech before a group of confederates who administered favorable, neutral or unfavorable feedback to the Speaker. The experimental setting seemed appropriate to test a proposition concerning how individuals would reSpond to feed- back that was either consistent or inconsistent with their prior attitude toward performance on the experimental task. Balance or consistency theories (e.g., Festinger, 1957; Osgood and Tannenbaum, 1955) would predict maximum attitude change toward the BXPerimental task in conditions where subjects received feedback that was inconsistent with their prior attitudes concerning their Speaking abilities. Someflhat less attitude change would be predicted in conditions where feedback from an audience was consistent with their prior attitudes toward Performance on the experimental task. The hypothesis derived from the balance or consistency notions was: S... E? intrlin) 17 Ha: Speakers receiving feedback inconsistent with their prior attitudes toward their performance on the experimental task will change their attitudes concerning this performance more than speakers receiving feedback consistent with prior attitudes of Speaking performance. Independent Variables In this study, type of feedback was operationally determined by the color of cards presented by the audience members. Three types of nonverbal feedback were used: favorable, neutral and unfavorable. In order to present the feedback 3 1/2" x 7 3/8" IBM cards were used. A GREEN card represented a favorable reSponse; a WHITE card, a neutral response; and a RED card, an unfavorable response. Six paid confederates served as the audience, and administered the appropriate feedback treat- ment to Speakers during the delivery of a Speech. Digendent Variables Two asPects of encoding behavior of Speakers were measured; encoding rate and nonfluencies. Encoding rate was defined as the total number of utterances per minute. Nonfluencies were Operationally defined as the total number of repetitions, vocalized pauses, or revisions which interfered with the fluency of utterance. Hethod Subjects Thirty-seven male and twenty-four fernale Students enrolled in an intmductory course in communication at Michigan State University served as subjects. Subjects were randomly assigned to one of three feedback 18 Favorable, Neutral or Unfavorable. Each feedback condition conditions: contained subjects with approximately equal amounts of prior Speaking experience: (Favorable, n_ = 20: 5 with a minimum of one college course in public speaking, It with high school experience and 11 with none; 9 with one college course, 5 with high school experience Neutral, n_ = 21: and 7 with none; Unfavorable, _n_ = 20: 6 with one college course, u with high school experience and 10 with none). Procedure Subjects were asked to participate in an oral communication study No subject knew the nature as part of their regular course assignments. Subjects of the experiment prior to reporting to the testing situation. reported individually to a preparation room where they received initial instructions concerning the experimental task.1 The experimental task consisted of preparing a 10 minute Speech on the topic "What factors Should be considered in the choice of a college or university?" When subjects entered the preparation room, they were assigned a number for identification purposes and assured that they would not be given a performance grade on the experimental task. Each subject was furnished with scratch paper, a copy of the topic, and a list of suggestive points to consider in his outline. Subjects were given approximately twenty-five minutes to prepare their speeches for delivery. At the end of each subject's preparation 1Appendix D contains the complete preparation instructions. 2The preparation sheet appears in Appendix E. 19 time, the experimenter escorted the subject to another room where he subsequently delivered his Speech to three male and three female con- federates. Upon his arrival, the subject was greeted by an assistant who read the following instructions: The Speech you are about to deliver will be recorded in order to reduce the amount of time each Student must spend here this evening. A panel of judges will serve as your audience. They will listen to what you. have to say and provide you with some nonverbal feedback while you are delivering your Speech. As you are Speaking, the judges will hold up a GREEN card if they agree with a point you are making in your Speech; a RED card if they disagree; and a WHITE card if they neither agree or disagree with you. At the end of your alloted time, one of the judges will stand and ask you to stop. Remember, a GREEN card indicates a favorable reSponse, a RED card an unfavorable one, and a WHITE card a neutral response. Are there any questions before we begin? The subject, assistant, and confederates were arranged in the testing room as illustrated in Figure l. The subject stood directly behind the podium. To help the Subjects remember the meaning of the cards, a set of three cards was situated in the upper half of the podium with "agree", "neutral" and "disagree" labels appearing on the appropriate color cards. The confederates were located approximately 12 to 15 feet from the podium. They were seated in a semi-circle so that the Speaker could easily observe their reSponses. One of the confederates served as a timer for the Speeches. The assistant recorded the Speeches, using a monophonic tape recorder and a lavalier microphone. This recording equipment was located behind the Speaker. 3Appendix F contains these instructions. 2O Assistant ‘* Tape recorder Subject i. 7‘ Confederates Figure l. Arrangement of subject, assistant, and confederates in the testing room. Prior to the beginning of a testing session, the experimenter furnished the confederates with packets of fifteen green, white, and red IBM cards along with a list of the kinds of feedback each Speaker was to receive. In order to obtain baseline measures of rate and nonfluency, no feedback was administered during the first minute of each Speech. At the end of the first minute and approximately every 20 seconds thereafter, the confederates held up the appropriate colored card for approximately 10 seconds. This feedback administration procedure was repeated fifteen times, three times a minute for a total of 5 minutes for each subject. The timer cued the other confederates by holding up the appropriate card. At the end of 6 minutes and 20 seconds, the timer asked the speaker to stop. The assistant eXplained to the subject that his alloted 21 time was up and thanked him for his presentation.“ The subjects were then instructed to go to another room and fill out a questionnaire. This questionnaire contained items concerning reactions to the audience and experimental task and was designed to provide data relevant to Hypotheses 3 through 8.5 Finally, all subjects read and signed a statement requesting them not to disclose their participation in the experiment to anyone, since the success of the project depended upon students coming into the test situation without prior knowledge of the procedures.6 Total utterance and nonfluency measures Each subject's Speech was transcribed from the tapes to a typed manuscript. The total number of utterances was determined by counting the total verbal output of a subject from the manuscript. The total utterance included both fluencies and nonfluencies for £91111 minutes of delivery time. The four minute unit of analysis was chosen, Since this interval contained the maximum number of subjects speaking the maximum number of minutes for each of the feedback conditiars. Fifty-seven subjects met the four minute criterion, but two subjects in the Favorable feedback condition and one in each of the other two conditions did not Speak longer than 2 minutes each. These subjects were therefore eliminated from the analysis . l’The complete explanation appears in Appendix F. 5 scoring. Appendix G contains a copy of the questionnaire and method of 6 . . A copy of this statement appears in Appendix H. 22 Nonfluency totals were obtained from the manuscript in a similar manner. It was stipulated that a nonfluency included "any repetition, vocalized pause, orrevision which interfered with the fluency of utterance." Continuous sequences of nonfluencies were counted as one nonfluency, while a nonfluent sequence separated by a fluent sequence was counted as two nonfluencies. Therefore a phrase such as: "I uh... uh...uh...cou1d" was counted as one nonfluency. On the other hand, a phrase such as: "I uh...think that uh..." was counted as tw0 nonfluencies. In those instances where doubt existed, the taped Speeches were used in conjunction with the typed manuscripts to determine the total number of nonfluencies. The accuracy of the total utterance and nonfluency counts was dependent upon the accuracy of the transcription of the taped speeches to manuscript form. Three typists, working independently, transcribed a random sample of ten Speeches. A total of 8,679 utterances were noted in the ten Speeches. The manuscripts used in the experiment deviated from the manuscripts of the other typists in nine instances. In those instances where deviations occurred, the experimental manuscripts differed from the others by no more than four utterances in one case and no more than one utterance per Speech in the other eight. Within the 120 portions of the manuscripts containing nonfluencies, the experimental manuscripts never differed from the manuscripts of the other typists. Use of Ebel's method for estimating reliability7 resulted in individual coefficients of greater than .99 for both total utterance and 7Robert L. Ebel "Estimation of the Reliability of Ratings," figychometrika, XVI, 1951,‘pp. “07-92% 23 nonfluencies. Therefore, the exPerimenter felt confident of the accuracy of the manuscripts used to obtain measures of the dependent variables. Reliability checks were also conducted on the experimenter's count of total nonfluencies. To determine the total nonfluencies, separate counts were made for total repetitions and total vocalized pauses from the manuscripts. Three judges, working independently, counted the number of repetitions and vocalized pauses from a random sample of ten manuscripts. Application of Ebel's method for estimating reliability resulted in individual reliability coefficients of greater than .97 and .99 for repetitions and vocalized pauses reSpectively. Results For all statistical tests, the .05 level of significance was required. Analysis of the dependent variables yielded the following results. Total Utterance Total utterance was broken down into twelve 20 second intervals for four’minutes of delivery time. The means and standard deviations for subjects in the Favorable, Neutral, and Unfavorable conditions appear in Table '7. The first three 20 second intervals provided baseline measures fer the total utterance, Since no feedback was administered to subjects during these intervals. A simple analysis of variance indicated that the groups' baSeline total utterance measures were not significantly different (P = 1.17, QEL= 2/54); thus, any subsequent differences between group 29 measures of total utterance would not be due to differences present before the administration of feedback. Table 7. Total utterance means and standard deviations by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions; n_ = 18, 20, 19 reSpectively. Interval Feedback Condition Favorable Neutral Unfavorable Mean §_D_ Mean _S_.__D_._ Mean S.D. 1 59.22 11.56 57.35 9.18 53.16 10.69 2 59.33 9.18 51.35 6.35 97.26 9.39 3 99.50 7.71 53.20 9.12 53.68 9.11 1-3 163.22 9.98 161.90 8.21 152.26 9.72 9 98.17 11.89 98.95 10.61 98.26 16.93 5 51.00 11.90. 52.05 11.96 93.37 12.15 6 99.83 7.03 98.85 10.09 95.58 10.88 7 59.17 11.89 98.90 11.22 97.31 11.15 8 52.39 9.92 99.95 9.77 95.16 19.10 9 50.28 11.89 95.50 15.31 95.58 19.08 10 53.33 12.73 51.85 11.16 99.89 9.61 11 50.50 11.19 97.90 9.36 97.00 11.79 12 99.89 19.97 99.65 9.93 92.10 11.05 9-12 960.11 10.50 938.10 10.28 916.31 10.93 25 Intervals 9 through '12 consist of measures obtained while subjects were being administered feedback. The means and standard deviations for subjects in the Favorable, Neutral, and Unfavorable conditions appear in Table 7. A Simple analysis of variance of the total utterance data for the entire eight feedback intervals was not significant (F = 1.92, _d_f_ = 2/59); however, the means differed in the predicted direction. A type I mixed analysis of variance8 was performed on the data from intervals 9 through 12 in an effort to account for more of the variability within groups. This particular analysis is useful when controlling for individual differences over a series of trials. Basically, the analysis consists of a mixture of sinnle-randomized and treatment-by- subjects analyses for each time interval. The total components of varimce are divided into tw0 types: those due to between-subjects variation and those due to within-subjects variation. The respective error terms are also separated. Table 8 indicates the results of this analysis. The F tests of the feedback and time intervals alone were not Significant. However, the feedback-by-time intervals interaction was Significant (F = 10.07, if = 16/932). This Significant interaction Suggests that both the kind of feedback and the £12: interval was important in determining differences in total utterance betwaen groups. 8E. F. Lindquist, Design and Analysis of Experiments in Psycholog and Education. Boston: Houghton Mifflin Corpany, 1953, pp. 266-273. 26 Table 8. Analysis of variance of total utterance for subjects in the Favorable, Neutral, and Unfavorable feedback conditions. Source df ss ms F Between Subjects (56) (27216.01) Feedback (A) 2 2116.56 1058.28 2.27 Error Between 59 25099.95 969.96 Within Subjects (956) (99553.17) Time Intervals (B) 8 1029.91 128.79 1.75 A x B 16 11816.79 738.55 10.07* Error Within 932 31696.52 73.37 Total 512 71769.18 *critical value = 1.69, p (.05 Figure 2 illustrates how both time intervals and feedback Operated ‘10 produce the significant interaction. In general, the mean total utterance for the Favorable grow increased with increases in time, while the mean total utterance for the Unfavorable grow decreased. Also, these differences appear to be consistent for the Favorable and Unfavorable EI‘Oups over time. During about one-half of the time intervals, the Neutral group's performance paralleled that of the Favorable grow. For the remaining time segments, the Neutral grow's mean total utterance “as Similar to the Unfavorable grow. The somewhat inconsistent performance Mean Utterance 27 55 5 3 . Favorable 52 9 I 51 50 I 99 ' \ Neutral 98 97 ‘ Unfavorable ,; / 96 95 an 93 92 9 5 6 7 8 9 10 11 12 Interval Figure 2. Mean utterance by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions. 28 of the Neutral group over time suggests a possible reason for not obtaining Significance on the Simple analysis of variance of total utterance. Therefore the results of the analyses of total utterance provide only qualified swport for Hypothesis 1: that Speakers receiving positive feedback from an audience will exhibit higher encoding rates than Speakers receiving neutral feedback, and that Speakers receiving negative feedback will exhibit lower encoding rates than speakers receiving neutral feedback. In this particular experiment, feedback in conjunction with time intervals was important in producing differences in the total utterance between groups. While the differences obtained in mean total utterance between groups were consistent with predictions, they were not significant. Suhjects receiving favorable feedback had somewhat higher total utterance Scores than subjects receiving neutral feedback, and subjects receiving unfavorable feedback had somewhat lower total utterance scores than SUbjectS receiving neutral feedback. Non fluencies The vocalized pause and repetition measures were combined to analyze the total number of nonfluencies. In order to control for the possibility that the nonfluencies might be due in part to the rate of Speaking, a ratio of total nonfluencies to fluencies was computed for each subject. The means and Standard deviations of these ratios for subjects in the Favorable, Neutral, and Unfavorable grows appear in Table 9. 29 Table 9. Means and standard deviations of nonfluency-fluency ratios by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions; n_ = 18, 20, 19 reSpectively. Interval Feedback Condition Favorable Neutral Unfavorable Mean S.D. Mean _S._D_._ Mean; _S_.1_)_._ .01 .01 .09 .01 .09 .02 .03 .02 .05 .03 .05 .03 a .05 .01 .06 .02 .06 .03 1-3 .10 .01 .19 .02 .12 .03 9 .05 .01 .06 .05 .06 .06 5 .02 .01 .05 .09 .10 .07 6 .03 .01 .06 .05 .09 .05 7 .03 .03 .06 .06 .10 .06 8 .02 .06 .06 .06 .09 .09 9 .09 .02 .09 .05 .09 .06 10 .02 .03 .05 .09 .06 .09 ll .02 .07 .09 .05 .07 .05 12 .03 .08 .05 .09 .08 .05 9-12 .25 .06 .98 .05 .78 .05 30 Intervals 1 through 3 contain measures of nonfluencies during the time in which no feedback was administered to subjects. These measures served as a baseline for comparison with subsequent measures of non- fluencies made during intervals 9 through 12, when feedback was being administered to subjects . A simple analysis of variance of the baseline data indicated that the nonfluency-fluency ratios did not differ significantly for the Favorable, Neutral, and Unfavorable groups (F = 1.92, d: = 2/59). Therefore, the experimenter assumed that any subsequent differences between grows could not be attributed to initial differences present before the administration of feedback. Intervals 9 through 12 consist of measures obtained while subjects were under the influence of feedback. Table 9 contains the means and Standard deviations for these nonfluency measures. A Simple analysis Of variance of the nonfluency data for the entire eight feedback intervals was significant (F = 9.57, _d_f_ = 2/59). The Newman-Keuls 9 13991191999 was applied to the data in order to determine whether HyPothesis 2 could be substantiated. The results of this analysis aPpear in Table 10. The obtained q values for all mean comparisons exceeded the critical value in every comparison, indicating that the direction of the differences obtained was consistent with predictions. 9 In: B. J. Winer, Statistical Principles in Experimental Design. New York: McGrai-Hill Company, 1962, pp. 80-85. 31 Table 10. Summary of Newman-Keuls Tests of all possible mean comparisons of nonfluency-fluency ratios for subjects in Favorable, Neutral, and Unfavorable feedback conditions. Favorable Neutral Unfavorable Favorable 5.1.0“ 10.140* Neutral 5 . 30* *Significant at < .05. 1 for the .05 level for “to steps apart is 1.614; for- three steps apart, 1.97. A Type I mixed analysis of variance was again used to break the data into 20 second intervals. The results of this analysis appear in Table 11. F tests for the feedback and time effects were significant. The feedback-by-time intervals interaction was also significant. Figure 3 illustrates the pattern of these differences in mean nonfluency-fluency ratios during the time when subjects were under the influence of feedback. Notice that the pattern of differences is more consistent and clearer than was the pattern for total utterance measures. I‘lean ratios of nonfluencies to fluencies for the Unfavorable grow are consistently higher than the ratios for the Neutral grow. Mean ratios Of nonfluencies to fluencies for the Favorable grow are consistently lower than the Neutral grow's ratios. The significant main effects for both feedback and time further swport this clear pattern. 32 Table 11. Analysis of variance of nonfluency to fluency ratios for subjects in the Favorable, Neutral, and Unfavorable feedback conditions. Source df 88 ms F Between Subjects (56) (.522u) Feedback (A) 2 .1618 .0809 12.12* Error- Between 5». .3606 .0066 Within Subjects (v.56) (.5976) Time Intervals (B) 8 .0366 .0046 5.19* A x B 16 .1807 .0113 12.83* Error Within u32 .3303 .0009 Total 512 1.1200 *Indicates significance, _p (.05 \ Two factors seem to be contributing to the significant feedback—by- tiIne intervals interaction: (1) the progressive increases in mean ratios of nonfluencies to fluencies for the Unfavorable grow between intervals it and 7, and the corresponding decreases in mean ratios for the Favorable .‘gl‘ow; and (2) the increases in mean ratios of nonfluencies to fluencies for the Favorable grow between intervals 8 and 9, and the decreases in I“ear: ratios for the Unfavorable grow between intervals 8 and 10. 33 . 10 .09 J«Unfavorable o .08 ---I #3 g, .07 E” .06 Neutral % / a; .05 E” .0», :3, ‘H .03 8 4—Favorable z .02 2?. E .01 t456789101112 Interval Pigut-e 3. Mean nonfluency-fluency ratios by 20 second intervals for subjects in Favorable, Neutral, and Unfavorable feedback conditions. 3n Therefore, the results for the nonfluency data swport Hypothesis 2: that subjects receiving negative feedback from an audience will exhibit a greater number of nonfluencies than speakers receiving neutral feedback, and that speakers receiving positive feedback will exhibit fewer non- fluencies than Speakers receiving neutral feedback. . . 10 Responses to Questionnaire Itens The results reported in this section were used to test Hypotheses 3 through 8. Data relevant to each hypothesis are reported chrono- 1(>gic:ally. The means and standard deviations for reSponses to all questionnaire items appear in Table 12. Only one of the analyses for these item yielded significant differences. Hypothesis 3: that Speakers receiving positive feedback from an a“dishes will make shorter estimates of their actual Speaking time than sPeaker's receiving neutral feedback, and that Speakers receiving negative feedback will make longer estimates of their actual Speaking time than sPeaker-s receiving neutral feedback was not swported. A simple analysis Of variance indicated that the differences obtained were not Significant (F = <1, 51; = 2/58). Data obtained from Questionnaire Item 9 were consistent with HInrpothesis It: that speakers receiving positive feedback from an audience will rate themselves as less nervous while delivering their Speech than sPeakers receiving neutral feedback, and that Speakers receiving negative “— 10 . A copy of the questionnaire and method of scoring appears in Appendix G. Tab 1e 12 . 35 Means and standard deviations of reSponses of subjects in Favorable, Neutral, and Unfavorable feedback conditions to Questionnaire Items: 2,3,“,5,6,7,8, and 9° Questionnaire Item Feedback Condition 2*: (estimate of actual speaking time) 3 = (attitude toward experimental task) pro-test posttest “* ‘ (personal reaction to recorder) 5* = (personal reaction to audience) 6 * = (reaction before delivering speech) 7* = (reaction while delivering speech) 3* = (reaction after delivering Speech) 9 * : (nervousness while speaking) Favorable Mean {§;2; 7.70 2.15 3.70 1.72 “.15 1.65 2.05 .93 1.90 .30 3.35 1.18 2.60 1.09 3.30 1.03 3.05 1.27 Neutral Mean §;2;_ 7.85 2.17 u.n1 .92 3.57 1.38 2.09 .88 2.90 .25 3.1% 1.01 3.00 .30 3.09 1.0a 3.50 1.03 Unfavorable Mean S.D. 7.u5 2.u5 3.25 1.35 3.12 2.13 2.10 .7u 3.70 1.20 3.23 1.54 3.35 1.66 3.60 l.u6 3.75 1.23 * A lower value indicates a more favorable reaction; except number 3 where a higher value indicates a more favorable reaction. 36 feedback will rate themselves as more nervous while delivering their speech than Speakers receiving neutral feedback. Subjects in the Favorable feedback condition rated themselves somewhat less nervous than subjects in the Neutral feedback condition, while subjects in the Un- favorable feedback condition rated themselves somewhat more nervous than Subjects in the Neutral feedback condition. Though the differences obtained were in the predicted direction, they were not significant (P = 1.81, _df = 2/58). Therefore, Hypothesis It was not swported. Hypothesis 5: that Speakers receiving positive feedback from an andience will rate the assistant recording their presentation more favorably than Speakers receiving neutral feedback, and that Speakers re