DETECTING DECEPTIONI THE ROLE OF VARIOUS FORMS OF INFORMATION Thesis Ior Hm Degree of M. A. MICHIGAN STATE UNIVERSITY Joyce Ellyn Bauchner I 976 PLACE II RETURN BOX to mnovothb mum your record. TO AVOID FINES Mum on o: baton duo duo. [—_————-———-——————————————- . DATE DUE DATE DUE DATE DUE MSU Is An Affirmatm Action/EM Oppoflunly Institution mam ABSTRACT DETECTING DECEPTION: THE ROLE OF VARIOUS FORMS OF INFORMATION BY Joyce Ellyn Bauchner The research examined the relationship between untrained observers' accuracy in detecting deception on the part of strangers, and available total and nonverbal information as a function of transmission channel. Twelve subjects were put through a deception inducing manipulation procedure almost identical to the one used by Exline gt §l° (1970) and Shulman (1973). This procedure yielded six subjects lying and six subjects telling the truth in a post-procedure interview. Eighty observers viewed these subjects either live through a one-way mirror, saw them on a videotape, heard them on an audiotape, or read a transcript of the interview. Observers reported whether they thought each subject was lying or telling the truth. Trained coders provided ratio-scaled estimates of how much total and nonverbal information was available when viewing each subject through each channel. The results indicate no significant relationship between amounts of available nonverbal and/or total information and the accuracy with which untrained observers detect deception on the part of strangers. DETECTING DECEPTION: THE ROLE OF VARIOUS FORMS OF INFORMATION BY Joyce Ellyn Bauchner A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1976 Accepted by the faculty of the Department of Communication, College of Communication Arts and Sciences, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree. flea/mgr? j? Ara/4,, Director of Thesis Guidance Committee: /%%é¢11&¥ ii Ekafizgf’ , Chairman AC KNOWLEDGME NT S This research was possible due to the work and com- mitment of a large number of peOple. Special thanks goes to Dr. Gerald R. Miller whose ex- penditure of time, energy, and resources made this whole project possible. Dr. Edward L. Fink and Dr. Katrina W. Simmons contributed endless hours considering both methodo— logical and theoretical issues involved in this research. Several of my colleagues also served as resource persons, volunteering both time and ideas. David Randall Brandt shared many hours of pain as the study developed. Dr. John Hocking deserves special recognition for instilling a true appreciation of excellence in research in a co-worker. Frank Boster contributed both insight and psychological support at times when they were both sorely needed. My parents, Seena and Arthur Bauchner, deserve much credit for creating a home where scholarship was valued, and achievement equally rewarded, regardless of sex. Not to be overlooked is the moral support coming from a number of friends and colleagues, Terrance Albrecht, James Taylor, and Laura Henderson. These people gave of themselves in ways far beyond the call of duty. ii Finally, special thanks goes to Dr. Richard W. Budd. His guidance and concern has made this and countless other efforts possible. He has earned the respect few educators ever receive. I owe whatever knowledge I have gained from this en- deavor to these individuals. Unfortunately the faults of this work are my own. iii Chapter I II TABLE OF CONTENTS REVIEW OF THE LITERATURE . . Introduction . . Literature Review and Problem Conceptualization . . The Scope of Concern . . Presentation of the Problem Procedural Problems with Past Creation of a Stimulus Additional Problems . Conclusion . . . . PROCEDURES . . . . . Overview . . . . . Definitions . . . . Subject-sources . . Subject-receivers . Deception Condition . Nondeception Condition Channel Condition . Nonverbal Information Total Information . Coders . . . . Experimental Design . . Coder Training . . . Experimental Procedure . Research Subject- source introduction . Decision task . Implication procedure Interviewing procedure . . Debriefing . . Data Gathering . . . Subject- -receivers Live condition . Other channel conditions . . Potential Procedural Problems Accuracy probabilitie S . . Independence of judgments . iv Page 10 22 22 26 31 33 33 35 36 36 36 36 37 37 37 37 38 42 45 46 48 50 53 54 55 55 56 57 S7 57 58 Chapter Page III RESULTS . . . . . . . . . . 60 Introduction . . . . . . . . 60 Descriptive Analysis . . . . . 71 General descriptive statistics . 71 Descriptive statistics by condition . . . . . . 72 Analysis of variance . . . 72 LISREL: Testing the total models . . . 78 ZSLS: Testing of the total model . . . 85 Compairson of LISREL and ZSLS . 89 IV DISCUSSION . . . . . . . . . 93 Summary of Findings . . . . . . 93 Implications . . . . . . . 96 Problems With the Study . . . . . 100 Future Research . . . . . . . 103 APPENDICES . . . . . . . . . . . 106 A. Inter-coder Reliability for Non- transformed and Transformed Variables . . 106 B. Instructions and Questionnaire for Standard Interval Pretest . . . . . 107 C. Questionnaire for Coder Information Estimates . . . . . . . . . 112 D. Introduction to Subject-sources . . . . 113 E. Sample Task Card . . . . . . . . 115 F. Subject—Receiver Instructions and Questionnaire . . . . . . . . 116 G. Transcripts . . . . . . 125 H. ANOVAs for Transformed Variables . . . . 151 I. Lisrel Matrices: Models I, II, and III . . 153 REFERENCES . . . . . . . . . . . 168 10. 11. 12. 13. LIST OF TABLES Methods Used to Create Lying and Truthful Stimuli Materials in Studies . . . . Descriptive Statistics for Non-transformed Variables . . . . . . . . . . Descriptive Statistics for Transformed Variables . . . . . . . . . . Means for Non—transformed Variables by Condition . . . . . . . . Means for Transformed Variables by Condition . . . . . . . . . . Analysis of Variance of Accuracy by Channel Condition . . . . . . . . . . Individual Comparisons of Channel Condition Accuracy Means . . . . . . Analysis of Variance of Available Total Information by Channel Condition . . . Analysis of Variance of Available Nonverbal Information by Channel Condition . . . Analysis of Variance of the Ratio of Available Nonverbal to Available Total Information by Channel Condition . . . . . . . Individual Comparisons of Channel Condition Available Total Information Means . . Individual Comparisons of Channel Condition Available Nonverbal Information Means . . Individual Comparisons of Channel Condition Ratio of Available Nonverbal Information to Available Total Information Means . vi Page 24 66 67 68 69 76 76 76 77 77 77 78 78 Table 14. 15. 16. 17. 18. 19. A1. H1. H2. H3. H4. II. 12. I3. I4. I5. 16. Page Key for Presentation of All LISREL and ZSLS Models . . . . . . . . 80 First Stage, First Equation of Two-Stage Least Squares Model . . . . . . . 87 First Stage, Second Equation of Two—Stage Least Squares Regression Model . . . . 88 Second Stage, Third Equation of Two-Stage Least Squares Regression Model . . . . 90 Comparison of Standardized Results of LISREL and ZSLS for Model III . . . . . . 91 Summary of Analyses' Relation to Hypotheses . 94 Inter-coder Reliability for Non-transformed and Transformed Variables . . . . . 106 Analysis of Variance of Available Total Information by Channel Condition Using Transformed Variables . . . . . . 151 Analysis of Variance of Available Nonverbal Information by Channel Condition Using Transformed Variables . . . . . . 151 Individual Comparisons of Channel Condition Available Total Information Means Using Transformed Variables . . . . . . 152 Individual Comparisons of Channel Condition Available Nonverbal Information Means Using Transformed Variables . . . . . . 152 Model I: Input Data . . . . . . . 153 Model I: Initial Solution . . . . . 154 Model I: Maximum Likelihood Solution . . 156 Model I: Standardized Solution . . . . 158 Model II: Input Data . . . . . . 160 Model II: Initial Solution . . . . . 161 vii Table Page I7. Model II: Maximum Likelihood Solution Center . . . . . . . . . . 162 18. Model II: Standardized Solution . . . 163 I9. Model III: Input Data . . . . . . 164 110. Model III: Initial Solution . . . . 165 Ill. Model III: Maximum Likelihood Solution Center . . . . . . . . . . 166 112. Model III: Standardized Solution . . . 167 viii LIST OF FIGURES Figure Page 1. First Model of the Accuracy of the Truth/ Deception Attribution as a Function of Available Information and Transmission Channel . . . . . . . . . . l9 2. Second Model of the Accuracy of the Truth/ Deception Attribution as a Function of Available Information and Transmission Channel . . . . . . . . . . 20 3. Counterbalancing of Observers in Latin Square Design . . . . . . . . 39 4. Layout for Laboratory Facilities . . ‘ . . 47 5. Model of the Relation Between Accuracy, Available Nonverbal Information, Avail- able Total Information, and Channel . . . 62 6. Model of the Relation Between Accuracy, Available Nonverbal Information, Avail- able Total Information, the Ratio Between Available Nonverbal and Available Total Information, and Channel . . . . . . 63 7. Mean Curves of Nonverbal Information, Total Information, Nonverbal/Total Infor- mation and Accuracy of Non-transformed Variables by Channel Condition . . . . 73 8. Mean Curves of Nonverbal Information, Total Information, and Accuracy for Trans- formed Variables by Channel Condition . . 74 9. LISREL MODEL I. The Effects of Condition Channel on Accuracy with Available Nonverbal, Total, and the Ratio of Nonverbal to Total Information as Intermediate Endogenous Variables: Standardized Solution and Non- transformed Variables . . . . . . 81 ix Figure 10. 11. LISREL MODEL II. The Effects of Channel Condition on Accuracy with Available Non- verbal and Total Information as Intermediate Endogenous Variables: Standardized Solution and Non—transformed Variables . . . . LISREL MODEL III. The Effects of Channel Condition on Accuracy with Available Non- verbal and Total Information as Intermediate Endogenous Variables: Standardized Solution and Transformed Variables . . . . . Page 82 83 CHAPTER I REVIEW OF THE LITERATURE Introduction Regardless of whether social scientists View communica- tion as the process by which peOple relate themselves to their environment (Budd and Ruben, 1972), "an intentional, transactional, symbolic process" (Miller and Steinberg, 1975, p. 33), or as "anything to which peOple can attach meaning" (Berlo, 1960, p. 2), the concept encompasses some form of attribution by the receiver concerning the source and/or the message. Whenever people engage in communication, they make evaluations about other participants in the interaction. Attributions of wealth, intelligence, intent, and guilt af- fect the interactions in which they take part (Maselli and Altrochi, 1969; Miller and Steinberg, 1975). One attribution present in all interactions concerns the honesty or credibility of the source. Regardless of the information being transferred, the receiver decides: 1) whether or not the information in the message is reliable, and 2) whether or not the source believes the information (s)he has sent is credible, i.e., whether or not the intent of the source is deception. Sometimes these evaluations l 2 take place at an non-conscious level, but nevertheless they are made. In extreme forms this attribution entails the re- ceiver accepting both the message as reliable and the source as honest (a truth attribution) or rejecting the message as unreliable and the source as intentionally providing false information (a deception attribution). Various degrees of this truth/deception attribution face individuals in all as- pects of daily life: however, in its extreme form, this at- tribution has severe impact on the communication behaviors in a number of crucial situations: marital affairs, job interviews, labor negotiations, international diplomacy, and litigation. These broad implications have lead behavioral scientists to study the truth/deception attribution in var- ious contexts and under various names. Studies have examined characteristics of credible sources (Berlo, Lemert, and Mertz, 1969), the ability to detect deception during one-to- one interviews (Maier and Thurber, 1968), and characteristics of deceivers and non-deceivers in dyadic interactions (Knapp, Hart, and Dennis, 1974). Due to the variety of unrelated situations, measurement techniques, and experimental manipu— lations employed in examining deceptive communication, knowl- edge of the phenomenon remains limited, disorganized, and ambiguous. Knapp et 31. recently noted that "given the pervasive nature and potential influence of our penchant to fabricate, it is surprising how little we know about lying. At this point in time, almost any systematic study of 3 deception must be labeled 'exploratory'" (p. 16). The purpose of this study is to add to the growing body of knowledge concerning the communication process sur- rounding truth/deception attributions by beginning to answer the question: What effect does the type and amount of infor- mation provided by a source have on the accuracy of the truth/ deception attribution made by the receiver? The first chap— ter of this paper discusses the theory and literature sur- rounding this question and presents the model of the process which was tested. The second chapter presents experimental procedures used in answering this question. The third chap- ter presents the results and analysis of the experimental data. The final chapter discusses the findings and their implications in terms of future communication research. Literature Review and Problem Conceptualization The efforts of behavioral scientists who have attempted to examine deceptive communication empirically can be divided into three basic study groupings. The first_group of_studies represents attempts to determine the degree to which physio- logical indicators can accurately herald deception (Landis and Wiley, 1926; Dearman and Smith, 1963; Gustafson and Orne, 1963b; Davidson, 1968). Since only the physiological differ- ences between deceivers and non-deceivers were examined, the findings of these studies offer little information concerning the accuracy of and basis for the truth/deception attribution. 4 Therefore, this first group of studies will be considered only in terms of procedural problems of dealing empirically with deception. The second group of studies represents attempts to determine the behavioral correlates of lying (English, 1926; Berrien and Huntington, 1943; Ekman and Friesen, 1969; Matarazzo, Wiens, Jackson, and Janaugh, 1970; Mehrabian, 1971; Horvath, 1973; Knapp et_al., 1974). This second group, though also falling short of dealing with the truth/deception attribution in terms of an interaction between sender and receiver, offers some indication as to what type of informa- tion provided by a sender might lead to an accurate truth/ deception attribution. This literature will also be consider- ed in terms of procedural and conceptual problems connected with examining deception empirically. The third group of studiesrepresents attempts to iden- tify the degree to which and under what conditions an un- trained observer can accurately make a truth/deception attribution (Fay and Middleton, 1941; Hildreth, 1953; Maier and Thurber, 1968; Shulman, 1973; Ekman and Friesen, 1974; Hocking, Bauchner, Miller and Kaminski, 1976). This third group provides the most useful findings concerning an inter- action between a sender and receiver in terms of the truth/ deception attribution, in addition to procedural and experi- mental guidance in dealing with deception experimentally. 5 Based on these three groups of studies, attribution theory and some information processing considerations, the basis for this study is presented as follows: 1) an expli— cation of the domain to which the study refers as the truth/ deception attribution; 2) on the basis of previous work, a discussion of the problem and resultant model which is tested; and 3) a discussion of some procedural problems with previous truth/deception research which served as a basis for estab- lishing procedures in the present study. i The Sc0pe of Concern Before examining further the problem of the truth/de- ception attribution, a clear definition of "deceptive com- munication" is needed. For the purposes of this study, deceptive communication refers to the withholding of spon- taneous behavior and/or the substitution of simulative be- havior by a source, with the intention of creating beliefs in a receiver which the source recognizes as false or invalid;! the source must consider the success of the creation of this false belief as important to his/her well being. The cen- trality of the intentionality and importance criteria to the definition can be explained through attribution theory and the work of Ekman and Friesen. As mentioned earlier, truth/deception evaluations are one possible attribution which receivers make concerning sources. According to Jones and Davis (1965) the attribution 6 process can be understood in terms of a number of fundamen- tal components. Jones and Davis state that before a receiver makes an attribution (s)he must perceive intention on the part of the source. This presupposes that the receiver be- lieves the source was aware of his/her actions and that they would have the resultant effect; Jones and Davis refer to these presuppositions as knowledge and consider them neces- sary factors for intention. Another factor necessary for in- tention is ability, or more precisely, the receiver's judgment of the source's capacity to bring about the observed effects. In this thesis these preconditions exist for all situations involving deceptive communication. Sometimes sources unknowingly provide false information; as a result the receiver may feel that an honest source is just misinformed. In the sense that false information has been provided the source has lied. Yet, in terms of the pre- conditions set down for the attribution process, this form of "lying" is not deceptive communication. At a practical level this distinction means that certain cases are beyond our focus: the behavior of the negotiator who deliberates in good faith only to have management stray from the agreed-upon policy, the politicians who obtain votes by making inaccurate statements and giving false impressions which they themselves believe, or the diplomat who deals with his colleagues in good faith, only to have government officials surprise him with new policies. Even if, in these cases, the receiver 7 attributed deception to the sources, the phenomena would still not be deceptive communication; lack of intention eliminates them from the domain being considered. Factors which lead to the truth/deception attribution may take various forms: past knowledge on the topic of dis- cussion, personal distrust of the class, race or sex of the source, past experience with the source, etc. What concerns us is communicative behavior, i.e., the words or actions of the source on which the receiver bases an attribution. Ekman and Friesen (1969) suggest that the importance of an interaction to a source has a direct effect on his/her ability to control behaviors upon which receivers base truth/deception attributions. They contend that individuals can easily lie successfully about something unimportant. The rationale for this position is clearer when viewing deceptive communication in terms of the source withholding spontaneous behavior and/ or engaging in simulative behavior to intentionally create a false belief on the part of the receiver. While one can claim all communicative behavior is Spontaneous, some be— haviors are not carefully considered by the actor before proceeding (spontaneous) as opposed to those which are addi- tionally monitored in terms of suppressed behavior (simula- tive). Specifically, simulative behavior in deception involves: 1) the substitution and/or addition of behaviors similar or parallel to those which are suppressed, i.e., those which draw attention of the receiver to the fact that 8 false or irregular information may be being presented; and 2) the self-monitoring by the source of the simulative be- havior in terms of his/her spontaneous behavior. This could be consciously suppressing a movement of the foot while try- ing to capture the receiver's confidence by looking him/her straight in the eye. The conflict and concentration involved when the source attempts to withhold spontaneous behavior which might reveal the "truth" and to substitute simulative behavior does not take place in inconsequential interaction. Unless the interaction is important the source may not go through those processes which differentiate truthful and de- ceptive communication; the need to hide and simulate behav- ior stemming from nervous anxiety over detection is absent. One final factor that limits the scope of deceptive com- munication is the relational history of an interaction. "Relational history" refers to the interactive experience the receiver has had with the source that enables him/her to perceive behavior on the part of the source in terms of the source's known behavioral idiosyncracies. More precisely, the distinction is being made in terms of the degree to which the history of the relationship allows for the use of psychological data (Miller and Steinberg, 1975) in making attributions concerning the source. This distinction could seemingly place a 50-year marriage, for instance, at one end of the continuum, and an interaction between strangers at the other end of the continuum. 9 The relational history between interactants can have a marked effect on the process surrounding the truth/deception attribution. Consider interactions between a job interviewer and a series of job applicants: the interviewer has no rela- tional history for the applicants and knows they all want work. The interviewer must determine if the applicant possess- es the qualifications claimed during the interview; some form of a truth/deception attribution thus takes place. While talking with the first applicant the interviewer decides that (s)he is lying. However, since the interviewer had no past contact with this applicant, the attribution had to be based on a stereotype of what constitutes deceptive behavior. In this case, let us say that the interviewer based the attri- bution on two behaviors: the applicant consistently avoided eye-contact and fidgeted throughout the interview. By contrast, the second applicant is a friend of the interviewer's brother. The interviewer knows nothing about this friend's job qualifications, but has had a chance to interact with him/her in other situations. The interviewer decides the second applicant is honest even though (s)he also consistently avoided eye—contact and fidgeted. In this second case, based on information from their relational his- tory, the interviewer knew the applicant avoided eye-contact and fidgeted whenever asked a lot of questions. The dif- ference in relational history provided information on which different attributions could be based. 10 In order to avoid an interaction between accuracy and various levels of relational histories among participants, the study presented here deals only with stranger dyads. To deal with relationships of a more interpersonal nature, measures for relational history would have to be developed. Also, any such research would have to develOp a typology to equate various levels of relational history based on the type of information interactants used when communicating. The complexity of these tasks, compounded with research consider- ations yet to be discussed, led to the limitation of this research to pairs of strangers. Presentation of the Problem Given the preceding framework for deceptive communica- tion, past research offers a body of inconclusive findings. A number of studies have found nonverbal cues which signifi- cantly distinguish deceivers from non-deceivers (English, 1926; Berrien and Huntington, 1943; Ekman and Friesen, 1969a; Matarazzo, gt_gl,, 1970; Mehrabian, 1971; Horvath, 1973; Knapp, gt gl., 1974; McLintock and Hunt, 1975). Even if the weak variety of deception manipulations (an issue to be dis— cussed in the following section) and various procedural prob- lems are ignored, the studies which deal with behavioral cor- relates of lying do not indicate what behavior receivers examine when making accurate attributions of veracity nor do they address the issue of interaction between a source and a ll receiver. Differences in the behavior of deceivers and non- deceivers are useful to receivers only if they notice the differences and process them as information useful in making a truth/deception attribution. No research indicates that receivers even notice, let alone base truth/deception attri- butions upon the behavioral correlates of deception reported in these studies. In fact, accuracy scores from studies ex— amining the ability of untrained observers to detect decep- tion on the part of strangers indicate that if receivers see the behavioral correlates of deception, they do not use them to make accurate truth/deception attributions (Fay and Middleton, 1941; Maier and Thurber, 1968; Ekman and Friesen, 1974; Hocking gt gt., 1976). Two studies have examined untrained receivers' ability to detect deception when a maximum range of behavioral corre- lates of deception are available. Maier and Thurber (1968) found an accuracy score of 58.3% in a live interaction, and Hocking gt gt. (1976) found an accuracy score of 58.5% when the receiver viewed a videotape of the source. Ekman and Friesen (1974) found a slightly higher accuracy score, 63.5%, by blocking out the head shot and audio track of the video— tape viewed, thus limiting the behavioral correlates avail- able for examination. Even using the arbitrarily chosen 50% as a criterion for chance accuracy of receivers detecting deception, neither score (58.5% or 58.3%) seems extremely high. In fact, the highest accuracy score using a visual 12 channel, 63.5%, found by Ekman and Friesen (1974), occurred in an experimental condition where a large number of the sources of behavioral correlates of deception were eliminated. These accuracy scores, in combination with the sender- orientation in behavioral correlate research, limit the utility of their findings. Hocking (1976) observed that "research on visual, paralinguistic, and verbal correlates of lying and truthful behaviors offers little in terms of iden- tifying specific cues on which accurate judgments of decep- tion may be based" (p. 29). One might add that the value of attempts to identify such cues may itself be questionable be- cause, according to Maier and Janzen (1967), judgments of deception seem to be based on impressionistic and intuitive ‘/.r grounds, rather than on the basis of specific behaviors. In / fact, it seems unlikely that verbal and nonverbal behavioral cues function independently in signaling or "leaking" clues to deception (Ekman and Friesen, 1969b), but rather in con— junction. If so, what may be indicated is the need for meth- odology which treats these behaviors holistically, perhaps in terms of amount of information. Such a methodology is also supported by further review of the truth/deception attribution findings. The greater part of the research dealing with the ability of untrained observers to make accurate truth/deception at- tributions has found higher accuracy scores in conditions which remove sources of behavioral cues, specifically 13 nonverbal cues. Maier and Thurber (1968) found the highest accuracy score, 77.3%, in a condition where receivers read transcripts of students role-playing deceivers and non- deceivers. Comparatively, Maier and Thurber (1968) reported lower accuracy scores for receivers in the audio-only condi- tion, 77%, which adds paralinguistic cues for examination, and in the live condition, 58.3%, which offers the maximum range of behavioral cues for examination. Hocking gt gt. (1976) report a similar pattern of higher accuracy the fewer behavioral cues available, with accuracy scores of 62.5% among receivers reading transcripts of deceivers and non- deceivers, 61.8% among receivers hearing audiotapes of de- ceivers and non-deceivers, and 58.5% among receivers viewing videotapes of deceivers and non-deceivers. Studies by both Ekman and Friesen (1974) and Fay and Middleton (1941) offer no internal information concerning accuracy patterns in re- lationship to the quantity of available behavioral cues, since they are basically single-channel studies; the former uses just an audiotape and the latter a videotape minus the audio track. Taken as a whole, the findings of studies dealing with the ability of untrained receivers to make truth/deception attri- butions concerning strangers indicate that accuracy increases as the transmission channel limits the range of information available for examination. One possible explanation offered by Maier and Thurber (1968) is that visual cues provided by 14 sources serve as distractors rather than helpful aides. This distraction explanation stems from research on the effects of distractive stimuli on persuasion and source credibility ratings. It has been argued that distraction may facilitate persuasion and perceptions of credibility by dividing the attention of a person toward whom a persuasive attempt is directed, reducing the person's ability to scrutinize care- fully the oncoming communication, and thus increasing his/her susceptibility to influence (Breitrose, 1966; Dorris, 1967; Osterhouse and Brock, 1970; Keating and Brock, 1974; Brandt, 1976). Given that a deceiver attempts to convince the re- ceiver that his/her deceptive performance typifies normal communicative behavior, persuasive and deceptive settings are analogous. Increasing the amount of available cues places greater demands on receiver attention, perhaps reducing the ability to scrutinize any specific behavior or set of be— haviors. If so, then behavioral cues which are extraneous to deception may distract attention from cues which are potential indicators of its occurrence, thus resulting in inaccurate truth/deception attributions. The conception of the attribution process presented by Jones and Nisbett (1971) and supported by a number of research findings (Jones and Harris, 1967; McArthur, 1970, 1972; Nisbett, Caputo, Legant, and Marechek, 1973; Storms, 1973) also lends theoretical support to the contention that a trans- mission channel will affect the attribution process, and 15 especially the truth/deception attribution. Jones and Nisbett proposed a theoretical framework for explaining dif- ferences in attributions made by sources and receivers, based on divergent perspectives. Attribution theorists contend that sources give considerable weight to external environ- mental causes when making attributions, while receivers place more emphasis on internal personal causes for the source's behavior. Jones and Nisbett posit that this difference be- tween source and receiver perception can be attributed to the difference in information available to the two participants. In addition to different past experiences and psychological perspectives, the roles of source and receiver force differ- ent external perspectives on the two participants. Sources and receivers must look--in a literal sense--hear, and con- centrate on different objects and additional interactions in the environment. Through the use of videotape, Storms (1973) found that attributions made by sources and receivers could be reversed by simply providing one with the physical view which the other had of the interaction; the visual perspec- tive significantly changed what participants thought and felt about the interaction. A change in transmission channel of deceptive communication presents receivers with a change in visual perspective. Seeing an event through a camera, read- ing about it, hearing it, or actually being there alters the amount of information available and the foci of concentration of the receiver. 16 As noted by Maier and Thurber (1968), traditional posi- tive effects attributed to various transmission channels by many researchers are not supported by the accuracy scores. Wiener and Mehrabian (1968) state, "it is hardly debatable that the greater the quantity and quality of sensory channels available in a communication link, the greater the informa- tion potentially put in, through and out of the system" (p. 82). Ryan (1976) expands on Wiener and Mehrabian to include a greater awareness and appreciation of other's (the source's) intentions based on the increase in sensory information pro- vided by face—to-face interaction (McGrath and Altman, 1966; Weston and Kristin, 1973; Champness, 1973; Turnbull, Strick- land and Shaver, 1976). In terms of the truth/deception at- tribution additional sensory information provided by face-to- face interaction does not seem to heighten awareness of the intention of the source on the part of the receiver. This traditional view of channel would suggest a positive effect of increased information due to effective knowledge utiliza- tion by the receiver, though truth/deception attribution ac- curacy findings reject this 1ine of reasoning. The rationale underlying this utilization explanation suggests that, to the extent that the richness of available cues is directly rela- ted to increased perceptual acuity on the part of a receiver, (s)he should be better able to detect signals of deceit. This rationale has been offered by researchers involved with the study of teleconferencing (e.g., Ryan, 1976) and is at 17 least implied by Ekman and Friesen (1969a, 1969b, 1974). An alternative explanation to distraction can more ad- equately explain the high accuracy score on the part of re- ceivers reading transcripts in comparison to those experienc- ing the interactions by means of audiotape, videotape, and live contact. Being distracted from verbal testimony by be- havioral cues which accurately distinguish between deceivers and non-deceivers should not result in lower accuracy scores on the part of receivers; as noted earlier, such behavioral cues are available which distinguish between deceivers and non-deceivers. Instead of distraction, a form of "information overload" may be leading to the inaccuracy of truth/deception attributions by receivers experiencing interactions through information-abundant channels. Danowski (1974) explains that when individuals receive more information than they can handle--than they have the capacity to process at one time-- they experience confusion which results in higher output of error. As visual and paralinguistic cues increase, the amount of data receivers must process also increases (in this case as a function of the transmission channel); with this increase in available information, receivers reach an information- processing threshold, and overload results. Filtering and chunking (Danowski, 1974) are two processing strategies by which receivers can adapt to the overload. Both strategies entail receivers processing information based on stereotypes of deceivers in order to avoid having to process all available data from the source; the inaccuracy of a stereotype leads to 18 the inaccuracy of a truth/deception attribution. The greater the overload as a function of increase in available data from a broad-spectrum sensory channel, the stronger the influence of inaccurate-~not highly generalizable--stereotypes on the truth/deception attribution. Regardless of whether the process which leads to inaccu- rate truth/deception attributions is distraction and/or over- load, the following two hypotheses result: Hl: As transmission channel(s) increase(s) the amount of nonverbal 1nformation available to a rece1ver, the accuracy of a truth/deception attribution concerning an unfamiliar source will decrease. H2: As transmission channe1(s) cause(s) an increase in the ratio of nonverbal to total 1nformat1on available to a receiver, the accuracy of the truth/deception attribution concerning an un- familiar source will decrease. These two hypotheses offer alternative causal models which would explain the relationship between transmission channel, available information, and the accuracy of a truth/deception attribution (Figure l and Figure 2). Nonverbal, total, and the ratio of nonverbal to total available information are hy- pothesized to be a function of the transmission channel. Transmission channel, for the purposes of this study, encom- passes: (1) live, (2) videotape, (3) audiotape, and (4) transcript presentations. In the first model (Figure 1) accuracy of the truth/deception attribution is hypothesized as a function of the amount of nonverbal and total informa- tion available, and in the second model (Figure 2) as a func- tion of these two forms of available information in addition 19 Available Nonverbal Information 4 . Transmission Channel Accuracy (Live) (Video) (Audio) (Transcript) (I Available Total Information Figure 1. First Model of the Accuracy of the Truth/ Deception Attribution as a Function of Available Information and Transmission Channel. 20 Available Nonverbal/Total Information Transmission Channe1///////::::I::::\\\\\\\\\\\A (Live) -—————9Nonverbal ) Accuracy (Video) Information (Audio) (Transcript) J. Available Total Information Figure 2. Second Model of the Accuracy of the Truth/ Deception Attribution as a Function of Available Information and Transmission Channel. 21 to the ratio between the two. Verbal information is not in— cluded, since it will not vary across the transmission channels and therefore cannot explain the difference in ac- curacy scores between channels of transmission. Both models also predict that total information will vary as a function of changes in nonverbal information. In considering these hypotheses, a number of additional comments are appropriate. First, the relationships between amounts of information available and accuracy should hold for changes in amounts of information caused by other factors in addition to transmission channel. However, other factors, such as idiosyncracies of sources as related to the amount of information they provide, may not produce large enough dif- ferences in the amount of information available to achieve the effect obtained by moving from a live to a transcript presentation. Also, procedurally, there is little difficulty in making a transcript or a videotape of a live presentation. Getting the same source to provide identical verbal presen- tations, while carefully and realistically varying amounts of nonverbal information presents a difficult task for all but the adept actor, and using actors in this type of nonverbal study presents serious questions of generalizability. These two considerations lead to the direct concern with channel of transmission as a key factor; this, however, does not imply that the relationship between amount of information and ac- curacy may not also be mediated by some other factor. In 22 addition, for the present study, a relationship between amount of information and accuracy has been proposed for the truth/deception attribution. Past research indicates that such a relationship is most appropriate to the deception do- main; nevertheless, the relationship may hold for other types of attributions. Finally, the hypothesized relationships would hold for interactions where participants have a long relational history ggty_if the mediating process were dis— traction. If information overload is the process mediating the inaccurate truth/deception attribution, receivers who have interacted with the source over a long period of time would be able to determine what available data were relevant to whether that specific source was lying or telling the truth; as a result of information from the past relationship with the source, overload would not lead to inaccuracy on the part of the receiver. Procedural Problems with Past Research All of the studies examining deceptive communication faced the task of creating a stimulus containing samples of deceivers and non-deceivers. A number of problems with the majority of stimuli created call for a careful examination of the previously proposed hypotheses under different conditions. Creation of a Stimulus The majority of researchers studying deception told sources to lie or tell the truth in situations that had few 23 and trivial consequences for the liar (see Table l). Fay and Middleton (1941) told the sources to reply to questions concerning their own physical characteristics according to a card which would be flashed in front of them which read "lie" or "tell the truth." The receivers, who could hear but not see the speakers, had to determine who was lying and who was telling the truth. Methods similar to those used by Fay and Middleton lack saliency for participants, a problem already discussed. "Saliency" refers to the degree to which sources are consciously concerned with the deception, and to the extent to which avoiding detection is important to them. Davis (1961) suggests a "punishment theory" which states the greater consequences of detecting the greater physiological response during lying. In addition, the research of Gustaf- son and Orne (1963) supports the notion that a person's per— ception of the consequences of detection of deception affects the likelihood that detection will occur. With minimal con- sequences for being detected, sources may aid co-participants in this experimental task by purposely behaving as they believe liars behave. The applicability of findings from the studies using a method that has few or trivial consequences connected with detection of deception thus remains questionable. A second method of generating samples of "liars and truthers" involves having individuals advocate a position con- sistent with or counter to their attitude (see Table l). Knapp gt gt. (1974) provide an example of counter-attitudinal 24 Table 1. Methods Used to Create Lying and Truthful Stimuli Materials in Studies Counterattitudinal Advocacy Role Playing Hildreth (1953) Maier (1965) Mehrabian (1971) Maier and Janzen (1967) Experiment I Maier and Thurber (1968) Experiment II Knapp, gt gt. (1974) Few or Trivial Consequences Serious Consequences Marston (1920) Ekman and Friesen (1969) English (1920) Mehrabian (1971) Goldstein (1923) Experiment III Landis and Wiley (1926) Ekman and Friesen (1974) Fay and Middleton (1941) Shulman (1973) Berrien and Huntington (1943) Davidson (1968) Matarazzo, gt gt. (1970) Hocking, gt gt. (1976) Cutrow, Parks, Lucas, and Thomas (1972) Motley (1974) McClintock and Hunter (1975) Mehrabian (1971) Experiment II Mehrabian (1971), Experiment II, employed two methods. advocacy in their recent study which presents a content an- alysis of verbal and nonverbal deceptive and non-deceptive behavior. Veterans argued for or against the initial stance they had expressed on the topic, "V.A. Educational Benefits Should Be Increased To Finance The Cost Of An Education Today." The first problem with the counterattitudinal advocacy method stems from the lack of any theory or data supporting the notion that deceiving is the same as stating the 25 advantages of a position with which one is in disagreement. The advocate may believe the advantages of the Opposite view are real, but not strong enough to sway his own opinion. Second, past research indicates that the counterattitudinal process may alter the initial Opinion of the advocate (Miller and Burgoon, 1973). If this change in Opinion takes place before the end of the speech, a process which thus far has not been tied to any specific point in time, some of the supposed deceivers may really be proadvocating their newly acquired attitudes. A final problem with this method is that when individuals counterattitudinally advocate, they may in- tentionally generate some cues to let the audience know they are not espousing a position they believe. The third method of generating samples Of "liars and truthers" involves the technique of role-playing lying and telling the truth (see Table 1). Maier and Janzen (1967) asked students to play the role of a student being cross- examined by a professor who suspects him/her Of having cheat- ed On an examination. The possibility exists that individ- uals engaged in this process are merely "acting" as they think deceivers or non-deceivers would act, i.e., attempting to emulate what may be inaccurate personal stereotypes of deceivers and non-deceivers. Since conclusive data on dif- ferences in behavior between liars and truth tellers have not been established, no reason exists to believe these stereo- types resemble behaviors resulting from actual deceptive 26 communication. The final method of generating samples of "liars and truthers" has individuals engaged in ego-involving tasks which provide important consequences for deceivers and non- deceivers. Shulman (1973), Ekman and Friesen (1974), and Hocking gt gt. (1976) use such a procedure (see Table 1) which will be described in detail later in this thesis. Since importance and involvement are two important criteria for examining deceptive communication, the greatest confi- dence may be placed on the results of studies using methods which arouse them. Additional Problems Next to the study by Maier and Thurber (1968), which faced the problem of using role-playing sources, Hocking gt gt. (1976) Offer the most convincing findings in terms of the proposed hypotheses. However, a close examination of the Hocking gt gt. design affords an alternative explanation for the findings; i.e., that they were a function of the experi- mental procedures used in creating the stimulus tapes. Hocking gt gt. presented a cover story to criminal justice majors designed to create an important task in which they would agree to serve as "liars and truthers." The stu- dents received a letter from the director of the School of Criminal Justice asking them to take part in research which would help "identify certain characteristics Of individuals which may contribute to their successful performance as 27 police Officers" (p. 6). When the students arrived they were told by both the researcher and a dectective, who served as the interviewer to add credibility to the procedure, that the research had been designed to test a possible screening pro- cedure for prospective policemen to be used by police depart- ments. They were told that policemen Often face situations where it is useful to provide false impressions and lie; in addition, that the School of Criminal Justice was very interested in the results of this project, particularly how well they as individuals performed the task. The students were asked to lie about feelings while view- ing slides of burn victims and to tell the truth about feel- ings while viewing slides of beautiful scenery and children playing. Students also lied and told the truth concerning a videotape they saw of a man receiving sentence for a crime. However, the matter was further complicated by telling the students that when lying about the factual information con- cerning the sentencing, they had to describe what happened in terms of a version Of the videotape they had not seen; a description of this version was supplied. This was done to solve the problem of possible varying descriptions by the 1y- ing students as opposed to relatively consistent descriptions by students not lying, therefore enabling almost foolproof identification of liars. Each student also gave a sample of truthful communication. The students were videotaped in both color and black and white to check for differences in the 28 format Of the visual and audio channels. In addition, a transcript and an audio-only tape of the interaction was made. Finally, in order to control for the possibility that some individuals might always look as though they were lying and others as though they were always telling the truth, two stimulus tapes were made. Tape I was the inverse of Tape II; so, i.e., if person 1 was lying on Tape I, the same person was telling the truth on Tape II. By summing across tapes in order to obtain accuracy scores, the study eliminated the problem of inflating or deflating accuracy scores by choosing a sample of lying behavior from individuals who always looked as though they were lying, or a sample of truthful behavior from individuals who always looked as though they were telling the truth. Under each condition subjects were asked to judge whether the student they Observed was lying and to indicate the degree of confidence with which the judgment was made. The major problem with the Hocking gt gt. design is that students always told the truth about the pleasant slides and lied about the unpleasant slides. The overload which Hocking gt gt. posit as a function of transmission channel may have merely been a confusion on the part of the subjects concerning nonverbal cues due to having to view unpleasant slides with nonverbal cues of deception. This confusion was amplified since some students did not really look at the slides while lying, due to the unpleasant content, while others with poorer vision strained to see the slides (which were shown on a 29 screen on the floor in front of them). All these additional nonverbal behaviors, which were a function of the design rather than the transmission channel, would have been missing in the written channel and only partially present in the audio-only channel in the form of paralinguistic cues. The only attribution that subjects had to explain the additional nonverbal behavior was the truth/deception attri- bution. The work of Schachter and Singer (1962) offers a theoretical framework for how subjects may have wrongly attri- buted the behavior. Schachter and Singer conducted an ex- periment designed to support a two-component theory of emotion. According to the theory, the first component Of emotion is physiological arousal. This arousal is identical regardless Of the specific emotion; it varies in intensity, not in kind. What differentiates these emotional states phenomenologically is a cognitive or labeling component. Individuals Observe the characteristics of a situation and differentially attri- bute arousal according to their interpretation of their ob- servations of how they think they should be responding emotionally, based on these Observations. Subjects in the Schachter and Singer study were injected with either the drug epinephrine, a synthetic adrenaline, or a placebo (under a doctor's supervision). Those who were in- jected with the drug would suffer from symptoms of physi- ological arousal (sweaty palms and increased heart rate, etc.), while recipients Of the placebo would not experience 30 these symptoms. A third Of the recipients Of the drug was accurately informed concerning its effects. A third Of the recipients of the drug was told that it produced effects un- related to arousal. The final third Of the epinephrine group was told nothing. The subjects were then put in a waiting room with a confederate, who they believed was a fellow par- ticipant in the experiment who had received an injection identical to their own. The confederate then proceeded to exemplify one of two types Of behavior: half the time the confederate exemplified euphoric behavior and half the time the confederate exemplified angry behavior. The subjects who received the real drug and accurate information did not in- dicate any real emotional experience; according to Schachter and Singer these individuals did not have to use the con- federate as a source of attribution for arousal, since the experimenter had already provided them with a plausible ex— planation for the physiological state. The recipients of the drug who were misinformed or not informed as to the ef- fects of the drug indicated emotional experiences similar to those depicted by the confederate; they attributed their physiological state to the social situation to which they were exposed. Subjects receiving the placebo reported little or no emotion due to the lack of the first component, arousal. The Schachter and Singer findings support the theory that arousal will be attributed in terms of the appropriate explanations provided by the social context in which it is 31 experienced. If these results are extended, arousal, when observed rather than experienced, will be attributed in terms appropriate to the social situation in which it is observed; thus, this allows for an alternative explanation of the attributions made by receivers in the Hocking gt gt. (1976) study. The unpleasant stimulus which only accompanied lying could have caused cues Of arousal due solely to the un- pleasant content Of the slides. The receivers Observing the criminal justice students would explain these cues in terms of the truth/deception attribution, since the situation pre- sented to them was one which called for such an attribution. Since the arousal cue occurred only in those lying (since sources lie only about unpleasant slides), these arousal cues could have inflated the accuracy scores Of receivers in all conditions. Conclusion Examining the models and hypotheses previously discussed calls for the creation Of a stimulus which overcomes as many as possible of the problems noted. This requires dealing with all four transmission channels within the same manipula- tion procedure. The fact that no previous research has done this makes comparisons of accuracy across transmission chan- nels difficult. The manipulation procedure must be both ego— involving and important to sources, yet not equate role-playing or counterattitudinal advocacy with deception; specifically it must Operationalize the details of the definition Of 32 deceptive communication presented in this thesis (see p. 4). A deception manipulation procedure used by Exline gt gt. (1970) and Shulman (1973) meets these criteria and therefore is used in this study. Details of the procedure appear in the following chapter. CHAPTER I I PROCEDURES Overview Data gathering involved three stages, each centering around a different group of participants. Stage I involved training coders to reliably estimate the amount of total and nonverbal information available in a given stimulus. Stage II involved taking twelve subject-sources through an experi- mental procedure which induced half of them to lie when inter- viewed by the experimenter. Stage III involved subject- receivers and coders seeing, hearing, or reading the inter- views of the twelve subject-sources. Subject-receivers filled out questionnaires reporting whether they felt each subject-source was lying or telling the truth, and the coders reported estimates of the amount of total and nonverbal in- formation available, Observing each interview in each chan- nel-condition. In Stage I, eight undergraduates at Michigan State University received eight weeks of training in making ratio- scaled estimates as to the amount Of total and nonverbal in- formation available in a given stimulus. Coders were trained in and made estimates by means of a direct interval estimation procedure (Silverman and Johnson, 1975). This procedure 33 34 called for the experimenter providing coders with two sample segments of interviewing and telling the coders the amount of available total and nonverbal information in these seg- ments. Coders then based all future estimates of amounts of nonverbal and total information on the values assigned to these segments. These two segments were called the "standard interval" throughout the procedure. Segments and values for the standard interval were chosen on the basis Of the results from a pretest using the sixteen interview segments in the Hocking gt gt. stimulus. In Stage II, twelve freshmen at Michigan State University participated in an experimental procedure in which a confeder- ate implicated half the subject-sources in cheating on the experimental task. After the task was completed the experi- menter questioned each of the twelve as to what took place during the task. Since none of the twelve reported the cheat- ing, half of them were lying. An experimental pretest was run to ensure standardization of procedures prior to the actual experiment. In Stage III, undergraduate subject-receivers and coders experienced the interviewing of the subject-sources by watch- ing through a one-way mirror, seeing a videotape, hearing an audiotape, or reading a transcript. Subject-receivers report— ed whether they thought subject-sources were lying or telling the truth. Coders estimated the amount of available total and nonverbal information as they experienced the interviews. 35 Sex was controlled for at all times so there were an equal number of male-male, female-female, male-female, and female-male dyads in each channel condition. Past research indicates that sex of either the source or receiver may affect the ability Of sources to control and/or receivers to Observe nonverbal cues (Fay and Middleton, 1941; Maier, 1965; Mehrabian, 1969, 1971; Shulman, 1973). The researcher controlled for acquaintanceship since the relationship between the relational history Of participants and detection of deception may be subject to confounding effects (see pages 8-9). Thorndike and Hagen (1961) and Woodworth and Schlosberg (1954) found that ideal detectors are those highly familiar with the source's past behavior. Freeberg (1969) agrees, citing Wherry as deriving a theorem that "raters will vary in the accuracy ratings given in direct proportion to the relevancy Of their previous contact with the ratee" (p. 10). However, Hastorf, Schneider, and Poleka (1970) suggest the Opposite, that a long relational history results in inaccurate detection. Due to these am- biguous results, in addition to the artificial experimental environment in which much of the deception data has been gathered, only strangers to subject-sources served as subject- receivers. Definitions The following terminology is used in the design and implementation Of the research. 36 Subject-sources Subject-sources refers to those subjects who were assigned to complete an experimental pretask with a confeder- ate. After the procedure was completed subject-sources were interviewed by the experimenter as to the details of the pro- cedures used to complete the task. Subject-receivers Subject-receivers refers to those subjects who attempted to identify which subject-sources were lying during the in- terview. Deception Condition When the confederate implicated the subject—sources in cheating on the experimental task, if the subject-sources were not willing to admit the cheating to the experimenter, they had to engage in deception during the interview. There- fore any subject-source which the confederate was to impli- cate in cheating was said to be assigned to the "deception condition." Nondeception Condition The confederate did not implicate half the subject- sources in cheating during the experimental task. These in- dividuals engaged in the task as instructed and therefore did not have to lie during the interview. Therefore individ- uals not implicated in cheating were said to be assigned to the "nondeception condition." 37 Channel Condition Subject-receivers and coders experienced the interviews via four different channels: (1) live, (2) videotape, (3) audiotape, and (4) transcript. Nonverbal Information Nonverbal information is amount of information available from nonverbal behaviors such as facial expression, eye con- tact, nodding, hand and body movement, posture, pausing, the ums and ahs people say; anything beyond the actual words an individual uses. Nonverbal information refers to Egg people say things, not what they say. Total Information Total information is a holistic estimate of all available information provided by a stimulus. It is the kind of judg- ment an individual would make if asked which of two books (s)he had just read provided the most total information. Total information is not necessarily the sum of verbal and nonverbal information. Some nonverbal behaviors may be re- dundant with some words in terms of the information that each provides. This overlap in information provided by the non- verbal behavior and words would cause total information to be less than the sum of the nonverbal and verbal information. Coders Coders are trained individuals who experienced the inter- view Of each subject-source in all channel conditions for the 38 purpose of estimating the amount of available total and non- verbal information. Experimental Design Since the present study included a "live" condition, a rather serious procedural problem had to be overcome by em- ploying a Latin square experimental design. The time required for briefing, engaging in thentask, the interview, and de- briefing of one subject-source was at minimum an hour. Given twelve subject-sources, this would have required subject- receivers and coders in the live channel condition to get through twelve hours of experimental procedures. Because of fatigue and its potential contaminating effect on experimental results, this was deemed impractical. On the other hand, the time actually needed to Observe and judge the veracity of one subject-source was approximately ten to fifteen minutes. Thus, in the videotape, audiotape, and transcript conditions, sub- ject-receivers would only have to spend ten to fifteen minutes per subject-source. TO minimize the time required of each subject-receiver and still ensure that all twelve sub- ject-sources were judged, subject-receivers were counter- balanced across conditions and subject-sources using a simple Latin square design (Lindquist, 1953; see Figure 3). Lindquist (1953) explains that simple Latin square de- signs are used Often when experimental administration becomes unmanageable or impractical. In addition to the twelve hour 39 cmflmmo mumsqm swung OH mum>uwmoo mo mcflocmamououcooo .m whomflm .COHqucOO cofiummomo mop Op omcmflmmm xaeoocmn mum3 mmousOmluommosm on» no mam: moo AHHOO comm Ou omcmflmmm masoocmn mv wum>HOOOH DOOmQSm om u o AHHOO some Op omcmflmmm Maeoocmu my mnm>flmowuluommo5m om u o AHHOO some 0» owcwflmmm maeoocmu my mum>flmomnluomn95m om u m AHHOO some on omcmfimmm mafioocmu my mum>wmowuluomflo5m om u d m>Hq emHmOmzame oHesc omoH> o OmQH> m>HA BmHmUmz¢MB OHQD€ U rm oHes< omoH> m>Hq BmHmOmzame m a emHmOmzcme oHooa omoH> m>Hq « ma Ha oa m m n o m v m m a monsomluomflmsm JeAIeoaJ-qoefqns 40 problem discussed in the previous paragraph, the design of the room from which one-way mirror Observation took place prohibited using more than six subject—receivers and three coders at one time without blocking someone's vision. Also, the one-way room was not soundproof; any noises made by subject-receivers and/or coders would have cued subject- sources to the deception-inducing procedure. In order to keep subject-sources unaware of the confederate's true ident— ity and the fact that they had been manipulated into lying, it was necessary to keep the Observation group small, and the observation time short. By using a Latin square design only five subject—receivers and three coders Observed any subject—source at one time. In the design used, 20 subject-receivers Observed three different subject-sources in each condition channel, yield- ing 240 judgments for use in analysis. Four independent randomsamples of subject-receivers, A-D, were counter— balanced across the four channel conditions and groups of three subject-sources (see Figure 3). Five members from each sample watched three subject-sources in each channel condition. SO, in the live condition, subject-sources 1-3 were Observed by subject-receivers 1-5 of sample A, subject—sources 4-6 by subject-receivers 1—5 of sample B, subject-sources 7-9 by subject-receivers 1-5 of sample C, and subject-sources 10-12 by subject—receivers 1-5 Of sample D. In the videotape con- dition, subject-sources 1—3 were observed by subject-receivers 6-10 Of sample D, subject-sources 4-6 by subject-receivers 41 6—10 Of sample A, subject-sources 7-9 by subject-receivers 6—10 of sample B, etc. Thus, "the comparison of overall treatment means for any one classification would then appear to be completely balanced so far as the effects of super- imposed treatments from other classifications are concern- ed" (Linguist, 1953, p. 258). Parallel groups of three coders were assigned to each of the four samples and counter- balanced in the same fashion. This procedure enabled three subject-sources to participate in the experimental procedure on four separate nights. Five subject-sources and three coders Observed each night; each night the observers were from the assigned sample. Experimental control called for two additional adjust- ments in sampling for this design. First, conducting the experimental procedure over a period of four nights increased the possibility Of interactions between subject-sources that would cue future subject-sources concerning the manipulations and Observations. In order to limit the possibility Of such a problem, all subject-sources were freshmen Obtained from different introductory communication classes; the goal was that no subject-source from any one evening would be familiar with a subject-source going through the procedure on any other evening. Debriefing of subject-sources confirmed that those from any one evening were not familiar with those who went through the experimental procedure on other evenings. Second, sampling had to allow for control of sex on all 42 factors. Six females and six males, freshmen volunteers from introductory communication classes, were randomly as- signed to the deception and non-deception conditions so that three females and three males were in each condition. Ten males and ten females, volunteers from each of four differ- ent introductory communication classes, served as subject- receivers; each class served as one sample (A-D) in the Latin square design. Assignment to each condition was done so that there were an equal number of male—male, male-female, female- female, and female-male subject-source and subject-receiver dyads within each condition and in terms Of each A-D sample. This produced the design represented in Figure 3 with sex controlled for along all factors. Coder Training Four male and four female undergraduates were trained over an eight-week period in providing holistic estimates of nonverbal and total information available in a given stimulus. Reliabilities for coder estimates were computed for nonverbal information, total information, the ratio of nonverbal to total information, and the log transformation data estimates of nonverbal and total information using Cronbach's alphas (Cronbach, 1951). The coefficients for reliability are .98, .96, .99, .99 and .98 respectively (p<.05). Log transforma- tion reliabilities are reported since some of the analyses were done on transformed data (see Appendix A). 43 All information measures were made by coders trained by means of ratio-scaled direct interval estimation. In direct interval estimation coders are shown two stimuli possessing different amounts of the attribute being estimated; in the present study nonverbal and total information are the attri- butes. Each stimulus is assigned a number Of points for each attribute, which represents the amount Of that attribute the stimulus contains in terms of the other stimulus. For ex- ample, if the first stimulus had half as much nonverbal and total information as the second, the first stimulus would be assigned half as many units of nonverbal and total information as the second stimulus. The stimuli could be assigned any number of information units as long as the units represented the actual ratio of information between the two. These two stimuli, together with the assigned attribute values, serve as the standard interval, the "psychological ruler" in terms of which all further estimates are based (Silverman and Johnson, 1975). The first problem researchers faced in using direct interval estimation centered around choosing two stimuli and values for nonverbal and total information which would allow for accurate and reliable estimates by coders, and would validly Operationalize the constructs as defined by the ex- perimenter. For this purpose, a pretest was conducted using the sixteen segments of interviews in the edited Hocking gt gt. (1976) stimulus. Twenty undergraduates from an 44 introductory communication course at Michigan State University were shown videotapes of sixteen students being interviewed. After each segment the students wrote down how much nonverbal and how much total information they felt were available in that segment. Nonverbal and total information were defined for all students. In addition, they were told that a blank screen represented 0 units of all types Of information and the first interview represented 100 units of nonverbal infor- mation and 150 units Of total information (see Appendix B, Pretest Instructions and Questionnaire). The mean for non— verbal information and total information was calculated for each of the 16 segments of the Hocking gt gt. (1976) stimulus. Based on these means two segments were chosen for the stand- ard interval such that Segment 1 was estimated in the pre- test to have half the amount of total information and half the amount of nonverbal information as Segment 2. Segment 1 was assigned the values Of 100 units of nonverbal informa- tion and 150 units of total information. Segment 2 was as- signed the values of 200 units of nonverbal information and 300 units of total information. These two segments served as the standard interval throughout coder training and the actual experiment. Each coder attended a minimum of two 2 1/2 hour training sessions each week for eight weeks; sessions always involved at least three coders. Coders were shown the standard in- terval throughout the session and asked to estimate the 45 amount Of nonverbal and total information available in a number of different stimuli. The training stimuli included the remaining 14 from the Hocking gt gt. (1976) videotapes, audiotapes from another stimulus used by Hocking gt gt. (1976), short interviews done live with Michigan State Uni- versity faculty and graduate students, transcripts from var— ious other studies, and videotape segments of simulated court- room testimony developed for other studies funded by a National Science Foundation grant. To avoid the fatigue which usually accompanies long time periods of estimating non- verbal behaviors (Harrison and Knapp, 1972), coders received a fifteen-minute break every forty-five minutes. At the end of the eight weeks coders participated in a pretest of ex- perimental procedures, estimating the amount of nonverbal and total information through the one-way mirror. After the ex- perimental procedure was standardized the actual experiment began, with coders taking estimates for all subject-sources in all channel conditions (see Appendix C, Questionnaire Used by Coders to Make Information Estimates). Experimental Procedure The experimental procedure used on subject-sources was almost identical to that used by Exline gt gt. (1970) and Shulman (1973). Twelve volunteers were asked to participate in a study to examine group problem-solving strategies. They were all told that they would be working with other students who had volunteered and that all students participating would 46 receive extra credit in an introductory communication course. Subject-source introduction. When each subject-source arrived in a designated waiting room, (s)he found another student, a female confederate, waiting. The confederate was always the same female, since Shulman (1973) found that con- federate sex had no significant effect in terms of this pro- cedure; however, changing confederates has been shown to have confounding effects on experimental results (Rosenthal, 1967; Rosenthal, 1968; Barber and Silver, 1968). Soon after the arrival of the subject-source the two students were escorted into a corridor of small rooms across the hall. The experi- menter explained that they would be asked to solve a problem and afterward answer some questions. She continued: We have had some problems lately with in- dividuals misunderstanding questions on the questionnaires, answering them wrong, and then leaving. After they are gone, we have a hard time getting peOple to come back and remember what happened during the problem- solving. As a result I would like to inter- view you afterwards and videotape the interview. Is it all right with you? Both the subject-source and confederate agreed. They were then shown the room where the interview would take place and told that a videotape camera was recording them through the one-way mirror. The dyad was then taken into another room in the corridor in order to engage in the task (see Figure 4 for Layout of Laboratory Facilities). The two students were seated at a table next to the experimenter and given two pieces Of scrap paper. The 47 mmfiuflaflomm muoumnoomq How usowmq .6 ECCm. IJ OZHBH<3 \/ mOB€>mqm Ii Jx 200m muzmmmmZOU /\ 200m ZOHB<>mmmmO mm>Hm0mmlmumDOm .e musmflm vfidz zmbfim ZOOM mmdh SJOIJIm Rem-euo v 48 experimenter and given two pieces of scrap paper. The ex- perimenter explained that they were participating in a grant funded by the National Science Foundation and designed to examine the relationship between group size, group problem- solving strategies, group success, and the sex combinations of group members. The experimenter added that four-, three-, and two-person groups, as well as individuals, were being asked to engage in identical content-free tasks. She said afterwards that the group members would be interviewed con- cerning the problem-solving strategies used in completing the task. They were told that the government was very inter— ested in using the results of these experiments for guidance in the formation of various task forces throughout the bureau- cracy, and as a result had provided funds for the research. Since the task they would be asked to complete was rather boring, in order to sustain interest in it, the group in each size category which performed the best would receive $50 to divide among its members. All subject-sources were informed that they had randomly been assigned to a dyad group and matched with a student from another class (see Appendix D, Introduction to Subject-sources). Decision task. The task required the dyad agree upon the number of dots in figures on nine 8 1/2 x 11-inch cards. The experimenter would show each of the cards to the dyad for fifteen seconds. They then would have to agree on one number which they felt represented the correct number Of 49 dots for that card. They were told they were to function as a team, take as much time as needed, use whatever strategy they wish, and use the pieces of paper they had been given to take down whatever notes they needed or to do calcula- tions. Before exposing the cards, the experimenter stated that she would try to provide the dyad with feedback as to their accuracy after each set Of three cards. At this point the confederate always asked if she could "talk to my partner and is there any trick or pattern in the answers that we are supposed to figure out." The experimenter emphasized that the two students should discuss what was going on whenever they felt it necessary and that as far as she knew there was no pattern or trick involved in the task. The experimenter then proceeded to show the dyad a prac- tice card, let them deliberate, record their answer, and give them the correct answer. After the practice card was completed and any further questions answered the actual task began. During the early trials the experimenter manifested some impatience with the length Of time the dyad took to complete the task. Though nothing was said concerning the amount Of time, the experimenter looked at her watch, squirm— ed during the discussion, and carried out her role in a hurried manner. This was done so that after the third card, when the confederate asked for some feedback as to how the dyad was doing, the experimenter could logically respond, 50 "It's not really important now and this is taking a lot longer than usual. Let's go on through a couple more cards and then I'll give you the answers for half; that way we won't keep the next group waiting too long (see Appendix E, Sample Task Card). Imptication procedure. The subject—sources randomly assigned to the deception condition were implicated in cheat- ing to create a strong motive to conceal information during the post—task interview. The procedure began when somewhere between the fourth and sixth card a second experimenter, who had been listening tO the interaction from an Observation room, interrupted the experimental session to inform the first experimenter that she had an "important telephone call from the director Of the research project.” The first ex- perimenter then left the room to take the alleged call. Actually, she went into the waiting room until the implica— tion procedure was completed. If the subject-source was in the nondeception condition the confederate just engaged in normal conversation with him/ her during the experimenter's absence. However, if the sub- ject-source was in the deception condition the confederate went through a procedure to implciate him/her in the act Of cheating. The confederate Opened the cheating phase by getting up to walk around in order to stretch her legs. On returning to her seat, the confederate noticed the experimenter's 51 folder which she had left on the floor beneath her chair. The confederate wondered aloud if the folder might contain the right answers, and complained that the experimenter had rushed them and had not provided the promised feedback. "She was supposed to give us the answers. How are we sup- posed to know how close we are, how we are doing?" Next, the confederate suggested that they look and see what was in the folder, and mentioned that she "really could use that prize money." Toward the end of the speech the con- federate went over to the experimenter's chair and picked up the folder, which contained the answer keys typed on 5" x 8" cards. Each of the answer key cards had the task number on it, followed by the number of dots for that card. The con- federate continued, "I think these are the right ones. I'll copy these down--we can look at them as we go along." Re- gardless of the subject-source's reaction, the confederate wrote down the information on the piece of scrap paper which the dyad had used to make notes and estimates for previous cards. The confederate always read the information aloud as she wrote, assuring that the subject-source heard the answers. She also always used the notes she got from cheating to make the remainder of the estimates. The confederate then re- placed the answer keys and folder just before the experimenter returned to the room. Whether or not the subject-source had helped the confed— erate list the answers (s)he knew that the confederate had 52 cheated. At minimum, the subject-source had been implicated as an accessory. Four courses Of action were thus Open to the subject-source: (l) (s)he could have prevented impli- cation by not allowing the confederate to cheat; (2) (s)he could have undone the implication by reporting the cheating to the experimenter upon her return; (3) (s)he could have sat quietly, inactively accepting the implication of cheat- ing; or (4) (s)he could have helped the confederate. All subject-sources chose either (3) or (4) as their course of action. It was important that the experimenter not know if the subject-sources were assigned to the deception or nondecep- tion condition, so she would not differentially question subject-sources based on her knowledge that they had been cheating. It was also important that the experimenter not return before the confederate had completed the implciation procedure and thus catch the dyad in the act of cheating. Therefore, a means had to be develOped for monitoring the whole implication procedure. A second experimenter listened from the Observation room to the conversation of the confed- erate and the subject-source. After the confederate had im- plicated the subject-source in the act Of cheating, the second experimenter told the first experimenter she could return from the "alleged" phone call. However, in order that the first experimenter could not tell who was in the decep- tion condition due to a longer period of time before she was 53 allowed to return, the duration of the experimenter's ab- sence was held constant across conditions. The timing also served to protect the confederate's cover in that the sub- ject-source had little time to question the confederate be- fore the first experimenter returned. Upon the return Of the experimenter the final task cards were completed. As the dyad discussed each card, the confederate attempted to get the subject-source to agree to report almost the exact number of dots listed on the answer key. If the subject-source resisted, the confederate was instructed to push as hard as seemed feasible, and then to hold out for an answer close to the one originally proposed. Interviewing procedure. After the task was completed, the experimenter took the dyad into another room to inter- view them concerning the strategies used to arrive at ans- wers to the task. The experimenter always began by inter- viewing the subject-source first under the pretense that the confederate would next be asked the same questions. The questions were as follows: Please state your name. Year in school? What are you majoring in? Have you ever been in any research before? How many communication courses have you had? Could you describe the strategy your group used to get their answers? 7. Could you be a little more specific? You did really well, especially toward the end. 8. If you had to describe to the next group what they should do to do as well as you did, what would you tell them, in two short sentences? 9. If you could choose what size group you could do the task over again in, what size would you 0101.5me 0 54 choose, 4, 3, 2, or alone? 10. Why? 11. Is there anything else you could add about the strategy you used? The first five questions provided subject-receivers with a sample of the subject-source's truthful behavior, as well as providing demographic information for future exam- ination. If the subject-source were in the implication pro- cedure, the remainder Of his/her answers was; untruthful, since no subject-source had mentioned that either (s)he or the confederate had cheated. Debriefing. Due to the deceptive nature of the impli- cation procedure, a one-to-One debriefing was conducted with each subject-source. After the interview was complete the dyad was taken into another room in the corridor. The first experimenter left and the second experimenter came in to de- brief the subject-source. After the identity of the confed- erate had been revealed, she left and went to the waiting room to begin the procedure again. The debriefing included a long explanation of the research grant funding the study, the true purpose of the study, a review of the literature which led up to the procedural choices made, an explanation of the actual procedure, and the answering of any further questions. All subject-sources were told that no subject- source in the deception condition had revealed the cheating to the experimenter. They were further assured by the fact that this was also true in the Exline gt gt. (1970) and 55 Shulman (1973) studies. Subject-sources were Offered a chance to see themselves on videotape, asked if it were still all right to use the videotape, and asked if they would like to receive the results of the study. NO one ob- jected to the use of the videotape; however, most subject- sources did not want to see it. Most subject-sources were enthused about the study, thought the procedure was clever, and wished to receive results. The debriefing usually took fifteen to twenty minutes. Debriefing continued until the experimenter was sure the subject-source understood the ex- perimental procedures, and understood the reason for the study and the procedures. The experimenter also questioned the subject-source as to whether (s)he had suspected his/ her partner was a confederate. Data Gathering Subject-receivers. Eighty students from introductory communication courses served as subject-receivers in the present study. Each student received extra credit for his/ her participation. Twenty students were taken from four different classes; ten from each class were females and ten were males. Five students from each class were assigned to Observe three of the subject-sources in each channel con- dition (refer to experimental design section for details). Each subject filled out a questionnaire telling whether they felt the subject-source was lying or telling the truth. Prior to Observing the subject—source the experimental 56 procedure was explained to them in detail (see Appendix E, Subject-receiver Instructions and Questionnaires). Live condition. The live condition called for somewhat unique data gathering procedures, since the subject-receivers had tO remain in the experimental area unobserved while the subject-sources were taken through the experimental proced- ure. A conference room next to the room from which one-way mirror Observation took place housed the subject-receivers for the three hours of experimental procedures (see Figure 4). Upon their arrival, the experimenter explained the pro- cedure to subject-receivers, and stressed the importance that no subject-source ever be aware of the subject-receivers. For this reason subject-receivers were asked to remain in the conference room with an experimental assistant until right before each interview. When the final phase of the task began, the assistant took the subject—receivers into the one-way mirror Observa- tion room. From there they observed the five minutes of interviewing. Since the one-way mirror was not sound-proof, it was emphasized that there must be total silence during the Observation. Questionnaires were left in the conference room and filled out when subject-receivers returned after each subject-source. NO discussion took place among the subject-receivers concerning their judgments. The assistant experimenter monitored subject-receivers throughout the entire 57 three hours and a television, cards, and other recreational materials were provided to divert the subject-receivers' attention away from the judgments while they were waiting. NO subject-receiver was allowed to travel from room to room in the experimental area until the assistant had checked that the halls were free of any subject-sources. Other channel conditions. For the videotape, audio- tape, and transcript conditions no such complex procedure was necessary. Subject-receivers were shown the stimulus in a room, and after exposure to each subject-source filled out the questionnaire. They were monitored during this period to make sure all judgments were made independently. All stimuli were shown once. The subject-receivers had no time limit as to how long they could take to make a judgment. In the audiotape condition all subject-receivers listen— ed to the stimulus with their eyes closed in order to cut down on environmental distraction. In the transcript con- dition (see Appendix G, Transcripts) subject-receivers were only allowed to go through the transcript once; however, as in other conditions, they could take as long as they wished. Potential Procedural Problems Accuracy probabilities. In this study, as in other studies examining ability to detect deception, half the sub- ject-receivers did not lie. Also, as in other studies, the 50% accuracy criterion was used for evaluating chance accur- acy. This criteria has been criticized since individuals do 58 not normally expect sources to be lying 50% of the time, and in other studies (Hocking gt gt., 1976) receivers were told that the probability might be 100% or 0; in such cases 50% may not be a realistic criterion. However, in this experi- ment each subject-receiver saw three subject-sources; these subject-sources were assigned to deception and nondeception conditions randomly. Therefore, some subject-receivers saw three deceivers, some saw three nondeceivers, and some saw a combination of the two. Subject-receivers were informed as to this variable probability of seeing deceivers and nonde— ceivers. The 50% criterion applies here since each subject- receiver had two choices which they understood were equally likely. Indgpendence of judgments. The judgment procedure as- sumes the independence of the three judgments of each indi- vidual. All subject-receivers were told of the random assignment of subject-sources to conditions and the necessity of independence Of judgment was emphasized. The range of combinations of responses of subject-receivers in the data did not indicate judgments were dependent, and since analy- sis was basically done between channel conditions, random assignment to channel conditions should make any problem with dependency of judgments equally likely in all channel conditions, and thus have no major effect on results. Likewise, running subjects together could be said to bias the independence of judgments. Again, in this 59 experiment all precautions were taken to ensure independence of judgments. CHAPTER III RESULTS Introduction Before discussing the results Of this study, a few statements concerning data handling and the analytical tools are necessary. In the two suggested models (see Figure 1 and Figure 2) judgment accuracy is the dependent variable, available nonverbal information, available total information, and the ratio of available nonverbal to available total in- formation are the intermediate endogenous variables, and various channel conditions are the exogenous variables. In testing these models four data handling procedures need ex- planation: 1) conditions are dummy-coded throughout all the analyses; 2) some analyses are performed on logarithmically transformed data; 3) based on the requirements of the analyt— ic method, coders are handled as multiple—indicators of an underlying variable or collapsed to the mean which functions as a single—indicator; and 4) usage of a dichotomous depend- ent variable. Three options were available concerning the four channel conditions: 1) channel condition could be the single true ex— ogenous variable with each channel condition coded as a dif- ferent level Of the variable (0 = transcript, 1 = audiotape, 60 61 2 = videotape, 3 = live); 2) channel condition could be the single exogenous true variable, with the condition channels functioning as multiple indicators; and 3) each channel con- dition could be a true exogenous variable with all judgments falling within the respective condition coded as 1, and all excluded coded as 0. The third option offers the benefit of discovering any pattern between the independent channels and the endogenous variables, as well as handling the linear hy- potheses suggested; for this reason channel conditions are dummy-coded. Dummy-coding the exogenous variables explains why only three conditions appear in Figure 5 and Figure 6. By coding three channels 0 or 1, the fourth, transcript, be- comes embedded in the coding scheme. More precisely, if Y = bOU + le1 + b2X2 + b3X3 + E, K II one of the intermediary endogenous variables in the models = Live = Videotape — Audiotape C><><>< NH I = Transcript "we show that unique estimates are possible for the bi's if a side condition bO = 0 is imposed. . . Consequently, the inclusion or exclusion of the unit vector causes no change in expected values or error terms. Nor will the degrees of freedom be modified as b0 is not counted in the model, for it is not an independent (of the other bi's) unknown weight" (Namboodiri, Carter and Blalock, 1975, pp. 138-139). Communication Channel (True Variable) Communication Channel (True Variable) Communication Channel (True Variable) Available Nonverbal Information (True Variable) Available Total Information (True Variable) Ability to Attribute Truth or Deception (True Variable) Distrubance Term for n1 Disturbance Term for n2 Disturbance Term for n3 Live Condition (Indicator of £1) Videotape Condition (Indicator of £2) Audiotape Condition (Indicator of £3) First Coder's Estimate of D3 Second Coder's Estimate of n3 Third Coder's Estimate of n3 First Coder's Estimate of n2 Second Coder's Estimate of n2 Third Coder's Estimate of n2 Observer Judgmental Accuracy Figure 5. Model of the Relation Between Accuracy, Available Nonverbal Information, Avail- able Total Information, and Channel Commnication Channel (True Variable) Commnication Channel (True Variable) Comnunication Channel (True Variable) Available Nonverbal Information (True Variable) Available Total Information (True Variable) Available Nonverbal/Available Total Information (True Variable) Ability to Attribute Truth or Deception (True Variable) Disturbance Term for n1 First Ooder' s Esti- Disturbance Term for 112 mate of n 2 Disturbance Term for n3 Disturbance Term for n4 y5 Live Condition (Indicator of £1) ‘< b II Second Coder's Estimate of n2 Videotape Condition (Indicator of £2) y6 = Third Estimate of I12 Audiotape Condition (Indicator of £3) y7 = First Estimate of n 3 First Coder's Estimate of I11 y8 = Second Estimate of 113 Second Coder's Estimate of n1 y9 = Third Estimate of n3 Third Coder's Estimate of n1 y10 = Subject-receiver Judg- mental Accuracy Figure 6. Nbdel of the Relation Between Accuracy, Available Non- verbal Information, Available Total Information, the Ratio Between Available Nonverbal and Available Total Information, and Channel 64 For some analyses the data are logarithmically trans- formed as a means of dealing with multicollinearity in the model. Analyses assume that underlying relationships among variables are linear and additive, and the effects of vari- ations in available information via communication channel were not expected to be necessarily additive. Logarithimic transformation seemed appropriate because it makes non- linear, non-additive relationships linear and additive (Namboodiri, Carter and Blalock, 1975, p. 489). The remain- der of the effect of multicollinear variables is accounted for by the path from nonverbal to tal information. The transformation formula used is: transformed variable = (variable + .05)log10. Because most of the analyses involved these transformed variables, results from all procedures are reported for both transformed and non-transformed variables; the reader may Observe that the transformations in no way distorted the relationships among variables, no had signifi- cant effects on initial reliability coefficients. The high inter-coder reliability (see Appendix A) for the informational variables allowed for analysis using the mean of the estimates from the three coders as a single indi- cator of informational variables without distorting the data. For three reasons available information estimates were orig- inally gathered to be used as multiple indicators of the underlying true variables: 1) the measurement procedure and conceptualization of these informational variables is new; 65 without previous work for comparison, multiple-estimates were necessary to establish reliability and allow for their use if reliability was too low for single estimation or averaging; 2) the prOposed models are tested using the LISREL program (JOreskog and Van Thillo, 1972) and optimal use of the analy— tic capabilities Of this program calls for multipbe indicators of unmeasured variables; and 3) multiple indicators allow for overidentification of the models, thus providing "some (of the) excess information (which) may then be used to test the adequacy of the model, since not all sets of empirical data will satisfy the model" (Namboodiri, Carter and Blalock, 1975, p. 448—450, 496-505; also see Stein, 1976). The LISREL analysis is reported in this chapter; however problems en- countered with the procedure made it necessary that a two- stage least square analysis (ZSLS) (Namboodiri, Carter and Blalock, 1975) be performed to clarify findings. The ZSLS, in addition to the analysis of variance of judgmental accur- acy by experimental condition (ANOVA), and g posteriori com- parison of cell means utilizing the Newman-Keuls procedure, are more appropriately handled using the mean estimates of the three coders as the informational variables. High inter- coder reliability (see Appendix A) allows for these procedures with little need to correct for attenuation. Therefore, a comparison of averaged and multiple-indicator handling of descriptive statistics is presented for coder estimates and averaged data (Tables 2, 3, 4 and 5), along with the LISREL 66 .H I o cmm3umo on omumocsuu ma Oflumu o no omcmu maowmmom on» umcu ouoz t. ovmuz mow. ooo.o awn. emm. vmv. m uwooo mam. ooo.o mam. mam. omv. m uwooo mam. ooo.o mam. mam. mmv. a nooou mam. ooo.o mam. omm. mme. Amuoooo m HON cameo «coflumEHOMcH Hmuos\amoum>coz mo ofiumm ooo.omv ooo.om ooo.oov mma.moa mam.nwm m umooo ooo.mmm ooo.oHH ooo.mmv mmm.maa mmm.mm~ m Hmoou ooo.omv ooo.~ma ooo.vmm nmm.moa mam.om~ a uwooo now.nam ooo.vaa www.moe omm.moa mea.amm Amumooo m Eoum Gooey cOHumEHOMcH Hmuoa ooo.oem ooo.o ooo.ovm wem.eoa mnm.mma m umooo ooo.mmv ooo.o ooo.mmv mmm.~ma mnm.moa N nooou ooo.mmm ooo.o ooo.mmm mmo.oma mmo.~ha H Hmooo mmm.mmm ooo.o mmm.mmm one.HHH mem.vma Amuocoo m Eoum cmmEV coflumEHOMcH Hmoum>coz Eseflxmz EdEHcHz mmcmm coHumH>mo oumocmum cams maowaum> mmHQMHum> om8u0mmcmuulcoz How mOHumHumum O>Hudflnomoo .N wanna 67 .OHOOHAmo. + OHQMHHM>V n maomwum> ooEH0mmcmup "we wasEHOM coflumfiu0mmcmne a. ovmuz Hmm.m mom.H mun. mmH. Hmm.N m Hmoou mh>.N avo.m mmn. «ma. mav.m N Hmoou www.m mmo.m ooo. vma. hmv.m H Hmoou NHB.N oeo.m mom. aha. NH¢.N Amhmooo m EOHM COOEV coflumEHOmcH Hmuoe mo coflumEuommcmua moq Hmm.m Hom.HI mmm.m mom.a mmm.H m Hwoou mmm.~ Hom.HI mmm.m vnm.a mov.H N Hmooo hmm.m Hom.HI mam.m mmm.a hN¢.H H HOUOU Nmm.m Hom.al mmm.m vsm.a oav.a Amumooo m EOHM GMOEV coflumEHOMcH Hmonm>coz mo coHumEuOmmcmuB moq EDEAxmz EdEHcHz omcmm cowumfl>wo Unaccepm coo: Odomflum> mmaomaum> omEuowmcmua MOM moaumflpmum O>HuQHMOmOQ .m manna k. 68 .coHuHocoo ODMHHmouddm mop cH mno>Hooouuu06no5m Bonn mucmfioosn HOOHHOO mo ommucmouom mo onoom some n mH momusood fi ovmuz o >H¢.mmH o m Hoooo o th.HmH o m HOUOU o mmm.mMH o H Hmooo new. 0 mmm.NvH o Amnmooo m m0 cameo umHuomcmua mom. omh.>mm mmm.mvH m Hmoou mvm. omh.HmN omh.mvH m Hmooo mom. mmm.mwm oom.hmH H Hocoo nHm. omm. eem.¢m~ mmo.mvH Amumooo m «0 cameo OdmuoHo5¢ mHh. om~.Hmm coo.HmN m Hoooo «we. omn.mmm mmm.mmm N Hoooo mmh. ooo.>mm me.mmm H Hmooo new. mmh. mmm.m>m hmH.H>~ Amuoooo m m0 gooey OmmaomoH> Hoe. om>.mmm mmo.n- m “meow mmm. nmm.Hmm hmH.HmN m Hwooo New. mmm.mvm nHm.wmm H Hoooo mom. ewe. hHm.vvm mmm.mm~ Amumooo m m0 cameo O>HH «momusood coHumEHOwcH Hmuoe :oHumEHOmcH coHumEHomcH "chHuHocoo\mmHomHHm> \Hmono>coz Hmuoa Hmouo>coz coHuHocoo mo mmHomHum> oweu0mmcmuuncoz How memos .q OHoma 69 .coHuHocOo OUMHHQOHQQM on» cH muO>HmomH|u0Onosm Eoum mucwEmosn uoouuoo mo ommucmouom mo muoom some O OH >0mnsoo¢ .OHeoHAmo. + memenm>c mHomHHm> omfiuommcmuu «« ”mH MHDEHOM coHumEHommcmue a. HHOO Hoe oouz mNH.N Hom.HI m umooo vhH.N Hom.HI N Hwooo mvH.N Hom.H| H Hmoou ewe. an.N Hom.H| Amumooo m m0 gooey umHHOmcmHB nov.N me.N m Hmoou mo¢.N va.N N umoou hmv.N omH.N H umooo nHm. eHv.N mmH.N Amumooo m m0 emcee OmmuoHo5¢ omm.N mmm.N m umoou omm.N mov.N N Hoooo mmm.N omv.N H umoou new. mmm.N va.N Amnmooo M NO cmeV OmmuoooH> mom.N mvm.N m umooo mvm.N omm.N N uwooo nmm.N mom.N H umooo mom. nNm.N mom.N Amuoooo m m0 cmoev O>HH «rxomusoog coHumEuOmcH Hobos coHumEu0mcH Hmouo>coz mcoHpHoc00\mOHomHum> mo coHumEuommcmuB 004 m0 coHumEHOmmcmue mOH rcoHuHocou we moHQMHum> ooEHOMOCMHB How memo: .m OHQMB 70 and ZSLS findings. The dependent variable, accuracy of the truth deception attribution, is a dichotomously dummy-coded variable: 1 if the subject—observer accurately indicated that the subject- source was lying or telling the truth, and 0 if the subject— observer inaccurately indicated that the subject-source was lying or telling the truth. Dichotomous dependent variables violate the assumptions of a number Of the analyses performed (ANOVA, LISREL, ZSLS). However, due to the power of these statistical procedures the substance of the results should not be affected. Performance of an arc-sine transformation upon data reduces this problem to some degree. However, it is ques- tionable how such a transformation would result in further distortion of the data. Hocking gt gt. (1976) performed an arc-sine transformation on similarly coded data. Results from the transformed and untransformed data were the same. Hocking gt gt. (1976) chose to use untransformed data due to the distortion problem. In this study an arc-sine trans- formation was not performed. The results from this study are presented in four parts: 1) descriptive statistics, 2) analysis of variance, 3) LISREL, and 4) ZSLS. The first two analyses provide in- formation preliminary to the analysis. Two stage least squares follows the LISREL presentation, since many of the choices involving ZSLS were based on problems with the 71 maximum likelihood program. Descriptive Analysis The presentation and discussion of the descriptive sta- tistics for each variable are divided into two general areas: 1) overall statistics for transformed variables and non- transformed variables, and 2) means of informational vari- ables by condition in terms of mean accuracy by condition. General descriptive statistics. Tables 2 and 3 pre— sent the means, standard deviations, and ranges for all transformed and non-transformed informational estimates, re- spectively. Means for coder estimates and summed variables may reflect the same relationship between informational var- iables. A comparison of Table 2 and Table 3 illustrates no major effect of transformation of variables. Standard devi- ations are stable across coders as well as means, reflecting the high inter-coder reliability reported earlier (see Appendix A). More importantly, however, are the wide ranges which can be observed, especially for the non-transformed variables. Often high inter-coder reliability figures reflect the un- conscious agreement among coders to not estimate "big" num— bers in order to avoid negative reinforcement from the experimenter. The procedure used for training coders in this experiment was extremely vulnerable to this flaw, since esti- mates were highly abstract, i.e., did not involve counting or 72 using a stOpwatch, thus preventing a close empirical check on estimates. Experimenters were aware of this problem of trun- cated estimates producing false reliabilities throughout the training. The broad ranges reported in Table 2 indicate that high reliabilities (see Appendix A) were not a function of truncated estimates on the part of coders. Descriptive statistics by condition. Tables 4 and 5, and Figures 7 and 8, indicate a nonlinear relationship be- tween condition and all exogenous variables. Means for in- formational variables indicate parallel curvilinear relation- ships by condition. However, mean accuracy scores seem to produce a unique u-shaped curve (see Figure 7 and Figure 8). Though little can be said based on descriptive statistics concerning the significance of the differences between con- ditions, clearly the relationships between condition and information, and condition and accuracy, are not linear or parallel. The means by condition indicate a rejection of both hypotheses prOposed in Chapter I. However, further analysis is necessary to determine what significant differences exist among conditions. The question also arises as to whether the information curve and the ac- curacy curve can be predicted on the basis of conditions. In other words, the overall model still needs to be tested, even though the descriptive statistics indicate that the re- lationship between the variables is nonlinear. Anatysis of variance. One of the major aims of the present research was to examine the ability to make accurate 4001 375~ 350- 325- 200s 275‘ 2504 225i 2004 175- 1504 1254 100- 75- 50. 25-1 Ratio x 100 Accuracy x 100 Total Information Nonverbal Information Curve 73 Figure 7. A * A * A 0 [j o D D * o Videotape Audiotape Transcript » D [j 0 Mean Curves of Nonverbal Information, Total Information, Nonverbal/Total Information and Accuracy of Non—trans- formed Variables by Channel Condition 74 3.00- 2.75~ 2.50- 2.25‘ 2.00d 1.75j 1.50~ 1.25d 1.00~ .75N .soH ‘3 E] -1.00~ Live Videotape Audiotape Transcript Nonverbal * Total A Accuracy 0 Figure 8. Mean Curves of Nonverbal Information, Total Information, and Accuracy for Transformed Variables by Channel Condition * Transformation formula is: transformed variable = (variable + .05)1og10 75 attributions of veracity under "live," videotape, audio- tape, and transcript conditions. While the descriptive sta- tistics indicate a nonlinear relationship between accuracy and condition, it is not Obvious whether conditions differ significantly in the way they affect accuracy alone. To shed further light on this issue, an analysis of variance of accuracy scores was conducted. The results (see Table 6) were significant at the .05 level. 5 posteriori comparisons of cell means utilizing the Newman-Keuls procedure (Winer, 1971) indicated (see Table 7) that subject-receivers in the live condition are signifi- cantly more accurate in attributing truthfulness or decep- tion than Observers in the audiotape condition (p>.05). NO other comparisons are significant. Identical procedures were also conducted to see to what degree channel conditions significantly differ in terms of available nonverbal information, available total informa- tion, and the ratio of available nonverbal to available total information. The results of these analyses (see Table 8, 9 and 10) are significant at the .05 level. 5 posteriori comparisons of cell means utilizing the Newman-Keuls procedure (Winer, 1971) indicate that all four channel conditions differ significantly in terms of the available information measures (see Tables 11, 12 and 13). 76 Table 6. Analysis of Variance of Accuracy by Channel Condition Source Sum of Squares df MS F p Total 59.496 239 —- --- --- Between 1.913 3 .638 2.613 .05 Within 57.583 236 .244 --- —-- Table 7. Individual Comparisons of Channel Condition Accuracy Means* Channel Condition Mean Live .567b Video .467a Audio .316c Transcript .467a * Means having different subscripts differ significantly at the .05 level of confidence. The higher the mean, the greater the judgment accuracy. Table 8. Analysis of Variance of Available Total Information by Channel Condition Source Sum of Squares df MS F p Total 2654564.340 239 -- --- --— Between 19l387l.238 3 637957.080 203.266 <0.001 Within 740693.102 236 3138.530 --- —-- 77 Table 9.* Analysis of Variance of Available Nonverbal Information by Channel Condition Source Sum of Squares df MS F p Total 3142674.896 239 -- -__ _-_ Between 2648214.016 3 882738.005 421.320 <0.001- Within 494460.880 236 2095.173 --— --- * Above calculations are non—transformed. See Appendix H for ANOVAs for Transformed Variables. Table 10.* Analysis of Variance of the Ratio of Available Nonverbal to Available Total Information by Channel Condition Source Sum of Squares df MS F p Total 20.505 239 —- --- ——— Between 20.2894 3 6.763 7415.560 <0.001 Within .2152 236 .001 --— —-- * Above calculations are non-transformed. See Appendix H for ANOVAs for Transformed Variables. Table ll.** Individual Comparisons Of Channel Condition Available Total Information Means* Channel Condition Mean Live 372.333e Videotape 344.917f Audiotape 264.944g Transcript 142.389h * Means having different subscripts differ significantly at the .05 level of confidence. ** Above calculations are on non-transformed variables. See Appendix H for ANOVAs for Transformed Variables. 78 Table 12.** Individual Comparisons of Channel Condition Available Nonverbal Information Means* Channel Condition Mean Live 238.389i Videotape 271.167j Audiotape 149.028k Transcript 0.0001 * Means having different subscripts differ significantly at the .05 level of confidence. ** Above calculations are on non-transformed variables. See Appendix H for ANOVAs for Transformed Variables. Table 13.** Individual Comparisons of Channel Condition Ratio of Available Nonverbal Information to Available Total Information Means* Channel Condition Mean Live .6870m Videotape .7232n Audiotape .5599O Transcript .0000p * Means having different subscripts differ significantly at the .05 level of confidence. ** Above calculations are on non-transformed variables. See Appendix H for ANOVAs for Transformed Variables. LISREL: Testing the total models. LISREL, a program which estimates a linear structural equation system involving multiple indicators of unmeasured variables, was used to solve for the models presented in Figures 5 and 6. LISREL 79 allows for the use of the coder estimates as multiple indi— cators of underlying available information measures, for both errors in equations and Observed variables, and produces estimates of the disturbance variance—covariance matrix, measurement error variances, and unknown coefficients in the structural equations. Overall, it uses all the data to pro- duce maximum likelihood estimates of parameters and then tests the goodness of fit of the whole system of variables at one time. Other available methodologies (ordinary least squares, two stage least squares, etc.) do not yield as much informa- tion or allow for testing of the goodness of fit of the model as a whole. Given that the descriptive statistics and ANOVAs indicate a rejection of both hypotheses, but a strong rela- tionship between available information measures and condition channel, the added information provided by LISREL is crucial to understanding where and why the proposed models are inadequate. The model in Figure 5 was tested using LISREL with non— transformed variables; Model I in Figure 9 contains the solu- tion of this test (see Table 14 for a Glossary of Variables in LISREL and ZSLS Models). The model in Figure 6 was tested with LISREL once using non-transformed data (Figure 10, Model II) and once using transformed data (Figure 11, Model III). Model I was not tested with transformed data, since the logarithmic transformation changes the ratio relationship Table 14. Key for Presentation of All LISREL and 2818 Nbdels £1 = Conrmmication Channel (True 64 Measurement Error of y 4 Variable) 6 Measurement Error of y :2 = Communication Channel (True 5 Me t E f 5 Variable) E6 asuremen rror o y6 53 = Communication Channel (True 67 masurement Error Of y7 Variable) 58 masurement Error of y8 n1 = Available Nonverbal Infor— 69 Measurement Error of y9 mation (True Variable) ”2 = Available 'Ibtal Informa- €10: Measurement Error of y10 tion (True Variable) 61 Measurement Error of XI n3 = Available Nonverbal/Avail- (52 Nbasurenent Error of x2 able Total Information 6 Me t E f x (True Variable) 3 asuremen rro r O 3 n 4 = Ability to Attribute Truth Y1 fmari‘léfizréfif’eli‘gfi Of or Deception (True Variable) Information C1 = Disturbance Term for n1 Y2 Mean Coder Estimate of C2 = Disturbance Term for n2 Available Total Informa- ;3 = Disturbance Term for n3 A tion . p ' Wwemmn. Y1 sfifmxfi’i. x = Live Cmdition Nonverbal Information x = Videotape Condition '92 Predicted Mean Coder _. udio - - Estimate of Available X3 — A tape Condition Total Information y1 = First Coder's Estimate of n1 y2 = Second Coder's Estimate of n1 y3 = Third Coder's Estimate of NI y4 = First Coder's Estimate of n2 y5 = Second Coder's Estimate Of n2 y6 = Third Coder's Estimate of Hz y7 = First Coder's Estimate of U3 y8 = Second Coder's Estimate Of n3 y9 = Third Coder's Estimate of U3 Y10= Subject—receiver Judgmental Accuracy 6 = Measurement Error of y1 £2 = Nbasurenent Error of y2 s3 = Nbasurenent Error Of y3 81 .H fig mom mooHuumE Hmsuom mom .A V mOmeCOHmQ CH oouuommn who mcoflmHmuuoo H2 R. mmHanHmS H6953 mcmuulcoz one :03 nsHom oONHoudqum "moHanHmS msocwmoofim OHMHUOEHOHEH mm cOHumEHOm .5 H58. 8 H3652 mo 03mm mfi ent . H309 . Hmnegeoz 63332 8N u 2 fit. Sansone co Trend 8388 no Bowman 65. .L deg .Hmmfl 3 SEE mm u be Hood v e an me be 32.3% n x T..- .7..- e... N as as e» m H QZH / Ho . H Wm: S w 9 H M Hon: 5.0: m of A an . 4k: 5.: «N NC v: I) .I a; 82 ovm H Z cm H MC 306 v m HHem.mmHH u mx v N QZH OH 0 0H C‘QNH'Q) O QM!) .H gamma own 8339: Hmflbbm mom .A v mommfiaua ca. ©698me one mcoHumHmuuoo H2 3.. «mOHQMHHm> BEOMOCMHTGOZ one coHusHOm ooNHoHdoqmum "mega-Hume, meoemmoerm mumemeumueH mm eoHumeuoeeH Henge but Hmbum>coz memHHmem ruH3,NomAsuea_eo eoHuHeeoo Homemru mo muoemmm ere .HH ammo: seamen “OH benefit mu mm m m mow: momm e» s e» m .2: m.mm Homm OHOOHAmo. + OHbmHHnBv u OHQOHHQV ogommsmb "mH mHsEHOM coHuEEowmcmne .H Hoodoo/H mow mmoHuumE ngom mom .A V mwmgcouwm CH oouhommn mum chHumHouHOo H2 eem n z om n we Hoo.o v m «mbemHnm> emeHOHmcmne new eposHom eeNHeAmeembm "mmemHHm> msocomoocm 338535 mm :oHumEHOmcH Hmpoe one Hmougcoz wHomHHm>< Hme.OHe u Nx ruHe somueou< co epoHbcoo Hmeemru Lo mbommmm ere .HHH ammo: ammmHa "HH musmHe NHQZH IIIXToe 84 to a subtractive relation, i.e. the hypothesized relation- ship is destroyed in the data. For all LISREL testing the raw data served as input, and a variance-covariance matrix was calculated upon which the remaining procedures were per— formed (for details concerning LISREL analyses, see Appendix I). Due to a problem with LISREL an exact solution is not reported for any of the models tested; all estimated para— meters are approximate and therefore the chi-square does not represent the goodness of fit of the final models. In test- ing the models, the LISREL program consistently produced IND # 0. Jereskog and Van Thillo (1972) indicate the follow- ing concerning this problem: If IND is 1, 2, or 3, "serious problems" have been encountered and the minimization of the function cannot continue. One reason for this may be erroneous input data. Another reason may be that insufficient arithmetic precision is used (in the program) (p. 33). The IND obtained for the reported models are contained in the corresponding figures. INDs of 4 and 5 indicate that the program has run out of the allocated time or has com- pleted the maximum number of minimizations allowed in one cycle. The models reported with these INDs (Figures 9 and 10) were resubmitted; the program yielded IND=2, but no new solution. Given the nature of the solutions, it was questionable whether the reported parameters were reliable. All solu— tions indicate the model did not fit (p < 0.001). However, 85 due to the problems encountered, it was decided to solve for the best Of the three models (Figure 11) using ZSLS. Model III was chosen since it was the only solution obtained by pushing the LISREL beyond its time and/or minimization con- straints, and this model had the lowest chi-square value (610.91), i.e., was the most likely model to fit. The low correlations between the residuals in Model III (-.004, .008, .009) suggest that any failure in the fit of the model is not due to any missing relevant variable, but rather to inadequate predictive power among the variables in the model. A small R2 in a ZSLS solution would confirm this interpretation of the LISREL solution. The paths from the ratio Of available nonverbal to total information and available nonverbal information to ac- curacy support the directional indications of Hypotheses 1 and 2 (Model I: NV + ACC = -.11, NV/TOT + ACC = .99; Model II: NV + ACC = -.97; Model III: NV + ACC = -.51). However, little can be said about the highly unstable paths in a LISREL solution which is a poor fit; at best we would hope that the ZSLS yields path coefficients of similar magnitude and direc— tion to LISREL results. ZSLS: Testing of the total model. The first stage of the ZSLS procedure consists of ordinary least squares regres- sion. In this case, two separate equations had to be esti— mated; the first to determine the path coefficients between available nonverbal information and the exogenous variables, 86 and the second to determine the paths between available total information and the exogenous, as well as available nonverbal information variables. Table 15 illustrates the results Obtained from estima- tion of the first stage, first equation. It was assumed earlier that variations in the communication channel would result in covariations in coders' perceptions of the amount of available nonverbal information. The results strongly support this assumption, with variations in the channel ac- counting for .997 percent of the variance in perceived avail— able nonverbal information. These results also serve as an indirect check of the success of the experimental procedure for manipulating available information (i.e., in terms of communication channel). It was also assumed that channel variations, as well as available nonverbal information, would result in variations in coders' perceptions of available total information. Table 16 illustrates the results pertaining to this assump- tion. Again, the results are overwhelmingly supportive (R2 = .969), and also serve as an indirect check of the ex- perimental procedure for controlling the availability of in— formation, a crucial variable in the present study. The results pertaining to channel variation and informa- tion availability are fairly straightforward and not particu— larly surprising. Of greater importance are the results pertaining to information availability as a predictor of the 87 (SEEso mo conmsomHo o How .mmINMH .mm .Amanv xoonHm ocm umuumo .HHHooooEmz mom k. .mummmon MOM «H mHome Op ummom .mHmemcm conmoumwu oHomHum> k. .OHmOHAmo. + OHomHum>v u OHQMHHM> omEuowmcmHu ”mH mHsEuom coHumEHmecmHB ¥ oem u 2 we» o.meHHm emm\m be.m mx we» ~.ommmm emm\m Ne.m mx we» ~.Hemem em~\m ee.m Hx mo.vm M mm m memHnm> mxee.m + Nxmk.m + ere.m + om.Hu u He ham. mam. mm» e.emmem m + Husxauxb + mxmb + Nxmb + Hbe + bob + m u H» m m mo.vm m «erooz Housuosuum OHomHum> wfifiao m oHeHuHez «Hoooz moumsqm unmoq mmmumsoze mo coHumsqm umHHm .mmmgm bmHHm .mH mHomB 88 .oamoHAmo. + manmflum>v u mHQMflum> UmEH0mmcmuu "ma mH5Eu0m coflumEu0mmcmuB k. ovm u 2 mm» h¢.bvaa mmm\v vmm. H» mm» H~.mmm mm~\v mm~.- mx mm» mH.nmm mmmxv mo~.- mx mmm no.mnm mmm\v vom.u Hx mo.vm m mm m manmflum> H»¢N.m . mxmm.m . mxmm.~ . Hxvm.m . m~.m n my mom. «mm. mm» mm.mwma m + H-xxaumn + van + mxmn + Nxmn + son + m u my m m mo.va m Hmuoz Hausuosuum manmflum> wasso m mamfluasz «H6602 coflmmmummm mmumsvm ummmq mmmumno3e mo coflumsvm Ucoomm ~mmmum umuflm .wH wanna 89 ability to make accurate attributions of veracity. By using the two-stage least squares procedure, the endogenous vari- ables in the structural model could be "purified" in such a way that their correlations with disturbance terms were elim- inated. Thus, given that we had little measurement or sam- pling error, we should get a fairly accurate estimate of the relation between information availability and subject-receiver accuracy. Table 17 illustrates the results obtained from this pro- cedure. Examination of these results suggests that varia- tions in availability of informational cues as a function of communication channel do not predict judgmental accuracy very well. The multiple R was only .064, accounting for less than one percent of the variance in accuracy scores. However, it should be noted that the path from available nonverbal infor- mation to accuracy is both stable and negative as hypothesized in the first hypothesis. Compairson of LISREL and 2 SLS: Table 18 compares the results of the 2 SLS with the LISREL results for Model III. All paths are of similar magnitude and in the same direction (positive or negative). The goodness of fit test of LISREL does not support the model (x2 = 610.091, p < 0.001). Exam- ination of the three ZSLS equations indicates that the first stage of the model, which explains over 99.7% of the variance in the first stage, has relatively stable paths (low standard errors) and yields a significant F, is not the source of the 9O .oamoHAmo. + mHQmHHM>V u manmflum> meHOMmcmuu umH MHSEpOM :oflumEu0mmcmuB .1 ovm u 2 oc omo. em~\m omm.u mm on mmm. nm~\m ooo.- Hm mo.vm M mm m moanmflum> mmomm. + Hmooo. . mmo.- n m» ooo. ooo. on How. m + mmmo + Hosp + o n m» m m mo.vm m Hopoz amusuosupm m mamfluasz «Hobo: Gmemmummm mmumsvm ummmq mmmumnoze mo coflumsom phase .mmmum pcoomm .na magma 91 Table 18. Comparison of Standardized Results of LISREL and ZSLS for Nbdel III Standard Error Items LISREL 2815 of 2818 Beta lst Equation Paths: ‘vaE—4»N0nverbal Infbrmation 1.014 1.010 .002 Videotape + NCnverbal Information 1.032 1.025 .002 Audiotape + NOnverbal Information .961 .955 .002 R2 —- .997 -- F -- 27956.603 -- Significance level -- p<0.05 2nd Equation Paths: lave + Total Information -8.200 -6.407 .009 Videotape + Total Information -8.257 -6.438 .009 Audiotape + TOtal Information -7.978 -6.273 .008 NOnverbal Infonmation - Tbtal Information 9.020 7.254 .002 R? -- .970 -- F -— 1883.917 -- Significance level -- p<0.05 3rd Equation Paths: NOnverbal Infbrmation + Accuracy -.507 —.128 .004 Tbtal Information + Accuracy .551 .138 .399 R2 -- .004 —- F -- .48176 -- Significance level —- p<0.05 MODEL III: x2 610.091 -- —- Probability p<0.001 -- —— 92 poor fit of the data to the model. Examination of the third equation indicates that the path from total available information (§2) to accuracy (§3) is unstable (large stan- dard error). The second stage of the ZSLS, which explains less than 1% of the variance in accuracy, suggests the reason for poor fit of the maximum likelihood estimate of the model. As mentioned earlier, the direction of the available nonverbal information to accuracy path supports Hypothesis 1. CHAPTER IV DISCUSSION The purpose of this chapter is to interpret the results of the experiment, to discuss practical and theoretical im— plications, to evaluate procedures, and to suggest future research. Summary of Findings Even though channel conditions systematically vary in terms of information, the differences have little effect on accuracy. In general, the results of this experiment do not support the hypotheses proposed in Chapter I: H1: As transmission channel increases the amount of nonverbal information available to a receiver, the accuracy of a truth/deception attribution concerning an unfamiliar source will decrease. H2: As transmission channel causes an increase in the ratio of nonverbal to total information available to a receiver, the accuracy of the truth/deception attribution concerning an un— familiar source will increase. Channel conditions with significantly differing amounts of available nonverbal information and the ratio of available nonverbal to total information-—videotape and transcript-- yielded comparable levels of accuracy. 93 Table 19. 94 Summary of'Analyses' Relation to Hypotheses H1: As transmission channel causes H an increase in the ratio of non- verbal to total information available to a receiver, the accuracy of the truth/deception attribution concerning an un- familar source will decrease 2: As transnussion channel causes an increase in the ratio of non- verbal to total information available to a receiver, the accuracy of the truth/deception attribution concerning an un- fanular source wd11.increase Analysis Relationship to H1 Relationship to H2 Descriptive Statistics General Table 2 Table 3 By Condition Table 4 Table 5 Figure 7 Figure 8 Analysis of variance Table 6 and 7 Table 8 and 11 Table 9 and 12 Table 10 and 13 LISREL Figure 9: Nbdel I Figure 10: Mbdel II Figure 11. Model III ZSLS Table 15 Table 16 Table 17 None NOne Not supported Not supported Not supported Not supported Not supported None NOne NOne Generally inconclusive; direction of NV'* ACC supports Hl Generally inconclusive; direction of NV'a'ACC path supports Hl Generally inconclusive; direction of NV’a'ACC path supports H1 None None Direction of NV 4*ACC path supportive of H and stable; variance explained is not significant None NOne Not supported Not supported Not supported Not supported Not supported None None NOne Generally inconclusive; direction of NVVTOT + ACC supports H2 None None None None None 95 In Chapter I, when presenting the two hypotheses, three theoretical perspectives concerning information processing were discussed as possible explanations for the truth/decep— tion attribution process: 1) a traditional information util- ity hypotheses, 2) the distraction hypotheses, and 3) the overload hypothesis. The former predicts an increase in ac- curacy as more information becomes available, while the latter two predict a drop in accuracy as information increases due to two different processes. All three of these perspectives were not supported by these results. The high accuracy score in the transcript condition (46.7%) is counter to the information utility hypothesis and rules out any linear relationship between available nonverbal and/or total information and the ability of untrained observers to detect deception on the part of strangers. The compar- atively high mean accuracy observed in the transcript condi- tion suggests that an attribute of that channel, distinct from type and amount of information, may provide an explana- tion. Amount of time an observer has to examine the stimulus and the ability of the observer to reexamine the stimulus may be two such attributes of transcripts of interest in future research. Both these sets of findings are supported by past research: (1) the low accuracy scores found by Maier and Thurber (1968), 58.3%, and Hocking et al. (1976), 58.5%, in the conditions where information was most abundant; (2) the conclusion of Maier and Janzen (1967) that judgments 96 of accuracy "seemed to be based upon impressions rather than logic" (p. 105); and (3) the high accuracy scores found by Maier and Thurber (1968) and Hocking gt al.(l976) in tran- script conditions. The distraction and the overload hypothesis are called into question by the high accuracy scores in the live and the videotape conditions (56.7%, 46.7%). These hypotheses would predict that channel conditions which provided the most available information (videotape) and offered the great- est amount of distracting external cues (live) would yield the lowest accuracy scores. Again, the comparable accuracy scores of the videotape and live channel condition, which represent the most extreme differences in amounts of all types of available information, suggests that these informa- tional hypotheses offer inapprOpriate explanations for the truth/deception attribution. Implications Upon close examination of the results a series of find- ings calls into question any causal relationship between amount of information and accuracy of attributions. First, coders consistently judged videotape as providing more infor- mation than live presentations (see Tables 3 and 4). Recall- ing that coders were measuring the amount of information available for them to examine, this finding seems somewhat strange. Videotape limits the amount of space the receiver 97 can visualize at one time, and often causes a loss of detail due to poor pictures, glare, and camera shot. Likewise, on videotape, the source is smaller. Could it be that in a live situation, the receiver is distracted by external stim- uli which have no informational value, thus missing many of the informational cues given by the source? Second, even though the live condition was judged to provide less infor- mation than the videotape, accuracy in the former condition was higher (M = .467, videotape .567). If the live Mlive = condition suffers from distraction, does this contribute to accuracy? Finally, what unique characteristics of the audio- tape condition produce such a significantly lower accuracy score? Overall these disjointed findings suggest that other variables beside information affect accuracy of judgments; it seems that a big difference among channels makes almost no difference in terms of accuracy of the truth/deception attribution. Nonverbal communication has been carefully examined over the years. The importance of the nonverbal component of messages has been emphasized in both scholarly journals and popular paperbacks. Mehrabian (1971b, p. 43) has even said that 93% of the impact of a message comes from nonverbal communication. A long list of studies (see Table 1) dating back to 1926 suggests that researchers feel nonverbal infor- mation may be an important influence in the truth/deception attribution process. These findings suggest that if nonverbal 98 communication has an influence on detecting deception, it is not in terms of accuracy. Hocking (1976) comes to a similar conclusion when he states, "the results of the present experi— ment suggest that for accurately detecting deceptive commun- ication, the nonverbal component is less important than the verbal component" (p. 120). In fact, considering the relatively low accuracy scores reported in all conditions--56.7% for the live, 46.7% for the videotape and transcript, and 31.6% for the audiotape-- it is highly questionable whether untrained observers can ac- curately detect deception on the part of strangers. None of the mean accuracy scores were much higher than the 50% cri- terion researchers have defined as chance accuracy in these studies. It should be noted that this criterion is arbitrary in the sense that all people may not expect sources to be lying 50% of the time. However, in the present study subject- receivers knew that there was a 50/50 chance that each of the subject-sources was lying. A few studies have obtained accuracy scores significantly above the 50% criterion. Specifically these were in Maier and Thurber's (1968) audio-only and transcript conditions, Ekman and Friesen's (1974) body-only condition, and Hocking gt gt.'s (1976) audio-only and transcript conditions. How- ever, the two types of deception-inducing procedures used in these studies can be criticized for problems which inflate accuracy scores. Maier and Thurber (1968) had students role- 99 play deceivers. When role-playing, lying behavior is not inconsistent with the matters of known fact to the subject; (s)he acts as he believes someone who is lying acts. When playing the part of a liar the tendency is to emphasize "lying behavior." The subject has no real motivation to look honest, as in the normal lying situation; rather (s)he wants to look like a liar if (s)he is to do an effective job. If an individual role-playing a liar looked honest, would anyone think (s)he was playing the role well? Such a tech- nique, at worst, inflates the accuracy scores of observers, while at best has been seriously questioned as a research technique, since no one seems to know whether role—players "know" how real life liars behave (Freedman, 1969). In both Ekman and Friesen (1974) and Hocking gt gt. (1976) individuals always lied while observing a very un- pleasant stimulus and told the truth while viewing a pleasant stimulus; this systematically increased the cues of discom- fort and arousal coming from the group of liars. These cues of arousal would be attributed by observers to lying rather than any extraneous stimulus, since that was the explanation offered by the social context in which observers made their attributions, i.e., a detecting deception experiment (of. Schachter and Singer, 1962). The arousal cues stemming from the unpleasant stimulus, thus, would have made it easier for observers to identify liars. 100 The deception—inducing procedure used in this study was chosen to overcome some of the criticism of past deception— inducing techniques. The authors realized that a more gen- eralizable deception-inducing technique might logically produce lower accuracy scores than role-playing or the tech- nique involving the viewing of an unpleasant stimulus; the resultant accuracy scores (56.7%, 46.7%, 46.7%, 31.6%) were lower than, but we believe more generalizable than, past scores. Given the criticism of past deception-inducing tech- niques, the generally low scores found under these past techniques, and the low scores found in the present study, the claim that untrained observers can accurately detect de- ception on the part of strangers is highly questionable. Problems With the Study As with all research, this study has a number of prob- 1ems which must be discussed and examined in terms of their implications. Most importantly they should be viewed as an argument for replication and refinement before drawing any final conclusions. Two basic principles of scientific experimentation are random assignment of subjects to conditions and random sam- pling of subjects from the population; the former safeguards internal validity, the latter external validity. Unfortu- nately, neither one of these principles was strickly adhered to in this research. 101 All subject-receivers, subject—sources, and coders were undergraduate students in the Department of Communication at Michigan State University. Individuals signed up to partici- pate in the research on a voluntary basis in exchange for extra credit in an introductory communication course. Statistically these results are only generalizable to a student population. The sample used is not representative of the general population and may possess unique characteristics which affected the outcome of the study. Students may be more/less suspicious of their peers than the average individ- ual engaged in conversation. This would affect the degree to which they have attempted to detect deception in the past, and therefore their present accuracy scores. This study should be replicated among a more heterogenous sample and measures should be taken of general levels of suspicion. Every attempt was made to randomly assign individuals to conditions; however, due to the nature of volunteer student subjects, individuals often had to switched from one con- dition to another. These minor adjustments, however, should not have had any major impact on the results of the study. Subject-observers experienced interviews over a three-week period. Therefore it is possible that individuals who had already participated had discussions with others who had not yet participated. The knowledge gained through these conversations could have influenced future subject-receivers' performances. However, since the cells were assigned to days 102 randomly, this should not have systematically affected results. Due to the constraints of the manipulation, the live con- dition was run under different circumstances and in a differ- ent laboratory setting than the other three conditions. This might cause the subject-receivers in the live condition to behave differently for reasons separate from the nature of the channel condition itself. However, given the overall pattern of the results, this change in environment does not seem to have had a significant effect. In this study a number of participants were engaging in various differing roles simultaneously. Although random as- signment was adhered to in all conditions, a number of system- atic interactions of individuals roles, etc. could have con- taminated results. Ideally correlations between all possible contaminating variables should be reported. However, given the precautions taken and the intricacies of such analyses, and the degree to which past research supports conclusions these correlations do not appear. Replication under less complex circumstances is thekxxflzconfirmatory evidence for these findings. Overall, the problems encountered in this experiment do not seem to be of the type which would casue a major differ- ence in the outcome. However, replication using a random sample from a more heterogenous population, greater controls, and strict random assignment would add considerable confidence to the findings reported here. 103 Future Research These findings suggest a number of areas of future re— search which may prove fruitful. All research in the area of detection of deception thus far has examined the process in terms of stranger dyad. Perhaps we should investigate de— ception detection in established relational settings. Miller and Steinberg (1975) suggest that when an individual engages in interpersonal communication the accuracy of predictions about the other increases. Accuracy increases because inter— personal communication involves knowledge on the part of the observer concerning the idiosyncrasies of the other. Pre- diction dominated by "stimulus discrimination" based on this knowledge should be more accurate than "stimulus generaliza- tion" based on stereotypes, which characterizes noninter- personal communication (Miller and Steinberg, 1975). Miller W and Steinberg's conception of interpersonal communication 3 would predict higher accuracy on the part of observers who communicate interpersonally with the source, due to the in- creased knowledge those observers have concerning the source's) lying and "truthing" behavior. Examination of accuracy in 1 detection of deception between source and receivers who have interpersonal relationships may prove fruitful in terms of the work of Miller and Steinberg. One way to test the degree to which type of relationship affects the accuracy of the truth deception attribution is to do a field study where members of dyads of various degrees of familiarity attempt to tell if the other is lying or 104 telling the truth. Various levels of the relational variable could be operationalized through using strangers, couples who have been dating about a month, engaged couples, and couples married over 10 years. A parallel laboratory experiment-- same manipulation, confederate, experimenter, and question- naire--could be done recontrolling for the number of times subject-observers would be allowed to view interviews of strangers; this gives researchers information as to whether 3 amount of time of contact, alone, or some other qualitative} component of relationship contributed to accuracy. Hocking (1976) also suggests that lying behavior may not be the same across individuals, but rather is distinguishable from "truthing" behavior only within individuals, based on differences between each individual's own lying and "truthing" behavior. In that case, detailed knowledge available to indi- viduals in an interpersonal relationship as to the "truthing" behaviors of the source would be necessary in order to notice deviations. Possible research in this direction calls for careful cue analysis of videotapes of the samples of the same: individual's lying and truthing behavior. Hocking's (1976) ) hypothesis that lying behavior is a deviation from the indi- E vidual's idosyncratic truthing behavior could be examined by comparing the cue analysis of lying and truthing segments within each subject, rather than across subjects. Though nonverbal behavior does not seem to be related to the accuracy of the truth/deception attribution, nonverbal 105 stereotypes may produce inaccuracy in the attribution. People may develop stereotypic conceptions of how certain nonverbal behavior indicates deception. In other words, based on non- verbal behavior, people may think they are being lied to even though they are not. Knowledge as to the stereotypes people have of liars could be obtained by comparing the cue analyses of segments observers judged as lying with segments observers judged as truthing. APPENDIX A Inter-coder Reliability for Non-transformed and Transformed Variables APPENDIX A Table A1 Inter—coder Reliability for Non-transformed and Transformed Variables* Unstandardized Relia— Standardized Relia- Variables bility Coefficient bility Coefficient Nonverbal Information .98157 .98302 Total Information .96285 . .96032 Nonverbal Information/ Total Information .99287 .99287 Log Transformation of Nonverbal Information .9995? .99958 Log Transformation of Total InfOrmation .96780 .96781 N = 240 Coders = 3 "Cronbach's Alpha p < .05. Transformation formula is: transformed variable = (variable +.05)loglo. 106 APPENDIX B Instructions and Questionnaire for Standard Interval Pretest APPENDIX B Instructions and Questionnaire for Standard Interval Pretest Michigan State University Department of Communication Winter 1976 We all have heard the saying that "How you say something is just as important as what you say." Over the years social scientists have been interested in what are the sources of information people use when they communicate. As you are probably aware, NONVERBAL behaviors (the look on your face, the way you move your hands and legs, how loud and fast you speak, whether you look the other person in the eye, etc.) often express a great deal about the way we feel, as well as what our words are intend- ed to mean, when we talk to others. In fact, the popularity of books like Julius Fast's Body_Language shows that the general public wants in— formation on tgg_and gtgg_"actions speak louder than words." You are participating in part of an ongoing research project being conducted by the Department of Communication here at Michigan State University, which deals with NONVBRBAL communication. We want you to help us find out how much people use NONVERBAL behaviors of others, in relation to the TOTAL amount of information the others provide, in order to interpret the others' messages. We need to know how much of the TOTAL INFORMATION you get when you watch someone speak, comes from their NONVERBAL BEHAVIORS (facial expression, eye contact, nodding hand and body movement, posture, pausing, the ums and ahs, anything besides their words). In order to do this, we will show you a series of short videotaped 107 108 interviews with students. You have to tell us two things about these interviews: (1) HOW MUCH NONVERBAL INFORMATION did the interview pro— vide? and (2) HOW MUCH TOTAL INFORMATION did the interview provide? HERE IS HOW WE WANT YOU TO GIVE US THE ABOVE INFORMATION: We measure the amount of NONVERBAL INFORMATION provided by a person in "NVs." An "NV" is a conceptual quantity, much as an ounce is a unit of physical quantity or weight. The greater the amount of nonverbal in- formation present in the interview, the greater the number of "NVs" you should assign it. Similarly, we measure the TOTAL AMOUNT OF INFORMATION provided by a person in an interview in "TOTs"; a "TOT" is like an "NV"--it is a unit of conceptual quantity, much as an ounce is a unit of physical quantity or weight. The greater the amount of total information in the interview, the greater the number of "TOTs" you will want to assign to that inter— view. The first interview you will see will serve as the basis for all other judgements about information that you make; 80 WATCH THE FIRST IN- TERVIEW CAREFULLY. You will see the first interview three times. We are telling you that there are 150 TOTs (units of total information) in the first interview. Now look at the first interview and think of these num— bers: TOTs = 150, NVs = 100. WATCH SCREEN After you see this interview a couple more times we are going to ask you to use it to estimate the amount and kinds of information in other inter- views. So when you see interview #2 you will be answering the following 109 two questions: (1) If interview #1 had 150 TOTs, how many TOTs are there in interview #2? (2) If interview #1 had 100 NVs, how many NVs are there in interview #2? Now look at interview #1 two more times. Remember TOTS (total informa- tion) = 150. NVs (nonverbal information) = 100. WATCH SCREEN Now here is interview #2. Try to estimate the following: NVs = TOTS = If you think there was twice as much total information in #2 as in #1 you should have written TOTs = 300. If you think there was 1/2 as much non- verbal information, you should have written NVs = 50. Now we are ready to start. Are there any questions? You will see the first interview on the screen to your right before every other esti- mate you make. REMEMBER the first interview TOTs = 150 and NVs = 100. Upon these numbers you base all other estimates. When you are finished, we will collect your questionnaires, and you are free to ask any questions you like. We appreciate your help and hope it will contribute to your knowledge of communication. WORK FAST. DO NOT DISCUSS YOUR ANSWERS WITH NEIGHBORS. DO NOT GO BACK AND CHANGE AN- SWERS. WE ARE INTERESTED IN YOUR SPONTANEOUS PERCEPTIONS AND JUDGEMENTS. 110 3A INTERVIEW #3 TOTs = NVs INTERVIEW #h TOTs = NVs INTERVIEW #5 TOTs = NVs INTERVIEW #6 TOTs = NVs INTERVIEW #7 TOTs = NVs INTERVIEW #8 TOTs = NVs INTERVIEW #9 TOTs = NVs INTERVIEW #10 TOTs = NVs INTERVIEW #11 TOTS = NVs INTERVIEW #12 TOTs = NVs INTERVIEW #13 TOTs = NVs INTERVIEW #Iu TOTs = NVs INTERVIEW #15 TOTs = NVs INTERVIEW #16 TOTs = NVs INTERVIEW #17 TOTs = NVs INTERVIEW #3 lll BB NVs = TOTs = INTERVIEW #u NVs = TOTs = INTERVIEW #5 NVs = TOTs = INTERVIEw'#6 NVs = TOTs = INTERVIEW #7 NVs = TOTs = INTERVIEW #8 NVs = TOTs = INTERVIEW #9 NVs = TOTs = INTERVIEW #10 NVs = TOTs = INTERVIEWT#11 NVs = TOTs = INTERVIEWT#12 NVs = TOTs = INTERVIEW #13 NVs = TOTs = INTERVIEW #14 NVs = TOTs = INTERVIEW #15 NVs = TOTs = INTERVIEW—#16 NVs = TOTs = INTERVIEW #17 NVs = TOTs = APPENDIX C Questionnaire for Coder Information Estimates APPENDIX C Questionnaire for Coder Information Estimates Michigan State University columns Department of Communication 1-u Project Number: OM76 S Coder Number: 6 Coder Group: 1__) 2__) 3__) u___ 7 Condition: Live (1) Video (2) Audio (3) Transcript (H) ———- 8 Sex of Coder: (O)ma1e (l)female 9-10 lst Participant Viewed: 11 Sex of lst Participant: (0)ma1e (l)female 12-15 NVs: ____________ 16-19 TOTs: .____.______ 20—21 2nd Participant Viewed: 22 Sex of 2nd Participant: (0)male (l)female 23-26 NVs: ___________. 27-30 TOTs: ____________ 31—32 3rd Participant Viewed: 33 Sex of 3rd Participant: (0)male (l)female . 30—37 NVs: ._____... 38-Ul TOTs: __._________ 80 Card No. 112 APPENDIX D Introduction to Subject-sources APPENDIX D Introduction to Subject-sources As was explained to you in class, the National Science Foundation has funded a grant to examine the effects of group size and the sex of group members on group success and problem-solving strategies. We have brought together four-, three—, and two-man groups, in addition to indi- viduals; some of the groups are all one sex, while others are composed of various combinations of the two sexes. Each group will be asked to per- form the same task. The task is contentless in that it requires no specific knowledge of any subject matter or any specific skill. This was done so as to prevent anyone's past education or history from helping them do extremely well. We are interested in what size and sex combina- tion groups do the best. We also are interested in finding out if some problem—solving strategies help various size groups do well, while others work better for different size groups. The government hopes to use this information as guidance for what size work groups, of what combination of the two sexes, and what strate- gies will work best for various government task forces. Later experi- ments will vary the type of problem groups are asked to perform; first, however, we want to deal with simpler tasks. Since these results will have real effects for the way our govern— ment will be restructured, it is important you pay attention 1r1 take this seriously. Now the contentless task you will be asked to perform .3 rather simple, but also rather boring. So, in order to increase in- terest in the task, we are using some of the funds provided by the National Science Foundation to give prizes to the most successful groups. The group that does the best in each size category will receive $50 to 113 ll4 divide among its members. There are six groups of each size category. Since you were assigned to the dyad size category you will have to do better than five other dyads in order to get $25 apiece. In about four weeks you will receive a letter letting you know if your group won; if it did there will be a check enclosed with the letter. The task simply involves you working as a team to estimate the num— ber of dots on each card I show you. There are nine cards. I will show you each card for no longer than 15 seconds. After you see the card, you must together agree on one number which you believe represents the dots on the card. We will go through a practice card first and then start the nine. After each set of three cards I will try to give you feedback so you have some idea how far away your answers are from the correct answers. Any questions? APPENDIX E Sample Task Card APPENDIX E Sample Task Card ll5 APPENDIX E Subject-Receiver Instructions and Questionnaire APPENDIX F Subject—Receiver Instructions and Questionnaire MICHIGAN STATE UNIVERSITY DEPARTMENT OF COMMUNICATION You have been asked to come here to participate in research funded through a grant from the National Science Foundation, and designed to examine the ability of individuals, under varying conditions, to detect deception. If you choose to participate, you will be (seeing live, see- ing on videotape, hearing, reading a transcript of) 3 students being questioned concerning the means by which they completed a task. Some of these students will truthfully describe the means by which they and their partner completed the task. Other students will be lying about the method used to complete the task, since at least their partner, a confed- erate helping us with the experiment, cheated. Your job consists of tell— ing us whether you think the student is lying or telling the truth, and answering some additional questions concerning your decision. We will provide anyone interested with their accuracy score. If you would like to receive your personal accuracy score, please put your name and address here: Name: Address: Understand however that in a sense, if you fill out the above information your answers are no longer anonymous. However, all answers will be con- sidered confidential and will not be seen by anyone not associated with the grant. ll6 117 For each of the 3 people, you will first (see, hear, read) a segment during which they are always telling the truth. During this segment par- ticipants are asked their name, major, year in college, have they par- ticipated in research before, and the number of are taking. Immediately following the question communication courses, participants will answer cerning the solution to the task. This will be fOr which the participants will either be lying Based on this segment you are to decide whether communication courses they concerning the number of a series of questions con— the series of questions or telling the truth. the participants are lying or telling the truth. You will also be asked some additional questions concerning your decision; these will be more clearly explained as we go through the example. The length of the segments you will judge will vary, but is unre- lated to whether the person in the segment is telling the truth or lying. You will be given adequate time between each person to mark all answers, so do not worry about filling out the questionnaire while viewing; in— stead, pay attention to the individual being questioned. The 3 individuals you will be seeing may all be telling the truth, all be lying, or some may be lying and others telling the truth. Please try to judge each person without regard to your previous judgments. Once you start to view the questionee, please do not talk to anyone. This request is extremely important if you are viewing the questioning live. In the case of live questioning, you will be viewing the inter- action through a one-way mirror which is not soundproof. The lights will be off in the viewing room and you will have to return to this room to fill out the questionnaire. Please save any questions for the experi- menter till returning to this room. Also please make all judgments ll8 independently; do not discuss them with your neighbor. Remember the first series of questions will always yield truthful responses. NOW WE WILL EXAMINE SAMPLE QUESTIONS Initially you will answer question 1. 1. Was the individual lying or telling the truth? lying telling the truth 2. If 0 represents no confidence in your judgment, and 10 the average amount of confidence you may have in a judgment, how confident were you in your judgment? OF COURSE IN ANSWERING THIS QUESTION 10 IS NOT THE LIMIT TO HOW HIGH YOU CAN GO. IF YOU ARE MORE CONFIDENT THAN AVERAGE YOUR ESTIMATE WILL BE HIGHER THAN 10. The third and fourth questions will be connected to the kind of information you used in making your judgment. There are three informa- tion measures in which we are interested. They are l) NONVERBAL INFOR— MATION, 2) VERBAL INFORMATION, and 3) TOTAL INFORMATION. HERE IS HOW WE WANT YOU TO GIVE US THE ABOVE INFORMATION: We measure the amount of NONVERBAL INFORMATION used in making your decision in "NVs." An "NV" is a conceptual quantity, much as an ounce is a unit of physical quantity or weight. The greater the amount of non- verbal information present in the interview, the greater the number of "NVs" you should assign it. NONVERBAL INFORMATION refers to the information you get from some— one's NONVERBAL BEHAVIOR (facial expression, eye contact, nodding, hand and body movement, posture, pausing, the ums and ahs, anything besides their words). 119 We measure the amount of VERBAL INFORMATION used in reaching your decision in "V's." Like "NV's," "V's" are conceptual units, much like ounces are units of physical quantity or weight. The greater the amount of verbal information present in the interview, the greater the number of "V's" you should assign it. VERBAL INFORMATION refers to infOrmation coming from the specific words a person says; what he says, not how he says it. Similarly, we measure the TOTAL AMOUNT OF INFORMATION provided by a person in an interview in "TOTs"; a "TOT" is like an "NV"--it is a unit of conceptual quantity, much as an ounce is a unit of physical quantity or weight. TOTAL INFORMATION does not have to equal the sum of VERBAL + NON— VERBAL INFORMATION. We can conceive of an instance where NONVERBAL and VERBAL INFORMATION are redundant, in which case you would only count 1 in coming up with TOTAL INFORMATION. SO FOLLOW YOUR INTUITION in making estimates and estimate each type of information without concern for the others. We are going to provide you with a kind of ruler to do this estima- tion. We will show you two sample interviews and tell you how many "NVs," "Vs," and "TOTs" there are in these two. Based on this sample "RULER" you will make your estimate. It will work the same way as if I gave you two pieces of paper. I told you the first piece was six inches long and the second piece was three inches long. Based on this information, I would ask you to estimate the length of the third piece of paper. Here are your two sample interviews. WATCH THE TV SCREEN. SAMPLE 1 contains 100 ”NVs," 50 "Vs," and 150 "TOTs." LOOK AT SAMPLE l AGAIN. SAMPLE 2 has 200 "NVs," 100 "Vs," and 300 "TOTS." NOW LOOK AT SAMPLE 2. These samples will be available to you for replaying on request and will be played before viewing each participant. You make your estimates in answer to the following questions. 3. If SAMPLE 1 has 100 "NVs" and SAMPLE 2 200 "NVs," how many "NVs" did you use in making your judgment? 4. If SAMPLE 1 has 50 "Vs," and SAMPLE 2 has 100 "Vs," how many "Vs" did you use in making your judgment? 5. If SAMPLE 1 has 150 "TOTS" and SAMPLE 2 300 "TOTs," how may "TOTS" did you use in making your judgment? You may feel uncomfortable with this procedure at first. Rgtgt. We are interested in your estimates. There are no right or wrong answers, nor do all your answers have to agree with other people's or each other's. Your estimates can be as low as 0 and as high as you wish. Do not con— sider the sample ruler as boundaries. Finally, we will ask you if you can list any specific behaviors that caused you to make the judgment you did. 6. List anything you can remember about the interview which caused you to make the judgment you did. NOW WE WILL DO A DRY RUN USING AN INTERVIEW ON THE VIDEOTAPE MONITOR. HERE ARE THE TWO SAMPLE SEGMENTS AGAIN. REMEMBER: SAMPLE l SAMPLE 2 NVs = 100 NVS = 200 Vs = 50 VS = 100 TOTS = 150 TOTS = 300 HERE IS THE TRUTHFUL SEGMENT. HERE IS THE SEGMENT YOU JUDGE. 121 1. Was the individual lying or telling the truth? lying telling the truth 2. If 0 represents no confidence in your judgment, and 10 the average amount of confidence you may have in a judgment, how confident were you in your judgment? 3. If SAMPLE 1 has 100 NVs and SAMPLE 2 200 NVs, how many NVs did you use in making your judgment? u. If SAMPLE 1 has 50 Vs and SAMPLE 2 has 100 Vs, how many Vs did you use in making your judgment? 5. If SAMPLE 1 has 150 TOTs and SAMPLE 2 300 TOTs, how many TOTs did you use in making your judgment? 6. List anything you can remember about the interview which caused you to make the judgment you did. NOW, if there are no questions, the questionnaire for the first person begins below the solid line. columns 1. Project Number: Ou76 1-4 5—7 2. Subject Number: 8 3. Sex of Viewer: (O)male (l)female 122 columns u. Condition: (1) live (2) video (3) audio 9 (4) transcript 5. Group: 1 2 3 u 10 6. Participants being viewed: 11-16 7. First participant viewed: 17-18 19 8. Sex of participant: (0)male (l)female ___ 20 9. Was the individual lying or telling the truth? ___ lying telling the truth right (1) wrong (0) FOR CODER USE ONLY 21—24 10. If 0 represents no confidence in your judgment, and 10 represents the average amount of confi- dence you may have in a judgment, how confident ____________ were you of this judgment? 25—28 11. If SAMPLE 1 has 100 "NVs" and SAMPLE 2 has 200 ”NVs," how many "NVs" did you use in making your judgment? 29-32 12. If SAMPLE 1 has 50 "VS" and SAMPLE 2 has 100 "VS," how many ”Vs" did you use in making your judgment? 33-36 13. If SAMPLE 1 has 150 "TOTS" and SAMPLE 2 has 300 "TOTS," how many "TOTS” did you use in making your judgment? 123 lu. List anything you can remember about the interview which caused you to make the judgment you did. 15. Second participant viewed: 16. Sex of participant: (0)ma1e (l)female 17. Was the individual lying or telling the truth? lying telling the truth right (1) wrong (0) FOR CODER USE ONLY 18. If 0 represents no confidence in your judgment and 10 represents the average amount of confi- dence you have in a judgment, how confident were you of this judgment? 19. If SAMPLE 1 has 100 "NVs" and SAMPLE 2 has 200 "NVs," how many "NVs" did you use in making your judgment? 20 If SAMPLE 1 has 50 "Vs" and SAMPLE 2 has 100 "Vs," how many ”Vs" did you use in making your judgment? 21. If SAMPLE 1 has 150 "TOTS" and SAMPLE 2 has 300 "TOTs," how many "TOTS" did you use in making your judgment? 22. List anything you can remember about the interview which caused you to make the judgment you did. columns 37-38 39 HO u1-uu 45-48 H9—52 53—56 124 columns 23. Third participant viewed: 57-58 24. Sex of participant: (O)male (l)female 59 25. Was the individual lying or telling the truth? lying telling the truth 60 right (1) wrong (0) FOR CODER USE ONLY ___ 26. If 0 represents no confidence in your judgment, 61—6u and 10 represents the average amount of confi— dence you may have in a judgment, how confident ____________ were you of this judgment? 27. If SAMPLE 1 has 100 "NVs" and SAMPLE 2 has 200 65-68 "NVs," how many "NVs" did you use in making your judgment? ____________ 28. If SAMPLE 1 has 50 "Vs" and SAMPLE 2 has 100 "Vs," 69—72 how many "Vs" did you use in making your judgment? ____________ 29. If SAMPLE 1 has 150 "TOTS" and SAMPLE 2 has 300 73-76 "TOTs," how many "TOTS" did you use in making __ __ __ ___ your judgment? 30. List anything you can remember about the interview which caused you to make the judgment you did. 77-78 31. Age: ._____ 32. Year in school: Freshman (O) Sophomore (l) 79 Junior (3) ___ Senior (n) has participated in hours of research Joyce Bauchner APPENDIX G Transcripts INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: APPENDIX G Transcripts MICHIGAN STATE UNIVERSITY PERSON 1 State your name.* And your year in school. Freshman And your major. No pref. No pref. Have you ever been in research before? No. How many communication courses have you taken? This one, I am enrolled in now. What's that? Communication 100. O.K. Could you describe the strategy your group used to get their answer? Any strategy. Just estimating by counting the number in one certain amount. Could you be more specific about how you did it at all. If you're not sure how she (partner) did it, at least how you did it, some kind of . . I just estimated from the first picture you showed us. And what kind of conferring did you do? Compromise. O.K. If you had to describe to the next group that was go— ing to come in what they should do to do as well as you did, in one or two sentences what would you tell them to do? "All responses to this question have been removed to protect the interests of participants. 125 126 page 2 — Person 1 PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: Guess high. Alright. If you could choose to work in any one of the fOur size groups I talked about, alone, with one other person, with two other people, with three other people, which one would you choose? Two people. How come? Because three people, they never agree on anything. And with the two people you just end up compromising between the two. Alright. Is there anything else that you can remember about how the group got the scores they got, especially at the end when you did better. I found we were doing the same as we did at the beginning, but guessing higher. O.K. That's it. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: 127 MICHIGAN STATE UNIVERSITY Transcript - Bauchner PERSON 2 Could you state your name? And your year in school. Freshman. And your major. Forestry. Forestry. Alright. Have you ever been in research before? Yes. I did . . . I worked with a girl in education class. We were video taped and then we watched outselves on video tape and we commented on our behavior and stuff. O.K. Then you are used to being on video tape. It was weird. How many communication courses have you taken? My first. Your first. And what course is that? 100 O.K. Alright. Could you describe the strategy that your group used to get their answers? We took a corner, like a 1/u or 1/8, and counted it, and then multiplied it by H or 8, and then like if there was a big whole we subtracted, and if it looked like we missed the big corner we added a little. Could you be specific, especially at the end, you started to do really well. Did you change the strategy at all? We compared them with how the other ones looked. You know like if we guessed really high or guessed really low, we could kinda tell if there were more dots and less dots or if they were bigger or smaller. O.K. If you had to describe how you did when you did well, alright, to the next group that was going to come in, tell me how you would describe it in a couple of short sentences. 128 page 2 - Person 2 PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: I'd look at the size of the dots and how far apart they were, any big spaces, and I'd divide it into 1/4 and l/8's and I think I'd keep a running total of like if one looked particularly alot of dots, and I'd remember which sections had more and which sections had less and just add them all together. O.K. If you could choose which size group you would work in, alone or with two people or three or four, which would you choose? Two. What kind of benefits did you think you got from working in this size group? Well, you can count that much faster. Plus I like your own judgment your not too sure about sometimes it just looks like more or looks like less. If you had three, there'd be too many. You couldn't all decide. But with two there would be two people to share and you could. Is there any way else that you could describe what you do that you did, in more detail? Especially when you felt you were doing well. Just I think comparing it to previous dots and like if you guessed a number it just sounded like there were too many or too few. O.K. Thank you very much. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: 129 MICHIGAN STATE UNIVERSITY Transcript - Bauchner PERSON 3 State your name. Year in school. Yes. Year in school. Not now. Year in school. Oh, year in school. Freshman. O.K. Major? No pref. No pref. Have you ever been in research before? No. O.K. How many Com courses have you taken? My first one. And I'd taken it in high school, public speaking class. And this is.... 100 O.K. Could you describe the strategy your group used to get the answers that they got? Sure. Mostly estimation. Guessing. Be more specific. O.K. I took a, I compared it to other pictures like the first practice one we had, and if there were less dots then I'd bring it down a couple hundred or so. And then I'd converse with her, figure out what she had, a kinda divided in half and agree on that. O.K. Could you be more specific by which you mean when you say converse? We made our decision on one number and then that number, if we agree on it, that was what we got. 130 page 2 — Person 3 INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: O.K. If you had to describe to the next group, that was going to come in, what they should do to do as well as you did, in two or three sentences, what would you tell them? You know what are, boil down to the secret to your success? Secret to our success. I think its totally guess. I wouldn't even know how to explain to another person how to do it better. Unless, maybe they could count faster. If you could choose any of the size groups we talked about to work in, to work alone, or with one other person like we did now, or two other or three other people which would you like to work in? Maybe four people. How come? Split it into quarters. It would be faster. O.K. Is there anything else you could describe how you did it? You know, how you did so well. Basically, when you did well, what did you attribute that to? Looking at the pictures. Comparing the pictures. Like if there were alot of dots in one of, one of them, I'd go by that. That's all I can tell you. O.K. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIAPNT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: 131 MICHIGAN STATE UNIVERSITY Transcript — Bauchner PERSON 9 Could you state your name? You want my name? Yes. It's O.K. G—E-N-E. After my mother. O.K. Year in school? Yes mame. Year in school. Oh, year in school. you're in school . . I come from the country see, 0 O.K. That's the second time this has happened. O.K. O.K. Major? It's my accent. O.K. New York? That's alright! Year in school is a freshman. New Jersey. Oh, New Jersey. O.K. O.K. Have you ever been in research before? My major is journalism presently. No mame. O.K. Well, I, they had a questionnaire, my ATL prof sent a lady out this winter asking me some questions. O.K. But like this before? No mame. O.K. Home many Com courses have you had? My first. 132 2 - Person u INTERVIEWER: Which is? PARTICIPANT: Com 100 INTERVIEWER: O.K. PARTICIPANT: Keith Adler INTERVIEWER: Keith Adler. O.K. Alright now, could you describe the kind of strategy, as you went along, that your group used in solving the task? PARTICIPANT: Mine was hunt and peck. In all truth it was. INTERVIEWER: Could you be more specific? PARTICIPANT: Well, what I did first of all is I looked. I had no, I'm very mechanically inapt. So, I looked, down the page as best I could and determined well, gee whiz, what is this, and I just came up with a number. And the dear lady here, straightened me out. And that's all there was. And in truth, in the first part as it progressed. . . Do you want the whole thing? INTERVIEWER: Yes. PARTICIPANT: O.K. Well as it went along and, then 3rd or 9th or what- ever it was, I started to count. I counted like in the top line, I counted 10, as I said in there. And then I tried to proportion that as in terms of the page, and then I also tried to take in account up here that some of the page was sparse in terms of dots, with larger dots with L— shaped. And then down it was more concentrated and I tried to take that into account. And then we debated. INTERVIEWER: O.K. If you had, if you had to, if I was to let you talk to the next group that was coming in, and I told you in two or three sentences to describe, you wanted to describe to them how to do well. What would you tell them? PARTICIPANT: I would say, do as we did. Communicate, and don't be hurried about the time. Try to make up a system. Try to be in groups or make, and count and then try to proportion in groups. But the main thing would be, communicate. And listen to the other person, and don't rely on yourself too much. INTERVIEWER: O.K. If you had to choose, now, if you could do it again and you could choose what size group to work in whether it's one, two, three or four man, What size would you choose? PARTICIPANT: Two. 133 page 3 - Person u INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: Why? Because if you get three, if you get one your going to have, at least I would anyway, have many, many self doubts. I'm sure. In there, there were numbers that I threw out that were hundreds off and my partner here straightened me out. And with three you would have three divergent opinions, probably. And it would seem to me, anyway, you would argue more than you should. And you would just get frustrated, it would seem to me. And you would just come up with a wild number that perhaps wouldn't be. . O.K. One more question. O.K. And it's simply is there anything else that you want to do to describe, anything I could have left out, or you could have left out in your description? My description. Your description of what, your strategy of how you No. O.K. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: 134 MICHIGAN STATE UNIVERSITY Transcript — Bauchner PERSON 5 State your name. Alright. What year are you in school? This is my first year. O.K. "yes." I keep on saying year in school and people go Major? Major is pre—med. High aspirations. Good luck! Yeah well, thank you. Have you ever been in research before? No I haven't. O.K. How many Com courses have you taken? This is my first. And this is. . . This is 100. O.K. Alright now . Could you describe as best as you can, the kinds of strategies that your group went through, to get their answers. For this experiment? That's right. Well . It took a couple tests before we realized that we were very much different in our answers. But, mainly we tried to, our big trouble was trying to compare all the pictures. Cuz when you see all these dots, sometimes, you're mainly trying to compare the different ones and you're trying to figure out in segments, like in groups of ten. Then we always added a few more cuz we did count so hard. That was our main strategy, though. O.K. Could you be more specific how it changed as it went along? Like you know, like what kind of changes in strategy you think you had at the end. You did really well. p 2 - Person 5 PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: 135 Well in the beginning we were mainly just trying to count them. . one, two, three, four, five. And we thought we didn't have time. So as it went along we were mostly com- promising, because Sandra was always alot higher. And I was mostly lower. And we were sort of trying to compare them more often. . O.K. Trying to come up and be more reasonable with each other. If you. If I. . . there's another group that is going to come in soon, if I say, O.K. I want you to go out and I want you to describe in two or three sentences, to that group, how to do as well, what would you tell them to do as well as you did? What would I tell them? What would you tell them? I would tell them not to worry too much about counting them, just to get, to keep a basic in mind. How many the comparison, . . . mostly to compare them, and to keep in mind what each thing looked like. And tell them that there is a pat . . no I won't tell them there's a pattern. Just tell them, mostly that, estimate higher than what they come up with. I mean, O.K. Alright now . . . remember I told you that there was groups of one, two, three and four that were doing this. If you would choose to do it again, let's say I told you this was a test run, now you get to choose what size group to work in, what size group do you think you'd pick? I think I'd pick three. Why? Well, I knew that if I'd do it by myself it would have been really. And I knew like when Sandra and I were doing it we still had major difficulties. If we had a third person there, they would be closer to one of the other two, which they could take their answers jointly and figure out their estimations on how many dots there were. O.K. That's fine. Alright that's it. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: 136 MICHIGAN STATE UNIVERSITY Transcript - Baucher PERSON 6 State your name. O.K. What year in school are you? Freshman. Major? Forestry. Forestry. Have you ever been in research before? No. None. None. How many Com courses have you taken? One Com, but I've taken an advertiSing course that counts as Communication in my major. O.K. Can you describe the kind of strategy in which your group used to get your answers. What kind of strategy? A huh. Just group the dots in groups of ten, and then count across, and then multiply that by ten. Is that what you did the whole time? That's what I did the whole time. Is there any kind of strategy between you two that you used to come up with one number? Yeah. O.K. If you had to describe like in two or three sentences to the next group that is going to come in, how to do as well as you did, let's say that was the question, what would you tell them to do? We usually raised it up a little. There's more numbers there than you think there are. Alright, if you could choose the size group you would work in, let's decide you were to do this again. Which would you choose, one, two, three, or four—man group? Probably, three. 137 page 2 - Person 6 INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: Why? Four I guess you would come up with too many conflicts. Two we, one would never have done it cuz it takes more than one person cuz somebody could underestimate. Three just seems like one person might say, well it seems like a little more. And two people might come up with, you know, not the, a little lower or something. Is there anything else about the strategy, we're particu- larly interested about the way you went about getting your answers, that you think. Can you differentiate when you did well and when you didn't do well? Could you repeat that? Just is there anything more about the strategies, you know like you think of the different kinds of strategies that were used when you did well and didn't do well? Well, just like, when we got shown, when we had small dots, there are a heck of a lot more. They can put a lot more on those papers with small dots. And you've just got to think big with those. You can do the groups of ten a lot easier with the big dots than you can with the small dots. We should of added a lot more to the small dots. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: 138 MICHIGAN STATE UNIVERSITY Transcript - Bauchner PERSON 7 Could you state your name? Year in school? Freshman. And your major? General Business. Have you ever been in research before? No. None? No. No I haven't. How many communication courses have you taken? Zero. Zero. Are you taking one now? Well, yeah, now. That's the only one? Right. O.K. Could you describe the kind of strategy that your group used to get your answers? We used . I'm asking you now, then I'll ask her. Oh. We used a lot of them, that's for sure. How will I say it, let's see. But, her strategy was right, I'll admit that. How would you describe it? What was her strategy? Well, we just figured there was 10, we had to figure out from the ten cards. Then we said that, somehow like we started with, I think it was like four numbers, and each one of those was like first, the practice one was 1000, then the next one was like 200, and the next one was like 200. So we said that there's maybe going to be like two of each number, cuz like the fourth one looks like 300, and we thought the number would start keep going higher and higher, you know like the next one would be 300, no the next one would be like MOO. But it wasn't. that wasn't 139 page 2 — Person 7 (cont'd.) PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: true. But, the basic part of it was right. There was going to be two of each number. Is there anything else that you, any other procedure that' you went through besides? Well, then we tried to measure it, you know, measure the card, and measure the area across, of how much was in each area, and that got to be too complicated. Then we tried, then each of us, you know, just would count like half of it, half of each side and we figure the discrepancy between the two. What strategy were you following when you found that you were really doing well? I think the best way was to figure which half, count our half and then discuss it from there. O.K. If you had to describe, let's say I want you to talk to the next group that was coming in, and I wanted you in two or three sentences, I was going to let you talk to them for a couple of minutes, and you would have to tell them how you do as well, what they would have to do to do as well as you did, what would you tell them? We did well? Yeah. But go ahead, I want to know how you would tell them. In one or two sentences what they would have to do to do as well as you did. I'd say, each of you count half the side, then just discuss it from there, how many you think was on each side. And count the total up. If you had to do this again, and I let you choose what size group you'd be in, one, two, three or four, what size group would you choose? One. Why? It might be because of my personality, and I think it might be because, I think it would work out better, too. O.K. Is there anything else that you could think of about the strategy that you used, that you haven't told me yet, that helped you in your answers? No. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: 140 MICHIGAN STATE UNIVERSITY Transcript — Bauchner PERSON 8 State your name. O.K. And your year in school? Freshman. Major? No pref. No pref. Alright, have you ever been in research before? Psychology experiments. Like this? Not like this. O.K. How many communication courses have you had? My first one. First one. O.K. Could you describe the strategy that your group used? Random. I guess. We tried to count them and then organize space, How many dots we thought were in the spaces. Could you be more specific? along? Did it stay the same? Did it change as you went Sort of took a small spot and tried an count the dots in the spot and then tried and figured how many spots there were like that and how many dots we found. Kinda of random there. How did you interact between you. strategies between you? Were there any new I don't know . O.K. If you had to describe, if I let you walk out, and I said I want you to go to the next group and tell them in two or three sentences what they have to do to do as well as you did, what would you tell them? Think fast. I think we just kinda guessed. Don't underestimate, probably. 141 page 2 - Person 8 INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: O.K. If you could choose, I told you there were four size groups that were going to be this, if you could choose the size that you are going to work in, one, two, three, or four, what size would you choose? Probably a smaller group. Could you be more specific? I have to pick one, right? Right. We'd like you to pick one. This is after you've done it once? Right. And now we said this didn't count, and you get to do it again, but this time you get to do it and pick what size. I'd probably go myself. Why? I don't know, I just . . . with more people it's harder to figure out, you know the different opinions. Is there anything else that you haven't told me that would help describe the kinds of decision-making processes you went through, you know, especially when you did well? Because that's what we are interested in. I guess, I don't know. Just try to figure out the Space. Some pictures look like they've got a lot of space and some had a little. O.K. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: 142 MICHIGAN STATE UNIVERSITY Transcript - Bauchner PERSON 9 Could you state your name? And the year you are in school. Freshman. O.K. Major? No pref. Have you ever done any research before? No. None. Whatsoever? No. O.K. You've had to count dots before, huh? No, not really, How many communication courses have you taken? This is the only one. O.K. Can you describe the kind of strategies your group used to get the answers? Only one. Well, we had to figure out something where we could utilize two people so that . How did you do that? Like taking certain areas and each one of us counting, and then taking the average of what we counted. O.K. Any more specifically how it changed as you went along? Well our, we found out that at the end, like at the beginning we were each taking halfs and counted them, when there were less dots, but more dots we'd take a little space, count them, and each one of us would count all the spaces, then we'd take what the average of those. O.K. If I told you there's another group waiting in the waiting room and that you could talk to them, in two or three sentences, tell them what they'd have to do to do as well as you did, what would you tell them? .143 page 2 - Person 9 PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: They wouldn't want to do as well as we did. I don't know. Just to, you know, try and make two people count their own thing and then average the two decisions. Alright. If I told you that this time didn't count, and that now you've got a chance to do the thing again, in any size group you wanted to do, one, two, three, four, what size would you take? Four. Definitely! How come? Cuz I think you'd get the closest estimate. Like what a lot of people see. O.K. Anything else you can tell me about the kind of strategy you used, especially when you were doing well at the end? Not really. You could take it a little easier at the be- ginning just by counting all the dots. We realized that we were under counting everything, so you know, that kinda made a difference in what we were saying. That's about it. O.K. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: 144 MICHIGAN STATE UNIVERSITY Transcript — Bauchner PERSON 10 Could you state your name? What year in school are you? Freshman at Michigan State University. What's your major? Packaging/management. Have you ever been in on research before? No. Nothing. Zero. O.K. What kind of communication, how many communication courses have you taken? Well, this is my second time around in Com 100. out last time cuz I didn't like the professor. Oh. O.K. O.K. I dropped We won't ask you who the professor was. Could you describe the strategies that you thought your group was using as you went through the task and tried to figure out. It was pretty inadvertent. I suppose, I just guessed. Well I guessed within a range of what would be the possible answers. I mean why did you guess what you guessed? If the dots were fairly dense, I'd say between 500 and 1,000, that the dots . . I didn't really guess. It was guessing in the range that was in my mind, you know. Could you, can you give me any idea of what kind of process... If the dots were really dense, I'd pick a wide range where the dots could be and I'd just close in a number of the dot that came to my head and I'd pick the number. How did you interact with your partner? kind of strategy between you? Was there any What happened between you? 145 page 2 — Person lO PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: Well, yeah because, if she thought I made a bad guess we'd get closer to her guess, and they were usually good cuz she helped me. But she was definitely doing better at the end. O.K. If I was to let you go out and talk, there's yet another two people waiting to come in, I told you there was another dyad, and I asked you to describe in two or three sentences to them what they have to do to do as well as you did, what would you tell them? I'd say you can tell them two or three things. Tell them. I wouldn't tell them anything, because it's just guessing. That's all. I also told you there were groups of one, two, three and four, and you were in a group of two. If I told that this time around didn't count, and you could do it again, you're going to do it again tonite, and you could choose what size group having gone through it once, you wanted to be in for the time it counted, what size group would you choose? Four. Why? More opinions, but not so many that there were would be confusion. If you only have a certain amount of time to look at the cards, four people would be a good amount to work with in anything. Oh, alright. Is there anything else about your strategy that you can think of that would describe how your group worked, that you felt helped you do well? I might have done better if the situation was more crucial. Like if you wouldn't let me out of here in the next five years, unless I got within 10 of every one, I might have thought much harder. O.K. Or if the offer was for $500. $500? INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: lNTERVIEWER: 146 MICHIGAN STATE UNIVERSITY Transcript — Bauchner PERSON 11 Could you state your name? My name? My whole name is O.K. And year in school? Right. Your year in school. Huh? What year in school? What year? Junior. O.K. Speak louder so the mike can pick you up. O.K. Your major? Communications O.K. Have you ever been in research before? Yes. I have. Like this? No. Nothing like this. The kind of research was, I did it up at Northern two years ago, it was for a kind of a random sampling of a presidential election, something like that. O. K. How many communication courses have you had? Two. What . Business Communications and one other one, writing reports, something like that, it was a business report class. O.K. And this one. Ah huh. O.K. Alright. Describe the kind of strategies that your group was using to get their answers. 147 page 2 - Person ll PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: Well, we did it this way . . . my type of strategy, first of all, was to count the number of dots going horizontally and that way I could determine that score vertically, cuz they're just about the same. It was one square. The square was about the same proportion, the vertical and the horizontal were just about the same proportion. So, what we did, what I did was multiply what I counted horizontally four times cuz there are four sides to a square and we counted down, it's hard to explain. Go ahead, just describe. I counted them horizontally and whatever number I came up with, multiply that by four. O.K. And we came really close. I don't know how, but we seemed to do alright cuz we were close to most of the answers. How did you interact as a group? You said that was your strategy. What did you do, I know you had to come up with one number. What did Sandra do? Well, we were supposed to, we originally, we didn't do it until later on. I was supposed to count horizontally and she was supposed to count vertically. So we wouldn't both be counting the same ones and waste time, you know. While she could be doing something else, that way we could come up with an answer. O.K. If I was to take you out into the other room and there's another group waiting to go in and I said O.K. I want you to tell, in two or three sentences, how to do as well as you did, what would you tell them? What strategy to use? The best strategy? Yeah, you have to describe in two or three sentences what to tell them to do as well as you did. O.K. I'd tell them instead of trying to count all the dots, cuz you only have 15 seconds, there's no way, you've got to start at the surface of the square, you know and kinda figure, by looking at the number of dots if there's a big open space in the middle, you subtract 100 dots or whatever, so what you start out by doing is counting the outside part of the square and that way, by doing that you kind of have an idea. Oh, it's hard to explain. I can't say it in words. I know what I'm talking about, but I can't say it in words. If I told you there were individuals doing this alone, or groups of two, like you, or threes, or fours, if I told you that what you just did didn't count and you got a chance to 148 page 3 - Person ll INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: do it again and you could pick what size group you would do it with, what size group would you pick? I would pick two. Because if you have a whole bunch of people, or maybe that's better, if you have a whole bunch of people cuz that way everybody can try out a different strategy and get all mixed up. Everybody would have dif- ferent numbers and everything. If you just have one person, another person that can help you, you know, I think that's better or if I do it by myself, I think it would be harder. You're doing it by yourself, you don't have somebody else helping you. So I think two people is best. ' O.K. Is there anything else that you can think of, any other ways to describe what you did especially when you really did well? Other ways to describe it? Yes. Anything you can think of that you left out? I'm just checking back. Youknow, anything that you felt helped you do really well? Help me do well? I guess it was I had some motivation because you mentioned I was competing. Whenever I'm com- peting, I get under pressure. I know every time I'm under pressure, I do well cuz the past experiences when I'm under pressure I do well. O.K. INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: 149 MICHIGAN STATE UNIVERSITY Transcript - Bauchner PERSON 12 State your name? Your major? No pref. Year in school? Freshman. O.K. And, have you ever been in research before? No. Not at all? No. How many communication courses have you taken? This is my first one. That is. Com 100. O.K. Now, I want you to describe for me the kind of strategy that your group used in order to come up with the numbers that you gave me. Well, by myself or both of us? Your perception of what you as a group used. Well, I counted dots that looked like were grouped to— gether and then I counted over in the square, like how many groups of those dots I thought there would be and then I counted down and multiplied those, and I don't know what she did, but then we put together, compromised and come up with an answer. O.K. Did it change at all as you went along? Did it change as time went on or did you use the same thing every time? Well after we started getting feedback we tried to, we realized that we were under cutting it, so we tried to make our numbers higher. 150 page 2 — Person l2 INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: PARTICIPANT: INTERVIEWER: O.K. Now, if there is another group out there, alright, and I said O.K. I want you to go to tell them in two or three sentences what they should do to do as well as you did, what would you tell them? I don't know. Don't get nervous, I suppose and just try your best. I don't know. O.K. Now also, let's say that I just told you that this time didn't count, alright, you were going to go in the other room now and do it again. If you had your choice, this time, you could either work alone, or with two like your working now, or say in a group of four, which size group would you prefer? A group of four. Why? Cuz you'd have more answers and more to compromise, with out of 4 there's likely to be someone right. Is there anything else that you can think of telling me about the kind of strategy, we're interested in group strat- egy, that you used to come up, especially when you started to do well? You want me to tell you something about the strategy? I think ours was good. Use that one. O.K. APPENDIX H ANOVAs for Transformed Variables APPENDIX H ANOVAs for Transformed Variables Table Hl Analysis of Variance of Available Total Information by Channel Condition Using Transformed Variables* Source Sum of Squares gt MS 3 p_ Total 7.637 239 —- —- _- Between 6.278 3 2.093 363.508 ‘<0.001 Within 1.359 236 .006 -- —- Table H2 Analysis of Variance of Available Nonverbal Information by Channel Condition Using Transformed Variables Source Sum of Squares if. M_S t _p_ Total 591.908 239 -- —- -— Between 590.2u8 3 196.799 27956.603 ‘<0.00l Within 1.661 236 .007 —- -- nTransformation formula is: transformed variables (variable + .05)loglo. 151 152 Table H3 Individual Comparisons of Channel Condition Available Total Information Means Using Transformed Variables* Channel Condition Mean Live 2.528q Videotape 2.558r Audiotape 2.418S Transcript 2.1'47t Table H4 Individual Comparisons of Channel Condition Available Nonverbal Information Means Using Transformed Variables* Channel Condition Mean Live 2.364u Videotape 2.1416v Audiotape 2.16Mw Transcript -l.301X * Means having the different subscripts differ significantly at the .05 level of confidence. Transformation formula is: transformed variable = (variable + .05)loglo. APPENDIX I Lisrel Matrices: Models I, II, and III 153 :mo.l moo. mmo. 0H arm. 0H www.ml omm.om mHH.:H m mma.m mmm.hmmoa Hmo.©| mNb.ow www.md m Hmm. mo:.mmm0H ham.HmmmH m:u.m| www.mm mmh.:a b bmm.a www.mmmm Hmo.mmzaa boo.mmmHH www.ml www.mm www.5H w mmb.m m:m.mHmOH bmm.Hbmo. omm.mmmoa mom.mm:HH Hmm.m| www.mm mmm.am m moH.m bom.mmmaa o:a.uo>ma mmm.wmhma 000.5NNNH mHm.mHm:H puma psch "H ameoz H xHszmm< m:w.wn mm:.om mom.ma .2 mow.m www.mioaa www.mmsmH on.nHmmH oaw.mmama oom.wmaza mmo.mwmsa .HH canoe mma. moo.| mma. m m mac. :Ho. mmo. mmo. Nmo. omo. m m woo. woo. www.mm NmH.:m mmH.mm :om.mm oom.bm mam.mm m:m.mm Ho:.mm mHm.om Nmb.am bmH.Hm Hm:.Nm wmo. :wo. bmo. m m oma mmH n HmHMBHH u moo.l mmo.| wma. mHo. owo. mro. zoo. HHO.mN mom.mm www.mm mmm.wm 0:0.Hm Nao.mm :mo. omo. bmo. a m 3 m 0H (Ii-«O’EZCL. (\IC'O xxm (V 5 O H r-Ide'LDLOFCDO') www mocooom cw mafia UOPMEHpmm maoPMOHpCH memucH maopMOHch Hmowwoq 03m 2 154 cofluoaom HmMpHcH 000.H 000.0 000.0 000.0 000.H 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 3 ”H Homo: 000.0 000.0 00:. H03. 00:. 01m. 0am. 000. m N 000. 000. 000.H mma.al 000.0 000.H 000.0 000.0 m m 000.H 000.0 000.0 000.H 000.0 000.0 m N 000.0 000.0 000. 000.0 :HH.H 000.0 000.H 000.0 000.0 000.H 000.0 mmm.a 000.0 000.H 000.0 000.0 000.0 000.0 000.0 000.0 m 0 .NH OHQMH 000.0 00:. mn:. 0m:. 000.: 000.0 000.0 000.H 000.0 000.0 000.H 000.0 000.0 000.0 000.0 000.0 000.0 000.0 H00. 0mm. 000.H meemo mumm x masses .1 > panama HNC‘OJ r—lC‘JC‘OII NCO r-‘lCVmJ'LDtOFQDOWt—i 155 000.0 0H 03:.H 0 05:.H 0 00H.H b Nom.H 0 000.H m 050. mum. H 3 000.0 mmo. 0N0. omm. 00H. 000.0 020. H00.| mbz. mma.m 000.: 00H. 000.0 a mpama wedge >30. mom nudge 0m0. >00. H00.| 00m. ".1 U) Q. 000.: 000.: 00H. HLm #1ij (\l Ap.poooc cospoaom Hoapaca 156 000.0 000.0 000.0 : 00N.0HI b0H.hHI 0b0.0N| 0 000.00 000.00 000.0: N 00>.HI N00.H| 00:.HI H 0 N H mesmw 000.H 0b0. :b0.| SHO.::| 3 000.0 000.H 0N0.0| 000.0 0 000.0 000.0 000.H 000.0 N 000.0 000.0 000.0 000.H H 3 0 N H mpom 000.H 000.0 000.0 0 000.0 000.H 000.0 N 000.0 000.0 000.H H 0 N H x MUQEQH 000.H 000.0 000.0 000.0 0H 000.0 000. 000.0 000.0 0 000.0 0NH.H 000.0 000.0 0 000.0 000.H 000.0 000.0 b 000.0 000.0 0N0.: 000.0 0 000.0 000.0 NbH.0 000.0 0 000.0 000.0 000.H 000.0 3 000.0 000.0 000.0 00H.NI 0 000.0 000.0 000.0 0N0.NI N 000.0 000.0 000.0 000.H H z 0 N H w mwnflmq 0 u 02H coHpsHom ooonHmeHq Esemez ”H H0002 .mH mHnme 157 000.0 0H 000.H0 0 .ooo.o u Ho>oq speaaomoopo mmema.o:mo on eoooopm 0:H.0: 0 HNb.0: b 00:.0N 0 H00.:0 0 :00.N0 : 000.:H 000.0 000.0 0 N H30.1 030.1 HbH.HHI 000.5: H00.:NNI :Hm.0N 0N0.0: 00H. 000.1 00H. 000.0 H mpaoo mpoee 00:. mam mpmne :00.N 0N0.0HI 000.H H00.H H amm 000.1 000.: 00H. Hcm mo mmopmmp 00 nqu opmsomano H P4010021 (\l Ao.ucoov coszHom eooaHmeHq enemee 158 coHPSHom pONHUpmpcmpm 000.0 000.0 000.: Nb0.l :00. :00.H HH0.1 000.1 0 N 000.H 0HH. :0H.I 000.0 000.H 000.H: 000.0 000.0 000.H 000.0 000.0 000.0 : 0 N :0:. 000.0 000.0 :0:. 000.0 000.0 0 N :00.00 000.0 000.0 000.0 000.00 000.0 000.0 5:0.bHH 000.0 000.0 0:0.00H 000.0 000.0 000.0 0:0.NOH 000.0 000.0 :00.0HH 000.0 000.0 NNO.NN 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 : 0 N "H Hopoz .:H mHQMB 000.0 000.1 0H0. 0H:.: 01ij mEEmo 000.1 000.0 000.0 000.H r-iCNICOJ’ muom 000.0 0 000.0 :0:. H CV x manMH 000.0 000.0 000.0 000.0 000.0 000.0 000.0 :0N.0| 0b0.0| 0H0.H H O H HNO'J:LDLOI\oa Hpnanoooopo mmama.o:mo we eoeoopm mo moopmoo mm apflz opooom-fleo 000.0 0 000.0 000.H0 0:H.0: HNb.0: 00:.0N H00.:0 :00.N0 H:0.I 0H 0 0 b 0 0 : 0 000. N00.1 HN0.I 000.H 000.0 N 0:0.1 000.: HHO. 000. 000.: 000.H 000.0 H opaoo pupae 00:. mam cpmze NNO. :0H.I H00. b0b. Hmm 000.1 000.: 000.H Hnm H P401003 (NCO Ap.pcoov coHchom UONprmpcmHm 160 :00.1 000. 0N0. 0:N. 000.N: 000.0N 0HH.:H 0 00H.0 N00.h000H HN0.0I 0Nb.0N 000.0H 0 H00. 00:.N000H h00.H000H 0:0.0: Nb0.0N 00b.:H .3 >00.H N00.0N00 H00.00:HH 500.000HH memo poacH ”HH proz 00H. 000.NI 005.0N 0hb.bH 0 005.N 0:0.0HO0H b00.Hb00H 000.0000H 000.00:HH .0H 000.1 00H. H00.0I 550.0N 0N0.HN N 00H.N 500.00NHH 0:H.>0>0H N00.0NbNH 000.NNNNH 0H0.0H0:H 00 633 Hmkmhhh 000.1 000.1 00H. 0:0.01 00:.00 00N.0H H 000.N 005.0:0HH 000.00:NH 0H0.bHONH 0HN.00HNH 00N.50H:H 000.N00:H H 00H 0:N 0 0 0 b (34053.2. (\l xxm N wxm HNC‘OJ‘LOKOF >>m mpcoomm pH mEHe empmermm mLOHmOHch pwwmucH mpOHmoHch Hmonoq mz MP \lOlU'l-F-‘OOMH NH M NH NH Table 16. Model 11: Lambda Y 1 2 3 1.000 0.000 0.000 4.729 0.000 0.000 412.5 0.000 0.000 0.000 0.000 0.000 0.000 1.116 0.000 0.000 .988 0.000 0.000 0.000 1.000 Lambda X 1 2 3 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 1.000 Beta 1 2 3 1.000 0.000 0.000 -4.266 1.000 0.000 .058 -.016 1.000 Gamma l 2 3 55.507 59.681 34.026 —30.944 -29.000 -22.966 0.000 0.000 0.000 Phi l 2 3 .188 -.O63 .188 —.063 -.063 .188 Psi l 2 3 52.454 94.548 —22l.433 -2.079 1.947 .247 Theta Eps 1 2 3 107.427 29.364 19.365 Theta Delta 1 2 3 0.000 0.000 0.000 161 Initial Solution u 29.990 5 40.390 6 26.163 7 0.000 Table I7. IND = 4 Lambda Y 1 1 1.000 2 2.553 3 2.238 4 0.000 5 0.000 6 0.000 7 0.000 Lambda X 1 1 1.000 2 0.000 3 0.000 Beta 1 1 1.000 2 -l.6l2 3 -.011 Gamma 1 1 101.271 2 42.814 3 0.000 Phi l 1 .188 2 —.063 3 -.063 Psi 1 1 123.334 2 424.863 3 1.775 Theta Eps 1 1 79.958 Theta Delta 1 1 0.000 Chi-Square with 30 degrees of freedom is 1453.9611; Probably Level Model II: 0.000 0.000 0.000 1.000 1.026 .913 0.000 0.000 1.000 0.000 0.000 1.000 .005 2 112.531 59.739 0.000 .188 -.063 -365.338 —3.059 2 38.214 2 0.000 162 Maximum Likelihood Solution Center 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 1.000 0.000 0.000 1.000 3 63.577 30.885 0.000 .188 .228 3 33.692 3 0.000 n 35.526 5 42.890 6 35.973 7 0.000 = 0.0000. M NH \IC‘IU'I-F-‘QJMH M NH NH 1 Table 18. Model 11: Lambda Y 1 2 3 45.207 0.000 0.000 115.431 0.000 0.000 101.169 “0.000 0.000 0.000 99.062 0.000 0.000 101.636 0.000 0.000 90.468 0.000 0.000 0.000 .499 Lambda X l 2 3 .434 0.000 0.000 0.000 .434 0.000 0.000 0.000 .434 Beta 1 2 3 1.000 0.000 0.000 -.736 1.000 0.000 —.981 .975 1.000 Gamma l 2 3 .971 1.079 .609 .187 .261 .135 0.000 0.000 0.000 Phi 1 2 3 1.000 -.335 1.000 -.335 -.335 1.000 Psi 1 2 3 .060 .095 -.037 .079 -.062 .914 Theta Eps 1 2 3 79.958 38.214 33.692 Theta Delta 1 2 3 0.000 0.000 0.000 163 Standardized Solution Center 4 5 6 7 35.526 42.890 35.973 0.000 Chi-Square with 30 degrees of freedom is 1453.9611; Probability Level = 0.0000. 164 00H. 000.: 000.1 0 00H. 000.: N 00H. H 0 N H xxm :00.: :00. N00.1 N00. 00H. :0H. HOH. 0 000. 000. 000. H:0. b:N. 00N. me. N 0N0. 0N0. H00. 5N0. 00N. 0:N. :0N. H b 0 0 : 0 N H wxm 0:N. :00. H00.: N00. :00.: 000.1 000.1 b :00. 000. H00. 00N. HON. 00N. 0 :00. H00. HON. me. :0N. 0 :00. 00N. 00N. 00N. : 0::.N 00:.N 00:.N 0 00:.N 00:.N N >0:.N H b 0 0 : 0 N H rwm 00N n mpcoowm CH wEHH popMEHpmm 00H u mHOHMOchH smmmch ememeee u msOmeHecH HmOHwoq o:N n 02 m u z m u z m n o e u m puma usoaH ”HHH Hoeoz .oH oHooe }—‘ \lO‘U’I-F-FOOMH M NH MP to NH Lambda Y Lambda X Beta Gamma Phi Psi Theta Eps Theta Delta Table 110. Model III: 1 2 3 1.000 0.000 0.000 .993 0.000 0.000 .992 0.000 0.000 0.000 1.000 0.000 0.000 1.000 0.000 0.000 .961 0.000 0.000 .892 0.000 0.000 0.000 1.000 1 2 3 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 1.000 1 2 3 1.000 0.000 0.000 -.415 1.000 0.000 -.415 -.490 1.000 1 2 3 .518 .519 .514 .439 .396 .441 0.000 0.000 0.000 1 2 3 .188 —.063 .188 -.063 -.063 .188 1 2 3 .525 —.103 .319 —.107 .235 .388 1 2 3 .057 .135 .056 1 2 3 0.000 0.000 0.000 165 Initial Solution .058 .068 .072 0.000 166 Table 111. Model 111: Maximum Likelihood Solution IND = 2 Lambda Y 1 2 3 1 1.000 0.000 0.000 2 .994 0.000 0.000 3 .988 0.000 0.000 4 0.000 1.000 0.000 5 0.000 .943 0.000 6 0.000 .934 0.000 7 0.000 0.000 1.000 Lambda X 1 2 3 1 1.000 0.000 0.000 2 0.000 1.000 0.000 3 0.000 0.000 1.000 Beta 1 2 3 1 1.000 0.000 0.000 2 -l.040 1.000 0.000 3 .160 —l.513 1.000 Gamma 1 2 3 1 3.680 3.744 3.486 2 —3.429 -3.453 ~3.336 3 0.000 0.000 0.000 Phi l 2 3 l .188 2 -.063 .188 3 —.063 -.063 .188 Psi 1 2 3 1 .006 2 -.001 -.000 3 —.007 .002 .252 Theta Eps 1 2 3 4 5 6 7 1 .037 .059 .066 -.028 .064 .073 0.000 Theta Delta 1 2 3 1 0.000 0.000 0.000 Chi-Square with 30 degrees of freedom is 610.091; Probability Level = 0.0000. 167 Table 112. Model 111: Standardized Solution Lambda Y 1 2 3 1 1.573 0.000 0.000 2 1.564 0.000 0.000 3 1.554 0.000 0.000 4 0.000 .181 0.000 5 0.000 .171 0.000 6 0.000 .169 0.000 7 0.000 0.000 .498 MMMaX l 2 3 l .434 0.000 0.000 2 0.000 .434 0.000 3 0.000 0.000 .434 Beta 1 2 3 1 1.000 0.000 0.000 2 -9.02O 1.000 0.000 3 .507 -.551 1.000 Gamma 1 2 3 1 1.014 1.032 .961 2 -8.200 -8.257 -7.978 3 0.000 0.000 0.000 Phi 1 2 3 1 1.000 2 -.335 1.000 3 -.335 -.335 1.000 Psi 1 2 3 l .002 2 —.004 -.009 3 —.009 .018 1.017 Theta Eps 1 2 3 4 5 6 7 1 .037 .059 .066 -.028 .064 .073 0.000 Theta Delta 1 2 3 1 0.000 0.000 0.000 Chi—Square with 30 degrees of freedom is 610.091; Probability Level = 0.0000. REFERENCES REFERENCES Anderson, K., & Clevenger, T. A summary of experimental research in ethos. Speech Monographs, 1963, 30, 59-78. Barber, T. Y., & Silver, M. J. Fact, fiction, and the ex- perimenter bias effect. Psychological Bulletin, 1968, 70, 1-29. Barrien, F., & Huntington, G. An exploratory study of pupillary responses during deception. Journal of Ex- perimental ngchology, 1943, 32, 443-449. Berlo, D. K. The process of communication. New York: Holt, Rinehart, and Winston, 1960. Berlo, D. K., Lemert, J. B., & Mertz, R. J. Dimensions for evaluating the acceptability of message sources. Public Opinion Quarterly, 1970, 33:4, 563-576. Brandt, D. Listener prOpensity to counterargue, distraction, and resistance to persuasion. Paper presented to the Convention of the International Communication Associa— tion, Portland, Oregon, April 1976. Breitrose, H. The effect of distraction in attenuating counterarguments. Unpublished doctoral dissertation, Stanford University, 1966. Budd, R. W., & Ruben, B. D. Communication education: Bastard child and intellectual Frankenstein. Unpublish- ed manuscript, Rutgers University, 1972. Champness, B. The scope for person-to-person telecommunica- tion systems in government and business. Communication Studies Group, London, England, 1973. Cronbach, L. Coefficient alpha and the internal structure of tests. Psychometrika, 1951, 16, 297-334. Cutrow, R., Parks, A., Luscas, N., & Thomas, K. The objective use of multiple physiological indices in the detection of deception. Psychophysiology, 1972, 9, 578-588. 168 169 Danowski, J. Organizing and communicating. In J. Shubert (Ed.), Human communication: Concepts, ptinciples and skills, Michigan State University, Department of Com- munication, 1974, H1-H39. Davidson, P. Validity of the guilty-knowledge technique: The effects of motivation. Journal of Applied Egy- chology, 1968, 52, 62-65. Davis, R. Physiological responses as a means of evaluating information. In A. Biderman and H. Zimmerman (Eds.), The manipulation of human behavior. New York: Wiley, 1961, 142-168. Dearman, H., & Smith, B. Unconscious motivation and the polygraph test. American Journal of Psychiatry, 1963, 119, 1017-1020. Dorris, J. Persuasion as a function of distraction and counterarguing. Unpublisfied manuscript, University of California, Los Angeles, 1967. Cited in P. G. Zimbardo, M. Snyder, J. Thomas, A. Gold, and S. Gurwitz, Modify- ing the impact of persuasive communication with external distraction. Journal of Personality and Social ng- chology, 1970, 16, 669-680. Eisenberg, A., & Smith, R. Nonverbal communication. New York: Bobbs-Merrill, 1971. Ekman, P., & Friesen, W. Nonverbal leakage and clues to deception. Psychiatry, 1969, 63, 88-106, a. Ekman, P., & Friesen, W. The repertoire of nonverbal be- havior: Categories, origins, usage, and coding. Semiotica, 1969, 1, 49-68, b. Ekman, P., & Friesen, W. Hand movements. Journal of Com- munication, 1972, 22, 353-374. Ekman, P., & Friesen, W. Detecting deception from the body and face. Journal of Personality and Social Psychology, 1974, 29, 288-298. English, H. Reaction-time symptoms of deception. American Journal of Psychology, 1926, 37, 428-429. Exline, R., Thibaut, J., Hickey, B., & Gumpert, P. Visual interaction in relation to machiavellianism and an un- ethical act. In R. Christie and F. Geis (Eds.), Studies in machiavellianism, New York: Academic Press, 1970, 53—75. 170 Fay, P., & Middleton, W. The ability to judge truthtelling, or lying, from the voice as transmitted over a public address system. Journal of General Psychology, 1941, 24, 211-215. Freeberg, N. E. Relevance of rater-ratee acquaintance in the validity and reliability of ratings. Journal of Applied Psychology, 1969, 56:6, 518-524. Freedman, J. Role playing: Psychology by consensus. Journal of Personality and Social Psychology, 1969, l3, ‘6 / 107—114. eizer, R., Rarick, D., & Soldow, G. Deception in judgment accuracy: A study in person perception. Unpublished manuscript, University of Minnesota, 1975. Goldstein, E. Reaction times and the consciousness of de- ception. American Journal of Psychology, 1923, 34, 562-581. Gustafson, L., & Orne, M. The effects of task and method of stimulus presentation on the detection of deception. Journal of Applied Psychology, 1946, 48, 383-387. Gustafson, L., & Orne, M. Effects of heightened motivation of the detection of deception. Journal of Applied Psy- chology, 1963, 47, 408-411. Gustafson, L., & Orne, M. Effects of perceived role and role success on the detection of deception. Journal of Applied Psychology, 1965, 49, 412-417. Hastorf, A. H., Schneider, D. J., & Polefka, J. Person per- ception. Reading, Massachusetts: Addison-Wesley, 1970. Hildreth, R. An experimental studygof audience' ability to distinguish between sincere andiinsincere speeches. Un- published doctoral dissertation, University of Southern California, 1953. Hocking, J. Detecting deceptive communication from verbal, visual, and paralinguistic cues: An exploratory experi- ment. Unpublished doctoraIidissertation, Michigan State University, 1976. Hocking, J., Bauchner, J., Miller, G., & Kaminski, E. 2g- tecting deceptive communication from verbal, visual, and paralinguistic cues. Unpublished manuscript, Michigan State University, 1976. 171 Horvath, F. Verbal and nonverbal clues to truth and decep- tion during polygraph examinations. Journal of Police Science and Administration, 1973, 1, 138-152. Jones, E., & Davis, K. From acts to dispositions: The attribution process in person perception. In L. Berko- witz (Ed.), Advances in experimental social psychology, New York: Academic Press, 1965, 2, 219-2661 Jones, E. E., & Harris, V. A. The attribution of attitudes. Journal of Experimental Social Psychology, 1967, 3, 1-24. Jones, E., Cancus, D., Kelley, H., Nisbett, R., Valens, S., & Weiner, B. (Eds.). Attribution. New Jersey: General Learning Press, 1972. J6reskog, K., & Van Thillo, M. Lisrel: A general computer program for estimatingta linear structural equation system involving multiple indicators of unmeasured variables. New Jersey: Educational Testing Service, 1972. Keating, J., & Brock, T. Acceptance of persuasion and the inhibition to counterargumentation under various distrac- tion tasks. Journal of Experimental Social Psychology, 1974, 10, 301-309. ' Knapp, M. Nonverbal communication in human interaction. New York: Holt, Rinehart and—Winston, Inc., 1972. Knapp, M. L., & Harrison, R. P. Observing and recording non- verbal data in human transactions. Unpublished manu- script, Michigan State University, 1972. Knapp, M., Hart, R., & Dennis, H. An exploration of deception as a communication construct. Human Communication Re- search, 1974, 1, 15-29. Landis, C., & Wiley, L. E. Changes of blood pressure and respiration during deception. Comparative Psychology, 1926, 6:1, 1-9. Lindquist, E. Design and apalysis of experiments in psychol- ogy and education. Boston: Houghton, Mifflin, 1953. Maier, N., & Thurber, J. Accuracy of judgments of deception when an interview is watched, heard, and read. Personal Psychology, 1968, 21, 23-30. Maier, N. Sensitivity to attempts at deception in an inter- view situation. Personnel Psychology, 1965, 19, 55-65. 172 Maier, N., & Janzen, J. The reliability of persons used in making judgments of honesty and dishonesty. Perceptual and Motor Skills, 1967, 25, 141-151. Marston, W. Reaction-time symptoms of deception. Journal of Experimental Psychology, 1920, 3, 72-87. Maselli, M. D., & Altrocchi, T. Attribution of intent. Psy- chological Bulletin, 1969, 71, 445-454. Matarazzo, J., Wiens, A., Jackson, R., & Janaugh, T. Inter- viewer speech behavior under conditions of endogenously present and exogenously-induced motivational states. Journal of Clinical Psychology, 1970, 26, 141-148. McArthur, L. Appropriateness of the behavior and consensus and distinctiveness information as determinants of actors' and observers' attributions. Unpublished manu- script, Yale UniVersity, I970. McArthur, L. The how and what of why: Some determinants and consequences of causal attribution. Journal of Personality and Social Psychology, 1972, 22:2, 171-193. McClintock, C., & Hunt, R. Nonverbal indicators of affect and deception in an interview setting. Journal of Applied Social Psychology, 1975, 5, 54-67. McGarth, J., & Altman, I. Small group research. New York: Holt, Rinehart and Winston, I966. Mehrabian, A. Nonverbal betrayal of feelings. Journal of Experimental Research in Personality, 1971, 5, 64-75, a. Mehrabian, A. Silent messgges. California: Wadsworth Publishing Company, Inc., 1971. Mehrabian, A. Some referents and measures of nonverbal be- havior. Behavior Research Methods and Instrumentation, 1969, 1, 203-207. - Miller, G., & Burgoon, M. New techniques of persuasion. New York: Harper and Row, 1973. Miller, G., & Steinberg, M. Between people: A new analysis of interpersonal communication. Chicago: Science Re- search Associates, Inc., 1975. Motley, M. Acoustic correlates of lies. Western Speech, 1974, 37, 81-87. 173 Namboodiri, N., Carter, L., & Blalock, H. Applied multi- variate analysis and experimental designs. New York: McGraw-Hill, 1975. Nie, N., Hull, C., Jenkins, J., Steinbrenner, K., & Bent, D. Statistical package for the social sciences. New York: McGraw-Hill, 1970. Nisbett, R., Caputo, C., Legant, P., & Marechek, J. Be- havior as seen by the actor and as seen by the observer. Journal of Personality and Social Psychology, 1973, 27, 154-164. Osterhouse, R., & Brock, T. Distraction increases yielding to propaganda by inhibiting counterarguing. Journal of Personality and Social Psychology, 1970, 15, 344-358. Rosenthal, R. Covert communication in the psychological experiment. Psychological Bulletin, 1967, 67, 356-367. Rosenthal, R. Experimenter expectancy and the reassuring nature of the null hypothesis decision procedure. Ps - chological Bulletin, 1968, 70, 30-47. Ryan, M. The influence of teleconferencing medium and status on participants' perception of the aestheticism, evalua- tion, privacy, potency, and activity of the medium. Human Communication Research, 1976, 2:3, 255—261. Schachter, S., & Singer, J. E. Cognitive, social and physio- logical determinants of emotion state. Psychological Review, 1962, 69, 379-399. Silverman, F., & Johnston, R. Direct interval-estimation: A ratio scaling method. Percgptual and Motor Skills, 1975, 41, 464-466. Shulman, G. An experimental study of the effects of receiver sex, communicator sex, apgiwarning on the ability of receivers to detect deceptive communicators. Unpublished masters' thesis, Department of Speech Communication, Purdue University, 1973. Stein, C. A. A simple user's guide to LISREL. Unpublished manuscript, University of Massachusetts, 1976. Storms, M. Videotape and the attribution process. Journal of Personality and Social Psychology, 1973, 27, 165-175. Taguiri, R. Person perception. In G. Lindzey and E. Aronson (Eds.), The handbook of social psychology, Vol. 3. Reading, Massachusetts: Madison Wesley, 1969, 395-449. 174 Thackray, R., & Orne, M. A comparison of physiological indices in detection of deception. PsychOphysiology, 1968, 4, 329-339, a. Thackray, R., & Orne, M. Effects of the type of stimulus employed and the level of subject awareness on the detection of deception. Journal of Applied Psychology, 1968, 52, 234-239, b. Thorndike, R. L., & Hagen, E. Measurement and evaluation in psychology and education. New York: Wiley, 1961. Trovillo, P. A history of lie detection. Journal of Criminal Law and Criminology, 1939, 29,4848-881. Trunbull, A., Strickland, L., & Shaver, K. Medium of com- munication, differential power, and phasing concessions: Negotiating success and attributions to the Opponent. Human Communication Research, 1976, 2:3, 262-270. Weston, J., & Kristen, C. Teleconferencing: A comparison of attitudes, uncertainty and interpersonal atmospheres in mediated and face-to-face group interaction. Pre— pared for the Department of Communications, Ottawa, Canada, 1973. Wiener, M., & Mehrabian, A. Language within language: Immediacy, a channel iq_verbaI communication. New York: Appleton-Century-Crofts, 1968. Winer, B. StatiStical principles in experimental desigp. New York: McGraw-Hill, 1962. Wile, I. Lying as a biological and social phenomenon. The Nervous Child, 1942, 1, 293-313. Wolk, R., & Henley, A. The rtght_tc lie. New York: Wyden, 1970. Woodworth, R. S., & Schlosberg, H. Egperimental psychology. New York: Holt, 1954.