Part V: Control

Large file: Patience please.

Intrapersonal communication is used in managing communication interaction. Sense-making, compliance gaining, control and communication apprehension are areas of study which relate well to intrapersonal processing. In this section, you can examine some of the ways that we manage interaction through intrapersonal communication.

Interpersonal Deception : XIII, Suspicion and the Truth-Bias of Conversational Participants (David B. Buller and Frank G. Hunsaker)

Making Sense of Interpersonal Conflict: Interpersonal Communication Effects on Intrapersonal Sense-Making Processes (Denise Haunani Cloven and Michael E. Roloff)

The Role of Nonverbal Communication in Compliance Gaining: Research and Uses (Andrew F. Hayes and Judith A. Barnes)

Topics, Turns, and Interpersonal Control: Using Serial Judgment Methods (Mark T. Palmer and Abby M. Lack)

The Effect of Information Processing Objectives on Persuasion (Lori A. Roman and Mark A. deTurck)

The Impact of Communication Apprehension, Gender, and Time on Perceptions and Behavior in Extended Interactions (Joe Ayres, Tim Hopf, Kevin Brown, and Julia Suek)

The Effects of Planned and Spontaneous Delivery on Familiar and Unfamiliar Lies: When Does it Pay to Plan? (Patrick diBattista)

Compliance-Gaining Behavior: An Analysis of Cognitive Processes Underlying Message Production (Elizabeth M. Goering and Melissa M. Spirek)

Interpersonal Deception: XIII. Suspicion and the Truth-Bias of Conversational Participants

David B. Buller and Frank G. Hunsaker

About the Authors: David B. Buller, Ph. D. is associate professor of Communication at the University of Arizona, Tucson, AZ 85721, telephone (602)621-1366, where Frank G. Hunsaker, M. A. is a doctoral candidate This project was funded in part by the U. S. Army Research Institute (Contract #MDA903-90-K-0113). The views, opinions, and/or findings in this report are those of the authors and should not be construed as an official Department of the Army position, policy, or decision. See the reference list for the series of articles on interpersonal deception written by these and other authors.

Abstract: Past experiments showed that (a) conversational participants attribute more honesty to conversational partners than do observers, (b) cognitive demands of conversational management may impair their deception detection abilities, and (c) participants rely on facial cues when evaluating honesty. This experiment investigated whether suspicion would alter these differences between participants and observers. Ninety-two interactions from a prior experiment examining suspicion and probing were shown to 92 observers. As expected, participants attributed more honesty to both truthtellers and deceivers than did observers. This truth-bias persisted even when they were informed that deception had been manipulated in the experiment. Participants, though, were no less accurate at detecting deception, perhaps because sources' anti-detection strategies mislead observers. Suspicion had little impact on participants' truth-bias, but it did produce a lie-bias in observers. Participants showed a facial primacy when evaluating source veracity, but channel reliance was not affected by suspicion.

Perspective:

1. How do conversational tasks such as impression management, turn-taking, encoding and decoding messages affect the focus and perceptions of conversational participants compared to outside observers of the conversation?

2. Which strategy is better for catching liars, confronting them yourself or watching as a friend confronts them?

3. Does suspicion improve your ability to detect deception?

Most research on deception detection has focused on the detection skills of observers rather than on participants in deceptive conversations (Brandt, Miller & Hocking, 1980, 1982; deTurck & Goldhaber, 1988; Ekman & Friesen, 1969, 1974). While these experiments have provided many interesting and informative results, they have ignored the mutual influence of both source and receiver. Buller, Burgoon, and their colleagues (Buller & Burgoon, in press; Buller, Comstock, Aune, & Strzyzewski, 1989; Buller, Strzyzewski, & Comstock, 1991; Buller, Strzyzewski, & Hunsaker, 1991; Burgoon, Buller, Dillman, & Walther, 1992; Burgoon & Buller, 1994a, 1994b; Hunsaker, 1991) have developed an Interpersonal Deception Theory that views deceptive messages as typically embedded within a mutually negotiated interaction context of moment-to-moment strategic moves and countermoves reflecting each participant's motives for engaging in interaction. Judgments of observers, who are not entangled in this web of mutual influence, may have limited generalizability to conversational participants.

Interpersonal Deception Theory (IDT) (see Buller & Burgoon, in press; Burgoon & Buller, 1994a and b; Buller, Strzyzewski, & Comstock, 1991; Burgoon et al., 1992) holds that interactive deception differs fundamentally from noninteractive deception. In particular, messages related to credibility are a central feature of interpersonal exchanges and deviations from truthful behavior cause receivers to become suspicious. This suspicion is manifested in a combination of strategic and nonstrategic behavior which can be detected by senders (i.e., truthtellers or deceivers). When senders perceive suspicion, they alter their behavior to convey a truthful demeanor and allay suspicion. Thus, deceptive interactions contain strategic moves and countermoves by both deceiver and deceived. Consequently, research on observers of deceptive conversations may not be entirely applicable to conversational participants (Buller, Strzyzewski, & Hunsaker, 1991; Hunsaker, 1991).

The Truth-Bias of Participants

A potentially important difference between conversational participants and observers is participants' tendency to attribute more honesty to another than observers (Buller, Strzyzewski, & Hunsaker, 1991). Participants may base their judgments on conventions or heuristics rather than on the behaviors manifested by senders (Fiedler & Walka, 1993; McCornack & Parks, 1986; Stiff et al., 1989). Truthfulness is a fundamental and essential maxim of conversation, allowing communicators to infer indirect meaning rather than relying entirely on what is made explicit (Clark & Clark, 1975; Grice, 1975; Kraut & Higgins, 1984). It may be particularly useful to participants as they manage on-going conversations. Participants have many pressing conversational tasks: They must encode and decode messages, provide relevant feedback, stay on topic, negotiate relational dynamics, manage their images, and maintain conversational continuity through appropriate turn-taking (Buller, Strzyzewski, & Hunsaker, 1991; Burgoon & Newton, 1991; Street, Mulac, & Wiemann, 1988). By comparison, observers have few conversational responsibilities, save the decoding of incoming messages. Thus, they are more capable of focusing on the sender's performance when judging veracity, while participants choose to assume truthfulness in order to cope with conversational demands (Buller, Strzyzewski, & Hunsaker, 1991; Ekman & Friesen, 1969).

Fiedler and Walka (1993) have identified several conventions that people rely on to form veracity judgments including infrequency (infrequent or unexpected responses are judged as more dishonest), falsifiability (falsifiable information is evaluated as more dishonest than information that cannot be easily falsified), and conspicuousness (more conspicuous nonverbal cues are incorporated in judgments of honesty than subtle cues). Buller, Strzyzewski, and Hunsaker (1991) suggest that the fundamental attribution erroróthe tendency of participants to attribute their own behavior to situational factors and observers to attribute participants' behavior to their dispositionsómay account for the higher honesty judgments of observers is consistent with Fiedler and Walka's conspicuousness heuristic. Environmental stimuli may be more conspicuous to participants than to observers, making situational attributions more plausible (Jones & Nisbett, 1971; Nisbett, Caputo, Legant, & Maracek, 1973; Storms, 1973). Participants also may hesitate to make negative dispositional evaluations because the relationship with the sender is more important, at least temporarily, to them than to observers (Buller, Strzyzewski, & Hunsaker, 1991; Burgoon & Newton, 1991).1

Buller, Strzyzewski, and Hunsaker's (1991) finding that participants overlook or ignore deception cues, while observers may recognize deception cues and feel unobliged to avoid attributions of deceit is consistent with this reasoning. Participants, as a result, attribute more honesty to sources than observers. Buller, Strzyzewski, and Hunsaker (1991) also found that observers were more accurate detectors than conversational participants. They speculated that the conversational demands on participants instantiated the truth-bias heuristic and caused them to overlook cues to deception. By contrast, the fewer conversational demands on observers, the lack of the truth-bias heuristic, and the observational setting may have made observers more sensitive to behavioral changes linked to deception. Notably, participants relied more on facial cues, which may be more conspicuous due to their close physical proximity to the sender (Hall, 1973), whereas observers relied more on vocal cues when judging veracity.

Recently, however, Hunsaker (1991) failed to confirm the general truth-bias of conversational participants and their insensitivity to deception cues, perhaps due to his low power to detect the relatively small effect sizes (.03 to .06). A small effect size questions whether the differences between participants and observers are meaningful and sufficient to consider deception detection during conversational participation dissimilar from detection while observing a conversation. These small effects correspond to a minimum binomial difference of .42 versus .58 and a maximum binomial difference of .38 versus .62 (Rosenthal & Rosnow, 1984) and appear to be too large to ignore (Buller, Strzyzewski, & Hunsaker, 1991).

These conflicting findings require another look at the honesty judgments of conversational participants and observers. This experiment, then, investigated the predictions that participants will have a stronger truth-bias (H1) and be less accurate detecting deception (H2) than observers.

Suspicion

In IDT, receivers' suspicions are an important aspect of deceptive conversations. Suspicion affects receivers' actions and reactions to messages from senders in interpersonal exchangesóaltering probing questions (Buller, Strzyzewski, & Comstock, 1991) and attendant nonverbal and verbal behavior (Burgoon et al., 1992; Buller, Burgoon, Buslig, & Roiger, 1992)óand honesty attributions, producing a lie-bias and greater uncertainty in honesty judgments (Buller, Strzyzewski, & Comstock, 1991; Burgoon, Buller, Ebesu, & Rockwell, 1994; Toris & DePaulo, 1985).

Conversational participants may be at a greater disadvantage relative to observers when their suspicions are aroused. Suspicion places additional cognitive demands on participants who already have many conversational responsibilities. Consequently, participants may be less capable of surveying in-coming communication and judging its veracity while maintaining a smooth, coherent conversation than are observers. Buller, Strzyzewski, and Comstock (1991) reported that suspicious participants had longer response latencies, were more deliberate in their verbal presentation, and made more speech errors than nonsuspicious participants, all signs of increased cognitive activity. Similarly, Burgoon et al. (1992) found that suspicious conversational participants made a poorer impression and did not manage their conversations as well as nonsuspicious participants. These competing demands placed on conversational participants by suspicion and the resulting uncertainty caused by their limited ability to judge veracity may cause participants to abandon their truth-bias and rely instead on a heuristic that assumes dishonesty, thereby manifesting a lie-bias.

By contrast, observers should have considerable untapped cognitive reserves to analyze in-coming communication and determine its validity. Therefore, suspicious observers should find a lie-bias less necessary. Further, their detection abilities may improve substantially as suspicion triggers closer surveillance of the partners' communication and increase the salience of deception as a possible explanation for unexpected or abnormal behavior, resulting in greater detection accuracy.

It is possible, though, for suspicion to aid conversational participants. Suspicion may elicit detection strategies unavailable to outside observers, like exerting more control over the conversation (Burgoon et al., 1992; Toris & DePaulo, 1985) or attempting to mislead the partner into believing that the participant is not suspicious of them (Buller, Strzyzewski, & Comstock, 1991). Such detection strategies provide less, or erroneous, feedback on deception success to the sender and reduce the sender's ability to hide deceit. However, such moves increase conversational complexity, nullifying their benefits to participants.

Another potential advantage of suspicion is that it may attenuate participants' facial primacy (i.e., over-reliance on facial cues) when judging deceit (Buller, Strzyzewski, & Hunsaker, 1991) in favor of more telltale vocal cues (Buller & Burgoon, 1994; Zuckerman, Spiegel, DePaulo, & Rosenthal, 1982; Zuckerman & Driver, 1985). Suspicion also may make them sensitive to nonverbal channel discrepancies (Zuckerman et al., 1982).

The present experiment, then, investigated whether suspicion had differential effects on conversational participants' and observers' judgments of veracity, in terms of a truth- or lie-bias and detection accuracy (RQ1). It also investigated whether suspicion reduces participants' reliance on facial cues when judging deception (RQ2). Finally, the experiment attempted to replicate that nonsuspicious conversational participants generally rely more on facial cues than nonsuspicious observers when judging honesty (H3).

Method

Overview

Ninety-two videotaped interviews between strangers from a study investigating the effects of probing and suspicion were evaluated by observers (Buller, Strzyzewski, & Comstock, 1991). Half the interviewees were instructed to lie in the interactions. After one minute of interaction, half of the participants (i.e., interviewers) probed the interviewee for additional information to help clarify their responses. Further, half of the participants were induced to be suspicious of sources' answers. Likewise, half of observers were induced to be suspicious; suspicion condition was matched for participants and observers. Participants and observers judged the honesty of interviewees immediately following the interviews and again after being informed of the deception manipulation.

Subjects

The original study from which the interviews were obtained involved four hundred twenty-three (423) lower-division undergraduates from a large southwestern university. One hundred eighty-four (184) students were recruited individually and paired to form stranger dyads and one hundred eighteen brought a friend to form friendship dyads (118). Only the 92 stranger dyads were included in the present study. Ninety-two (92) undergraduate students from the same southwestern university were recruited from lower-division communication courses to act as observers. All students participated voluntarily for extra credit in their classes.

Participant and Observer Pretest

Subjects assigned to the participant and observer roles completed Wheeless and Grotz' (1975) Generalized Trust Scale (alpha reliability =.72), F(1,209)=304.69, p<.05, eta2=.59 (friend M=64.06; stranger M=47.55).

Interviewee Pretest

To provide topics for discussion, interviewees completed Crowne and Marlowe's (1964) Social Desirability Scale with a true or false answer format.

Participant Procedures

Upon arrival at the interaction laboratory, subjects were told that the study examined how people communicate with friends and strangers. They were randomly assigned to interviewer (the "participant" in this analysis) and interviewee roles and to deception and suspicion conditions.

In a separate room, the interviewee completed the pretest. Meanwhile, the participant was seated on a sofa approximately 3.5 feet in front of a swivel chair and completed the pretest. A video camera filmed the interviewer from across the room; a camera behind a one-way mirror recorded the interviewee.

In the nonsuspicious, probing condition, the participant was provided a blank copy of the Social Desirability Scale and instructed to question the interviewee about her or his responses. The participant was told that the study was concerned with how people respond to different styles of questioning. The participant was instructed to manipulate question-asking behavior. Initially, the participant was to ask simply whether the interviewee answered true or false to each question. Then, on a signal from the experimenter (i.e., the experimenter walking across a doorway to an adjoining room) that could only be seen by the participant, the participant was to probe for additional information as to why the interviewee answered each question true or false, until the participant felt she or he understood the reasons for the answer.

In the nonsuspicious, non-probing condition, the participant was told to ask for the true-false response to each question and not to ask any follow-up questions. The participant did not know the significance of the experimenter walking across the doorway although it still occurred to maintain equivalence.

In the suspicious, probing condition, the participant was told to begin the interview by asking the source about her or his answers in the same simple question-response format without follow-up questions. The experimenter then asked the participant to help with a problem in the study. The participant was told that the experiment was being conducted because data from an earlier study had been discarded because some interviewees lied during the interactions. To ensure that the same problem did not arise again and, if it did, to help the experimenter understand this behavior, the experimenter had the interviewee answer a second pretest, containing very similar but not identical questions. The experimenter would check to see if the interviewee's responses in the interview corresponded with the answers on the second pretest. If they did not, the interviewee might be lying. If this happened, the experimenter wanted the participant to ask in-depth questions following each response (i.e., probes) to find out whether the person was being honest or not. The experimenter would signal the participant to begin probing responses by walking across the doorway to the adjoining room. This cue induced probing and suspicion.

In the suspicious, non-probing condition, the participant was given the same suspicion induction, except that she or he was not told to ask any follow-up questions. Instead, on the experimenter's signal, she or he was to be careful of the interviewee's answers. Deception, then, appeared to be an interviewee decision, not an experimental manipulation, to increase external validity.

Meanwhile, a second experimenter instructed the interviewee to discuss their answers to each question on the pretest. In the truth condition, the interviewee's last instruction was to be honest in all answers to the participant. In the deception condition, the interviewee was told that the primary interest in the experiment was to see if people could tell whether someone was answering truthfully on personality questions. The interviewee was asked to lie about her or his responses to all the true-false questions and provide false answers to any questions posed by the participant.

Following this last instruction, the interviewee was immediately escorted into the room where the participant was waiting. The experimenter briefly reviewed the instructions and told the participant to start the interview. After one minute, the experimenter walked across the doorway. After another minute, the experimenter stopped the interaction. The interviewee and participant completed posttests in separate rooms.

During the interaction, the second experimenter recorded the interviewee's true-false responses and verified that the deception manipulation was successful. For the entire experiment, 62 interviewees in the truth condition told the truth on all true-false questions (the 46 who did not averaged only 1.85 false answers/interview) and 70 deceivers answered all true-false questions dishonestly (the 32 who did not gave 1.94 true answer/interview). Further, truthtellers reported that 97.1% of the information they delivered was true, while deceivers said that 16.7% of their information was true, F(1,208)=93.72, p<.05, eta2=.31.

Observer Procedures

Observers arrived at the communication laboratory in groups of five at 15-minute intervals. The study was introduced as an investigation of how people interact in conversations in an interview setting. Each observer was seated at individual carrels equipped with VCRs and color monitors and instructed to complete the pretest.

The observer was told that she or he would be watching two people discussing choices that people make in social situations and that she or he would see and hear the person being interviewed (i.e., interviewee) but only be able to hear the interviewer (i.e., participant). (The camera perspective gave observers a view of interviewees over the shoulder of participants.) The experimenters were interested in the observers impressions of the interviewee. In the nonsuspicious condition, the observer was told to watch the interview in its entirety and then complete the posttest about the interviewee. In the suspicious condition, the observer was told that the interviews were from an earlier study in which some interviews had to be discarded because some of the interviewees did not answer questions honestly. The observer was told that the experimenter had discovered inconsistencies in the interview she or he was to watch, which may indicate that the interviewee was lying. The observer was told to watch the entire interview and then completed the posttest. After completing the posttest, the deception manipulation was revealed and observers judged the veracity of the interviewee again.

Posttests

Participant and Observer Posttest. A 35-item version of Burgoon and Hale's (1987) Relational Communication Scale (RCS) was completed by participants and observers following the interview. Attributions of honesty were assessed by the sum of "she or he was honest" and "she or he was insincere" (hereafter referred to as the "honesty rating") (alpha reliability =.77). Ratings on the trust dimension (referred to as the "trust rating") (alpha =.76) of the RCS provided a second measure of honesty attributions. After completing the RCS, the deception manipulation was revealed to participants and observers and they were asked to indicate whether they thought honesty or deception had occurred to obtain a final dichotomous measure of honesty (the "honesty judgment").

Embedded in the RCS was a question that asked whether participants and observers were suspicious of the interviewee. Unfortunately, suspicion condition did not affect participants rating of suspicion, F(1,178)=.68, p>.05, eta2=.003. Consequently, a follow-up study was performed on 59 dyads (32 strangers, 27 friends) in which 22 dyads completed a posttest measuring expectation of deceit (alpha =.53), attention to interviewee's behavior (alpha =.59), and motivation to detect deception (alpha =.62) immediately after receiving the suspicion induction and 37 dyads interacted until just after receiving the suspicion cue, at which point they completed the posttest. Participants in the suspicion condition were more attentive to interviewee behavior than in the no suspicion condition, F(1,43)=3.52, p<.05, eta2=.06 throughout the interview. Expectation of deceit, F(1,14)=4.58, p<.05, eta2=.11, and motivation to detect deception, F(1,14)=3.32, p<.05, eta2=.15, was evoked initially but dissipated after one minute (Fs=.00 in delayed posttest group). While not as strong and persistent as intended, the suspicion induction initially evoked vigilance, expectation of deceit, and detection motivation, and it altered participants subsequent conversational behavior (see Buller, Strzyzewski, & Comstock, 1991).

Interviewee posttest. Interviewees also completed the RCS (with two items measuring perceived suspicion), and questions verifying the deception manipulation: (a) whether they told the truth or lied and (b) percentage of responses that were truthful.

Measurement of Interviewee Behavior

Five pairs of undergraduate students coded the interviewees' nonverbal behaviors identified in past research as consistent deception indicators (Buller & Aune, 1987; Buller & Burgoon, in press, Buller et al., 1989): frequency of speech errors (Ebel's intraclass correlation for interrater reliability, r=.71), speaking turns (r=.76), responses to a question (r=.58), gazes (r=.75), head nods (r=.74), smiles (r=.76), head shakes (r=.77), and laughs (r=.92); counts of illustrators (r=.91) and adaptors (r=.60-.86) using Friesen, Ekman, and Wallbott's (1979) measurement system; duration of talking time (r=.94), response latency (r=.60), and time spent gazing (r=.87); and 5-point bipolar ratings of facial animation (r=.73) and facial pleasantness (r=.67). Counts of interruptions, talkovers, and shrug emblems were omitted due to low interrater reliabilities (Buller, Strzyzewski, & Comstock, 1991). Behaviors were coded in 15-second intervals, 45 seconds prior to and 45 seconds after the first probe.

Statistical Analysis

The effects of perspective (participant v. observer) (H1 and H2) and suspicion (RQ1) on honesty attributions were tested by 2 (perspective) X 2 (suspicion condition) X 2 (honesty condition) X 2 (probing condition) analysis of variance. One ANOVA examined honesty rating, while another examined the trust or receptivity rating. The final honesty judgment, after subjects were told about the deception manipulation, was analyzed by chi-square tests.

The prediction that participants would rely more on facial cues than observers when making honesty ratings (H3) and the potential influence of suspicion on channel reliance (RQ2) were tested by multiple regressions on honesty and trust ratings and by a multiple discriminant analysis on final honesty judgment. Three blocks of predictors were entered in the following order: (a) main effects of perspective, suspicion condition, and the 19 nonverbal behaviors, (b) two-way interactions between nonverbal behaviors and perspective, nonverbal behaviors and suspicion, and perspective and suspicion, and (c) three-way interactions between nonverbal behaviors, perspective, and suspicion. Predictors whose t- or F-tests were less than 1.00 were pooled into the error term and reduced models produced the final parameters.

Results

Hypothesis 1 and Hypothesis 2: Truth-bias and Detection Accuracy

Hypothesis 1: Truth-bias. As predicted in hypothesis 1, conversational participants attributed more honesty (M=10.01), F(1,167)=13.62, p<.05, eta2=.07, and trustworthiness (M=36.07), F(1,167)=5.02, p<.05, eta2=.05, than did observers (Ms=8.14 and 33.69, respectively). This truth-bias persisted even when subjects were informed about the deception manipulation: Participants felt that 71% of sources were honest, while observers considered only 52% of sources to be honest, Chi2(1, N=171)=6.34, p<.05, eta2=.04.

Hypothesis 2: Detection Accuracy. Hypothesis 2, predicting that participants would be less accurate detectors of deception than observers, was not supported in either honesty rating, perspective by deception F(1,167)=.29, p>.05, or trust rating, perspective by deception F(1,167)=1.34, p>.05. The analysis of honesty judgments after being informed of the deception manipulation also failed to show that participants were inferior detectors. Both participants (81% of truthtellers and 60% of deceivers judged to be honest), Chi2(1, N=92)=4.82, p<.05, eta2=.05, and observers (62% of truthtellers and 42% of deceivers considered honest), Chi2(1, N=79)=2.87, p=.09, eta2=.04, (judged fewer deceivers to be honest [52%] than truthtellers [72%], Chi2[1, N=171]=7.50, p<.05, eta2=.041). This general accuracy was also reflected in the fact that participants and observers considered deceivers to be less honest (deception M=8.57), F(1,167)=4.60, p<.05, eta2=.02, and trustworthy (M=33.80), F(1,167)=4.38, p<.05,eta2=.02, than truthtellers (Ms=9.58 and 35.94, respectively), prior to being informed about the deception manipulation.

RQ1: Effect of Suspicion on Honesty Attributions

Truth-bias. Suspicion had no effect on participants' truth-biases. There was no significant interaction between suspicion and perspective on honesty ratings, F(1,167)=.37, p>.05, or trust ratings, F(1,167)=1.09, p>.05,

Table 1

Significant Behavioral Predictors of Honesty Attributions

*p<.05

and it did not affect the truth-bias of participants once they were told about the deception manipulation, participants Chi2(1, N=92)=.08, p>.05, observers Chi2(1, N=79)=1.47, p>.05.Detection Accuracy. Prior to informing deceivers of the deception manipulation, suspicion did not affect the detection abilities of participants or observers: three-way interaction on honesty ratings, F(1,167)=.74, p>.05 and on trust ratings, F(1,167)=.23, p>.05. However, when warned about the deception manipulation, suspicion appeared to make observers less accurate than participants, at least when evaluating the honesty of truthtellers. Specifically, suspicious observers considered 44% of truthtellers to be honest whereas nonsuspicious observers considered 76% of truthtellers to be honest, Chi2(1, N=39)=4.13, p<.05, eta2=.10. By contrast, suspicion did not affect participants' evaluations of truthtellers (suspicious participants =79% honest; nonsuspicious participants =83% honest), Chi2(1, N=47)=.09, p>.05, observers' evaluations of deceivers (suspicious observers =44% honest; nonsuspicious observers =41% honest), Chi2(1, N=40)=.05, p>.05, or participants' evaluations of deceivers (suspicious participants=63% honest; nonsuspicious participants=58% honest), Chi2(1, N=45)=.14, p>.05.

In sum, the effects of suspicion were disappointing. It did not reduce the truth-bias of participants or improve the detection accuracy of observers during the conversations. Interestingly, it produced a lie-bias in observers when they were warned about the deception manipulation.

Hypothesis 3 and Research Question 2: Channel Reliance

H3: Facial Primacy. As Buller, Strzyzewski, and Hunsaker (1991) found,

Table 2

Correlations Between Honesty Attributions and Nonverbal Behaviors

*p<.05

participants displayed a facial primacy when judging deception relative to observers.

Table 1 presents significant regression and discriminant coefficients for the reduced models. Specifically, when rating honesty, participants were affected more by head shaking than were observers (Table 2). Participants continued to rely on facial cues more than did observers after being informed about the deception manipulation. In particular, participants' honesty judgments were associated more strongly with head nodding, smiling, head shaking, facial animation, and facial pleasantness than were observers' ratings (Table 3). In all cases, increases in these cues lead participants to judge sources as more honest.

Participants did not completely ignore vocal and body cues, particularly once they were informed of the deception manipulation. Participants' final honesty judgment was more strongly associated with response latencies than were observers' judgments, with longer latencies producing more honesty judgments (Table 3). Further, this judgment was related to brief body and long body adaptors and postural shifts. What is particularly noteworthy about these relationships is that conversational participants seemed to overlook body movement by interviewees, while observers considered body movement a sign of deception (Table 3). The same contradiction was true for gazing; increased gaze produced judgments of honesty among participants but judgments of deception among observers (Table 3).

RQ2: Effects of Suspicion. Suspicion did not have a clear, consistent, or substantial effect on channel reliance. Suspicion seemed to increase participants' reliance on head shaking, when rating honesty, and decrease their attention to brief head and face adaptors, when rating trust. Suspicion also decreased observers' attention to turn length, when rating honesty, and brief head and face adaptors, when rating trust (Table 2). Suspicion when coupled with being informed of the deception manipulation enhanced participants' attention to brief object adaptors, whereas it altered observers' interpretations of these adaptors. When not suspicious, observers associated these adaptors with truthfulness, but when suspicious, they associated them with dishonesty (Table 3).

Discussion

Participants' Truth-bias

Conversational participants have a greater truth-bias than observers. Truth is a fundamental maxim of conversations (Grice, 1975; Kraut & Higgins, 1984), and the presumption of truth may be instrumental when participants are enmeshed in a conversation that requires adept

Table 3

Mean Scores on Behaviors Predicting Honesty Judgment for

Interactions Between Nonverbal Behaviors and Perspective and Suspicion

conversational management and formulation of messages, and creates a relationship between the participant and conversational partner. In such instances, a truth-bias becomes a valuable heuristic freeing cognitive capacity to manage conversations and a product of the favorable relational climate that develops between interactants (Buller, Strzyzewski, & Hunsaker, 1991; Fiedler & Walka, 1993).

The strength of this conversational truth-bias is debatable. The present data, along with Buller, Strzyzewski, and Hunsaker's (1991) and Hunsaker's (1991) estimates suggest that perspective accounts for 4% - 7% of the variation in honesty attributions. While not large in absolute terms, it was greater than the effect of deception on honesty judgments in the current experiment and represents a maximum binomial difference of .37 versus .63 (Rosenthal & Rosnow, 1984).

Further, this truth-bias persists even when participants are warned about deception. Kraut and Higgins (1984) contend that participants often assume that sources actually adhered to conversational maxims even when given reason to believe otherwise. The truth-bias also may persist because conversational participants remember and interpret behaviors differently than observers when warned after the fact. Participants continued to rely more on facial behaviors to judge deception after the warning than on cues in the body or voice. Facial cues are less likely to reveal deception than vocal or body cues (Buller & Burgoon, 1994; Ekman & Friesen, 1969; Zuckerman, DePaulo, & Rosenthal, 1981; Zuckerman & Driver, 1985) and may have contributed to participants' continued belief that the interviewee was honest. Further, participants considered random body activity, like adaptors and postural shifts, to be signs of honesty, but observers felt they indicated deceit. Participants may expect a certain degree of random movement and consider it a sign of relaxation, while observers may believe that such movement implies arousal or anxiety. Thus, they continued to rely on conspicuous nonverbal cues and may not have had the skills to accurately judge the authentic deception cues that were available (Fiedler & Walka, 1993).

To summarize, participation introduces a consistent, albeit small, truth-bias in honesty evaluations. It may arise (a) as a means of coping with conversational demands, (b) from differences in the frame of reference, or (c) from differences in attention to, and memory for, conversational behavior.

Detection Accuracy

By contrast, conversational participation did not affect overall detection accuracy (although by assuming more truth, participants were less accurate when it came to identifying deceivers but more accurate when detecting truthtellers than were observers). The cognitive and communicative tasks related to conversational management did not apparently impair participants' abilities to distinguish between truthtellers and deceivers in this experiment. They were able to make such distinctions just as much as observers, but they started with a greater presumption of truth than observers. This result may have stemmed in part from anti-detection strategies utilized by interviewees, such as reducing body movement when perceiving suspicion from the participant (Buller, Strzyzewski, & Comstock, 1991). It is possible that these changes misled observers, who were expected to be more sensitive to the interviewees' behaviors because of their fewer concomitant tasks, because observers focused more on body cues than participants. Consequently, they may have mistakenly felt that less active interviewees were more honest, reducing their detection advantage witnessed in Buller, Strzyzewski and Hunsaker's (1991) earlier study.

Effects of Suspicion on

Honesty Attributions

The effects of suspicion were disappointing but may not be too surprising given the earlier discussion of how communicators often ignore information suggesting that conversational maxims have been violated (Kraut & Higgins, 1984). During the conversations, suspicion neither aided observers by increasing their surveillance of interviewee behavior nor did it help participants enact successful detection strategies (although Buller, Strzyzewski, and Comstock, 1991, did report that suspicious participants in the original experiment encoded less skeptical probing questions).

Contrary to popular belief, past studies have provided little evidence for the benefits of suspicion in deception detection. It seems that the most common effect of suspicion on receivers is to evoke a lie-bias. It does not appear that greater detection motivation and vigilance (the two aspects of suspicion most strongly manipulated in this study) produce real improvements in people's abilities to evaluate a source's honesty or dishonesty. Instead, the expectation of deceit may have the most pervasive effect on attributions.

Notably, the truth-bias created by conversational participation overrode the lie-bias associated with suspicion. Participants were unaffected by the suspicion induction. It may be that the conversational tasks required them to assume truth in order to enact a smooth conversation. Or, it may be that the anti-detection strategies enacted by senders mislead participants.

Suspicious observers, however, showed the lie-bias characteristic of suspicion, but interestingly only when watching truthtellers. Perhaps, truthtellers engaged in fewer anti-detection strategies than deceivers (Buller, Strzyzewski, & Comstock, 1991) and, therefore, exhibited some behaviors that observers consider signs of dishonesty, like more random body movement, reinforcing the lie-bias. Deceivers, though, may have attenuated observers' lie-bias, because they muted their bodily activity.

Channel Reliance in Honesty Attributions

Participants' Facial Primacy. The facial primacy in participants' honesty attributions reported by Buller, Strzyzewski, and Hunsaker (1991) emerged again in this experiment, contradicting past studies that have reported a visual primacy by observers (Berman, Shulman, & Marwit, 1976; DePaulo, Rosenthal, Eisenstat, Rogers, and Finkelstein, 1978; Stiff & Miller, 1986; Stiff, Kim, & Ramesh, 1988; Storms, 1973). It may be that conversational demands make it harder for participants than observers to shift away from facial cues (Burgoon, Buller, & Woodall, 1989). Also, the physical immediacy of interaction may make facial cues more conspicuous because they take up more of the visual field (Hall, 1973). By contrast, observers of videotaped conversations see the sender at a farther distance in which facial cues do not overwhelm the observer's focus.

Reliance on facial cues is likely to mislead a receiver when judging deception, because they are well controlled by communicators (Buller & Burgoon, in press; Ekman & Friesen, 1969; Zuckerman et al., 1981; Zuckerman & Driver, 1985). Therefore, participants' truth-bias may be in part a perceptual bias as well as a conversational assumption or cognitive heuristic. Participants believe that most conversational partners are truthful because they focus on the channel which provides the least accurate cues to honesty.

Finally, observers did not rely on vocal cues as much as they appeared to in Buller, Strzyzewski, and Hunsaker's (1991) experiment. This may have occurred because fewer vocal cues were analyzed in this experiment due to reliability problems with measures of vocal behavior.

Effects of Suspicion. Suspicion had almost no effect on channel reliance, and the few relationships that emerged revealed no systematic pattern. Moreover, the suspicion manipulation did not mimic the postwarning that deception had been manipulated in the experiment. This is somewhat surprising, as the two situations increased the expectation of deception. Telling a receiver that a source may choose to deceive (i.e., the suspicion induction) may not evoke the same sort of suspicion as telling a receiver that a source was required to either deceive or tell the truth in the experiment (i.e., the postwarning). Deception is less certain with the former than the latter induction. This uncertainty may reduce suspicions' impact on cognitive processing of subsequent communication in the former induction.

This raises concerns over the generalizability of studies in which subjects are told ahead of the communication that the source will either be deceiving or telling the truth. Suspicion in many day-to-day situations is accompanied by some uncertainty, especially if it is triggered by unexpected or unusual conversational behavior. Deception probably is rarely considered a 50-50 chance, like it often is presented in deception experiments (and it is possible with such inductions receivers consider deception much more likely than that). Uncertainty actually may work against detectors, because it may not engender sufficient vigilance to accurately detect deception, allowing the truth-bias in conversations to override suspicion. Consequently, participants attribute honesty despite a nagging sense that all is not completely right with the senders' communication.

Summary

This study further substantiates that judging deception from the vantage of a conversational participant is different from judging it as an outside observer. Conversational participation creates a small but consistent truth-bias, arising from several cognitive, perceptual, and memory differences.

Thus, our understanding of deceptive communication should not be based entirely on evidence from studies of observers because the conversational interplay creates a different frame of reference and places different requirements on participants. This further confirms our belief that an interpersonal approach to deception, such as reflected in our Interpersonal Deception Theory, will enhance our understanding of this common communicative phenomenon. However, conversational participation may not harm detection accuracy. Suspicion may be of limited utility to either conversational participants or observers and any lie-bias it creates seems easily overpowered by the truth-bias created by conversational participation.

Notes

1In the original conceptualization of the fundamental attribution error, conversational participants actually are a form of observer, because they are evaluating the behavior of the conversational partner, not their own behavior. However, several studies on interpersonal interaction have shown that the fundamental attribution error also produces more favorable attributions by conversational participants than by observers (Buller, Strzyzewski, & Hunsaker, 1991; Burgoon & Newton, 1991; Street, Mulac, & Wiemann, 1988).

2Probing condition was also entered into the analyses. Only one significant effect involving probing emerged. Probing decreased all receivers' (participants' and observers') abilities to detect deception (probing: truth M=9.43, deception M=9.57; no probing: truth M=9.71, deception M=7.61), probing by deception condition F(1,167)=4.11, p<.05, eta2=.02.

References

Berman, H. J., Shulman, A. D., & Marwit, S. J. (1976). Comparison of multidimensional decoding of affect from audio, video, and audio-video recordings. Sociometry, 39, 83-89.

Brandt, D. R., Miller, G. R., & Hocking, J. E. (1980). The truth-deception attribution: Effects of familiarity on the ability of observers to detect deception. Human Communication Research, 6, 99-119.

Brandt, R., Miller, G. R., & Hocking, J. E. (1982). Familiarity and lie detection: A replication and extension. Western Journal of Speech Communication, 46,276-290.

Buller, D. B., & Aune, R. K. (1987). Nonverbal cues to deception among intimates, friends, and strangers. Journal of Nonverbal Behavior, 11, 269-290.

Buller, D. B., & Burgoon, J. K. (1994). Deception: Strategic and nonstrategic communication. In J. A. Daly & J. M. Wiemann (Eds.), Strategic interpersonal communication. Hillsdale, NJ: Erlbaum.

Buller, D. B., & Burgoon, J. K. (in press). Interpersonal deception theory. Communication Theory.

Buller, D. B., Burgoon, J. K., Buslig, A., & Roiger, J. (1992). Interpersonal deception: X. The language of interpersonal deception. Communication Theory.

Buller, D. B., Comstock, J., Aune, R. K., & Strzyzewski, K. D. (1989). The effect of probing on deceivers and truthtellers. Journal of Nonverbal Behavior, 13, 139-204.

Buller, D. B., Strzyzewski, K. D., & Comstock, J. (1991). Interpersonal deception: I. Deceivers' reactions to receivers' suspicions and probing. Communication Monographs, 58, 1-24.

Buller, D. B., Strzyzewski, K. D., & Hunsaker, F. G. (1991). Interpersonal deception: II. The inferiority of conversational participants as deception detectors. Communication Monographs, 58, 25-40.

Burgoon, J. K., & Buller, D. B. (1994a). Interpersonal deception: III. Effects of deceit on perceived communication and nonverbal behavior dynamics. Journal of Nonverbal Behavior, 18, 155-184.

Burgoon, J. K., & Buller, D. B. (1994b, July). Interpersonal deception theory. Paper presented at the 5th International Conference on Language and Social Psychology, Brisbane, Australia.

Burgoon, J. K., Buller, d. B., Dillman, L., & Walther, J. B. (1992). Interpersonal deception: IV. Effects of suspicion on perceived communication and nonverbal behavior dynamics. Manuscript in submission.

Burgoon, J. K., Buller, D. B., Ebesu, A., & Rockwell, P. (1994). Interpersonal deception: V. Accuracy in deception detection. Communication Monographs, 61.

Burgoon, J. K., Buller, D. B. & Woodall, W. G. (1989). Nonverbal communication: The unspoken dialogue. New York: Harper/Collins.

Burgoon, J. K., & Hale, J. L. (1987). Validation and measurement of the fundamental themes of relational communication. Communication Monographs, 54, 19-41.

Burgoon, J. K., & Newton, D. A. (1991). Applying a social meaning model to relational interpretations of conversational involvement: Comparison of observer and participant perspectives. Southern Communication Journal, 56, 96-113.

Clark, H., & Clark, E. (1975). Psychology and language. New York: Harcourt Brace Jovanovich.

Crowne, D., & Marlowe, D. (1964). The approval motive: Studies in evaluative dependence. New York: Wiley.

DePaulo, B. M., Rosenthal, R., Eisenstat, R. A., Rogers, P. L., & Finkelstein, S. (1978). Decoding discrepant nonverbal cues. Journal of Personality and Social Psychology, 36, 313-323.

deTurck, M. A., & Goldhaber, G. M. (1988). Perjury and deceptive judgments: How the timing and modality of witness deception affects jurors' deceptive judgments. Communication Quarterly, 36, 276-289.

Ekman, P, & Friesen, W. V. (1969). Nonverbal leakage and clues to deception. Psychiatry, 32, 88-105.

Ekman, P, & Friesen, W. V. (1974). Detecting deception from the body or face. Journal of Personality and Social Psychology, 29, 288-298.

Fiedler, K., & Walka, I. (1993). Training lie detectors to use nonverbal cues instead of global heuristics. Human Communication Research, 20, 199-223.

Friesen, W. V. Ekman, P. , & Wallbott, H. (1979). Measuring hand movements. Journal of Personality and Social Psychology, 29, 288-298.

Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and semantics: Vol. 3, Speech acts (pp. 41-58). New York: Seminar.

Hall, E. T. (1973). The silent language. Garden City, NY: Anchor Books/Doubleday.

Hunsaker, F. G. (1991). Honesty attributions and deception detection accuracy of friends and strangers in the roles of participant and observer. Unpublished Masters Thesis, University of Arizona.

Jones, E. E., & Nisbett, R. E. (1971). The actor and the observer: Divergent perceptions of the causes of behavior. Morristown, NJ: General Learning.

Kraut, R. E., & Higgins, E. T. (1984). Communication and social cognition. In R. S. Wyer, Jr. & T. K. Srull (Eds.), Handbook of social cognition, Vol. 3 (pp. 88-127). Hillsdale, NJ: Erlbaum.

McCornack, S. A., & Parks, M. R. (1986). Deception detection and relationship development: The other side of trust. In M. L. McLaughlin (Ed.), Communication yearbook 9 (pp. 377-389). Beverly Hills, Sage.

Nisbett, R. E., Caputo, C., Legant, P. , & Maracek, J. (1973). Behavior as seen by the actor and as seen by the observer. Journal of Personality and Social Psychology, 27, 154-164.

Rosenthal, R., & Rosnow, R. L. (1984). Essentials of behavioral research: Methods and data analysis. New York: McGraw-Hill.

Stiff, J. B., Kim, H. J., & Ramesh, C. N. (1988). Truth biases and aroused suspicion in relational deception. Unpublished manuscript, Michigan State University.

Stiff, J. B., & Miller, G. R. (1986). "Come to think of it. . . .": Interrogative probes, deceptive communication, and deception detection. Human Communication Research, 12, 339-358.

Stiff, J. B., Miller, G. R., Sleight, C., Mongeau, P. , Garlick, R., & Rogan, R. (1989). Explanations for visual cue primacy in judgments of honesty and deceit. Journal of Personality and Social Psychology, 56, 555-564.

Storms, M. D. (1973). Videotape and the attribution process: Reversing actors' and observers' points of view. Journal of Personality and Social Psychology, 27, 165-175.

Street, R. L., Mulac, A., & Wiemann, J. M. (1988). Speech evaluation differences as a function of perspective (observer versus observer) and presentational medium. Human Communication Research, 14, 333-363.

Toris, D., & DePaulo, B. M. (1985). Effects of actual deception and suspiciousness of deception on interpersonal perceptions. Journal of Personality and Social Psychology, 47, 1063-1073.

Wheeless, L. R., & Grotz, J. (1975, May). Self-disclosure and trust: Conceptualization, measurement, and inter-relationships. Paper presented at the annual meeting of the International Communication Association, Chicago.

Zuckerman, M., DePaulo, B. M., & Rosenthal, R. (1981). Verbal and nonverbal communication of deception. In L Berkowitz (Ed.), Advances in experimental social psychology, 14. (pp. 1-59). New York: Academic.

Zuckerman, M., & Driver, R. E. (1985). Telling lies: Verbal and nonverbal correlates of deception. In A. W. Siegman & S. Feldstein (Eds.), Multichannel integrations of nonverbal behavior (pp. 129-148). Hillsdale, NJ: Erlbaum.

Zuckerman, M., Spiegel, N. H., DePaulo, B. M., & Rosenthal, R. (1982). Nonverbal strategies for decoding deception. Journal of Nonverbal Behavior, 6, 171-187.

Response from Leonard J. Shedletsky

An intriguing finding in this study is the difference between participants' and observers' perceptions of conversational features. The authors speculate on the reason for the difference. But the finding, especially since it is not the only study of communication behavior that shows a participant/observer difference in perception, deserves further study. How might we pursue the issue of participant/observer difference? And what alternative explanation can you offer for this finding? What implications does this finding hold for communication research?

Making Sense of Interpersonal Conflict: Interpersonal Communication Effects on Intrapersonal Sense-Making Processes

Denise Haunani Cloven and Michael E. Roloff

About the Authors: Denise Haunani Cloven is on the faculty of the Department of Communication Arts, University of Wisconsin-Madison, Vilas Communication Hall, 821 University Avenue, Madison, WI 53706 (telephone 608-262-2543). Michael E. Roloff is on the faculty of the Department of Communication Studies, Northwestern University, 1815 Chicago Avenue, Evanston, IL 60208 (telephone 708-492-7179).

Abstract: This article clarifies how people use intrapersonal communication to make sense of interpersonal conflicts and examines how this process is shaped by the interpersonal communication context in which sense-making occurs. Differences between intrapersonal sense-making in the presence versus absence of prior communication with the conflict partner are described. We then define three effects of anticipating interpersonal communication about conflicts on intrapersonal sense-making activities: the "cognitive tuning effect," the "accountability effect," and the "prevailing affect effect." The research reviewed suggests that interpersonal communication experiences and intention shape intrapersonal communication and the conclusions individuals reach when making sense of interpersonal conflicts.

Perspective:

1. How do interpersonal communication experiences and intentions influence intrapersonal communication directed toward making sense of conflicts?

2. What are the pros and cons of avoiding and engaging in interpersonal communication about conflicts in terms of intrapersonal sense-making processes?

3. How might the effects of interpersonal communication on intrapersonal sense-making result in or perpetuate biases in perceptions of interpersonal conflict?

Close interpersonal relationships are defined by a high degree of interdependence (Berscheid, Snyder, & Omoto, 1989). As a result of this interdependence, relationship partners have ample opportunity to interfere with, as well as facilitate, the accomplishment of personal goals (Altman & Taylor, 1973; Berscheid, 1983). Hence, intimate associations are also characterized by a heightened potential for interpersonal conflict.

When faced with interference from a relationship partner, individuals are likely to engage in intrapersonal communication directed toward comprehending the threat posed by the partner's behaviors. Ample evidence suggests that negative life events (Goodhart, 1985; Sarason, Potter, & Sarason, 1986), adverse interpersonal experiences (Camper, Jacobson, Holtzworth-Munroe, & Schmaling, 1988; Holtzworth-Munroe & Jacobson, 1985), interruptions to desired behavioral sequences (Millar, Tesser, & Millar, 1988), and unresolved disputes (Klos & Singer, 1981) induce a high degree of sense-making or ongoing thought. Moreover, intrapersonal processes influence the form and intensity of reactions to potential threats (Berkowitz, 1990; Berkowitz & Heimer, 1989), as well as an individual's personal well-being (Goodhart, 1985; Kendall, Howard, & Hays, 1989; Rush & Weissenburger, 1986).

The goal of this chapter is to describe intrapersonal processes associated with interpersonal conflict and to explicate the effects of interpersonal communication experiences and plans on these sense-making processes. First, the role of intrapersonal communication in making sense of conflict is reviewed. Then, we examine how interpersonal communication contexts influence intrapersonal sense-making activities.

The Sense-Making Process

Our model assumes that sense-making is evoked by the perception of potential threats to the self embodied in the disruptive behaviors of a close relationship partner. We further assume that the intrapersonal process of sense-making takes the form of an internal dialogue similar to interpersonal searches for causal explanation (cf. Mead, 1934; Hilton, 1990). In particular, intrapersonal interrogation involves an internal dialogue wherein the same individual is both the provider and receiver of explanations (Hilton, 1990).

Once instigated, sense-making activities have a primary goal of determining whether a threat actually exists and requires action (Berkowitz, 1990; Berkowitz & Heimer, 1989). In addition, issues addressed during sense-making are relevant to devising solutions to relational problems, should they be defined as sufficiently threatening. For example, individuals often try to evaluate the severity and cause of interpersonal problems, and these evaluations shape reactions to relational difficulties (Cloven, 1992; Mitchell & Wood, 1980; Roloff & Cloven, 1990; Rusbult, Johnson, & Morrow, 1986; Showers, 1988; Sillars, 1980).

Hilton (1990) argues that the form of questions addressed during an internal dialogue depends in part on the subjective goals salient to the covert interaction. Moreover, the form of these questions will determine the causal explanations derived (Hilton, 1990; Showers, 1988; Strack, Schwarz, & Gschneidinger, 1985). Hence, the remainder of this chapter explores how intrapersonal sense-making and corresponding perceptions of conflicts can be influenced by the interpersonal communication context within which sense-making occurs.

Interpersonal Communication Effects on Intrapersonal Sense-Making

Responses to interpersonal threats can involve communicating with others about problems (Planalp, Rutherford, & Honeycutt, 1988; Witteman, 1988) or refraining from communication (e.g., Roloff & Cloven, 1990). Interactions about a partner's disagreeable behavior may further vary in cooperativeness and directness (Rusbult, Zembrodt, & Gunn, 1982; Sillars, Coletti, Parry, & Rogers, 1982). We argue that whether interpersonal communication occurs, whether individuals anticipate interpersonal communication, and the character of conflict-related interactions affect intrapersonal sense-making and the corresponding perceptions of the severity and cause of relational difficulties (cf. Pennebaker, 1989; Pennebaker, Hughes, & O'Heeron, 1987).

Sense-Making in the Presence or Absence of Interpersonal Communication

When a person does not anticipate communication about a relational problem, her or his own perspective may be the only view represented in the internal dialogue. Individuals thinking about situations where partners have negatively affected them have little motivation to consider the others' perspectives (cf. Gould & Sigall, 1977, Regan & Totten, 1975, and Wegner & Finstuen, 1977). Hence, this process may be generally self-affirming and prone to demean the partner (cf. Baumeister, Stillwell, & Wotman, 1990).

Given the one-sidedness of internal dialogue in the absence of interpersonal communication activity, ongoing thought is likely to have negative personal and relational consequences. Indeed, Sadler and Tesser (1973) demonstrated that thinking in isolation about a negative social encounter increases hostility toward an interaction partner. Furthermore, individuals who report thinking most about relational disputes are also likely to report feeling depressed and lonely (Harvey, Wells, & Alvarez, 1978). Finally, Cloven and Roloff (1991) found that the frequency of intrapersonal communication focused on a roommate conflict was positively associated with perceptions of problem severity and partner responsibility for the conflict, and this positive correlation was greatest among respondents who reported a low level of interpersonal communication activity.

In contrast, internal dialogue that occurs in conjunction with frequent interpersonal activity will be more likely to incorporate alternative perspectives on a conflict. Specifically, talking with partners can make new information and alternative views directly available and thereby reduce self-serving biases (e.g., Gioia & Sims, 1986). Accordingly, Cloven and Roloff (1991) found that the frequency of interpersonal communication about roommate conflicts was negatively associated with perceived problem severity and blaming the roommate. Moreover, engaging in intrapersonal sense-making did not have the negative effects noted previously, provided individuals reported frequent interpersonal communication activity.

Sense-Making in Anticipation of Interpersonal Communication

We believe interpersonal communication contexts influence evaluations of conflicts because interpersonal goals change how individuals think about conflicts during sense-making. Hence, the effects of interpersonal communication should also be apparent when individuals engage in intrapersonal activity in anticipation of interpersonal interaction.

When individuals anticipate interaction about conflicts, intrapersonal communication may take a form that incorporates the expected partner implicitly or explicitly in an imagined interaction or rehearsal for confrontation (Edwards, Honeycutt, & Zagacki, 1988; Stutman & Newell, 1990). Such internal discussions can serve to organize thoughts, to improve message selection, to reduce anxiety, to build confidence, and to provide an opportunity to practice or to plan for the ensuing encounter (Edwards et al., 1988; Honeycutt, Zagacki, & Edwards, 1990; Stutman & Newell, 1990). Furthermore, empirical investigations have revealed that imagined interactions are commonly about conflicts, typically involve intimate partners, and often occur prior to interpersonal exchanges (Edwards et al., 1988; Honeycutt et al., 1990). We extend this work by identifying three effects of interpersonal communication on imagined interaction about conflict: a "cognitive tuning effect," an "accountability effect," and a "prevailing affect effect."

The Cognitive Tuning Effect. Traditionally, cognitive tuning effects refer to information processing differences associated with transmitting versus receiving information. A "transmission tuning" was argued to promote resolution of contradictory information and to discourage the search for knowledge, whereas a "reception tuning" was thought to induce maintenance of contradictory information and information searches (Cohen, 1961; Zajonc, 1960).

More recent research suggests that cognitive tuning effects depend on whether individuals expect to encounter new information in an anticipated interaction, regardless of their roles as transmitter or receiver (Hoffman, Mischel, & Baer, 1984; Higgins, McCann, & Fondacaro, 1982). Specifically, when people anticipating communication about a target person do not expect new information, both senders and receivers construe the target's behavior in terms of global traits, dispositional attributions, and interpretations of actions (Hoffman et al., 1984; Higgins et al., 1982). In contrast, senders and receivers anticipating interactions involving new information about a target produce descriptive, neutral, and non-valuative messages (Higgins et al., 1982).

Individuals who communicate with others about events that produce relational uncertainty report that during such interactions they seek to gather information (Planalp et al., 1988). Hence, anticipating interaction about interpersonal problems should induce a "reception tuning" in which individuals think descriptively and examine or maintain a variety of ideas. Consistent with these arguments, Cloven and Roloff (1990) found that individuals anticipating communication with a friend or roommate about a roommate conflict reported a greater proportion of descriptive thoughts and a greater variety of thoughts during a think-aloud episode than did individuals who did not expect to communicate with anyone about their roommate conflicts.

The Accountability Effect. Anticipating interpersonal communication may also change the extent to which intrapersonal communication involves self-affirming versus self-deprecating thoughts. Venting frustration and gaining retribution are common goals associated with confronting partners (Newell & Stutman, 1991); therefore, intrapersonal communication preparing for interpersonal exchange may involve developing and rehearsing arguments. Conversely, imagined interactions may incorporate the partner's perspective in the form of explanations for behaviors, questions, or role reversal (Edwards et al., 1988; Honeycutt et al., 1990; Johnson, 1971).

Whether anticipating interpersonal communication about a relationship problem will evoke self-affirming rehearsal or self-critical perspective-taking will depend on whether individuals feel compelled to account for their views. Accountability effects refer to the tendency for people under social pressure to justify their views to think in a more complex, self-critical, or valuatively diverse manner (Tetlock, 1983; Tetlock, 1985a; Tetlock & Kim, 1987). Moreover, evidence that attribution biases are mitigated when people are made to feel accountable for those attributions (Tetlock, 1985a; Tetlock & Kim, 1987) suggests that accountability concerns promote consideration of a target person's perspective.

Although pressures to justify oneself are likely to be salient when people expect to communicate about conflicts (Tetlock, 1983), anticipating communication will not evoke accountability concerns equally for all persons. Accountability effects arise out of motivations to protect and enhance one's social image (Tetlock, 1981; Tetlock, 1985b). Thus, the accountability effect of anticipating interaction about conflict should be greatest among individuals who express concern with behaving in a socially appropriate manner, in general (e.g., Lennox & Wolfe, 1984).

Consistent with this view, Cloven and Roloff (1993) found that anticipating communication about roommate conflicts evoked self-deprecating thoughts about conflict situations contingent on a person's concern for social appropriateness. In other words, people who expressed concern for social appropriateness reported a greater proportion of self-deprecating thoughts when anticipating communication than did individuals who reported a low concern for behaving appropriately. In fact, anticipating interaction produced only a minimal accountability effect among individuals who expressed a low concern for following social rules.

The Prevailing Affect Effect. The two effects previously described emphasize the impact of impending interaction on intrapersonal communication; however, prior interpersonal experiences may also influence sense-making activities in anticipation of interaction. Sillars and colleagues (1982) identify three categories of behavior manifested in conflict interactions: Integrative behaviors emphasize cooperative problem solving, distributive behaviors are focused on competition and placing blame, and avoidant behaviors attempt to suppress the conflict interaction. Whereas Sillars (1980) has demonstrated that these conflict behaviors correspond significantly with conflict attributions and resolution, we extend his perspective by suggesting that the tenor of conflict interactions will also influence intrapersonal communication prior to further interaction.

People tend to judge integrative conflict behavior as more effective and more competent than distributive and avoidant tactics, and these judgments exert an indirect effect on relational qualities such as trust (Canary & Spitzberg, 1987). Moreover, people most prefer that partners express grievances to them in a calm and rational manner and least prefer complaints accompanied by yelling and personal attacks (Alberts, 1989). Similarly, when voicing their own complaints, individuals express a preference for partner responses characterized by acknowledgment or agreement, rather than yelling, arguing, or ignoring (Alberts, 1989).

We believe the affective climate of conflict interactions characterized by different conflict behaviors, coupled with evaluations of those experiences as desirable or undesirable, will have ramifications for intrapersonal sense-making processes. A person's mood can influence judgments and recall for interpersonal communication experiences (Forgas, Bower, & Krantz, 1984). Accordingly, prior conflict interactions should influence how individuals construct imagined interactions wherein they role-play conflict discussions with partners.

Although integrative and distributive behaviors both provide information about a partner's perspective, distributive strategies are perceived to be more competitive, nonsupportive, and critical (Sillars et al., 1982). Conversely, integrative conflict behaviors promote conflict resolution, minimize blaming processes, and involve positive or neutral affect (Sillars, 1980; Sillars et al., 1982). Not surprisingly, then, Cloven and Roloff (1991) found that individuals who characterized their conflict interactions with roommates as integrative defined their problems as significantly less serious. In contrast, the frequency of intrapersonal sense-making was positive correlated with perceived problem severity and blaming roommates for conflicts among individuals reporting distributive conflict interactions with roommates. Although Cloven and Roloff (1991) did not examine the effect of prior conflict interactions on the content of intrapersonal communication, the effects demonstrated for judgments of problem severity and cause suggest that interpersonal communication experiences may define the affective climate that prevails during intrapersonal sense-making.

Summary

The process by which individuals make sense of conflicts in close relationships exemplifies the conjunction of intrapersonal and interpersonal communication. Conflict experiences often stimulate extensive and sometimes uncontrollable intrapersonal activity ultimately oriented toward determining the nature and degree of threat posed by a partner's behavior. However, these intrapersonal processes do not operate in a vacuum. Rather, the frequency, likelihood, and character of interpersonal communication experiences can serve to change the very nature of intrapersonal communication and, accordingly, the evaluations of conflicts derived from the sense-making process. Thus, whereas intrapersonal communication is presumed by several scholars to be central to other communication processes, our perspective suggests that interpersonal communication reciprocally and importantly influences intrapersonal processes.

References

Alberts, J. K. (1989). A descriptive taxonomy of couples' complaint interactions. The Southern Communication Journal, 54, 125-143.

Altman, I., & Taylor, D. A. (1973). Social penetration: The development of interpersonal relationships. New York: Hold, Rinehart, & Winston.

Baumeister, R. F., Stillwell, A., & Wotman, S. R. (1990). Victim and perpetrator accounts of interpersonal conflict: Autobiographical narratives about anger. Journal of Personality and Social Psychology, 59, 994-1005.

Berkowitz, L. (1990). On the formation and regulation of anger and aggression. American Psychologist, 45, 494-503.

Berkowitz, L., & Heimer, K. (1989). On the construction of the anger experience: Aversive events and negative priming in the formation of feelings. In L. Berkowitz (Ed.) Advances in Experimental Social Psychology (Vol. 22, pp. 1-37). New York: Academic.

Berscheid, E. (1983). Emotion in close relationships. In H. Kelley, E. Berscheid, A. Christensen, J. J. Harvey, T. L. Huston, G. Levinger, E. McClintock, L. A. Peplau, & D. R Peterson (Eds.), Close relationships (pp. 110-168). New York: W. H. Freeman.

Berscheid, E., Snyder, M., & Omoto, A. M. (1989). Issues in studying close relationships: Conceptualizing and measuring closeness. In C. Hendrick (Ed.), Close relationships (pp. 63-91). Newbury Park, CA: Sage.

Camper, P. , Jacobson, N., Holtzworth-Munroe, A., & Schmaling, K. (1988). Causal attributions for interactional behaviors in married couples. Cognitive Therapy and Research, 12,195-209.

Canary, D. J., & Spitzberg, B. H. (1987). Appropriateness and effectiveness perceptions of conflict strategies. Human Communication Research, 14, 93-118.

Cloven, D. H. (1992). Conflict avoidance in personal relationships: An examination of motives. Unpublished doctoral dissertation, Northwestern University, Evanston, IL.

Cloven, D. H., & Roloff, M. E. (1990). Internal dialogue about interpersonal conflict: Communicative antecedents and relational outcomes. Paper presented at the annual meeting of the Speech Communication Association, Chicago, IL.

Cloven, D. H., & Roloff, M. E. (1991). Sense-making activities and interpersonal conflict: Communicative cures for the mulling blues. Western Journal of Speech Communication, 55, 134-158.

Cloven, D. H., & Roloff, M. E. (1993). Sense-making activities and interpersonal conflict II: The effects of communicative intentions on internal dialogue. Western Journal of Communication, 57, 309-329.

Cohen, A. R. (1961). Cognitive tuning as a factor affecting impression formation. Journal of Personality, 29, 235-245.

Edwards, R., Honeycutt, J. M., & Zagacki, K. S. (1988). Imagined interaction as an element of social cognition. Western Journal of Speech Communication, 52, 23-45.

Forgas, J. P. , Bower, G. H., & Krantz, S. E. (1984). The influence of mood on perceptions of social interactions. Journal of Experimental Social Psychology, 20, 497-513.

Gioia, D., & Sims, H. (1986). Cognition-behavior connections: Attribution and verbal behavior in leader-subordinate interaction. Organizational Behavior and Human Decision Processes, 37, 197-229.

Goodhart, D. (1985). Some psychological effects associated with positive and negative thinking about stressful event outcomes: Was Pollyanna right? Journal of Personality and Social Psychology, 48, 216-232.

Gould, R., & Sigall, H. (1977). The effects of empathy and outcomes on attribution: An examination of the divergent-perspectives hypothesis . Journal of Experimental Social Psychology, 134, 480-491.

Harvey, J. H., Wells, G., & Alvarez, M. (1978). Attribution in the context of conflict and separation in close relationships. In J. Harvey, W. Ickes, & R. Kidd (Eds.), New directions in attribution research (Vol. 2, pp. 235-260). Hillsdale, NJ: Lawrence Erlbaum.

Higgins, E. T., McCann, C. D., & Fondacaro, R. (1982). The "communication game": Goal-directed encoding and cognitive consequences. Social Cognition, 1, 21-37.

Hilton, D. J. (1990). Conversational processes and causal explanation. Psychological Bulletin, 107, 65-81.

Hoffman, C., Mischel, W., & Baer, J. S. (1984). Language and person cognition: Effects of communicative set on trait attribution. Journal of Personality and Social Psychology, 46, 1029-1043.

Holtzworth-Munroe, A., & Jacobson, N. S. (1985). Causal attributions of married couples: When do they search for causes? What do they conclude when they do? Journal of Personality and Social Psychology, 48, 1398-1412.

Honeycutt, J. M., Zagacki, K. S., Edwards, R. (1990). Imagined interaction and interpersonal communication. Communication Reports, 3, 1-8.

Johnson, D. W. (1971). Role reversal: A summary and review of the research. International Journal of Group Tensions, 1, 318-334.

Kendall, P. C., Howard, B. L., Hays, R. C. (1989). Self-referent speech and psychopathology: The balance of positive and negative thinking. Cognitive Therapy and Research, 13, 583-598.

Klos, D., & Singer, J. (1981). Determinants of the adolescent's ongoing thought following simulated parental confrontations. Journal of Personality and Social Psychology, 41, 975-987.

Lennox, R. D., & Wolfe, R. N. (1984). Revision of the self-monitoring scale. Journal of Personality and Social Psychology, 46, 1349-1364.

Mead, G. H. (1934). Mind, self, and society (Vol. 1). Chicago, IL: University of Chicago.

Millar, K. U., Tesser, A., & Millar, M. G. (1988). The effects of a threatening life event on behavior sequences and intrusive thought: A self-disruption explanation. Cognitive Therapy and Research, 12, 441-457.

Mitchell, T. R., & Wood, R. E. (1980). Supervisor's responses to subordinate poor performance: A test of an attributional model. Organizational Behavior and Human Performance, 25, 123-138.

Newell, S. E., & Stutman, R. K. (1991). The episodic nature of social confrontation. In J. A. Anderson (Ed.), Communication Yearbook 14 (pp. 359-392). Newbury Park, CA: Sage.

Pennebaker, J. W. (1989). Confession, inhibition, and disease. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 22, pp. 211-244). New York: Academic.

Pennebaker, J. W., Hughes, C. F., & O'Heeron, R. C. (1987). The psychophysiology of confession: Linking inhibitory and psychosomatic processes. Journal of Personality and Social Psychology, 52, 781-793.

Planalp, S., Rutherford, D. K., Honeycutt, J. M. (1988). Events that increase uncertainty in personal relationships II: Replication and extension. Human Communication Research, 14, 516-547.

Regan, D. T., & Totten, J. (1975). Empathy and attribution: Turning observers into actors. Journal of Personality and Social Psychology, 32, 850-856.

Roloff, M. E., & Cloven, D. H. (1990). The chilling effect in interpersonal relationships: The reluctance to speak one's mind. In D. D. Cahn (Ed.), Intimates in conflict: A communication perspective (pp. 49-76). Hillsdale, NJ: Lawrence Erlbaum.

Rusbult, C. E., Johnson, D. J., & Morrow, G. D. (1986). Determinants and consequences of exit, voice, loyalty, and neglect: Responses to dissatisfaction in adult romantic involvements. Human Relations, 39, 45-63.

Rusbult, C. E., Zembrodt, I. M., & Gunn, L. K. (1982). Exit, voice, loyalty, and neglect: Responses to dissatisfaction in romantic involvements. Journal of Personality and Social Psychology, 43, 1230-1242.

Rush, A. J., Weissenburger, J. (1986). Do thinking patterns predict depressive symptoms? Cognitive Therapy and Research, 10, 225-236.

Sadler, O., & Tesser, A. (1973). Some effects of salience and time upon interpersonal hostility and attraction during social isolation. Sociometry, 36, 99-112.

Sarason, I. G., Potter, E. H., Sarason, B. R. (1986). Recording and recall of personal events: Effects on cognitions and behavior. Journal of Personality and Social Psychology, 51, 347-356.

Showers, C. (1988). The effects of how and why thinking on perceptions of future negative events. Cognitive Therapy and Research, 12, 225-240.

Sillars, A. L. (1980). Attributions and communication in roommate conflicts. Communication Monographs, 47, 180-200.

Sillars, A. L., Coletti, S., Parry, D., & Rogers, M. (1982). Coding verbal conflict tactics: Nonverbal and perceptual correlates of the "avoidance-distributive-integrative" distinction. Human Communication Research, 9, 83-95.

Strack, F., Schwarz, N., & Gschneidinger, E. (1985). Happiness and reminiscing: The role of time perspective, affect, and mode of thinking. Journal of Personality and Social Psychology, 49, 1460-1469.

Stutman, R. K., & Newell, S. E. (1990). Rehearsing for confrontation. Argumentation, 4, 18-198.

Tetlock, P. E. (1981). The influence of self-presentation goals on attributional reports. Social Psychology Quarterly, 44, 300-311.

Tetlock, P. E. (1983). Accountability and complexity of thought. Journal of Personality and Social Psychology, 45, 74-83.

Tetlock, P. E. (1985a). Accountability: A social check on the fundamental attribution error. Social Psychology Quarterly, 48, 227-236.

Tetlock, P. E. (1985b). Accountability: The neglected social context of judgment and choice. Research in organizational behavior (Vol. 7, pp. 297-332). Greenwich, CT: JAI.

Tetlock, P. E., & Kim, J. I. (1987). Accountability and judgment processes in a personality prediction task. Journal of Personality and Social Psychology, 52, 700-709.

Wegner, D. M., & Finstuen, K. (1977). Observer's focus of attention in the simulation of self-perception. Journal of Personality and Social Psychology, 35, 56-62.

Witteman, H. (1988). Interpersonal problem solving: Problem conceptualization and communication use. Communication Monographs, 55, 336-359.

Zajonc, R. B. (1960). The process of cognitive tuning in communication. Journal of Abnormal and Social Psychology, 61, 159-167.

Response from Leonard J. Shedletsky

Cloven and Roloff argue persuasively that the extent and character of intrapersonal communication is influenced by life events. Anticipated interaction, style of conflict behavior, and absence or presence of interpersonal communication influence intrapersonal communication. They conclude: "Thus, whereas intrapersonal communication is presumed by several scholars to be central to other communication processes, our perspective suggests that interpersonal communication reciprocally and importantly influences intrapersonal processes." The line of evidence presented does not support the view that intrapersonal communication is not central to communication processes. Has a line been drawn between the twoó intra and interpersonal communicationó such that we can draw the theoretical lines of linkage between the two?

The Role of Nonverbal Communication in Compliance Gaining: Research and Uses

Andrew F. Hayes and Judith A. Barnes

About the Authors: Andrew F. Hayes is affiliated with the Department of Psychology, Cornell University. Judith A. Barnes is an assistant professor in the Department of Communication Studies, San Jose State University.

Abstract: Influence attempts have both a verbal and a nonverbal component. This article selectively examines literature on several nonverbal components of communication as related to increasing or decreasing compliance with requests: gaze, touch, proximity, physical appearance, and paralinguistics. Attention is focused first on basic laboratory and field research and then on several applied contexts, including surveying and opinion assessment, child rearing, education, and counseling. A call for process-oriented research is made to further our understanding as to the mechanisms underlying the relationship between nonverbal messages and compliance rates.

Perspective:

1. How do people feel when a stranger stares, touches, or invades one's space? What effects does this invasion have on evaluation of that person? Is positive or negative reaction a more likely behavior? Why?

2. Should teachers be allowed to touch their students? If so, why? If not, why not? Where does one draw the line between appropriate and inappropriate touch between teachers and their students?

3. In the Chaiken and colleagues (1976) study, attractive people elicited more agreement with the topic they discussed. Might physically attractive people be more persuasive? If so, why?

Interpersonal influence is a part of the daily lives of everyone. We are relentlessly inundated by television advertisers, salespeople, friends, lovers, teachers, and parents trying to influence our attitudes, opinions, behaviors, and preferences. One might even say that socialization itself is a continuous influence process as a society molds its individuals to think and behave according to the society's standards of appropriateness. A large portion of such influence attempts are verbal in nature: the advertiser telling you to buy their product, the politician requesting a vote, the mother telling the child to say hello to visiting neighbors. A less widely-recognized and acknowledged source of influence comes from the nonverbal cues emitted by people in their attempts to gain compliance. Their power lies in their subtlety. While we can often tell when someone is trying to persuade us to adopt their opinion or do them a favor by attending to their verbal messages, the nonverbal component of communication often operates below the level of awareness.

Many authors have written on the power of nonverbal communication in influence but most examine specifically the role of such messages in persuasion and attitude change rather than compliance gaining. This article explores selected research on nonverbal communication factors related to compliance gaining, helping, and when relevant, perceived persuasiveness. Not all possible categories of nonverbal communication are addressed and it is acknowledged that each category examined is incomplete. Entire books have been written on each specific class of nonverbal behavior (e.g., Montague, 1979). The specific studies presented here were selected because they represent the basic flavor of research conducted in the area and serve as a beginning guide to those interested in exploring this area of research in further detail. Indeed, much research is needed since, as the reader will discover, we still have much to learn about the processes by which nonverbal communication affects compliance rates.

Paralinguistics

Most influence attempts have some kind of vocal component. In the past, communication researchers have discounted the role of factors not related to verbal content in the communication process, believing that the delivery of a message was not important in effective communication (Pearce & Conklin, 1971). Now it is well known that how something is said is just as important as what is said. Researchers exploring vocalic or "paralinguistic" components of speech have concentrated on two factors in particular: speed of delivery and voice tone.

Miller, Maruyama, Beaber, and Valone (1976) presented audio-recorded speeches to subjects in a field setting and measured their subsequent agreement with the topic after listening to the speech, as well as subject perceptions of the speaker along the dimensions of knowledge, intelligence, and objectivity. The investigators manipulated both speed of delivery (110 versus 140 versus 190 words per minute) and message complexity (e.g., by manipulating the grammatical structure of the sentences, number of qualifying clauses). The results indicated that regardless of message complexity, faster delivery resulted in higher agreement with the speaker on the topic. Faster speakers were also rated as more intelligent, more knowledgeable, and more objective. Could this be due to the difficulty of forming counterarguments so quickly? This seems unlikely, since this effect existed even for simple messages on topics familiar as well as unfamiliar to the audience. The authors argue that such a speed effect will be most pronounced when the listener cannot rely on expert opinion and the speech is lengthy with numerous detailed facts, leaving the listener somewhat puzzled as to what they should conclude.

The extent to which vocal cues affect persuasion seems to depend on two major variables: perceived characteristics of the sender and the sensitivity of the receiver in nonverbal decoding. Pearce and Conklin (1971) demonstrated that targets rate the sender of a message using a conversational style rather than a dynamic style (with conversational defined as smaller vocal inflections, greater consistency in pitch and rate of speed, lower volume, and lower pitch) as more trustworthy, more attractive, and higher in education, income, occupational status, and generally more favorably. Hall (1980) found that senders who are trying to induce help naturally speak slower, calmer, and with a higher degree of "warmth" than senders who are trying to induce lower rates of helping behavior. Hall's experiment also revealed a peculiar interaction: requesters who tried to induce lots of help (measured in time volunteered), relying only on vocal cues, received more help from good nonverbal decoders and less from poor nonverbal decoders. However, targets who tried to induce little help through their use of vocal cues received more help from poor decoders (decoding ability was measured via the Profile of Nonverbal Sensitivity Test invented by Rosenthal et al.). Buller and Burgoon (1986) found a similar effect: good decoders volunteered more hours participating in research to a pleasant-sounding requester than did poor decoders, while poor decoders donated more hours to a hostile-sounding requester. These results demonstrate that just as is the case with verbal messages, it is important to consider how nonverbal messages are construed by the recipient.

Overall, it is difficult to assess just what type of voice style leads to more compliance. While speakers using a conversational style are perceived more favorably and more trustworthy, those vocal qualities that define "conversational" may be seen as less persuasive and are less spontaneously used by people who are trying to induce help. It appears that the vocal qualities related to high persuasiveness may not be the same as those related to high compliance-gaining.

Physical Appearance

One source of information about a person that becomes immediately salient in interpersonal interaction is the external, physical characteristics of the person. These characteristics fall into two broad categories: clothing and attractiveness.

Clothing. Numerous investigators have examined the relationship between attire and compliance. By far the most widely-known type of clothing effects are those involving the uniform. The classic study on uniforms was conducted by Bickman (1974) in which subjects were requested by experimenters wearing different uniforms to give a dime to a person whose parking meter had expired, pick up a piece of litter, or step away from the curb. The subjects were approached by an experimenter donning either civilian clothing, a milkman uniform, or a security guard uniform, who then made the request (e.g., "give him a dime!"). Compliance was much higher for the request made by the "guard" than by the "milkman" or "civilian" for both the dime request and the litter request. This result has been replicated in similar studies (e.g., Bushman, 1984).

The common explanation for this finding is that the status level of the requester is communicated by the clothing and that the subsequent power associated with that status leads to compliance. Joseph and Alex (1972) argue that uniforms exert their influence primarily through the legitimate power they project (French & Raven, 1968), authenticating or sanctioning the wearer as an authority. Bickman was able to rule out the possibility that this greater compliance was due to coercive or reward power. Interestingly, uninvolved subjects did not perceive the requests as legitimate for their roles, decreasing the likelihood that perceived legitimate power was responsible for the effect. The request made by the security guard may not have been perceived as legitimate because such a uniform conveys legitimate power only when the request falls within the domain of authority given to security guards. It is possible that perceived authority mediated this effect: subjects may have perceived the guard as a man of authority who was trustworthy and credible and therefore didn't question the request, whereas both the milkman and the civilian were perceived as low in authority so the request was dismissed or not taken seriously. Bushman's (1984) study lends support for this possibility. Essentially a replication of the Bickman study, compliance was high for requests coming from people dressed as "firefighters," moderate for requests from "businesspeople," and low for requests made by "bums." It seems likely that at least some uniforms communicate authority and topic-inspecific legitimate power and this is partially responsible for increased compliance.

Mixed effects have been found in investigations exploring the compliance rate for requests coming from experimenters who are dressed sloppily or deviantly (e.g., as a "hippie") versus neatly. Several studies indicate that a person who is approached by a neatly-dressed requester is more likely to lend a dime to the person (Kleinke, 1977a), participate in a survey (Walker, Harriman, & Costello, 1980), accept a leaflet (Darley & Cooper, 1972), or give directions (Schiavo, Sherlock, and Wicklund, 1974) than when approached by a sloppily-dressed requester. Elmswiller, Deaux, and Willis (1971), however, found that compliance was highest when the request was made by someone similar in appearance to the target person: a deviant-appearing person was most effective at eliciting compliance from someone dressed deviantly, while neatly dressed people received greater compliance from neatly-dressed people. Of course, this similarity effect may have been due not to the actual attire but by values, attitudes, and personality characteristics inferred about the person from their clothing. Indeed, several studies do demonstrate that clothing influences perceptions of the person (e.g., Harris, James, Chavez, Fuller, Kent, Massanari, Moore, & Walsh, 1983; Malandro, Barker, & Barker, 1989). Several of the previous studies can also be explained by this effect. Most people would perceive themselves as more similar to the "straight" dresser than the hippie deviant (either in terms of dress preference or values), so compliance would be higher for a straight, neatly-dressed requester.

There are, of course, alternative explanations for some of these effects. For example, many subjects were given two chances to comply before measurement of compliance. In some, if compliance was not gained by the first request, the subjects were given a reason why they should comply (e.g., "I [the experimenter] would give him a dime but I don't have any change," "it's kind of important"). Langer, Blank, & Chanowitz (1978) demonstrate that a reason, relevant or not to the request, is sufficient to increase compliance substantially, particularly for small requests. If initial resistance to the first request was not equal across experimental conditions, this represents a significant confound. It is unclear in many studies whether the experimenters were experimentally blind (e.g., Bushman, 1984; Elmswiller et al., 1971; Walker et al., 1980) and in some, the subjects were hand picked by the experimenter rather than some random procedure (Harris et al., 1983; Kleinke, 1977a; Walker et al., 1980). Interestingly, the one study that does report a significant experimenter effect failed to find a clothing effect (Harris et al., 1983). There are several studies, however, that appear experimentally sound that did produce substantial clothing effects (e.g., the classic Bickman study, 1974; Darley & Cooper, 1972). It would therefore seem that there is evidence that clothing, particularly uniforms, can increase compliance to the extent that they demonstrate the knowledge, credibility, power, or authority of the source, or communicate similarity between the requester and the recipient.

Attractiveness. Numerous studies on physical attractiveness have shown that those who have been endowed with a pleasant face receive many benefits from this gift (Berscheid and Walster, 1974). Attractive people are thought to have more desirable personalities, more likely to lead happy lives, and are generally more liked than are unattractive people (Dion, Berscheid, and Walster, 1972; Walster, Aronson, Abrahams, and Rottman, 1966). And one who appears to us as physically appealing may be more likely to induce compliance, receive help, and come off as more persuasive than those who are less physically appealing. Benson, Karabenick, and Lerner (1976), for example, found that lost graduate school applications were more likely to be mailed to their final destinations if a photo attached to the application indicated it was lost by an attractive person. In another study, Chaiken (1979) had 3 attractive and 3 unattractive student-confederates attempt to persuade target students that the dining commons on campus should stop serving meat at breakfast and lunch. Attractive persuaders elicited more reported agreement with the issue and got more signatures than did unattractive persuaders.

At least four explanations for these attractiveness effects are possible. First, the attractive people may really have been perceived as more persuasive (although all confederates were trained to give their persuasive speech in an identical manner). Second, perhaps attractive people elicit more attention and thus we are more likely to attend to requests from attractive people (cf. Faw & Nunnally, 1967; Kleck & Rubinstein; 1975). Third, complying with a request may be a form of self-presentation designed to elicit a favorable evaluation from the requester. Considering that attractive people are generally thought of as possessing more social desirable characteristics and are more likable, it makes sense that we would want to present ourselves favorably to them, perhaps even in the hope that this may increase the possibility of a initiating a friendship or, in the case of cross-sex interactions, romantic involvement. Indeed, some compliance studies have demonstrated higher compliance rates in cross-sex than same-sex dyadic interactions (e.g., Brocker, Pressman, Cabitt, & Moran 1982; Emswiller, Deaux, & Willets, 1971). Finally, attractiveness may influence compliance indirectly through its effect on liking, as attractive people are generally perceived favorably (and thus likable), and liking is related to compliance and helping (Cialdini, 1984).

Nonverbal Involvement

Behaviors which bring a person closer to a person physically or psychologically are said to enhance nonverbal "involvement" or intimacy. Once a person becomes nonverbally involved, it is difficult to escape through ordinary avoidance maneuvers such as ignoring the person or simply walking away. Nonverbal behaviors that enhance or decrease intimacy and involvement are also intertwined with affect and are natural responses in our interactions with others (Edinger & Patterson, 1983). For example, attraction toward someone is expressed nonverbally, often automatically and unconsciously, through increased eye contact, closer approach distances, increased touching, and positive facial expressions, all of which may induce positive affect in the target. At the other extreme, hatred or disgust is manifested through such behaviors as averted gaze, negative facial expressions, and increased distance. To the extent that nonverbal signals elicit some kind of affective response, behavior corresponding to that response may follow. There is abundance of research demonstrating that affect and such social behaviors as helping and cooperation are intimately associated (Isen, 1987). The three classes of nonverbal behavior discussed in this section may owe much of their effectiveness in increasing compliance rates through the induction of nonverbal involvement with the target or inducing different states of affect in the recipients.

Gaze. Kleinke (1977b) had experimenters approach subjects who took a dime found in a phone booth and stated "Excuse me. I think I might have left a dime in this phone booth a few minutes ago. Did you find it?" Half of the subjects were stared at by the experimenter, while the other half were not. Compliance was defined as returning the dime to the experimenter. A second experiment involved approaching subjects in a shopping mall and asking if she or he would "lend me a dime." In both studies subjects who were gazed at by the experimenter were more likely to comply. A replication by Brocker, Pressman, Cabitt, and Moran (1982) also yielded a marginally significant increase in compliance with increased gaze. Kleinke (1980) later found that gaze significantly increased compliance only when the request was legitimate ("Could you lend me a dime? I have to call someone to pick me up. ") but not when the reason was illegitimate ("Could you lend me a dime? I want to buy a candy bar."). Interestingly, a replication with a different illegitimate request ("I want to buy some gum.") resulted in a significantly greater compliance rate when the requester did not gaze at the subject. Compliance rate for a legitimate request was again greater when the experimenter gazed at the subject.

Why might gaze produce increased compliance? If you haven't made eye contact with a person, you are in some respects "anonymous" to the requester, involvement is minimal, and escape from the request is easy. Being stared at produces different interpersonal effects. Nichols and Champness report evidence that eye gaze does in fact lead to physiological arousal (see Ellsworth & Langer, 1976). Perhaps this arousal is then attributed to interpersonal involvement and intimacy felt toward the requester or is attributed to such emotions as sympathy or concern. In addition, when stared at, one is no longer anonymous to the requester and responsibility to respond to that person's plight is increased, particularly if both the situation and specific required action are made clear. This interpretation is consistent with Cialdini's (1984) recommendation for how to summon help in an emergency situation: make it clear to others that there is an emergency and request specific action from specific people by directly involving them in the situation through direct addresses or eye contact (e.g., "[Staring] You, sir, in the blue jacket, I need help. Call an ambulance!", p. 138). Alternatively, perhaps it is not gaze per se that increases compliance but a decrease in compliance that occurs with a lack of eye contact. Under some circumstances, gaze leads to increased feelings of attraction or liking, as increased eye gaze has in some cases been associated with more favorable socio-emotional evaluations toward the gazer (Goldberg, Kiesler, and Collins, 1969). Finally, it has been suggested that gaze increases compliance only when the appropriate course of action is clear and unambiguous (Ellsworth & Langer, 1976).

Touch. Tactile communication, also known as "haptics," involves the use of physical contact between people. The variety of experiments conducted on the effectiveness of touch in inducing compliance generally follow the same procedure as those on eye contact. An experimenter approaches a subject in a field setting or the laboratory and makes a request while lightly touching the subject on the arm or shoulder. Compliance with the request is then measured. It is difficult to assess whether increased compliance is actually due to the experience of being touched, however, because touching someone is also violation of the subject's personal space. Research on proxemic cues in compliance is discussed in the next section.

Several studies were discussed in the previous section that included a touch manipulation in addition to the gaze manipulation (Brocker et al., 1982; Kleinke, 1977b). The results indicated that subjects who were lightly touched on the arm or shoulder when the request was made were more likely to return the dime. Both of these studies suffered from one major problem: differential approach distances. Subjects in the touch condition were approached and touched at a distance of 1.5 feet whereas subjects in the no-touch condition were approached at 3 feet. The experimenter therefore intruded the subject's intimate zone in the touch condition, but only intruded the personal zone in the no-touch condition (Hall, 1969). Willis and Hamm (1980) attempted to correct this oversight by controlling for approach difference, maintaining it at 3 feet across conditions. In two experiments, subjects complied more to requests to sign a petition supporting the renovation of a railroad station and were more likely to participate in a rating task if they were touched by the experimenter making the request.

One final study conducted by Patterson, Powell, and Lenihan (1986) used a unique compliance request. Subjects were led into a room and were asked to fill out a variety of personality tests. On a table was a large pile of unscored personality tests that the experimenter casually mentioned she or he had to score by hand. The subjects were then led to a cubicle and filled out additional forms. Afterward, the experimenter approached the seated subject and either did or did not touch her or him on the shoulder while requesting assistance in scoring a number of the tests. Willingness to help was measured by the amount of time the subject sat and scored personality tests. Subjects who were touched spent significantly more time helping the experimenter in the scoring task than did those who were not touched.

Proximity.

Some thirty inches from my nose, the frontier of my person goes

And all the untilled air between is private pagus or demesne.

Stranger, unless with bedroom eyes, I beckon you to fraternize

Beware of rudely crossing it: I have no gun, but I can spit. (W. H. Auden)

Auden's limerick illustrates a psychological phenomenon that may not always be the case: people respond negatively to an encroachment of our personal space unless it is welcomed. As results from the touch studies indicate, we aren't always bothered by someone violating our space. In fact, we may be quite willing to help them out. Is there something about a touch from a stranger that makes such a violation situationally legitimate or is there evidence that the "body bubble" is more penetrable than commonly thought?

A set of experiments by Konecni, Libuser, Morton, & Ebbesen (1975) represent confirmations of intuition of the effects of space violations. Konecni and colleagues had confederate experimenters approach pedestrian subjects at a crosswalk. The confederates stopped at the crosswalk and stood next to the subject at distances of either 1, 2, 5, or 10 feet for a period of 10 seconds. For one set of studies, the experimenter then crossed the street at a speed faster than the subject and inconspicuously dropped either a set of keys or a pencil. In an attempt to maintain the space violation for a longer period of time in a different context (walking), an additional set of studies was performed where the experimenter walked next to the subject as they crossed the street, maintaining the same distance until an observer gave the cue to walk faster, and then dropped the keys or pencil. Helping was measured by the subject either returning the keys (pencil) to the owner or shouting out to the experimenter that she or he had dropped something. When the experimenter simply dashed out and then dropped an object, the space violation had no effect if the object was seen as important to the owner (keys); virtually everyone helped. Helping rates decreased as the space violation increased if the objects was of little importance (pencil). However, when the violation persisted after the light changed, the space violation decreased helping for both objects, with the effect largest for the unimportant object.

Baron and Bell (1976), however, found a very different effect of space violations. Subjects were asked if they would participate in a psychology experiment and if so, how much time would they be willing to devote, at distances of 12-18 inches versus 3-4 feet. Those who had their personal space violated volunteered significantly more time than did those who were asked from a distance. Interestingly, subjects in the former condition did report that the requester was standing "too close" while subjects in the latter condition felt she or he was standing at the "correct" distance. Ernest and Cooper (1974) also found greater compliance when personal space was violated, although procedural shortcomings make their findings difficult to interpret.

Can the discrepancies between these studies be resolved? It is generally thought that space violations place a person in very uncomfortable state. This uncomfortable state can best be resolved by escaping the situation. But there are two ways this can be done: either by sending off the violator as quickly as possible or by merely fleeing. Subjects in the Baron and Bell (1976) study may have complied more because it was the most effective way of eliminating the uncomfortable arousal resulting from the space invasion. Indeed, subjects in this study were actually sitting at a table and thus it would have been much more difficult physically as well as socially for them to just walk away. Alternatively, as Baron and Bell suggest, since inappropriate space violations run counter to U. S. American norms, perhaps the person was perceived as particularly desperate for assistance. Self-report measures confirmed this: subjects in the close condition perceived the requester as in greater need than did those in the far condition. The importance of perceived need is also highlighted in the studies by Konecni et al. (1975). Compliance was generally quite high when the experimenter dropped an important collection of objects (keys), and thus the experimenter may have been perceived as more in need. Space violations most greatly reduced compliance when the person was not perceived as greatly in need of assistance.

Summary of Nonverbal Research

Several of the nonverbal components discussed above all share a potential explanation: Those variables in interpersonal compliance requests that make the participants aware of the other, increase their physical or psychological involvement, or enhance the possibility that they will listen to and understand the request will lead to enhanced compliance. Attractive people or those who gaze at, touch, or violate the space of others may stand out in the environment and may be more likely to be given time and attention. It is also possible that such nonverbal behaviors influence the affective state of the target, resulting in increased helping or general cooperativeness (Isen, 1987). Of course, these are not the only explanations, as many other variables have been shown to be related to whether or not a person complies with a request. Studies demonstrate mixed effects of requester-target sex interactions, similarity, and the actual content of the request. Is it possible then to claim that these studies tell us anything certain about the role of nonverbal communication in compliance gaining? Certainly. First, the large majority of these studies were conducted not in the laboratory but in actual field settings with real requests that were allegedly perceived as legitimate by the subjects who did not know they were participating in any kind of investigation. The loss in control and explanatory power is balanced by the demonstration that these nonverbal messages effect real people in "real-world" situations. Second, no one is claiming that it is these variables alone that determine the outcome of a request. A touch on the arm, an unusually close approach, or a particular tone of voice can produce a variety of affective and cognitive responses in the targetóattempts at understanding why this person behaved the way they did or how they feel about this personóis he attractive, sincere, credible, trustworthy? Can I use this situation to make a personal friendship that might develop into something serious? Does this person really need my assistance? The person's responses to these questions will influence ultimately whether or not she or he complies or gives aid. In addition, research demonstrates that very different effects are found if the recipients of an influence attempt are not effective encoders or decoders of nonverbal messages (e.g., Buller & Burgoon, 1986). Finally, precise understanding is not always necessary in some situations to effectively employ a compliance-gaining method to achieve socially desirable ends. We now turn to some of those situations in which the compliance-increasing effects of nonverbal messages have been explored.

Nonverbal Communication in Applied Compliance-Gaining Contexts

Most of the research discussed thus far has addressed the use of nonverbal communication in gaining compliance to relatively meager requests. It would be much more convincing to demonstrate that these nonverbal behaviors play a significant role in influence, helping, and compliance in situations where such effects have more of an impact on both the requester and the target. Fortunately, research has been conducted exploring such situations. Although the contexts discussed in the next section represent only a few of those situations in which research has been done, hopefully the reader will agree that these are some of the important ones.

Research and Opinion Assessment. One of the major difficulties in conducting field research employing questionnaire or survey response measures is the difficulty in getting people to stop and take time out of their busy schedules to answer a few questions. Research on response rates have shown that not many more than half of those who are asked to participate in a survey agree to do so (Hornick & Ellis, 1988). Clearly, as the above studies demonstrate, experimenters and survey takers can greatly enhance their response rates by strategically employing nonverbal communication to induce compliance with such requests. There is evidence that compliance can be increased in response to requests to participate in opinion polls and experimental research through increased touch (Hornick et al., 1988; Willis et al, 1980), eye contact (Hornick et al., 1988), personal space violations (Baron & Bell, 1976), and dressing neatly (Walker et al., 1980). Political campaigners can increase the number of signatures they receive on petitions and acceptance of political leaflets by touching the targets (Willis et al., 1980), and by dressing neatly (Darley et al., 1972) or similarly to the target (Suedfeld, Bochner, and Matas, 1971). Poller attire has also been linked to target perceptions of a political candidate (Darley et al., 1972). These studies illustrate how researchers and pollsters can use nonverbal communication to their advantage both in inducing public participation and in presenting a political image.

Counseling. The ability for a counselor to influence a client is largely a function of how the counselor is perceived. A counselor who is perceived as warm, trustworthy, competent, and sincere will have much more luck in producing the desired changes in the client. Understandably, most research conducted on the nonverbal associates of effective counselors have focused on how such behaviors influence the attractiveness of a counselor, as attractiveness has been shown to influence client perceptions of counselor persuasiveness (LaCrosse, 1975) and client opinion change (e.g., Schmidt and Strong, 1971). In this context, counselor attractiveness is defined not in terms of physical appearance but as general counselor warmth, understanding, and trustworthiness. Nonverbal behaviors crucial to enhancing attractiveness (and liking) are head nods, smiling and other facial expressions, variety in voice tone, eye contact, and assorted hand gestures (Kerr, Claiborn, and Dixon, 1982).

General nonverbal expressiveness (high eye contact, forward body lean, hand movements without associated body movements) has been shown to influence both perceived attractiveness and perceived persuasiveness of a counselor (LaCrosse, 1975). Nonverbal responsiveness to the client, operationalized as facial expressiveness, eye contact, vocal variety, and gesturing, has also been shown to increase the perceived trustworthiness and attractiveness of counselors, particularly when coupled with verbal interpretations and reinstatements of client messages (Claiborn, 1979). However, in this study, influence ratings did not mirror these perceptions. More specific studies have explored the influence of single nonverbal behaviors on client reactions, finding that therapist warmth was related to more glances at patients, and it has been demonstrated that a counselor who occasionally touches the client comes across more favorably than counselors who do not touch, particularly in cross-sex counselor-client pairs (Alagna, Whitcher, Fisher, and Wicas, 1979). Of course, reactions to touch in such a context will vary, depending on whether the client perceives the touch as appropriate or higher in intimacy than that which she or he is comfortable. Indeed, Donnelly (1992) reported in one study that just as many clients perceived a therapist's touch negatively as did positively.

Unfortunately, much of the research conducted on nonverbal communication in the counseling interview (including those cited above) is difficult to interpret, largely because most studies have used simulated counseling situations in which a subject speaks to an ostensible psychotherapist about some topic of a nonpersonal nature. Clearly, there are external validity problems associated with such methods, but the studies do illustrate the potential of nonverbal behaviors as effective counseling tools.

Education. One factor associated with being a good teacher is a possession of the ability to manage the classroom effectively, motivate the students, and get them to complete their work. Clearly, interpersonal influence is an important factor in teacher effectiveness. Much of the research in nonverbal factors in teacher influence has revolved around teacher "immediacy." Nonverbal immediacy communicates such things as general friendliness, warmth, likability, and interaction availability, and can be achieved through smiles, eye contact, relaxed body postures, and close physical distances.

Surprisingly, few studies have empirically examined the effect of various nonverbal messages in the classroom. Chaiken, Gillen, Derlega, Heinen, and Wilson (1976) manipulated nonverbal immediacy cues communicated by teachers and measured students' perceptions of the teacher. In this study, the immediate teacher was trained to increase eye gaze, use affirmative head nodding, lean forward, and smile at the pupils. The nonimmediate teacher minimized eye contact, leaned away from the pupils, exhibited side to side head movements, and frowned at the pupils. The fifth-grade students used as subjects rated the immediate teacher as more friendly and more understanding. In addition, they felt that the immediate teacher spoke more interestingly, better liked and enjoyed being closer to the students, evaluated her as a better teacher, and would prefer her as an instructor more than the nonimmediate teacher. This same study also demonstrated that 9 and 13-year-old students were sensitive to physically attractive teachers. Students rated a physically attractive teacher as more likable, friendlier, a better teacher, more enjoyable, kinder, more understanding, and more interesting than an unattractive teacher. They also felt that the attractive teacher was more interesting to listen to and tried harder. Additionally, Borroughs (1990) found that college students at least stated they would be more likely to comply with the requests of an immediate teacher. Similarly, Kearney, Plax, Smith, and Sorenson (1988) found suggestive evidence that immediate teachers may receive greater compliance from students, although their study lacked real-world realism so it is difficult to interpret their findings.

Touch as an expression of teacher immediacy in the classroom is currently a hot topic under much scrutiny. What types of touch should and should not be used in the classroom? This would seem to depend on the age of the students involved. Since touch is one of the most immediate, intimate forms of communication (Montague, 1979; Morris, 1971; Thompson, 1973), tactile contact may be necessary for the teacher to convey love and affection to students at the elementary level. Since touch is a very personal form of communication, handshakes and shoulder touches can convey immediacy while rarely being interpreted sexually. Similarly, pats on the back and other nonthreatening forms of touch can serve as powerful reinforcers in a way that talk or high grades simply cannot. However, teachers who employ touch should be aware of the school and community norms governing such behavior. Furthermore, since many people in American culture are highly touch-avoidant (Anderson & Leibowitz, 1978; Hall, 1969) teachers should be aware that touch is not reinforcing or pleasurable to some students. Moreover, teachers who are themselves touch avoiders must find other ways of communicating immediacy.

Assuming that teachers who are evaluated more positively will be able to manage their students more effectively, increased compliance and greater classroom control could result through the use of nonverbal behaviors which communicate immediacy and enhance the psychological attractiveness of the teacher. Perhaps increasing teacher awareness to the influence of nonverbal communication in bringing about compliance would not only make the teacher's job easier, but would enhance the atmosphere of the educational environment as well.

Child rearing. One of the more common reasons a parent brings a child to a clinician is that the child will not comply with the demands or wants of the parent. Not only are parents referred to clinicians for this reason, but clinicians often consider noncompliance a symptom of hyperactivity. There is a variety of literature that addresses child noncompliance to parental requests. However, most of these investigations examine only the verbal aspects of the request. Very few researchers have examined the role of parent nonverbal communication in producing child compliance. This would seem to be an important component of the parent-child communication process, considering that a child does respond to the mother's commands before verbal understanding exists. The infant can detect when she or he has behaved against mother's wishes by attending to the tone of the voice. "Yes" or "Good" are both produced differently by the mother and sound different to the child than do "Bad" and "No." Similarly, it would seem that the response of a child to its parent would depend to some extent on whether the child perceives that she or he is being addressed. This could be determined by the parent's direction of gaze or her or his physical proximity to the child.

Unfortunately, there has been very little systematic research that is unequivocally informative conducted on this topic. One exception is a study conducted by Hudson and Blane (1985), who addressed this issue by examining child compliance to the parent in a variety of play situations. Hudson and colleagues compared the nonverbal behaviors exhibited by mothers whose children were presently under clinical care for noncompliance (clinic mothers) to mothers with children not under clinical care (non clinic mothers). In addition, they examined the differences in nonverbal behavior toward the child between successful and unsuccessful requests. In a laboratory setting, the mothers played with their children in a directive way, meaning they were to direct their children to engage in 3 specific activities for a period of 3 minutes each. Observers recorded the nonverbal behaviors (distance from child, body orientation, eye contact, and tone of voice) exhibited by the mothers when making the requests and whether or not the child complied with the requests. Based on the total number of instructions given to the children, compliance rates were highest when the mother was within 3 feet of the child, was squatting down, was looking at the child or there was mutual eye contact, and when a pleasant tone of voice was used. The comparison of the mothers indicated that clinic mothers were more likely than the nonclinic control mothers to issue commands from a distance and use a tone of voice that was either stern or neutral. Although this is not compelling evidence, it is interesting that children were more compliant when the mother's nonverbal behavior was more intimate or "immediate" and that mothers with children under clinical care were less likely than the control mothers to use intimate nonverbal behaviors when issuing commands.

There are other interpretations of these results however. The clinic mothers may expect noncompliance from their children and therefore use different types of nonverbal strategies when attempting to gain compliance. In addition, the clinic mothers gave roughly twice as many instructions than did the control mothers. It is unclear why this may have been so. Perhaps it was due to repeated attempts to obtain compliance from a stubborn child. Regardless, it appears that there is a relationship between the nonverbal channel of commands given to children and compliance with those commands. Research to establish whether this is a causal link would be helpful.

Conclusion

As the research discussed above clearly illustrates, our ability to influence others is affected by the nonverbal components of our communication. Unfortunately, there is little coherent theory underlying much of the research to date and rarely do investigators test alternative explanations. Thus, very few studies are able to tell us just how nonverbal communication affects compliance. Affective processes and the induction of nonverbal "involvement" or "intimacy" are popular explanations for these effects. Alternatively, some nonverbal components of communication may simply induce greater attention to the person and the surrounding context, the interpretations of which determine whether compliance increases or decreases. Another major problem associated with this research is how investigators have tended to use rather mundane requests in their experiments. The applied research, however, demonstrates that nonverbal behavior can also have an impact when the stakes are much higher and the desired behavior is of greater importance. Nevertheless, more research is needed to further understanding as to why these nonverbal messages produce the results they do. Are the mechanisms emotional, social, purely cognitive, or some complex combination of these? Are there situations in which one mechanism is more prominent? How much do these mechanisms vary across nonverbal channels?

Cultures of the world are much closer than they have been in past years, largely through advances in the technology of communication and a growing awareness of the interdependency of nations and the effects that the behavior of one nation has on another. As this occurs, it becomes increasingly important to understand and appreciate the differences in nonverbal communication patterns of people in different cultures. Much of our contact with other nations will include attempts at influence. Cooperation in solving problems of world-wide relevance will be necessary. Such cooperation will include communication centered on who should act in a particular way and who should change their present behavior in a particular direction. For example, how do we bring about compliance with international laws? Who makes these laws and who should be responsible for enforcing them? Answers to such questions will certainly involve one party trying to influence another. Considering the diversity in nonverbal language across cultures, we must be careful where we step, for it is easy to insult someone unknowingly with a poor choice of gestures or an inappropriate touch. Are there consistencies in particular nonverbal cues that will lead to greater acceptance, greater understanding, fewer misunderstandings, and greater cooperation and compliance across cultures? The answer would seem to lie in discovering those cues as well as exploring and understanding differences between cultures and being sensitive not only to what we are communicating, but how we are communicating it, and how this affects other people.

References

Alagna, F. J., Whitcher, S. J., Fisher, J. D., & Wicas, E. A. (1979). Evaluative reaction to interpersonal touch in a counseling interview. Journal of Counseling Psychology, 26, 465-472.

Anderson, P. A., & Leibowitz, K. (1978). The development and nature to the construct touch avoidance. Environmental Psychology and Nonverbal Behavior, 3, 89-106.

Baron, R. A., & Bell, P. A. (1976). Physical distance and helping: Some unexpected benefits of "crowding in" on others. Journal of Experimental Social Psychology, 6, 95-104.

Benson, P. L., Karabenick, S. A., & Lerner, R. M. (1976). Pretty pleases: The effects of physical attractiveness, race, and sex on receiving help. Journal of Experimental Social Psychology, 12, 409-415.

Berscheid, E., & Walster, E. (1974). Physical attractiveness. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 7). New York: Academic.

Bickman, L. (1974). The social power of a uniform. Journal of Applied Social Psychology, 4, 47-61.

Borroughs, N. F. (1990). The relationship of teacher immediacy and student compliance- resistance with learning. Dissertation Abstracts International, 51(6-A), 1825-1826.

Brocker, J., Pressman, B., Cabitt, J., & Moran, P. (1982). Nonverbal intimacy, sex, and compliance: A field study. Journal of Nonverbal Behavior, 6, 253-258.

Buller, D. B., & Burgoon, J. K. (1986). The effects of vocalics and nonverbal sensitivity on compliance: A replication and extension. Human Communication Research, 13, 126-144.

Bushman, B. J. (1984). Perceived symbols of authority and their influence on compliance. Journal of Applied Social Psychology, 14, 501-508.

Chaiken, A. L., Gillen, B., Derlega, V. J., Heinen, J. R., & Wilson, M. (1976). Students' reactions to teachers' physical attractiveness and nonverbal behavior: Two exploratory studies. Psychology in the Schools, 15, 588-595.

Chaiken, S. (1979). Communicator physical attractiveness and persuasion. Journal of Personality and Social Psychology, 37, 1387-1397.

Cialdini, R. (1984). Influence: The new psychology of modern persuasion. New York: Morrow.

Claiborn, C. D. (1979). Counselor verbal intervention, nonverbal behavior, and social power. Journal of Counseling Psychology, 26, 373-383.

Darley, J. M., & Cooper, J. (1972). The "Clean for Gene" phenomenon: The effect of students' appearance on political campaigning. Journal of Applied Social Psychology, 2, 24-33.

Dion, K., Berscheid, E., & Walster, E. (1972). What is beautiful is good. Journal of Personality and Social Psychology, 24, 285-290.

Donnelly, L. H. (1992). Effects of counselor touch and counsel gender on perceptions of client compliance. Dissertation Abstracts International, 53(4-B), 2057.

Edinger, J. A., & Patterson, M. L. (1983). Nonverbal involvement and social control. Psychological Bulletin, 93, 30-56.

Ellsworth, P. C., & Langer, E. J. (1976). Staring and approach: An interpretation of the stare as a nonspecific activator. Journal of Personality and Social Psychology, 33, 117-122.

Elmswiller, T., Deaux, K., & Willets, J. E. (1971). Similarity, sex, and requests for small favors. Journal of Applied Social Psychology, 1, 284-291.

Ernest, R. C., & Cooper, R. E. (1974). "Hey mister, do you have any change?": Two real world studies of proxemic effects on compliance with a mundane request. Personality and Social Psychology Bulletin, 1, 158-159.

Faw, T. T., & Nunnally, J. C. (1967). The effects on eye movements of complexity, novelty, and affective tone. Perception and Psychophysics, 2, 263-267.

French, J. R. P. , & Raven, B. H. (1968). The bases of social power. In D. Cartwright & A. Zander (Eds.) Group dynamics: Research and theory (3rd ed.). New York: Harper & Row.

Goldberg, G. N., Kiesler, G. A., & Collins, B. E. (1969). Visual behavior and face-to- face distance interaction. Sociometry, 32, 43-53.

Hall, E. T. (1969). The hidden dimension. New York: Anchor.

Hall, J. A. (1980). Voice tone and persuasion. Journal of Personality and Social Psychology, 38, 924-934.

Harris, M. B., James, J., Chavez, J., Fuller, M. L., Kent, S., Massanari, C. M., Moore, & Walsh, F. (1983). Clothing, compliance, and choice. Journal of Applied Social Psychology, 13, 88-97.

Hornick, J., & Ellis, S. (1988). Strategies to secure compliance for a mall intercept interview. Public Opinion Quarterly, 52, 539-551.

Hudson, A., & Blane, M. (1985). The importance of nonverbal behavior in giving instructions to children. Child and Family Behavior Therapy, 7, 1-10.

Isen, A. (1987). Positive affect, cognitive processes, and social behavior. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 20). New York: Academic .

Joseph, N., & Alex, N. (1972). The uniform: A sociological perspective. American Journal of Sociology, 77, 719-730.

Kearney, P. , Plax, T. G., Smith, V. R., & Sorenson, G. (1988). Effects of teacher immediacy and strategy type on college student resistance to on-task demands. Communication Education, 37, 54-67.

Kerr, B. A., Claiborn, C. D., & Dixon, D. N. (1982). Training counselors in persuasion. Counselor Education and Supervision, 12, 138-148.

Kleck, R. E., & Rubenstein, C. (1975). Physical attractiveness, perceived attitude similarity, and interpersonal attraction in an opposite-sex encounter. Journal of Personality and Social Psychology, 31, 107-114.

Kleinke, C. L. (1977a). Effects of dress on compliance to requests in a field setting. Journal of Social Psychology, 101, 223-224.

Kleinke, C. L. (1977b). Compliance to requests made by gazing and touching experimenters in a field settings. Journal of Experimental Social Psychology, 13, 218-223.

Kleinke, C. L. (1980). Interaction between gaze and legitimacy of request on compliance in a field setting. Journal of Nonverbal Behavior, 5, 3-12.

Konecni, V. J., Libuser, L., Morton, H., & Ebbesen, E. B. (1975). Effects of a violation of personal space on escape and helping responses. Journal of Experimental Social Psychology, 11, 288-299.

LaCrosse, M. B. (1975). Nonverbal behavior and perceived counselor attractiveness and persuasiveness. Journal of Personality and Social Psychology, 22, 563-566.

Langer, E., Blank, A., & Chanowitz, B. (1978). The mindlessness of ostensibly thoughtful action: The role of "placebic" information in interpersonal interaction. Journal of Personality and Social Psychology, 36, 635-642.

Malandro, L. A., Barker, L., & Barker, D. A. (1989). Nonverbal communication (2nd ed.). New York: Random House.

Miller, N., Maruyama, G., Beaber, R. J., & Valone, K. (1976). Speed of speech and persuasion. Journal of Personality and Social Psychology, 34, 615-624.

Montague, A. (1979). Touching: The human significance of the skin. New York: Columbia University.

Morris, D. (1971). Intimate behavior. New York: Random House.

Patterson, M. L., Powell, J. L., & Lenihan, M. G. (1986). Touch, compliance, and interpersonal affect. Journal of Nonverbal Behavior, 10, 41-50.

Pearce, W. B., & Conklin, F. (1971). Nonverbal vocalic communication and perceptions of a speaker. Speech Monographs, 38, 235-241.

Schiavo, R. S., Sherlock, B., & Wicklund, G. (1974). Effect of attire on obtaining directions. Psychological Reports, 34, 245-246.

Schmidt, L. D., & Strong, S. R. (1971). Attractiveness and influence in counseling. Journal of Counseling Psychology, 18, 348-351.

Suedfeld, P. , Bochner, S., & Matas, C. (1971). Petitioner's attire and petition signing by peace demonstrators: A field experiment. Journal of Applied Social Psychology, 1, 278-283.

Thompson, J. H. (1973). Beyond words: Nonverbal communication in the classroom. New York: Citation.

Walker, M., Harriman, S., & Costello, S. (1980). The influence of appearance on compliance with a request. Journal of Social Psychology, 112, 159-160.

Walster, E., Aronson, V., Abrahams, D., & Rottman, L. (1966). Importance of physical attractiveness in dating behavior. Journal of Personality and Social Psychology, 4, 508-516.

Willis, F. N., & Hamm, H. K. (1980). The use of interpersonal touch in securing compliance. Journal of Nonverbal Behavior, 5, 49-55.

Response from Leonard J. Shedletsky

The connection between nonverbal behavior in compliance gaining and intrapersonal communication is left implicit in this essay. Just what is the intrapersonal theoretical underpinning to nonverbal behavior and compliance? If it is not isolated variables that determine the outcomes of, say, a request, just how do the variables under study work in conjunction with one another? Is a variable approach the best way to conceptualize compliance gaining and nonverbal communication?

Topics, Turns, and Interpersonal Control: Using Serial Judgment Methods

Mark T. Palmer and Abby M. Lack

About the Authors: Mark T. Palmer is an assistant professor of Communication Studies, Northwestern University, 1815 Chicago Avenue, Evanston, IL 60201, telephone (708) 491-7530. Dr. Palmer is a 1990-1992, National Academy of Sciences, National Research Council Associate. Abby M. Lack is a graduate student in Communication Studies and was a 1991-1992, San Jose State University/NASA Research Assistant.

Abstract: Serial judgment techniques capture the on-line, dynamic judgments of participants or observers as they experience an interaction. Two studies are reported that apply serial judgment methods to demonstrate how moment-to-moment changes in behaviors are associated with moment-to-moment changes in observers' perceptions of interpersonal dominance. Study 1 (Palmer, 1989) demonstrated that judgments of increased dominance made at each utterance and aggregated by turns in a transcript conversation were significantly associated with ratings of decreased topic relatedness and increased turn duration. With each behavior, conversational partners were perceived to increase or decrease dominance. Study 2 attempted to replicate and expand these findings using four video taped conversations and computerized techniques to collect moment-to-moment measures of the same variables. Results of Study 2 somewhat supported the findings of Study 1. Complete tables available from the authors.

A major task of research attempting to establish relationships between conversational behaviors and interpersonal dominance is to describe how these behaviors map perceptions of dominance in conversational settings. Most studies have used methodologies that compare summary levels of behavioral activity and summary levels of judgments. Similar paradigms exist for encoding studies in which different subjects are induced to act more or less dominant and summary measures of their behaviors (rather than judgments) are compared as the dependent measures (Ellyson & Dovidio, 1985). However, there are conceptual and empirical reasons why summary level measures may be limiting, if not inappropriate. In this section, reasons are discussed for augmenting the typical summary level data collection and analyses methods with methods attempting to reflect the dynamic, changing nature of conversational interactions and relational development.

Relational Inferences

Conversational negotiations of relational definitions proceed through interactive patterns in real time (see Cappella, 1984; Street & Cappella, 1985; Rogers & Farace, 1975; Courtright, Millar & Rogers-Millar, 1979). For example, in a face-to-face interaction, Partner A performs behaviors attempting to create a relational effect on Partner B (e.g., increase A's interpersonal dominance relative to B). Partner B interprets A's behaviors (not necessarily as intended by A) and adjusts her or his perceptions of the relationship. At this point, Partner B is cognitively establishing the state of the relationship by engaging in an inferential process. Through an unknown (and perhaps unknowable) mental calculation, Partner B integrates the behavioral information observed in A's behaviors with general and specific knowledge about how individuals behave. The cognitive outcome of this process is an inference or judgment about the state of the relationship. This inference is stored in short term or working memory as a mental representation we may call a "mental model." Thus, the process by which relationships are developed in conversations is dynamic in nature, and modeling this process requires mapping the changing output of the cognitive system (i.e., mental models) onto changing displays of behaviors.

There has been much speculation and some research on the nature of mental models in various problem-solving or task-related situations (Johnson-Laird, 1983; Gentner & Stevens, 1983). It is generally agreed that mental models hold processors' cognitive reconstructions of an immediate problem space; that is, they hold necessary information about the interrelationship among task relevant focal stimuli or objects, continually up-dated by new information. Thus, mental models are temporary memory structures that represent the state of the current problem space. During interpersonal transactions these models represent the state of the relationship as it is developing. Therefore, understanding the process by which behaviors enact relationships in face-to-face conversations depends to some extent on the ability to map moment-to-moment products of mental inference processes (i.e., mental models) onto dynamic patterns of behaviors.

Tapping the cognitive output of the inference process is a complex, difficult task. Although processors may be aware of making relational judgments, they are rarely aware of the algorithms by which such judgments or inferences are made (Nisbett & Wilson, 1977; Lewicki, 1986). The mental algorithms that drive the process are acquired nonconsciously and remain, for the most part, unavailable to conscious appraisal (Lewicki, 1986). Furthermore, with age, many relational inferences and behavioral responses become automatic. Automatic processing relieves the processor of the burden of conscious attention and reduces cognitive effort, but occurs too rapidly for processors to consciously access it (Ericsson & Simon, 1984).

Given the dynamic nature of the cognitive-behavioral process, methods that attempt to map behavioral inputs to mental outcomes must demonstrate how the temporary and changing mental models of the relationship are associated with the changing states of observable behaviors. Summary level measures of behaviors and of mental representations across an entire interaction may not completely capture the true, dynamic nature of the cognitive-behavioral process. For example, in studies in which observers watch a complete interaction and then are asked which partner is more dominant, results may reflect the last mental model of the interaction (a recency effect) or the first impression stored in long term memory (a primacy effect), an average of the moment-to-moment impressions (an integration effect), or some other judgment based on a simple heuristic. Serial judgment methods, by contrast, collect responses over time, allowing comparisons of behavioral displays with inferential judgments on an act-by-act or moment-to-moment basis.

To illustrate the possible discrepancy between summary level and dynamic level analyses, Figure 1 presents data from a hypothetical study. The patterns of talk for partners A and B are described in the top part of the figure. A turn is here defined as a string of unilateral utterances. When the curve for each partner moves up, the data indicate that the partner is speaking or, a turn is "on." Adding all of the "ons" for each speaker shows that both partners held the floor approximately 50% of the time. In a common experimental paradigm, a group of observers might also have watched the tape and made summary level judgments of the level of dominance exhibited by one or both partners. Because no differences exist between the behaviors, it is not possible to statistically find a relationship among judgments and actions, even if there were an overall difference in perceived dominance. A similar situation arises when differences in behaviors exist but there are no differences in observers' judgments between partners' levels of dominance. The latter case could be the result even when using a carefully manipulated behavioral stimulus in which confederates or subjects are trained or induced to vary their levels of behaviors. Therefore, in both judgment and encoding studies that use summary level measures or manipulations, associations between judgments and behaviors can be found only when overall measures of a behavior and overall measures of a judgment exist together.

If methods were used that would produce continuous, over-time judgments of dominance as well as continuous measures of behaviors, different conclusions may result. For example, the lower part of Figure 1 shows hypothetical dominance judgments of A and B taken continuously across the entire conversation. A close look at these judgments shows that as observers experienced the conversation, their perceptions of interpersonal dominance changed. When a partner held the floor, the perception of dominance for that partner increased. Clearly, an association between dominance and talk time is indicated even if summary levels of behaviors or judgments do not agree. In other words, it is possible that moment-by-moment changes in behaviors may be associated with moment-by-moment changes in perceptions without any associations being indicated at the summary level.

There are compelling reasons for expanding the study of relational inferences onto dynamic patterns of behaviors and judgments. Presently it is unclear how summary level measures relate, if at all, to the process by which relationships are created in human interactions. However, mapping on-line, or moment-by-moment perceptions of relational dimensions (such as dominance) onto changing patterns of behaviors has the advantage of revealing the dynamic nature of the cognitive-behavioral process and presents a clear picture of the association of behaviors and cognitions.

Serial Judgment Methods

Serial judgment methods require conversational participants or observers to make judgments regarding the relational state of interactants at regular or selected intervals as they are observing the interaction. This technique assumes that judges' on-line responses approximate moment-by-moment inferences because responses are made with limited opportunity to access high

level, complex processing. Serial judgment methods also create issues that must be conceptually or empirically resolved: 1) how to control for autocorrelation or serial dependence in over-time data; and, 2) how to simultaneously analyze serial judgments and continuous streams of behaviors when they change at different rates.

Serial measures introduce a potential source of biasó serial dependence or autocorrelation. Serial dependence is the correlation between errors in measurement at time t with errors in the measure at time t + 1, and/or measures at time t + 2, t + 3 and so on. This condition arises out of the common finding that the state of some object or some phenomenon is not isolated from the influence or state of the same object or phenomenon at some earlier time. Therefore, changes in a judgment over time are the result of the effects of true changes in behaviors plus some autocorrelation or serial dependence due to the judgment influencing itself. In other words, the score for a judgment at time t + 1, is due to some "carry-over" effects of the score for the same judgment at time t, plus some effects due to other input variables such as topic relatedness and turn duration. Therefore it is essential in analyzing time series data of judgments to know to what extent there is serial dependence operating in the data so that it is possible to discriminate the direct effects of other forces (i.e., behavioral changes) on judgments. In order to know how behaviors may be driving the judgments, it is necessary to remove the effects of autocorrelation.

Autocorrelation in time series data has the added effect of violating assumptions of independence of observations. Serial dependence inflates T values and de-stabilizes coefficients in regression analyses. However, as described by Hibbs (1974), a General Least Squares procedure for the regression of time series data is possible if the degree of autocorrelation in the time series can be modeled. Once the degree and form of autocorrelation is diagnosed, its effects can be controlled in a regression model by entering the autocorrelation component as a variable in the model (thus partitioning the effects). This procedure is briefly described in the studies reported below.

Another potential problem with data collected using serial judgment methods arises when behaviors and judgments change over time at unequal rates. Time series regression analyses are constructed to match changes in one series (e.g., behaviors) with changes in another (e.g., judgments) at regular and even intervals, after removing autocorrelation effects. Thus, a significant effect is found when the judgment changes at the same rate and at the same levels as the behaviors. However, if behaviors show a more rapid changing pattern than judgments then it becomes more difficult for time series regression techniques to show a significant association. Figure 2 illustrates these effects.

Figure 2a shows two continuous measures over time, topic relatedness and dominance from a hypothetical data set. Both series are mapped onto the same time line (horizontal axis) and scales (vertical axis). The data appear to indicate that the two series are synchronous, both in term of levels and rate of change, i.e., both show similar peaks and valleys. The correlation of these variables is .94 over time. These data seem to reveal a very strong relationship between the variables as long as they are synchronized.

Closer examination of Figure 2a also demonstrates a lagged synchrony, that dominance scores tend to lag the topic relatedness scores by one or two time units. This means that changes in topic relatedness are followed by changes in perceived dominance at some later time, perhaps one or two units. A cross-correlation of the two variables (i.e., correlation of one series lagging behind the other) shows that the correlation of relatedness with dominance is .86 with dominance lagging one unit, and .76 with dominance lagging 2 units. As long as the lagged effects are consistent across time units, the relationship between the variables remains synchronized and is statistically significant.

A more realistic representation of time series data is shown in Figure 2b, however, in which both series show a high degree of variability. Dominance judgments seem to show the same general pattern as the topic relatedness data, but the latter appears to show more changes and greater variance (i.e., more peaks and valleys with a greater range of scores). A cross-correlation of the dominance and relatedness scores is .54 for these data with no lag, or .39 and .24 with lags of 1 and 2 units, respectively. Thus, the "eyeball" reading of the figure appears to show a general relationship, but the discrepancies between the units in terms of rates and levels of changes greatly reduces the chances of finding a significant association.

The solution to this problem is found in data "smoothing" procedures that effectively smooth out the peaks and valleys in a time series (Makridakis & Wheelwright, 1978). Briefly, smoothing reduces the number of small, sharp peaks and valleys without radically changing the general shape of the time series. Figure 2c shows the same hypothetical data after it has been smoothed over 5 time intervals. That is, any data point represents a moving average over 5 observations. The cross-correlations of topic relatedness and dominance at 5 intervals are .57 for no lags, and .42 and .25 for lags of 1 and 2, respectively. Thus, smoothing brings out operant relationships without grossly inflating the correlations.

At present, no empirical or conceptual guidelines are available to indicate how many intervals should be used in smoothing. We suggest the use of multiple analyses at various levels of smoothing. The most appropriate level of smoothing may be indicated when the largest number of time intervals used produces stable results across several interactions, and when results show small differences in direction or magnitude as the number of intervals increases. However, at this stage of development, the methods must be considered exploratory.

Topics, Turns and Dominance: Two Studies Using Serial Judgment Methods

The studies described here used techniques in which observers produced relational inferences (operationalized as serial judgments) of interpersonal dominance or control on an act-by-act or moment-by-moment basis. Study 1 is published elsewhere (Palmer, 1989) but is described briefly to provide background for the second study and to demonstrate a transcript-based serial judgment methodology. Study 2 demonstrates a computer-based, real-time data collection method. In both studies the hypotheses were the same:

Hypothesis 1: As the degree of topic relatedness of a verbal contribution increases, the perceived dominance of the speaker will decrease.

Hypothesis 2: As the length of a verbal contribution increases, the perceived dominance of the speaker will increase.

Controlling Conversations I: Transcript-Based Method

Method

Judgments of interpersonal control were gathered from undergraduates who read a transcript of a spontaneous, initial interaction between two male college students. A set of items used as a check on the conversation revealed that it was perceived as "normal." Topic relatedness was coded by groups of observers who rated each consecutive pair of utterances on a seven-point Likert-type scale. This scale described how related an utterance was to the preceding utterance, ranging from 1 = "not very closely related" to 7 = "exactly the same topic." Utterances, as the units of coding, were defined as simple declarative sentences or independent clauses. Means for the groups were taken as the independent measure. Length of talk was determined by counting the number of words in an utterance and words in a turn. Interpersonal control was rated by groups who recorded a judgment for each utterance on a scale ranging from 1 = "Person A totally controlled," to 4 = "neither controlled," to 7 = "Person B totally controlled." This scale forced judges to consider both partners but produced only a single judgment at each utterance. Because analyses were conducted at both utterance and turn levels, measures were also computed as the average relatedness of the utterances within each turn (i.e., a contiguous set of unilateral utterances).

Results

Analyses were conducted using the time series regression techniques described earlier. Space does not allow a complete, detailed account of the entire process by which the independent and dependent measures were tested for time series components. Briefly, the steps involved in the analysis were: 1) identification of the serial dependence model for the residuals of the independent and dependent measures; and, 2) performing regression using transformed variables and controlling for the identified serial dependence components (McCleary & Hay, 1980; Hibbs, 1974).

Identification of the serial dependence components in the serial judgments of dominance began with the plotting of the autocorrelations (AC) and partial autocorrelations (PAC). Autocorrelations and partial autocorrelations describe the degree to which each observation is correlated with subsequent observations at each lag (i.e., time t with t + 1, time t with t + 2, and so on). Plots of AC's and PAC's showed gradually decreasing AC's and single large spikes in the PAC's, thus indicating an autoregressive process. Spikes at interval 1 in the plots of the PAC's were indications that the autoregressive process was first order. This means that an observation of

Table 1

Time Series Regression: Perceptions of Dominance with Topic Relatedness and Turn Duration

Utterance Level Analysis

Variable	Beta	SE	T	df	p
length of utterance	.03	.02	2.10	1	.04
relatedness	-.15	.03	-4.58	1	.00
moving avg. component	-.48	.08	-5.68	1	.00
constant	.38	.18	2.16	149	.03

Note. Overall F (3,149) = 25.80, p < .05, R2 = .34, adjusted R2 = .33 and Durbin-Watson Test

Statistic = 2.0.

Turn Level Analysis

Variable	Beta	SE	T	df	p
length of utterance	.09	.04	2.58	1	.01
relatedness	-.17	.04	-3.86	1	.00
moving avg. component	-.54	.12	-4.57	1	.00
constant	.50	.19	2.52	149	.01

Note. Overall F (3,74) = 17.06, p < .05, R2 = .41, adjusted R2 = .39 and Durbin-Watson Test

Statistic = 1.98.

dominance at time t + 1 would be best predicted by a dominance judgment at time t. No other lagged relationships were apparent in the data. This effect was controlled for in time series regression by entering the first order, autoregressive effect as a variable in the equation along with the variables of interest. Thus, the effects of topic relatedness and turn duration on dominance could be directly assessed. Table 1 shows the results of the regression analyses.

As shown in Table 1, decreases in topic relatedness (i.e., moving the topic away from the current topic) and increases in time talking or holding the floor were associated with increases of interpersonal dominance. The significant Beta for the moving average component and the Durbin-Watson statistic, which was approximately 2.00, indicate that an appropriate time series component was diagnosed and controlled for.

An important implication of these findings is that interpersonal dominance is an emergent property of an interaction. With each utterance, topic changing and floor management have a relational impact. It was not possible to claim the relative importance of an individual behavior versus individual differences or social level factors. However, the effect of the verbal behaviors, irrespective of other potential factors, was clearly suggested.

Serial Judgment Versus Summary Level Results. In this study, subjects also were asked to provide overall judgments of dominance on the partners; that is, judgments were collected after the entire conversation was read and rated on the act-by-act basis. Person A was perceived by judges to have been more dominant in the conversation (72.2% said A was more dominant, versus 17.8% who said B), and to have had a more dominant personality (67.7% for A versus 16.1% for B). While Person A did speak longer than Person B (t(38) = -2.24, p < .05), Person A did not demonstrate a different mean level of topic changing behavior (t(38) = 1.25, p = .21). Thus, summary level data and over-time data did not produce similar results.

Controlling Conversations II:

Video Tape Based Method

Method

Study 2 expanded on the first experiment by increasing the number of stimulus conversations and developing a computerized method for making on-line, real time judgments of interpersonal control (see Palmer, 1990). The stimuli in this study consisted of four video taped female-female dyads who talked for approximately ten minutes.

Table 2

Results of Individual Regressions of Serial Correlation Components of Dominance, Turn Duration and Topic Relatedness for Each Conversation

Dominance

Conversation/Partner	Compo-nent**	Coef- ficient	Std. Error	T	p	Durbin Watson	Q
One/ Left Side	AR (1)	.41	.07	5.88	.00	1.89	17.79
One/ Right Side	AR (1)	.59	.06	9.85	.00	1.86	10.75
Two/ Left Side	AR (1)	.37	.07	5.24	.00	1.89	20.40
Two/ Right Side	AR (1)	.29	.07	4.04	.00	1.91	21.41
Three/ Left Side	AR (1)	.28	.08	3.64	.00	1.90	32.78
Three/ Right Side	AR (2) AR (1)	-.17 .47	.08 .07	-2.32 7.06	.02 .00	1.94	20.04
Four/ Left Side	AR (1)	.72	.07	9.77	.00	1.97	40.63*
Four/ Right Side	AR (2) AR (1)	-.22 .27	.07 .07	-2.96 3.76	.00 .00	1.96	18.26

Topic Relatedness

Conversation	Compo-nent**	Coef- ficient	Std. Error	T	p	Durbin Watson	Q
One							11.35
Two							24.27
Three							20.88
Four	AR (2)	-.16	.08	-2.04	.04		25.11

Turn

Conversation	Compo- nent**	Coef- ficient	Std. Error	T	p	Durbin Watson	Q
One	AR (1)	-.20	.07	-2.77	.006	2.05	35.05
Two	AR (1)	-.32	.07	-4.55	.000	2.08	37.07*
Three	AR (1)	-.27	.07	-3.61	.000	2.05	21.10
	AR (2)	-.25	.07	-3.42	.000
Four	AR (2)	-.28	.07	-3.88	.000	2.11	45.00*

Note. Because topic relatedness was judged serially without regard to speaker, there are no separate measures for partners on either side of the video screen. Because turn measures were complementary, one set of scores reflects both partners' floor holding. Also, * indicates significance compared to chi squared (df = 24) of 36.70 at alpha = .05. ** indicates AR (1) and AR(2) are first and second order autoregressive components, respectively.

Stimulus tapes consisted of split-screen images of seated partners from below the knee to top of the head. Cameras were arranged to provide full front views. Conversational partners were given no directions as to topics except to "try to get to know one another." The first three minutes of the conversations were analyzed.

Topic relatedness was coded by 44 undergraduates who participated in the study for extra credit. Sets of judges rated transcribed sections of the four stimulus conversations in the following manner. Judges were presented with a series of 4 x 6 inch cards, each showing two consecutive utterances from the same conversation. Judges were asked to "read the two sets of remarks and decide on a scale of 0 to 9 how closely related the topic of the second set of remarks is (no matter who made them) to the first set of remarks." Ratings were made on a scoring sheet on scales ranging from 0 = "not at all related" to 9 = "exactly the same topic." This technique resulted in a continuous series of ratings for each conversation for topic relatedness on the utterance-by-utterance level. Turn level data was computed by taking the average relatedness score for a set of unilateral utterances. Each turn was then marked with the time it occurred during the interaction.

Turn length was measured in two steps. First, using specially designed computer hardware and software (see Palmer & Cunningham, 1987), coders focused on only one partner and pressed buttons indicating start and stop times of speech. The computer sampled the button box at each .1 second and recorded the talk and silence codes in real time. Coding reliability, calculated using Cohen's Kappa, was K = .88 (averaged over 5 tests).

After coding, special computer programs transformed the talk and silence data into turn units based on definitions of turn-holding by Jaffe and Feldstein (1970). The rules of this system identify a turn as a unilateral utterance or a string of uninterrupted talk. Turn duration in time was calculated for each of these turn units.

A total of 60 undergraduates participated as dominance judges for extra credit. These judges were randomly distributed across the four stimuli such that 30 students judged each of the 8 conversational partners. Each observer judged only one partner in each conversation, alternating between partners on each side of the video screen. Judges were randomly assigned to one of 4 orders of presentation (i.e., 15 per order). One group judged the partner on the left of the video in conversation one, the partner on the right in conversation two, on the left in three, and on the right in four. Another group did the same partners but started with conversation four. A third group judged conversation one, right side; conversation two, left; three, right; and four, left. The fourth group judged the same partners as the third but started with conversation four.

Procedures for Collecting Dominance Judgments. Stimuli were presented on a large, 27" monitor screen labeled "LEFT" and "RIGHT" to help subjects focus on one partner. In front of each participant was a dial box that was connected to a computer that recorded their judgments at each .1 second (see Palmer & Cunningham, 1987). After being seated, participants listened to an audio tape containing instructions. Each judge was also provided with a written copy of the instructions.

The instructions directed participants to concentrate on the assigned partners and to "decide how dominant that person is trying to be in the conversation." Dominance was defined for the judges as "how much that person is trying to control how the other person acts or thinks during the conversation." Judges were then told to use the dial box by turning the dial to the right the more dominant they perceived the partner. The far left of the dial was labeled "0" for zero dominance.

The experimental phase began after subjects completed a 90-second practice run. Judges watched all four tapes in one of the randomly assigned presentation orders. Between conversations, subjects answered a set of questions about the conversational partners contained in their test booklets.

Results

Data Transformation. For the final analyses, data were reduced to 1 second intervals by taking the average of 10, .1 second data points. This transformation was used because it better reflected the duration of turns that were perceptibly greater than .1 second. Also, this reduction adjusted for possible error resulting from the time-coding of the transcript used in topic relatedness codes. The transcript was time coded from a video tape player that registered time only in 1 second intervals. The .1 second data collected on-line was aggregated to match the level of accuracy of the transcript data.

Check of The Stimuli. After coding the transcripts for topic relatedness, the 44 participants were asked to judge the conversations' naturalness. As in Study 1, this test ensured that the stimuli typified initial interactions and that the generalizability of the findings would not be limited. Naturalness was operationalized by three Likert-type scales (1 = "strongly agree," 3 = "mildly agree," 4 "= neutral," and 7 = "strongly disagree") measuring whether: (1) the partners acted as might be expected in an initial encounter; (2) the conversation was NOT like most initial encounters (reversed for analysis); and, (3) the conversation was "normal." Every mean on every item was significantly greater than scale level 3 (neutral) for all conversations except in two instances. In conversations one and two the means on item 2 were greater than 3, but not significantly so. However, all other scales concerning these conversations were greater than 3 thus indicating participants considered the stimuli "natural."

Tests for Order of Presentation Effects. The effects of order of presentation were tested using a repeated measures ANOVA (Rosenthal, 1987). No significant order effects were found indicating that subjects seeing conversation one through four did not use the scales differently than subjects seeing the same conversations ordered four through one.

Table 3

Time Series Regressions: Perceptions of Dominance with Topic Relatedness and Turn Duration

Set One

Variable

Coefficient

Standard Error

T

p

Adjusted R2

F

p

Conversation One, Partner on Left Side

Topic Relatedness	-.01	.03	-.18	.86	.15	11.42	.00
Turn Duration	-.004	.09	-.05	.96
AR (1)*	.41	.07	5.84	.000
Constant	.08	.05	1.42	.16

Conversation Two, Partner on Right Side

Topic Relatedness	-.002	.07	1.12	.26	.07	5.50	.000
Turn Duration	.07	.11	.57	.57
AR (1)*	.29	.07	3.96	.000
Constant	.08	.07	1.12	.26

Conversation Three, Partner on Left Side

Topic Relatedness	-.09	.08	-1.10	.27	.07	4.40	.002
Turn Duration	.15	.19	.78	.43
AR (1)*	.28	.08	3.60	.000
AR (2)*	-.15	.08	-2.01	.05
Constant	.06	.08	.70	.48

Conversation Four, Partner on Right Side

Topic Relatedness	-.01	.04	-.27	.78	.06	4.70	.003
Turn Duration	-.02	.11	-.16	.87
AR (1)*	.27	.07	3.74	.000
Constant	.09	.05	1.70	.09

Set Two

Overall

Variable

Coefficient

Standard Error

T

p

Adjusted R2

F

p

Conversation One, Partner on Right Side

Topic Relatedness	-.22	.14	-1.54	.12	.36	33.67	.000
Turn Duration	.32	.39	.82	.41
AR (1)*	.60	.06	9.92	.00
Constant	.17	.38	.46	.65

Conversation Two, Partner on Left Side

Topic Relatedness	-.03	.04	-.63	.53	.12	9.28	.000
Turn Duration	.06	.17	.39	.70
AR (1)*	.37	.07	5.15	.000
Constant	.06	.12	.50	.62

Conversation Three, Partner on Right Side

Topic Relatedness	.02	.04	.38	.70	.21	16.53	.000
Turn Duration	-.02	.09	-.28	.78
AR (1)*	.47	.07	7.05	.000
Constant	.06	.07	.88	.38

Conversation Four, Partner on Left Side

Topic Relatedness	.06	.04	1.76	.08	.39	28.60	.000
Turn Duration	-.17	.11	-1.57	.12
AR (1)*	.74	.07	9.87	.000
Constant	-.24	.08	-3.21	.002

Note. AR (1)*, AR(2)* indicate first and second order autoregressive components, respectively.

Reliabilities of Dominance Judges. Reliabilities of the dominance judges were calculated to assess homogeneity of the pooled perceptions of dominance. Initial calculations revealed high variability among the original 30 judges rating each of the partners. Although the analysis plan called for using means for each judgment group, it was decided that high variability around those means may have reflected artifactual effects. For example, judges unfamiliar with making explicit dominance judgments or with the on-line judgment methods might have used the scales in the extremes. Therefore, an iterative process was used in which judges were removed from a group until stable, high reliabilities were attained. Final reliabilities for the 8 judgment groups ranged from alpha = .89 to .95, with numbers of judges from 10 to 15 (Rosenthal, 1987).

Modeling The Time Series. This process required the detailed analysis of 3 time series (perceived dominance, turn time, and topic relatedness) for each of the 8 conversational participants. Time series identification of dominance judgments revealed a first order, autoregressive pattern in the autocorrelations (AC) and partial autocorrelations (PAC) for dominance judgments for all conversational partners except conversations three and four, right side which showed spikes at interval 2 (i.e., second order, autoregressive components). Also, high Q statistics for the variables indicated a large amount of serial dependence.1 Standard transformations of the data were carried out to reduce autocorrelation effects. As a final check on time series diagnostics, the first and/or second order autoregressive components were regressed on the transformed data. Table 2 shows the final diagnostic results of regressing a time series component on each of the variables that were differenced. A significant autoregressive component means that the diagnostic procedure has successfully identified the degree of autocorrelation in the time series and that entering this component in the final regression model (including the variables of interest) will control for autocorrelation effects. In all cases the Q statistic was greatly reduced and was non-significant or nearly so. Also, in each case the Durbin-Watson statistics was about 2.00, indicating the removal of serial dependence.

Topic relatedness showed no serial components (Table 2). However, these ratings were also subjected to differencing in order to transform the series at the same level of the dominance judgments. Turn Duration proved to have some autoregressive components, which are displayed in Table 2.

Tests On Unsmoothed Data. After preliminary diagnostics, linear models for each partner in each conversation were created by regressing topic relatedness, turn duration, and time series components with serial judgments of dominance. All relevant time series components were entered into regression analyses along with the independent and dependent variables. Where time series components were not significant in the regressions, they were deleted.

Although there were 8 separate and independent groups providing dominance judgments on one and only one partner, it was possible that perceived dominance for one partner might have affected judgments of dominance for the other. Also, one partner's behaviors were likely to influence the other partner's responses. To control for this possibility, analyses were organized into two sets. The first set grouped the analyses of conversations one, left side; two, right side; three, left side; and four, right side. The second set included the analyses of the remaining conversational partners. Convincing evidence would therefore depend on finding similar results in multiple conversations in each set of analyses.

As shown in Table 3, there were no significant effects of topic relatedness or turn

duration on perceptions of dominance in the first or second set. However, in the first set, all of the coefficients for topic relatedness were in the expected, negative direction. In the second set, coefficients for this variable were negative in two regressions (conversation one, right side and two, left side). Thus, there was weak evidence of a negative relationship between topic relatedness and perceptions of dominance. Turn duration coefficients were positive in two of the four conversations in the first and second sets. Thus, expectations of a positive association of dominance and turn duration were not strongly supported in these data.

Tests with Smoothed Data. A moving average algorithm was applied to the data using windows of 10, 20, and 30 seconds (Makridakis & Wheelwright, 1978). Significant or near-significant findings were found only at smoothing done with 20 or 30 second windows. Results of these analyses are shown in Table 4. Set one analyses showed significant or near-significant results for topic relatedness in conversations two, right side and three, left side. In one case,

Table 4

Time Series Regressions of Smoothed Data: Perceptions of Dominance with Topic Relatedness and Turn Duration

Set One

Overall

Variable

Coefficient

Standard Error

T

p

Adjusted R2

F

p

Conversation Two, Partner on Right Side**

Topic Relatedness	-.05	.02	-2.12	.04	.12	7.67	.000
Turn Duration	.03	.11	.30	.76
AR (1)*	.35	.08	4.50	.000
Constant	-.001	.003	- .19	.85

Conversation Three, Partner on Left Side**

Topic Relatedness	-.11	.06	-1.81	.07	.18	7.26	.000
Turn Duration	.26	.16	1.55	.12
AR (1)*	.31	.08	3.82	.000
AR (1)*	-.33	.08	-4.07	.000
AR (1)*	.32	.08	3.91	.000
Constant	-.002	.001	- .43	.67

Conversation Four, Partner on Right Side**

Topic Relatedness	.03	.034	1.32	.19	.04	2.88	.04
Turn Duration	-.05	.12	- .37	.71
AR (1)*	.21	.08	2.51	.01
Constant	-.002	.003	- .69	.50

Set Two

Overall

Variable

Coefficient

Standard Error

T

p

Adjusted R2

F

p

Conversation One, Partner on Right Side***

Topic Relatedness	- .20	.11	-1.76	.08	.42	38.77	.000
Turn Duration	1.05	.43	2.43	.02
AR (1)*	.65	.06	10.52	.000
Constant	- .001	.03	- .004	.996

Conversation Three, Partner on Right Side**

Topic Relatedness	-.08	.03	-2.56	.01	.12	7.67	.000
Turn Duration	.05	.09	.56	.58
AR (1)*	.51	.07	7.09	.000
Constant	.002	.004	.39	.70

Conversation Four, Partner on Left Side**

Topic Relatedness	.002	.03	.06	.95	.25	12.77	.000
Turn Duration	-.33	.14	-2.38	.02
AR (1)*	.56	.08	6.67	.02
Constant	-.001	-.004	-.23	.82

Note. AR (1)* indicate first autoregressive components. ** indicates data smoothed at 30 seconds. *** indicates data smoothed at 20 seconds.

conversation four, right side, the non-significant result in the smoothed data analysis also reversed the direction of the coefficient. Thus, smoothing strengthened the original findings for topic relatedness in two of four cases. Turn duration showed almost no change in the smoothed analysis.

The second set of conversational analyses of smoothed data also showed improvement. In conversations one, right side and three, right side, smoothing brought topic relatedness to significance or near significance with the signs of coefficients remaining in the expected direction. Turn duration coefficients also reached or approached significance in the expected direction. Smoothing reinforced the non-smoothed findings for conversation four, left side but as with the non-smoothed analyses, produced coefficients with signs opposite to predictions.

Therefore in many cases, smoothing at a large interval (20 to 30 seconds) helped: (1) to reinforce the findings of analysis done on the original, non-smoothed data; and, (2) to show significant or near significant relationships of topic relatedness and turn duration with perceptions of dominance. However, the improvements were small in several cases and were reversed in one conversation (four, right side). Furthermore, the smoothing intervals were selected without theoretical rationale and findings reported here mark only consistent patterns in the results. The 30 second smoothing (and in one case, 20 second) produced the only consistent "improvements" in the results.

Summary Level Versus Time Series Analyses. Although this study was not designed to test for differences between summary level and time series analyses, the data provide an opportunity to make comparisons. For this analysis, summary level dominance judgments were taken from a different study in which the same four conversations were used as stimuli (Palmer, Cappella, Patterson, & Churchill, 1992). In that study 56 undergraduates were randomly assigned to 8 groups of 7 subjects. Each judge viewed only one conversation and rated only one partner on 5 dominance subscales.2 Thus, the dependent measure was the mean of an overall judgment for each conversation. Topic relatedness was computed as the average topic relatedness score for each partner's utterances, and turn duration as the proportion of time a partner held the floor.

Discrepancies between serial judgments and summary level analyses were indicated in these data in several instances. Agreements between the two levels were fewer and more tentative. In conversation one, there were no differences in overall perceived dominance between partners (t = -1.11, df = 35, p > .05). Thus, it was not possible to find an association between behaviors and perceptions, even though topic relatedness and turn length differed between partners at the summary level ( t = 8.61, df = 128, p < .001 and t = 5.88, df = 228, p < .001, respectively). However, as reported above, serial level analyses did suggest associations between perceived dominance and behaviors for this conversation.

Conversation two did show differences in dominance between partners at the summary level (t = 2.39, df = 34, p < .05), but there were no differences between partners on topic relatedness (t = -1.15, df = 117, p > .05) or turn duration (t = -.59, df = 239, p > .05). However, as in conversation one, serial level analyses did reveal some evidence of an association between perceptions of dominance and these behaviors (Table 4).

In conversation three, the partner on the Left was perceived as more dominant than the partner on the Right (t = 3.90, df = 34, p < .05). The left partner also changed the topic more (t = -2.68, df = 127, p < .05) but talked less (t = -2.07, df = 229, p < .05). Therefore, at the summary level, it is suggested that decreases in both topic relatedness and turn duration are associated with increases in perceived dominance. The serial level analyses showed dominance to be weakly associated with decreases in topic relatedness (as at the summary level), but not at all with turn duration (Table 4). Furthermore, the positive direction of the coefficients for turn duration might be interpreted as contradictory to the direction of the findings at the summary level; that is, at the serial level increases in turn duration would be associated with increases in dominance (if at all) rather than with decreases in dominance as suggested by the summary level data. Thus, only partial agreement between levels of analysis was suggested for topic relatedness but not for turn duration and dominance.

Analyses of the summary level data for conversation four showed that the partner on the right was perceived as more dominant (t = -10.20, df = 36, p < .05). This partner was also found to change the topic more (t = 4.00, df = 128, p < .05). The serial level analyses (Table 4) showed no association of dominance with topic relatedness although the coefficient was positive (as expected). Therefore, there is no strong parallel between summary level and serial level analyses of this variable in conversation four.

The findings regarding turn duration in this conversation, however, produced the only instance in which serial level analyses agreed with summary level analyses. In this case the partner who spoke less was perceived as more dominant overall (t = 2.85, df = 228, p < .05). This finding is contradictory to the hypothesis that predicted increases in dominance associated with increases in talk time but agrees with the results at the serial level, which also showed a negative association of talk time and perceived dominance (Table 4).

Therefore, the general trend in these data indicated little or no agreement between results from serial judgment analyses and overall, summary level findings. In only one conversation was there any indication that significant findings from the two levels of analysis might agree.

Conclusion

Topics, Turns, and Interpersonal Control

Results of Study 1 strongly indicated that on an act-by-act basis, the more a partner changed a topic and the longer a partner spoke, the more that partner was considered to be dominant. The data in Study 2 offered only weak support for the hypotheses that topic relatedness was negatively related to perceptions of dominance; and little or no support for the idea that turn duration is positively related to dominance at the moment-to-moment level of analysis. However, smoothing the data at a large interval (i.e., 30 time units) did strengthen the findings.

Reasons for discrepancies in results between studies might be attributable to several factors. First, differences in modes of presentation of the conversational stimuli may have caused differences in the criteria by which inferences were made. Study 1 judges made ratings of dominance using a transcript which forced the raters to focus entirely on the verbal features of the communication. Study 2 used video taped stimuli and allowed judges to focus on any verbal or nonverbal features which may have drawn their attention while making judgments. Thus, the greater number of channels permitted by the video tape could have reduced or eliminated the importance of topic relatedness and turn duration when judgments were made on a moment-to-moment basis.

Second, in Study 2 judges were asked to focus on one and only one partner when making their ratings of dominance. In Study 1, judges were asked to compare the two partners when making judgments. The judges of Study 1 could have been more sensitive to the relational dimension of dominance because they were forced to compare both partners. Study 2 judges may have relied less on relational aspects in making judgments. Thus, the studies may represent two different kinds of dominance judgments and produced two different kinds of results.

Together, the results of these studies offer some indications that interpersonal dominance is an emergent property of a social interaction. Irrespective of individual differences in personality or status, conversational behaviors have relational impact at the time they are used. These findings might offer an explanation of how lower status individuals who may enter a conversation at a power disadvantage, are able to exert influence at least temporarily, on higher status partners at critical moments in the interaction. While these conclusions are based on the findings of only two studies using new methods, the indications are strong and heuristically provocative enough to warrant further investigation.

The relationship between the "temporary" and dynamic level of perceptions and behaviors and the more permanent images of a relationship are not apparent in these data. It is unclear whether the momentary effects of conversational behaviors and perceptions are stored in long term memory or influence the overall, permanent record of a relationship. However, the differences between the summary level and time series analyses in these data suggest that links may be weak.

Summary Level Versus Time Series Analyses

Both studies produced some indications that differences may exist between judgments gathered on the act-by-act or moment-to-moment basis and judgments gathered at the summary level. In only one case was there any parallel between the results at different levels of analyses. Such similarities and differences are only tentative in these data because of weak findings in Study 2 and because the studies were not designed to test them. However, enough indications of differences between the two modes of data collection exists to warrant future research. Such differences may be important if it can be shown that the two approaches, serial judgments and summary level judgments, represent different cognitive processes. For example, if on-line serial judgments represent the moment-by-moment, changing states of relational inferences, then results can be said to describe only effects operating during an interaction. Summary level judgments that may be based on more permanent representations of the relationship may have longer lasting effects. Thus, the two modes of judgment may tap different features of memory and the information processing system (see Hastie & Park, 1986). The important question would then be how the two modes are related; that is, how does the moment-by-moment processing affect the summary level decision and which kind of decision is stored in long term memory.

This research has attempted to describe the relationship between topic relatedness, turn duration and interpersonal dominance in the context of a dynamic, over time approach. The studies presented here have demonstrated the feasibility and some of the limitations of serial judgment methods in modeling the dynamic interaction of cognitions and behaviors. While the findings stand for themselves, the future applications of the methods remain a challenge. If communication can be thought of as a process in which two individuals interactively create a shared relationship, then increased attention should be given to over-time, serial analysis and data collection paradigms.

Notes

1The Q statistic tests whether the series is different than a random, white noise process. Q is compared to a Chi Square with degrees of freedom equal to k - p - q - P - Q; where k is number of observations, p is the order of the autoregressive component, q is the order of the moving average component, P is the autoregressive seasonal component, and Q is the moving average seasonal component. A significant Q means the presence of some autocorrelation.

2These subscales were part of a 30 item instrument composed of 17 items taken from the Relational Messages Scale (Burgoon & Hale, 1987) and 13 items composed especially for Palmer and colleagues (1992). The dominance subscales consisted of 7-point Likert-type scales assessing how much a partner tried to dominate their partner, control the interaction, be aggressive, influence, and persuade. Reliability for this subscales was alpha = .82.

References

Burgoon, J. K., & Hale, J. L. (1987). Validation and measurement of the fundamental themes of relational communication. Communication Monographs, 54, 19-41.

Cappella, J. N. (1984). The relevance of the microstructure of the interaction to the relationship change. Journal of Social and Personal Relationships, 1, 239-264.

Courtright, J. A., Millar, F. E., & Rogers-Millar, L. E. (1979). Domineeringness and dominance: Replication and evaluation. Communication Monographs, 46, 179-192.

Ellyson, S. L., & Dovidio, J. F. (Eds.) (1985). Power, dominance and nonverbal behavior. New York: Springer-Verlag.

Ericsson, K. A., & Simon, H. A. (1984). Protocol analysis: Verbal reports as data. Cambridge, MA: The MIT.

Gentner, D., & Stevens, A. L. (1983). Mental models. Hillsdale, NJ: Erlbaum.

Hastie, R. & Park, B. (1986). The relationship between memory and judgment depends on whether the judgment task is memory-based or on-line. Psychological Review, 93, 254-261.

Hibbs, D. A. (1974). Problems of statistical estimation and causal inference in time-series regression models. In H. Costner (Ed.), Sociological methodology (pp. 252-308). San Francisco, CA: Jossey-Bass.

Jaffe, J., & Feldstein, S. (1970). Rhythms of dialogue. NY: Academic.

Johnson-Laird, P. N. (1983). Mental models. Cambridge, MA: Harvard University.

Lewicki, P. (1986). Nonconscious social information processing. NY: Academic.

Makridakis, S., & Wheelwright, S. C. (1978). Forecasting: Methods and applications. NY: Wiley.

McCleary, R., & Hay, R. A. (1980). Applied time series for the social sciences. Beverly Hills, CA: Sage.

Nisbett, R. S., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84, 231-259.

Palmer, M. T. (1989). Controlling conversations: Turns, topics and interpersonal control. Communication Monographs, 56, 1-18.

Palmer, M. T. (1990). Mapping relational inferences onto conversational behaviors: Cross-sectional and time series analyses. Ph D dissertation, University of Wisconsin-Madison.

Palmer, M. T. & Cunningham, R. (1987, November). Computerizing the collection and coding of interaction data. Paper presented to the Speech Communication Association, Boston.

Palmer, M. T., Cappella, J. N., Patterson, M., & Churchill, M. (1992). Mapping relational inferences onto nonverbal behaviors: Summary level analyses. Unpublished manuscript.

Rogers, L. E. & Farace, R. V. (1975). Analysis of relational communication in dyads: New measurement procedures. Human Communication Research, 1, 222-239.

Rosenthal, R. (1987). Judgment studies: Design, analysis and meta-analysis. NY: Cambridge University.

Street, R . L. & Cappella, J. N. (1985). Sequence and pattern in communicative behavior: A model and commentary. In R. L. Street & J. N. Cappella (Eds.) Sequence and pattern in communicative behavior (pp. 243-276). Baltimore: Arnold.

Response from Leonard J. Shedletsky

At least three serious problems are engendered by serial judgment methods applied to conversation. First, as the authors note, a critical assumption is required, namely, that "judges' on-line responses approximate moment-by-moment inferences." Second, treating conversation as if it contains countable units that bear upon relational issues requires operationalization of concepts such as dominance, topic-relatedness, and turn. And third, conversation may well activate both local (moment-by-moment) information and global (e.g., relational history) information. That the studies reported here find "little or no agreement between results from serial judgment analyses and overall summary level findings" leaves the reader wondering if it is possible to simplify meaning to fit the constraints of the serial judgment technique. As the authors point out, an additional concern is the comparison between studies that varied in just how the conversation was presented to subjects, e.g., transcript versus video, focus on one or both interactants. We can also wonder whether participant versus observer status matters.

The Effect of Information Processing Objectives on Persuasion

Lori A. Roman and Mark A. deTurck

About the Authors: Lori A. Roman is a doctoral candidate at the State University of New York at Buffalo where Mark A. deTurck is an associate professor of communication. The authors can be contacted at the Department of Communication, State University of New York at Buffalo, 338 Millard Fillmore Complex, Buffalo, New York 14260, telephone (716) 645-2141. The authors thank Frank Tutzauer for his incisive comments on an earlier version of this article. This article is based in part on the first author's masters thesis. An earlier version of this article was a top paper presented at the Speech Communication Association's Commission on Intrapersonal Communication Processes and Social Cognition at the 1993 meeting in Miami, Florida.

Abstract: The present study was designed to determine the effects of information processing objectives on persuasion. Specifically, it looks at Wyer and Srull's (1986) model of cognitive processing and the inclusion of information processing objectives as an input variable. One hundred subjects were instructed to attend to a counterattitudinal message while using an assigned processing objective (impression-set, memory-set, comprehension-set, or default). Subjects' attitude toward the message and message recall were then measured. It was hypothesized that subjects using an impression-set objective would report the least amount of attitude change, whereas subjects with a memory-set objective would report the most attitude change. In addition, the study predicted that recall would be the greatest when an impression-set processing objective is used. A main effect for information processing objectives and attitude is reported. Implications of the results and suggestions for the directions of future research on information processing are addressed.

Perspective:

1. How does a person's reason for attending to a set of informationóthe information processing objectiveóaffect the way the information is cognitively processed?

2. Can specific processing objectives be externally imposed in order to maximize the persuasiveness of a counterattitudinal message?

Recent research suggests that the way a person cognitively processes a persuasive message may mediate the persuasive impact of the message (deTurck & Goldhaber, 1992; deTurck & Goldhaber, 1989). Specifically, a person's reason(s) for attending to a messageóthe information processing objective(s)ómediates the effectiveness of a persuasive message (deTurck & Goldhaber 1992; Wyer & Srull, 1986).

People lack the cognitive ability to process all of the sensory information available to them in the social environment. Likewise, when a person is faced with a persuasive message, they are unable to focus their attention on all of the elements in the social context. Hence, it is necessary for people to filter information available in the environment, and only attend to fragments of the information. The selection of certain stimuli over others is partially dependent on the person's processing objective for the communicative encounter (deTurck & Goldhaber 1992; Wyer & Srull, 1986). More specifically, the person's reason for attending to the social information can influence the selection of certain information over other aspects of the environment.

Processing objectives are often default, where people unknowingly invoke a processing goal while attending to information. However, research indicates that externally imposed information processing objectives exert considerable influence on the formation of attitudes (Wyer & Srull, 1986; Hastie & Park, 1986) and the effectiveness of counterattitudinal persuasive messages (deTurck & Goldhaber 1992). The present study will investigate the effect of externally imposed information processing objectives (instructions to evaluate, memorize, comprehend, or default) on attitude change and recall.

A Cognitive Model of Information Processing

Wyer & Srull's (1986) model of cognitive processing provides a general conceptual framework of how people receive information, process it, and eventually react to the information received. Arguably, the feature that distinguishes Wyer and Srull's (1986) model from other schematic representations of cognitive processing (Hastie & Carlson, 1980; Wyer & Srull, 1986) is its inclusion of information processing objectives as an input variable.

Because people have difficulty absorbing all of the information presented to them in any given social context, they must focus their attention on specific aspects of the environment. An individual's reason for attending to social informationótheir information processing objectiveódirectly affects the way the information is cognitively processed and eventually stored for retrieval.

The model can be viewed as a system of interrelated processing and storage units that process information in an organized pattern, and depends on the current information processing objective being used (Hamilton, Katz, & Leirer, 1980; Texter, 1989; Wyer & Srull, 1986). The information processing objective controls the type and amount of information that individuals process and store (Hamilton, Katz, & Leirer, 1980; Texter, 1989; Wyer & Srull, 1986).

Research on information processing objectives has focused on three objectives: 1) memory-set, 2) impression-set, and 3) comprehension-set (deTurck & Goldhaber 1992; Texter, 1989). With a memory-set objective, people attempt to memorize the information presented to them. With an impression-set objective, the person attempts to form a unit judgment, e.g., attitude, based on a set of information. Finally, a comprehension-set objective prompts a person to gain a basic understanding of a set of social information.

According to Wyer and Srull's (1986) model, when a processing objective is given for a set of information, it is placed in a cognitive goal specification box. Although the goal specification box can contain information about more than one objective at a time, its capacity is limited. Hence, an increase in the complexity of procedures needed to achieve an information processing objective requires a greater portion of the goal specification box to store it, and the number of other goals that can be pursued decreases (Wyer & Srull, 1986)

When activated, the processing objective informs the executor (a monitor that controls the flow of information) which units in the model to activate. Hence, the processing objective steers the entire cognitive process. If the goal specification box is empty because no designated goal is given with the information, the executor initiates a default objective to guide how the processing units address the information.

After information enters the model through the sensory organs, the information moves to a temporary storage unit called the work space. The work space acts as a "temporary repository" for all information "operated on in pursuit of specific processing objectives" (Wyer & Srull, 1986). Both the information placed into the work space from sensory system, and the way the information is processed are dependent on the person's information processing objective. Once the goal of the processing objective is met, the goal related information moves from the work space into permanent storage. Any non-goal oriented information not sent to permanent storage is discarded and the work space is cleared.

Permanent storage is a long-term memory unit composed of content-addressable information storage bins. The material in the bins may vary greatly and is largely dependent on the information processing objective used when the information was first received. The major difference between permanent storage and the work space is that once information is placed in permanent storage it is saved. The work space, however, is irretrievably cleared once the information is no longer needed to achieve the cognitive processing goal.

Because processing objectives direct the way information is processed and information not placed in permanent storage is lost after the processing objective is met, it can be argued that processing objectives influence persuasion. Specifically, a memory-set objective facilitates attitude change because the goal of the objective is to memorize information, not form evaluations of the arguments presented in the message. Therefore, when a person's objective is to memorize information for later use, and the task is accomplished, all non-goal oriented informationóevaluationsóis eliminated from working memory and never stored in permanent memory. In addition, the rehearsal process underlying memorization inhibits subvocal counterargumentation and attitude formation, allowing only for the memorized message to be available for permanent storage (deTurck & Goldhaber 1992). Hence, in order for a person to form an evaluation, they must first recall all of the information from permanent memory, and then form an impression.

Impression-set objectives inhibit persuasion because they prime message receivers to evaluate the arguments and evidence in the persuasive message (deTurck & Goldhaber 1992). Hence, they prime people to engage in subvocal counterargumentation by directing the receiver to critically evaluate information in the work space. When the work space is cleared, the attitude regarding the persuasive message is stored in permanent memory. Therefore, permanent memory holds the evaluation and not factual information. The information used to create the attitude (i.e. the message's arguments) is often irretrievably lost when the work space is cleared.

Comprehension-set objectives prompt people to attempt to gain a basic understanding of a message by attending to, and eventually, storing factual information (Lichtenstein & Srull, 1985, 1987). Thus, message receivers will not counter-argue and critically evaluate the message. Therefore, persuasion should be easier with a comprehension-set objective than with an impression-set objective. However, a memory-set objective should still be superior to a comprehension-set objective in enhancing a message's persuasibility due to the distraction caused by the repetition or memorization process.

Although difficult to ascertain a person's default objective, research indicates that a person's default objective closely mimics an impression-set objective when presented with a counter-attitudinal message (deTurck & Goldhaber 1992, Stafford & Daly, 1984). Wyer & Srull's (1986) study focusing on person perception, however, indicates that the default objective closely parallels a comprehension-set objective. The difference in the assumed default is most likely due to the differences in the task (i.e., judgment formation versus person perception). More research in this area will allow for greater insight in determining which processing objective is the default.

Effect of Information Processing Objectives on Recall

Research on the hypothesis that more factual information is available from memory when subjects process information using a memory-set objective than when using an impression-set objective has revealed different results. Specifically, deTurck and Goldhaber (1989) found greater recall of information in product warning labels when a memory-set objective was imposed than when subjects were asked to use an impression-set objective. In contrast, Hamilton, Katz and Leirer (1980) found greater recall under impression-set conditions than in memory-set conditions. According to Hamilton, Katz, and Leirer (1980), an impression-set objective prompts receivers to create links or networks among incoming information so as to form a unit evaluation, whereas a memory-set objective requires the majority of the energy devoted to processing the information to the task of memorization and not integrating the material into networks. Hence, a person can recall information from an impression-set objective by accessing the network. It is more difficult, however, for the person to recall the pieces of information stored under memorization (Hamilton, Katz & Leirer, 1980; Srull, 1983).

Relationship Between Attitudes and Recall

The relationship between the amount of information recalled and its effect on judgments is based on the information processing objective used (Lichtenstein & Srull, 1985). Specifically, when using an impression-set objective, only the goal related informationóthe evaluationóis stored for retrieval and non-goal oriented information is discarded. Hence, when a judgment is to be made, the retrieval of the stored evaluation is sufficient to form the judgment. In this vein, the amount of information recalled should not be related to the formation of attitudes.

When the initial processing objective is to memorize, the retrieval process for judgment formation requires the recall and computation of all remembered facts. The retrieval process is necessary due to the lack of attitude formation during the information processing stage. When processing information for memorization, information is rehearsed and sent to permanent storage until it is needed. When the person needs to form an attitude in the future, the person recalls the original information and then computes an attitude. Hence, the amount of information recalled should correlate with the formation of attitudes.

A comprehension-set objective also requires a person to store factual information rather than an attitude. Because an attitude alone is not stored, the formation of an attitude requires the recall of the information labeled a basic understanding. Thus, the attitude should be related to the information recalled.

Distraction Effects

Petty, Wells & Brock (1976) argue that distractions interfere with a message receiver's dominant cognitive response. In their study, a distraction stimulus increased the persuasiveness of an easily counter-argued message. In this vein, a distraction occurring simultaneously with the presentation of a persuasive message can be argued to hinder subvocal counterargumentation. The underlying rehearsal needed for the memorization of information should act as a distraction. Hence, the rehearsal process needed with a memory-set objective serves as a simultaneous distraction.

Hypotheses

As previously stated, the goal of this study is to examine how information processing objectives affect persuasion. Based on the above line of argument, the following hypothesis is posited. People who process the persuasive message with a goal to memorize the information will be most persuaded by the message. An impression-set objective will prompt a great amount of subvocal counterargumentation, and hence, result in the least amount of positive attitude change.

To the extent that a message receiver's default processing objective is to form an impression or attitude, the pattern of obtained results when using a default objective should closely parallel the findings obtained from an impression-set condition. However, if Wyer & Srull's (1986) argument is correct for judgment formation as well as person perception, then results in the default condition should be similar to the comprehension-set condition.

Specifically, it is expected that:

Hypothesis 1: Subjects using an impression-set objective should demonstrate less attitude change than those using a comprehension-set objective, and those using a comprehension-set objective should have less attitude change than those using a memory-set objective.

Hypothesis 2: If an impression-set objective is the default objective, subjects using a default objective should experience attitude change in line with the amount experienced in the impression-set condition. If a comprehension-set objective is the default, subjects using a default objective should experience attitude change in line with the amount experienced in the comprehension-set condition.

Hypothesis 3: Subjects should have the greatest amount of recall when an impression-set processing objective is used. The tested recall, however, will not affect the persuasive response to the message.

Method

Sample

One hundred first and second year full-time undergraduate students from an introductory course in communication at an Eastern University participated in partial fulfillment of a research requirement. Subjects were randomly assigned to the experimental conditions.

Design

The independent variable is the information processing objective (memory-set, impression-set, comprehension-set and none) and the dependent variables are attitude toward the message's issue advocated in the persuasive message, and recall.

Procedure

A standard written prebriefing was given to the participants when they arrived at the experiment site. The prebriefing informed the participants that the Department of Communication was conducting a study aimed at obtaining more insight about students.

Participants were then placed in one of four experimental conditions. First, the information processing condition was assigned. Each subject then received an instruction form that was specific to their processing objective condition. The instruction form was adapted from deTurck and Goldhaber (1992) and is listed below for each of the conditions.

Memory-set. In the memory-set condition, the instruction sheet read:

We would like you to try to memorize as much of the information as possible. After attending to the message, you will be tested to see how much of the information you can accurately recall. Remember, your goal is to MEMORIZE as much of the information as possible.

Impression-set. In the impression-set condition, the instruction sheet read:

We would like you to critically evaluate the information in the message. After attending to the message, you will be asked your opinion regarding the issue in the message. Remember, your goal is to CRITICALLY EVALUATE the information in the message so as to form an opinion.

Table 1

Cell Means and Standard Deviations for Attitude Toward Senior Comprehensive Exams

OBJECTIVEMEANSD

Memory-set14.091.72

Comprehension-set13.312.50

Default12.372.66

Impression-set12.23 2.56

Comprehension-set. In the comprehension-set condition, the instruction sheet read:

We would like you to get a basic understanding of the information in the message. After attending to the message, you will be asked questions regarding the gist of the message. Remember, your goal is to get a BASIC UNDERSTANDING of the message.

Default. In the default condition, the instruction sheet read:

We would like you to simply attend to the message.

After the subject finished reading the pre-briefing, the experimenter asked the subject if they had any questions about the study. Questions were answered by referring the subject back to the pre-briefing statement. When the subject felt comfortable with the instructions, the researcher removed the prebriefing statement.

Subjects were then presented with the persuasive message. The written message was typed on one sheet of standard white paper. The message was adapted from Texter (1989).

Subjects were permitted to attend to the message for as long as they wanted. Hence, the subject's reading time was not restricted. When the subject finished reading the message, the subject informed the research assistant. The researcher immediately gave the subject a distractor task which asked the subject to write down as many of the names of the 50 U. S. states that they could remember. The subjects were given three minutes to complete this task. The researcher then gave the subjects a questionnaire packet.

Measurements

Attitude Toward Issue. The attitudes of the participants toward required senior comprehensive exams prior to graduation were assessed by eight semantic differential scales.

Attitude Toward Message. Subjects indicated their attitude regarding the strength of the message by answering seven semantic differential scales.

Recall. The amount of information recalled was tested by using six multiple choice questions about the message with five foils per question.

Manipulation Check. A manipulation check was obtained for the information processing objective of each participant. Participants were asked to report their main goal with a multiple choice response with foils of memorize, critically evaluate, get a basic understanding, simply attend to, and don't know.

Results

Manipulation Check

Processing Objective. Results of the processing-objective manipulation indicate that subjects were aware of their processing objective. The participants in the memory-set objective condition (92.3%), the comprehension-set objective condition (88.5%), the impression-set condition (77.3%) and the control condition (73.1%) accurately identified their assigned information processing goal [Chi-square (12) = 204.049, p < .01]

Table 2

Correlations of Attitude and Recall

OBJECTIVECORRELATION

Memory-set0.26

Comprehension-set-.07

Impression-set0.33

Default0.41

Measurement Models

Attitude Toward Senior Comprehensive Exams. A factor analysis of the eight semantic differential scales resulted in a one factor solution with factor loadings ranging from .63 to .87. Thus, the eight scales were collapsed and tested as a single index measuring attitudes.

Attitude Toward Arguments in the Message. A factor analysis of the seven semantic differential scales revealed a one factor solution with each of the seven scales loading at or above the .60 level. The measures were collapsed to form a single indicator of attitude toward the argument.

Effect of Information Processing Objectives on Attitude Toward

Senior Comprehensive Exams

An analysis of variance indicated that significant differences existed between processing objective conditions (F 3,92 = 3.077 p < .04). As predicted, a Student-Newman-Keuls test of individual cell means yielded the most positive attitude change in the memory-set condition, followed by the comprehension-set condition, default condition, and impression-set condition (See Table 1.). In addition, the similarity between the attitudes in the default and impression-set conditions is consistent with the predictions made by deTurck and Goldhaber (1992) and Stafford and Daly (1984).

Effects of Information

Processing Objective on Attitudes Toward the Message

It was predicted that subjects using an impression-set objective would be more critical in their assessment of the quality of the arguments in the message. An analysis of variance failed to support the hypothesis that information processing objectives affect message ratings (F 3, 92 = 1.932, p < .130).

Effects of Information Processing Objectives on Recall

It was hypothesized that recall would be affected by the processing objective used, in that subjects should have the greatest amount of recall when an impression-set objective is used. In addition, it was argued that the greater recall would not correlate with the stated attitude toward the issue. Rather, the memory-set condition, though lower in recalled information, should have the highest correlation between attitudes and recall. However, no significant differences in recall were noted across processing conditions (F 3, 96 = .45, p < .720). Furthermore, recall and attitude had a higher correlation in the impression-set condition than in the memory-set condition, although the differences were not significant at the .05 level (See Table 2).

Discussion

This study attempted to examine the effect of information processing objectives on recall and attitude change. Specifically, the study set out to demonstrate the effectiveness of using a memory-set objective in order to achieve attitude change when faced with a counterattitudinal persuasive topic. The data were consistent with the hypothesized relationship between information processing objectives and attitude change.

In addition, this study was designed to learn more about the nature of the default processing objective. The similarities obtained in the findings between the default objective and the impression-set objective suggest that the default objective may be to form an impression. This finding is consistent with results obtained by Stafford and Daly (1984) as well as deTurck and Goldhaber (1992). However, Wyer and Srull (1986) suggest that the default objective is similar to a comprehension-set objective. The difference in the assumptions about the default objective is most likely due to the nature of the information to be processed. Whereas Wyer and Srull (1986) focused on person perception, deTurck and Goldhaber (1992), Stafford and Daly (1984) and this research have looked at situations where social influence is involved. Although the findings of this study suggest that an isomorphism of the default objective and the impression-set objective is, in fact, evident, the similarity alone provides necessary but not sufficient evidence to support the claim that social perceivers' objective is to form impressions. Additional research comparing different research paradigms (e.g., impression formation versus persuasive information) is needed to determine if the default objective varies across information types.

The focus of future research dealing with information processing objectives should also focus on other intervening variables that affect persuasion. These variables include the message's modality and the presentation rate of the information.

Modality has been shown to affect persuasiveness of messages in certain situations (Chaiken & Eagly, 1976; Chaiken & Eagly, 1983). One reason audiovisual material may exert more influence on judgments than written messages is its enhanced vividness (Taylor & Thompson, 1982). Vivid messages attract and hold attention because they are emotionally interesting, concrete and image provoking, and proximate in a sensory, temporal or spatial way (Nisbett & Ross, 1980). The enhanced vividness of an audiovisual message may interfere with a social perceiver's ability to memorize information. Audiovisual messages may prompt people to form impressions spontaneously.

Other factors may also enhance a messages persuasiveness when it is presented in audiovisual format. It may be the case that audiovisual stimuli distract social perceivers from forming judgments (Chaiken & Eagly, 1976; Petty, Wells & Brock, 1976). For example, Petty, Wells, and Brock (1976) obtained evidence indicating that an audiovisual stimulus serves to inhibit a subject's ability to critically appraise persuasive information. In a different vein, Eagly and Chaiken (1976) found that audiovisual messages are more persuasive than written messages, unless the messages are highly complex. When a simple or moderate message is presented in written form, it is easy for people to critically analyze the message and engage in subvocal counterargumentation. However, audiovisual stimuli appear to decrease the tendency for people to engage in subvocal counterargumentation.

The tendency for subvocal counterargumentation to decrease with an audiovisual mode may be due to the numerous non-message stimulió visual distractions that exist in audiovisual channels. Petty, Wells and Brock (1976) argue that visual distraction interferes with a message receiver's dominant cognitive response. In their study, a distraction stimulus increased the persuasiveness of an easily counter-argued message. When a person views an audiovisual message, the visual stimuli may inhibit the person's ability to engage in subvocal counterargumentation. To the extent that an audiovisual stimulus inhibits subvocal counterargumentation, it should be more persuasive than its written counterpart. Therefore, it follows that persuasiveness can be maximized when an audiovisual medium is used with a memory-set objective (deTurck & Goldhaber, 1992). In addition, persuasion should be inhibited when an impression-set objective is used with a written message.

The rate of information presentation may also mediate the persuasive effects of a message. If information is presented to social perceivers at a high rate of speed, then they may have difficulty memorizing the information or forming an impression ófulfilling either a memory-set or impression-set objective. More specifically, information presented rapidly will inhibit an individual's ability to rehearse the to-be-remembered information. Similarly, when information is presented at a high rate of speed, persons seeking to evaluate the information critically should have a difficult time processing and organizing the information into a unit judgmentó forming an attitude. Differences between the processing effects of information processing objectives should be maximized when information is presented at a moderate rate. Findings obtained from this line of research may be of particular interest to professionals in the broadcast media due to the fact that the electronic media control the rate of information transmitted.

In addition to the rate of information presentation, the specific nature of the content itself may affect how people process information. More specifically, messages may vary with respect to language intensity, the extent to which the information is congruous with receivers' pre-existing schemas, or the amount of information. Messages laden with high intensity language may prompt receivers to evaluate information, despite the fact they were instructed to memorize the information. Similarly, incongruent information may lead receivers to evaluate the disconfirming information so as to integrate the information into an already existing attitude framework (Wyer & Srull, 1986). The extensiveness of the information-set may also cause people to use a processing objective other than the one they were instructed to use. For example, if a receiver is instructed to memorize a set of information that is quite extensive and requires a substantial degree of processing, then she or he may abandon a memory-set objective in favor of a processing goal (e.g., comprehension-set) that requires less cognitive energy.

Research on information processing objectives would benefit from an examination of the effects of various message factors on receivers' processing. As mentioned above, certain linguistic message patterns may override receivers' processing objectives and prompt them to utilize a different processing strategy. To the extent that message content or presentation affect receivers' processing, it would be useful for future research to determine the specific effects of these message characteristics on receivers' information processing objectives, and ultimately, the persuasive effects of the message.

References

Chaiken, S., & Eagly, A. H. (1976). Communication modality as a determinant of message persuasiveness and message comprehensibility. Journal of Personality and Social Psychology, 34, 605-614.

Chaiken, S., & Eagly, A. H. (1983). Communication modality as a determinant of persuasion: The role of communicator salience. Journal of Personality and Social Psychology, 45, 241-256.

deTurck, M. A., & Goldhaber, G. M. (1989). Effectiveness of product warning labels: Effects of consumers' information processing objectives. Journal of Consumer Affairs, 23, 111-126.

deTurck, M. A., & Goldhaber, G. M. (1992). Effects of information processing objectives on persuasion. Speech Communication Annual, 6, 79-106.

Eagly, A. H. & Chaiken, S. (1984). Cognitive theories of persuasion. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 17, pp. 268-359). New York: Academic.

Hamilton, D. L., Katz, L. B., & Leirer, V. O. (1980). Organizational processes in impression formation. In R. Hastie, T. M. Ostrom, E. B. Ebbesen, R. S. Wyer, D. L. Hamilton, & D. E. Carlston (Eds.), Person memory: The cognitive basis of social perception (pp. 121-153). Hillsdale, NJ: Erlbaum.

Hastie, R. & Carlson, D. E. (1980). Theoretical issues in person memory. In R. Hastie, T. M. Ostrom, E. B. Ebbesen, R. S. Wyer, D. L. Hamilton, and D. E. Carlson (Eds.), Person Memory: The cognitive basis of social perception (pp. 1-54). Hillsdale, NJ: Erlbaum.

Hastie, R. & Park, B. (1986). The relationship between memory and judgment depends on whether the judgment is memory-based or on-line. Psychological Review, 93, 258-268.

Lichtenstein, M. & Srull, T. K. (1985). Conceptual and methodological issues in examining the relationship between consumer memory and judgment. In L. F. Alwitt & A. A. Mitchell (Eds.), Psychological Processes and Advertising Effects, (pp. 113-128). Hillsdale, NJ: Erlbaum.

Lichtenstein, M. & Srull, T. K. (1987). Processing objectives as a determinant of the relationship between recall and judgment, Journal of Experimental Social Psychology, 23, 93-118.

Nisbett, R. & Ross, L. (1980). Human inference: Strategies and shortcoming of human judgment. Englewood Cliffs, NJ: Prentice-Hall.

Petty, R. E., Wells, G. L., & Brock, T. C. (1976). Distraction can enhance and reduce yielding to propaganda: Thought disruption versus effort justification. Journal of Personality and Social Psychology, 34, 874-884.

Srull, T. K. (1983). Organizational and retrieval processes in person memory: An examination of processing objectives, presentation format, and the possible role of self-generated retrieval cues. Journal of Personality and Social Psychology, 44, 1157-1170.

Stafford, L., & Daly, J. A. (1984). Conversational memory: The effects of recall mode and memory expectancies on remembrances of natural conversations. Human Communication Research, 10, 379-402.

Taylor, S. E. & Thompson, S. C. (1982). Stalking the elusive "vividness" effect. Psychological Review, 2, 155-181.

Texter, L. A. (1989). The effects of information processing objectives and issue involvement on attitude formation latencies, recall, and attitudes. Unpublished Dissertation. State University of New York at Buffalo.

Wyer, R. S., Jr. & Srull, T. K. (1980). The processing of social stimulus information: A conceptual integration. In R. Hastie, T. M. Ostrom, E. B. Ebbesen, R. S. Wyer, Jr., D. L. Hamilton, & D. E. Carlston (Eds.), Person Memory: The Cognitive Representation of Social Perception (pp. 227-301). Hillsdale, NJ: Erlbaum.

Wyer, R. S., Jr. & Srull, T. K. (1986). Human Cognition in its social context. Psychological Review, 93, 322-359.

Response from Leonard J. Shedletsky

It feels intuitively right that information processing objectives affect various outcomes. What seems counter-intuitive, however, is that the memory-set produces the same recall as other processing conditions. Is it that the memory demand was simply insufficient to bring out differences? Six multiple choice questions may not be enough; recognition memory may not tax the retrieval process sufficiently.

The Impact of Communication Apprehension, Gender, and Time on Perceptions and Behavior in Extended Interactions

Joe Ayres, Tim Hopf, Kevin Brown, and Julia Suek

About the Authors: Joe Ayres is a professor (Ph. D., University of Utah) at the School of Communication, Washington State University, Pullman, WA 99l64-2520. Tim Hopf is a professor (Ph. D. Pennsylvania State University), School of Communication, Washington State University, Pullman, WA 99l64-2520. Kevin Brown (M. A., Washington State University), Graduate Student Department of Communication, University of Utah, Salt Lake City, UT 84112. Julia Suek (M. A., Washington State University) is a Graduate Student, Counseling Psychology, St. Martins, Longview, WA 98632. The authors thank Frances Ayres for her assistance in preparing this report. This article is not based on a thesis or dissertation.

Abstract: This study examined the impact of communication apprehension (CA), gender, and time on participants' perceptions and behavior in extended initial interactions. Perceptual and behavioral data were gathered after the first 5 minutes and after 20 minutes of these initial interactions. The results indicate that time was a relatively unimportant discriminator in this study. Gender and CA, singly and in combination, produced a wide variety of effects. In general, females appeared to display more behavioral involvement than males. High CAs' perceptions and behavior differed from low CAs' in a variety of ways. For example, high CAs were less attracted to their interaction partners and used different eye contact patterns than did low CAs. Although the interaction effects obtained here are complex, it appears that high CA people in opposite sex dyads are less attracted to their interaction partner and less satisfied with the interaction than low CA people in same or opposite sex interactions. High CA people in opposite sex dyads also disclosed more, talked less, and displayed less behavioral involvement than did those in other circumstances. These and other results are discussed at the conclusion of this report.

A prominent line of inquiry in communication research has been to identify factors that disrupt the communication process. One disruptive factor, communication apprehension (CA), has been found to have a negative impact on the communication process in a variety of contexts (Daly & Stafford, 1984).

Of concern here is the role communication apprehension plays in the formation of interpersonal relationships. A wide variety of studies indicate that CA has an important influence on perceptions of relationship development (Hurt & Preiss, 1978). These studies indicate that high CAs are perceived as less interpersonally attractive (McCroskey & Richmond, 1976; McCroskey, Richmond, Daly, & Cox, 1975), less immediate, less dominant, and less intimate (Burgoon & Koper, 1984), and less satisfying to interact with (Rubin & Rubin, 1989) than are low CAs. Further, high CAs perceive themselves to be more uncertain (Douglas, 1991), to disclose less and more negatively, to be less honest and more superficial (McCroskey & Richmond, 1977; Stacks & Stone, 1983), to interact less and have fewer dates, and to establish fewer close relationships in general (McCroskey & Sheahan, 1978) than do low CAs.

Most of these studies have provided respondents with descriptions of potential interaction partners, rather than having respondents engage in actual exchange with one another. Contrary to this trend, Richmond (1978) examined interaction in a naturalistic setting. Working in the context of a normal classroom, Richmond (1978) paired the same students for five different classroom exercises. These students filled out trait CA scales before being paired, then completed state CA and interpersonal perception scales after the first and fifth exercises. Scales filled out after the first exercise were interpreted to reflect early acquaintance, and those filled out after the fifth exercise were interpreted as indicative of later acquaintance. Richmond found very few perceptual effects that could be attributed to CA during initial interactions, though CA accounted for considerable variance at later stages of acquaintance.

Burgoon and Koper (1984), in contrast to Richmond, found that a number of perceptual and nonverbal variables were related to communication reticence. In the first of two studies, pairs of strangers and pairs of friends interacted for nine minutes. Half of these study participants were confederates who employed immediate, non-immediate , or normal interaction patterns. In the second study, participants engaged in interviews with male or female confederates, in which each confederate violated normal interaction protocol by asking highly personal questions. In general, more reticent participants "nodded less, showed less facial pleasantness and animation, displayed more anxiety and tension, leaned away more, and communicated greater disinterest" (p. 601). Burgoon and Koper also reported that strangers perceived reticents less positively than did friends. Burgoon, Pfau, Birk, and Manusov (1987) concluded that the effect of communication reticence on behavior in initial interactions is "neither so glaring nor so deficient that it is readily noticed by and denigrated by others" (p. 127). This conclusion was based on the Burgoon and Koper (1984). research, as well as on two studies contained in Burgoon and colleagues (1987). One of these studies investigated a dyadic advocacy situation and the other a persuasive speaking situation. Only the first study is of relevance here. In that study, students were given descriptions of a political candidate and told to persuade another person to support their candidate. "Reticents showed only four noticeable differences in nonverbal performances" (P. 124). Reticents were less animated, more tense, used fewer adapters, and were less flexible. "Partners' perceptions of perceived relational communication showed little impact due to reticence" (p. 124).

These studies of actual interaction patterns have several notable limitations. The Richmond (1978) study, for example, examined perceptions formed as a result of interaction but did not examine verbal and nonverbal behaviors. On the other hand, the Burgoon and Koper (1984) and Burgoon, et. al. (1987) studies included perceptual verbal, and non-verbal behavior, but manipulated variables in ways that depart quite markedly from normal interaction protocol.

Ayres (1989) attempted to fill this gap by examining interactions between high CA male and female partners in initial encounters. He found high CA males perceived their female interaction partners to be less physically attractive, less trustworthy, and less satisfying to interact with than did low CA males. There were no statistically significant differences for females' perceptions of the high and low CA males. High CA males talked less, displayed fewer disfluencies, and used less head nodding than did low CA males. Although Ayres' study takes a step in the direction of helping us understand the nature of high CAs' behavior and perceptions in initial interactions, his work was restricted to an examination of opposite sex dyads in which high CA or low CA males were paired with females whose CA level was unknown. Additional research needs to be undertaken to uncover CA's role in same sex interactions, as well as to determine how the CA level of one's interaction partner influences the behavior and perceptions of high CA men and women. The fact that Ayres found large perceptual differences but few behavioral differences may be related to the time the interactants were allowed to talk (5 minutes). It may well be that behavioral differences associated with CA do not emerge during the first few minutes of an interaction due to the ritualized nature (Goffman, 1963) of such interactions.

This study was designed to investigate some of these issues by examining interactions that were 20 minutes in length between dyads composed of same and opposite genders who were both high CAs, both low CAs, or one of whom was a high CA and the other a low CA.

Based on the preceding literature, it was hypothesized that high CA regardless of gender or time would be associated with lower levels of trust, satisfaction, and attraction than low CA. Further, high CAs were expected to talk less but disclose more than low CAs. High CAs were also expected to display less behavioral involvement than low CAs.

In accord with related literature, women were expected to display more behavioral involvement than men. It was also expected that high CAs would find interacting with opposite sex partners more disruptive than interacting with same sex partners vis a vis perceptual and behavioral indicators employed in this study. These effects were expected to be more pronounced at the end of 20 minutes than at the end of 5 minutes due to the less scripted nature of extended interaction. Engaging in less scripted interaction should exacerbate high CAs' proven difficulties in initial encounters.

Method

Participants

A total of 2,521 undergraduate students, enrolled in basic communication courses at a medium-sized western university, filled out the PRCA-24 at the beginning of each of four semesters. Participation was voluntary, but students received class credit for participation. Approximately three weeks into each semester, participants were contacted by telephone until 640 males and females scoring one or more standard deviations above and below the mean on the interpersonal subscale of the PRCA-24 agreed to take part in the study. These participants were assigned to conditions according to gender and CA. Thus 40 participants were examined under each of the following conditions: two high CA males, two high CA females, two low CA males, two low CA females, high CA male with a low CA female, low CA female with high CA male, low CA female with a high CA female, high CA male with a high CA female, high CA male with a low CA male, low CA male with a high CA male, low CA male with high CA female, high CA female with a high CA male, high CA female with a low CA male, high CA female with a low CA female, a low CA male with a low CA female and low CA female with a low CA male. Subjects were the first person listed in each dyad. For instance, in the high CA male low CA female cell the subject of interest was the high CA male. Thus, although there were 640 participants, data were gathered on the 320 subjects of interest. In cases where participants failed to appear as scheduled, additional participants were drawn from the pool allowing an equal number of participants to be maintained in each condition.

Data Collection

Participants were paired according to the conditions outlined above with a partner they did not know for a 20 minute initial interaction. Data were collected at the end of 5 minutes and again at the end of 20 minutes. At each of these points, Ss completed Wheeless & Grotz's (1977) trust scale, Hecht's (1978) satisfaction scale, and McCroskey and McCain's (1974) attraction scale (these instruments are discussed in more detail later). Participants were seated at normal conversational distance in comfortable chairs separated by a small table. A single, stationary camera was placed to obtain a clear view of the upper torso and face of the interactants. At the end of the interaction, participants completed Booth-Butterfield and Gould's (1986) state CA instrument.

Data Processing

Coding. The videotapes were processed by 40 undergraduate students enrolled for research credit. Coders were blind as to the purpose of the investigation. They had been trained to code the behaviors of interest at the .80 level of interrater reliability. In the case of nominal data (e.g., head nodding), intercoder reliability was estimated using Scott's (1955) formula. In the case of interval data (e.g., self-disclosure ratings), intercoder reliability was estimated using Pearson product-moment correlations.

Coders, trained for approximately 15 hours over the course of several days by the authors, were given detailed definitions and examples of target behaviors and then asked to code several training tapes (see Table 1 for relevant definitions). Coders' categorization of behavior on a training tape was used to determine if the .80 level of agreement had been reached. Training tapes were similar in nature to the tapes gathered for this investigation.

When training was completed, each videotape was assigned to two coders. Reliability coefficients for verbal behaviors were .94 for talk time, .83 for self-disclosure, and .88 for information seeking. Reliability values for nonverbal behaviors were .81 for back-channel, .92 for disfluencies, .86 for eye contact, .83 for head nodding, and .94 for lean.

These reliability estimates were based on coders' records of each turn. That is, coders were provided with forms that listed turns down on the side of the form and behaviors of interest across the top of the form. Interval data reliability estimates were developed by correlating the value the first coder entered for that item with the value the second coder entered for that item vis a vis each turn. Reliability estimates for the nominal variables were developed by examining whether coders agreed that a behavior of interest (e.g., head nodding) had occurred during the same turn and then applying Scott's formula (Scott, 1955). Coder's scores were averaged to arrive at a composite score used in the statistical analyses performed on these data.

Operational Definitions of Independent Variables

Communication Apprehension. Males and females scoring one standard deviation above and below the mean on the interpersonal subscale of the PRCA-24 (McCroskey, 1982) constituted the sample for this study. The obtained reliability of this subscale was .87. Since the PRCA-24 measures subjects' general disposition to experience anxiety, it cannot be relied on to assess an interactant's level of CA accurately in a particular setting. To assess whether participants experienced high or low CA in this setting, participants completed Booth-Butterfield and Gould's (1986) state anxiety measure immediately after these initial interactions. State anxiety scores for participants who placed in the high CA category were then compared with scores for those

Table 1

Operational Definitions of Verbal and Nonverbal Behaviors Analyzed in This Study

Verbal

ïTalk Time. Proportion of talk time was derived by taking the total number of seconds a participant talked, as determined by using a stop watch, and dividing that number by the total length of the conversation (i.e., 5 minutes, since all conversations were stopped after exactly 5 minutes had elapsed). Talk time began when the participant started talking and ended when she or he stopped. Pauses between turns were not added into the participant's talk time.

ïSelf-disclosure. Coders rated participants' disclosure in thirteen content areas (e.g., Taylor & Altman, 1987) on 1 to 7 scales, ranging from "Revealed No Personal Information About This Topic" to Revealed Extremely Personal Information About This Topic." Ratings were averaged across the thirteen topics to arrive at an overall disclosure score.

ïInformation-seeking. Information seeking was operationalized by counting the number of questions participants directed to their partners.

Nonverbal

ïBack-channel. The number of back-channel responses was arrived at by counting how many brief vocal comments participants uttered that acknowledged the partner's comment and served to confirm the partner's right to continue talking (e.g., I know, yeah, really).

ïDisfluencies. Hesitations, prolongations, repetitions, and errors (as revealed by corrections by the person of pronunciation or content) were totaled for each participant.

ïEye contact. Operationalized as the proportion of time a participant looked directly at the eye area of her or his partner (i.e., time looking divided by the total time one could have looked).

Operational definition of Verbal and Nonverbal Behaviors Analyzed in This Study

ïHead nods. Head nods were counted while a participant was listening. Nods were defined as discernible up and down movements of the head. A series of rapid nods that functioned as a nod was treated as a single nod.

ïLean. Lean was defined by reference to the participant's upper torso in relationship to the chair and table. A person whose back was firmly against the chair was scored as 1 (i.e., little or no forward lean) and those almost horizontal with the table were scored as 6 (i.e., leaning forward as much as possible). Scores of 2-5 were intermediate degrees of forward lean (about 15 degrees each). Lean was scored at the end of the 1st, 2nd, 3rd, and 4th minutes and averaged to derive a lean value.

who placed in the low CA category. The obtained t (478) test value was significant at well beyond the .01 level of significance (High CA = 57.2, SD = 6.3; Low CA - 17.1, SD = 5.2; omega squared = .87), indicating that participants in the low CA condition experienced much lower levels of state apprehension than participants in the high CA condition.

Gender. Gender was determined by reference to the category the subject checked when she or he filled out the PRCA-24.

Time. Time was operationalized as the first 5 minutes of a 20 minute interaction and the last 5 minutes of the same 20 minute interaction. Time began when the experimenter turned on the VCR and left the room and ended when she or he re-entered and shut off the VCR. The experimenter turned on the VCR, left, returned in precisely five minutes, had Ss fill out forms, restarted the VCR, left, returned in precisely 15 minutes, shut off the VCR, had Ss fill out forms, and then debriefed participants.

Table 2

Means and Standard Deviation Values for Physical Attraction, Task Attraction, and Information Seeking on Gender and CA

Physical AttractionTask Attraction Info. Seeking

NMSDM SDMSD

HMHM**2019.1a*05.0508.9a3.615.4c3.12

HMHF2024.5b06.0112.6b4.384.7b2.20

HFHM2024.9b07.4711.9b5.516.4d2.31

HFHF2024.3b07.9412.1b5.785.2bc2.38

HMLM2019.9a06.8812.9a6.875.7cd3.31

HMLF2022.1b06.3307.2a5.495.2bc3.09

HFLM2022.0b06.6510.6b4.244.4b3.45

HFLF2023.5b10.1210.9b5.057.2d4.48

LMHM2019.4a05.6809.6a4.355.2bc2.19

LMHF2022.0b07.0709.8a3.375.3bc3.79

LFHM2022.1b05.9210.9b4.393.8a2.30

LFHF2023.2b05.1512.5b4.524.2ab3.02

LMLM2019.2a05.0613.8c3.883.3a2.28

LMLF2023.5b08.7812.3b4.514.1a2.28

LFLM2023.2b08.9209.5ab5.076.3d3.37

LFLF2022.6b09.5413.3bc6.233.9a2.16

Note: The higher the score the higher the attraction.

H = high communication apprehensionL = low communication apprehensionM = maleF = female

In each case the data refers to the target person's behavior. Thus, in a HMHM cell the target is a high CA male who talked with another high CA male.

Means with the same superscript are not significantly different from one another.

Operational Definitions of Dependent Variables

Participants' perceptions of their interaction partners, as recorded on the trust (Wheeless & Grotz, 1977), satisfaction (Hecht, 1978), and attraction (McCroskey & McCain, 1974) scales, served as indicators of those variables in this study. McCroskey and McCain's attraction scale contains separate factors to tap social, task, and physical attraction, with reliabilities ranging from .83 to .91 in this application. Hecht's satisfaction scale is unidimensional; reliabilities were .88 at time one and .90 at time two in this study. For Wheeless and Grotz's unidimensional trust scale, reliabilities were .87 at time one and .88 at time two in this investigation.

Indicators of the participants' verbal and nonverbal behavior were derived by analyzing the videotapes. These behaviors are listed and defined in Table 1.

Analyses

Data were analyzed using 2 x 4 x 4 ANOVAS with repeated measures on the time factor. The three factors were time (first 5 minutes, last 5 minutes), gender (MM, FF, MF, FM), and CA (HH, LL, LH, HL). In each case the target is the first person. Thus, in a HMLF cell, the target was the high CA male who talked with a low CA female. Bonferroni's procedure was used to avoid alpha inflation. Significant effects were pursued with Duncan's Multiple Range Test.

Results

Three Way Interactions

No statistically significant time by gender by CA effects emerged in this study. Power to detect small, medium, and large effects was .14, .73, and .99 respectively.

Two Way Interactions

Several significant two way interaction effects emerged in this study. The first of these concerned a gender by CA interaction effect with regard to task attraction [F(9,288) = 2.65, p <.01, omega squared = .06]. As can be seen in Table 2, high CA males interacting with low CA females

Table 3

Means and Standard Deviation Values for Self-disclosure on Time and CA

1st 5 minutes2nd 5 minutes

NMSDMSD

HCA1601.8a*.643.4c.82

LCA1601.7a.592.2b.55

*Means with the same superscript are not significantly different from one another.

reported the lowest degree of attraction (high CA males interacting with high CA males reported the next lowest level of task attraction) while low CA males interacting with low CA males reported the highest degree of attraction (low CA females interacting with low CA females reported the second highest level of attraction).

Another two way interaction effect emerged for time and CA [F(9,288) = 4.22, p <.001, omega squared = .03]. As can be seen in Table 3, high CAs disclosed more during the last 5 minutes of interaction than did those in other circumstances. The ANOVA applied to the information seeking data yielded a significant interaction between gender and CA [F(9,288) = .484, p <.001, omega squared = .11]. As Table 2 reveals, low CA males talking to low CA males asked the fewest number of questions while high CA females talking to low CA females asked the largest number of questions.

There was a significant gender by CA effect with regard to disfluency [F(9,288) = 3.30, p <.001, omega squared = .06]. As can be seen by examining Table 4, high CA males interacting with high CA males displayed the highest number of disfluencies while high CA males interacting with Low CA females displayed the lowest number of disfluencies. An interaction effect for gender and CA [F(9,288) = 5.66, p <.001, omega squared = .11] with regard to eye contact also emerged. Low CA females interacting with low CA males looked at their partners (see Table 4 for relevant values) the most while high CA males interacting with high CA females looked at their interaction partners the least. Another interaction effect between gender and CA [F(9,288) = 5.99, p <.001, omega squared = .15] appeared with regard to the use of head nods. As Table 4 reveals, low CA females interacting with low CA males used head nodding the most while high CA males interacting with high CA females displayed the lowest amount of head nodding. Lastly, there was an interaction effect for gender and CA [F(9,288) = 4.20, p <.001, omega squared = .08] for body lean. As can be seen in Table 4, low CA females interacting with low CA females and males leaned toward their partners the most while high CA males or females did not display a great deal of forward lean.

No other two way interactions achieved statistically significant levels. Power for detecting significant two way interactions was .27, .96, and .99 for the time and gender interactions. Comparable power values were .14, .73, and .99 for the interaction between gender and CA.

Main Effects

Time. Only one main effect concerning self-disclosure emerged on the time variable [F(3,288) = 13.82, p <.001 omega squared = .13]. Understandably, self-disclosure was higher after twenty minutes than after five minutes of interaction. The meaning of this effect is somewhat unclear since time also interacted with CA to produce an effect on this variable. The power to detect significant time effects was .39, .99, .99 for small medium and large effects respectively).

Gender. A main effect for gender emerged for physical attraction [F(3,288) = 9.46, p <.001, omega squared = .09]. Not surprisingly, males interacting with males reported them to be less attractive than did those in any other gender combination (M = 19.4, SD = 5.67 and M = 22.9, SD = 7.52 respectively).

A gender effect also emerged on the satisfaction scale [F(3,288) = 2.71, p <.01, omega squared = .04]. Females reported higher levels of satisfaction than did males (M - 57.6, SD = 15.36 and M = 52.2, SD = 15.01 respectively.

Two other gender main effects appeared for head nods [F(3,288) = 5.41, p <.001, omega squared = .05] and body lean [F(3,288) = 6.49, p <.001, omega squared = .04]: but these effects are difficult to interpret since gender was also involved in significant two way interaction effects on these variables.

No other gender main effects were observed. Power to detect such differences was .27, .96, and .99 for small, medium, and large effects respectively.

CA. With regard to talk time, there was a main effect for CA [F(3,288) = 6.15, p <.001, omega squared = .05]. As Table 3 indicates, those with high CA talked less than those with low CA. Two other main effects emerged involving eye contact [F(3,288) = 18.19, p <.001, omega squared = .12] and head nods [F(3,288) = 14.69, p <.001, omega squared = .14]. However, neither of these effects can be interpreted unambiguously because they also appeared in interaction with the gender variable.

No other main effects appeared with regard to CA. Power to detect such differences was .27, .96, and .99 for small medium, and large effects respectively.

Discussion

This study was designed to examine whether perceptions and behavior would differ for various gender and CA combinations during the first 5 minutes and the last 5 minutes of a 20 minute initial encounter. The interaction of these three variables did not produce any significant differences on the dependent variables of interest in this study. Power was more than sufficient to detect large effects at this level but inadequate to detect medium and small effects. Thus, we can be quite confident any large three way interactions would have been detected but much less confident that medium and small effects would have been detected.

However, a large number of two way interaction effects emerged in this study. The interaction between gender and CA produced a variety of perceptual and behavioral differences. One such effect concerned task attraction. In particular, low CA males reported being most willing to work with other low CA males, while high CA males reported being the least willing to work with low CA females. This latter finding suggests that high CA males may find it difficult to engage in task relevant behavior with low CA females. It may be, as is suggested by Ayres' findings (1989), that high CA males may be particularly uncomfortable in this circumstance. Why low CA males reported the highest task attraction with low CA males is another matter. It may be that low CA males perceive low CA males to be more competent than others. Of course, these inferences must be considered speculative until relevant data can be gathered vis a vis these issues. It is a bit surprising that social attraction was not associated with gender and CA. The means are in expected directions. It maybe that the intrusive nature of the videotape equipment, repeated testing, and other factors dampened social attraction in this setting.

There were also a number of effects attributable to CA and gender with respect to verbal and nonverbal behavior. With regard to information seeking, low CA males interacting with low CA males or females used this tactic the least while high CA males and females interacting with low CA females used this tactic the most. Reliance on questions to initiate or maintain conversation suggests high CAs may be unaware of or unable to employ more subtle, sophisticated tactics for finding out desired information or changing the topic. This finding may be related to the talk time result (i.e., high CAs talked less than low CAs). It may well be that high CAs ask questions fully aware that this will require them to talk less. High CAs may be quite capable of using other conversational behaviors but choose this one to avoid having to talk.

The nonverbal data also produced a variety of differences. One puzzling result was that high CA males displayed the highest degree of disfluency when interacting with other high CA males but the lowest degree of disfluency when interacting with low CA females. Perhaps low CA females enabled these high CA males to carry on a smoother interaction. Females in general engage in more eye contact, forward lean, and head nodding than males. Perhaps these signs of encouragement assisted high CA males' fluency.

These generally facilitative behaviors on the part of females appears most prominently when low CA females are interacting with low CA males. Low CA females in this circumstance displayed higher levels of head nodding, eye contact, and forward lean than those in most other circumstances; high CA males generally displayed the lowest level of these indicators of behavioral involvement.

Another interesting interaction effect occurred with regard to CA and time. High CA males and females disclosed more during the last five minutes of interaction than did those in any other dyadic combination. This suggests that high CAs tend to disclose too much too soon in these interactions. However, since the perceptual data did not produce large differences, this tendency on the part of high CAs may not be particularly disruptive.

The only unambiguous main effects to emerge in this study concerned gender. Males reported being less physically attracted to other males than did those in any other circumstance. The general societal condemnation of males being physically attracted to males would seem to account for this effect.

The only other significant perceptual effect had to do with satisfaction. Females reported higher levels of satisfaction than did males. The nonverbal data indicates males are less involved in these interactions than females which may lead to inferences of this nature. Or, it could be that men are less able to interact freely in a laboratory setting than are women. Clearly, this is an area in need of further exploration.

Overall, these perceptual, verbal, and nonverbal data paint a reasonably clear picture. It does not seem to matter whether high CAs are interacting early or late in a 20 minute interaction. They encourage their interaction partner to talk by asking questions. They tend to discourage their partners to involve them in the conversation by presenting nonverbal displays of non-involvement. They lean back, they avoid looking at the other person, or nodding their heads. This pattern is most pronounced for high CA males. Given this pattern, it is surprising that more pronounced perceptual effects did not emerge. Nonetheless, it would seem that this non-involved, withdrawn pattern influences the development of relationships and in turn diminishes the high CA's chances of pursuing relationships. It seems obvious that people would prefer to interact with people who share and are involved. The fact that high CA males and females tend to disclose more than others may exacerbate this difficulty. To interact with a withdrawn person who blurts out personal information has to be trying.

Of course, this study has limitations that may have enhanced some of these effects. No data are available on this point, but it seems likely that high CAs do not often engage in uninterrupted 20 minute initial interactions (low CAs may not either). In shorter interactions, the effects obtained here may be less evident. Repeated testing may also have affected these results. A person answering the perceptual questionnaires may have remembered earlier answers and repeated them the second time. Selection procedures (i.e., volunteers) utilized in this study may also have influenced these findings. Zakahi and McCroskey (1989) point out that such procedures may bias samples toward those willing to communicate. It may be that our entire sample is more willing to communicate than the general population, which would tend to distort the data pattern. More sophisticated designs and sampling procedures need to be employed to account for such effects. A clear limitation of these data is that they reflect interaction patterns under laboratory conditions.

However, these data do mesh well with the general literature. High CA males displayed withdrawn, uninvolved behaviors relative to low CAs. These data suggest that skills training that stressed nonverbal displays to encourage involvement along with an understanding of appropriate verbal exchange might be particularly helpful to high CA males. A similar pattern exists for high CA females but tends to be more restricted. These effects are most pronounced when these women are interacting with low CA males. Intervention work for them may be most valuable if it is targeted to interactions involving males. Overall, this study has helped refine our knowledge of the impact of CA on initial interaction by taking into consideration the gender and CA level of one's interaction partner.

References

Ayres, J. (1989). The impact of communication apprehension and interaction structure on initial interactions. Communication Monographs, 56, 75-87.

Booth-Butterfield, S., & Gould M. (1986). The communication anxiety inventory and the validity of state and context communication apprehension. Communication Quarterly, 34, 294-205.

Burgoon, J. K., & Koper, R. J. (1984). Nonverbal and relational communication associated with reticence. Human Communication Research, 10, 601-626.

Burgoon, J., Pfau, M., Birk, T., & Manusov, V. (1987). Nonverbal communication performance and perceptions associated with reticence: Replications and classroom implications. Communication Education, 36, 119-130.

Daly, J. A., & Stafford, L. (1984). Correlates and consequences of social-communication anxiety. In J. A. Daly and J. C. McCroskey (Eds.), Avoiding communication: Shyness, reticence, and communication apprehension (pp. 125-143). Beverly Hills: Sage.

Douglas, W. (1991). Expectations about initial interaction: An examination of the effects of global uncertainty. Human Communication Research, 17, 355-384.

Duncan, S. Jr., & Fiske, D. W. (1977). Face to face interaction: Research, methods, and theory. New York: Wiley.

Goffman, E. (1963). Behavior in public places. New York: Free Press.

Hecht, M. L. (1978). The conceptualization and measurement of interpersonal communication satisfaction. Human Communication Research, 4, 253-264.

Hurt, H. T., & Preiss, R. (1978). Silence isn't necessarily golden: Communication apprehension, desired social choice, and academic success among middle-school students. Human Communication Research, 4, 315-328.

McCroskey, J. C., & McCain, T. A. (1974). The measurement of interpersonal attraction. Speech Monographs, 41, 261-266.

McCroskey, J. C., & Richmond, V. P. (1976). The effects of communication apprehension on the perception of peers. Western Speech Communication, 40, 14-21.

McCroskey, J. C., & Richmond, V. P. (1977). Communication apprehension as a predictor of self-disclosure. Communication Quarterly, 25, 40-43.

McCroskey, J. C., Richmond, V. P. , Daly, J. A., & Cox, B. G. (1975). The effects of communication apprehension on interpersonal attraction. Human Communication Research, 2, 51-65.

McCroskey, J. C., & Sheahan, M. E. (1978). Communication apprehension, social preference, and social behavior in a college environment. Communication Quarterly, 26, 41-45.

Richmond, V. P. (1978). The relationship between trait and state communication apprehension and interpersonal perceptions during acquaintance stages. Human Communication Research, 4, 338-349.

Rubin, R. B., & Rubin, A. M. (1989). Communication apprehension and satisfaction in interpersonal relationships. Communication Research Reports, 6, 13-20.

Scott, W. A. (1955). Reliability of content analysis: The case of nominal scale coding. Public Opinion Quarterly, 19, 321-325.

Stacks, D., & Stone, J. D. (1983). The effect of self-concept, self-disclosure, and type of basic speech course on communication apprehension. Communication, 12, 105-127.

Taylor, D. A., Altman, I. (1987). Communication in interpersonal relationships: Social penetration processes. In M. E. Roloff & G. R. Miller (Eds.), Interpersonal processes: New directions in communication research (pp. 257-277). Newbury Park, CA: Sage.

Wheeless, L. R., & Grotz, J. (1977). The measurement of trust and its relationship to self-disclosure. Human Communication Research, 3, 250-257.

Zakahi, W. R., & McCroskey, J. C. (1989). Willingness to communicate: A potential confounding variable in communication research. Communication Reports, 2, 96-104.

Response from Leonard J. Shedletsky

Although a 20 minute conversation may not be an "extended interaction," it may well account for some changes that we experience as we carry on an initial interaction. But what is particularly puzzling is the hypothesis that high CAs would disclose more than low CAs, in spite of the high CAs self perception to disclose less. Why would someone who is relatively withdrawn and uninvolved disclose more than someone who is relatively not withdrawn, or involved?

The Effects of Planned and Spontaneous Delivery on Familiar and Unfamiliar Lies: When Does It Pay to Plan ?

Patrick deBattista

About the Author: Patrick deBattista is on the faculty of the Department of Speech, University of Hawaii at Manoa, George Hall 321, 2560 Campus Road, Honolulu, Hawaii 96822. The study reported was part of a pretest for his dissertation and was presented at the annual conference of the Speech Communication Association, 1992, Chicago.

Abstract: Previous research has focused on how planning affects decoding of deception. This study examines what scripts and plans mean for encoding deception. It is reasoned that planning will not benefit the performance of white lies because, as routines, they are deployed mindlessly, but will benefit the performance of lies breaching another's trust. Results suggest that (a) women may be better at dissembling than men, (b) that women's advantage may reside in nonverbal expressivity, (c) that males' deceptive efforts benefit from planning regardless of whether the lie is familiar or unfamiliar, and (d) that planning enables lie's verbal component.

Perspective:

1. Do people have routines for performing certain common social transactions?

2. What would such routines mean for people's willingness to think about social transactions which, because of their repeated enactment, are familiar?

3. In social transactions that are less familiar, is planning beneficial to the interacts?

Since turning their attention to cognition's role in human interaction (Planalp, 1985; Planalp & Hewes, 1982; Roloff & Berger, 1982), communication scientists have examined cognition's role in action production (Greene, 1984; Greene & Lindsey, 1989), action comprehension (Schank & Abelson, 1977) and interpersonal influence (Berger, Karol & Jordan, 1989; O'Keefe & Shepard, 1987); yet, the domain of deceptive communication remains largely unexamined (Knapp & Comadena, 1979). This article considers the role cognitive plans, conceptualized as both procedural rules and specified routines, play in deceptive communication.

Routinization

Repeated enactment leads to routinization. Persons enact routines to satisfy situational-goal-states as disparate as awakening in the morning and preparing for work (von Cranach, Kalbermatten, Indermuhle & Gugler, 1982), getting the newspaper (Wilensky, 1981) or performing surgery (Hammond, 1989). While routinized, these courses of action may vary in the degree they are specified, deployed mindfully, and intended for modification. For example, routines for awakening in the morning (Bower, Black & Turner, 1979) or getting the newspaper may be enacted invariably and with little thought, unless, of course, one is running late and/or it is raining outside (Wilensky, 1981). In contrast, while a doctor may have a general routine for surgery, the routine may exist as a general principle or rule: monitor the patient's blood pressure while administering anesthetic (Hammond, 1989). The doctor must be mindful of the rule's enactment and must be prepared to modify actions taken: cease anesthetizing patient if blood pressure drops (Hammond, 1989). Thus, routines can be scripted actions (Schank & Abelson, 1977; Scholnick & Friedman, 1987), or more flexible plans (Berger, in press), which while varying in their degree of specification (DeLisi, 1987), nonetheless, can be ready-made or "canned" (Wilensky, 1983). In other words, a routine may be a way to approach an event that is potentially new each time, or it may be a specific response to a situation's sameness (Hammond, 1989; Wilensky, 1983). Scripts can account for routines of the latter sort. Plans can account for both the former and the latter (Berger, in press). The conceptualization of plans as both general structures monitored for anticipated modification and as specified routines deployed mindlessly has relevance for deceptive speech acts, or lies.

Routine White Lies

As speech acts, white lies are a routine part of our communicative repertoire. Their social acceptability and people's willingness to admit telling them provides evidence of their frequent deployment as means to avoid hurting others' feelings (Camden, Motely & Wilson, 1984). There are reports (Camden et al., 1984) that on the average, persons tell six white lies per week. Persons report telling lies to avoid hurting the feelings of their friends, romantic partners or spouses (Lippard, 1988; Metts, 1989). In addition, they report both "feigned willingness" (Lippard, 1988, p. 99) and dishonest excuses as means for achieving goals of tact and kindness in response to friendly requests to which they would rather not comply. In Lippard's (1988) categorization of lie types, giving the false impression of wanting to comply with a friend's request and giving false reasons for noncompliance to the same requests account for the largest number of deceptive acts in the sample.

Willingness to encourage false impressions resides in-part in concern for others' feelings. In Lippard's (1988) study, the desire to avoid hurting another's feelings was the second most frequently cited motive for lying. In addition, for Lippard's sample, lies told to avoid hurting another's feelings ranked third in prevalence, being closely behind lies told to parents and being more frequent than lies told to powerful others. As would be expected, deployment of lies in concern for others' feelings is more pronounced in intimate relationships. Research suggests that persons are willing to lie to help those to whom they perceive a relational debt. Metts' (1989) study of deceit in close relationships reveals that across relational categories, the most frequent reason for lying is to avoid hurting a partner's feelings. In particular, women, who may be socialized to have greater concern for others' feelings (Lippard, 1988), are significantly more likely than men to tell white lies (Camden et al., 1984; Lippard, 1988). In addition, research suggests that women are better at communicating feigned pleasure than their male counterparts, at least when being judged by older decoders (Parham, Feldman, Oster & Popoola, 1981). Finally, women may be more highly motivated than men not to make bad impressions (DePaulo, Stone & Lassiter, 1985). Women's intent to ingratiate themselves to another is more obvious in visual and audiovisual channels, as opposed to audio-only channels, while for men this is not the case (DePaulo et al., 1985), suggesting that women's attempts to ingratiate are accompanied by greater nonverbal expressiveness than the ingratiation attempts of men. In sum, telling lies in concern for others' feelings is routine communicative behavior.

Infrequent Lies as Unfamiliar

Knapp and Comadena (1979) argue that persons learn about lying by coming to understand the costs of enacting certain lies versus the rewards enactment brings. In general society punishes self-interested acts that hurt others. Lies rooted in self interest and which hurt others are negatively sanctioned. Negative conditioning which pairs the telling of such lies with punishment, potentially makes later performance of such lies enormously taxing because of the guilt, fear and trauma that their telling would arouse (Zuckerman, DePaulo & Rosenthal, 1981). Given this negative conditioning, it is likely that persons avoid telling these lies as a way to avoid negative affect associated with them. Telling such lies would be an unfamiliar experience, in contrast to the positively sanctioned telling of white lies.

In a number of studies subjects have told lies that are either unfamiliar and/or are told under unfamiliar conditions, such as lying when signaled to do so during an interview, (deTurck & Miller, 1990; Kraut, 1978; Miller, deTurck, & Kalbfleisch, 1983; McClintock & Hunt, 1975; O'Hair, Cody & McLaughlin, 1981; Streeter, Krauss, Geller, Olson & Apple, 1977) lying about important issues (Kraut, 1978; McClintock & Hunt, 1975; O'Hair et al., 1981; Streeter et al., 1977), and lying about the unpleasant content of a picture (deTurck & Miller, 1990; Miller et al., 1983), or about topics, such as favorite television programs, vacation plans and hobbies (Littlepage & Pineault, 1984) that because of their mundane nature would, under most circumstances, be associated with truthtelling. In such instances, behaviors that signal arousal, either through explicit nonverbal behavior or through lessened ability to evade detection, have accompanied the deceptive act. These include more self manipulations (McClintock & Hunt, 1975; O'Hair et al., 1981) increased pitch (Streeter et al., 1977), shorter, less plausible answers (Kraut, 1978) and more pauses and disfluencies (Miller et al., 1983).

The unfamiliarity or familiarity of a lie should have ramifications for the amount of effort needed to enact it. Situations infrequently encountered should require formulation of a lie. In contrast, for the familiar exigence of being tactful, ready-made, white lies should already be available. The ability to deploy a routine as opposed to the need to plan should influence a lie's performance.

Routines and Performance

While routinization leading persons to overlook relevant situational features can be detrimental (Grasesser, Woll, Kowalski & Smith, 1980; Langer, 1985), a match between situational demands and available routines enhances performances. As the product of learning, routines are premised on an understanding of task part's interrelatedness (e.g., Pea & Hawkins, 1987) and enactment as a sequence (e.g., Greene, 1984, experiment 2). Experts' advantage resides in schematized knowledge that allows the processing of routine task parts as "chunks" (Fiske, Kinder & Larter, 1983). For example, relating what they hear to the game's goal structure allows persons with expert knowledge of baseball to have better, more descriptive recall of a fictitious baseball game (Spilich, Vesonder, Chiesi & Voss, 1979; Chiesi, Spilich & Voss, 1979). Similarly, expert chess players' ability to formulate complex strategic moves stems in part from their routinization of the game's opening moves, with such knowledge being procedural (Anderson, 1974). Even, "expert" date-getters, persons with greater dating experience, display greater ability to organize events that occur in a standard date (Pryor & Merluzzi, 1985), assumedly because of their highly schematized knowledge of the date-getting sequence. In those instances where persons do not have fully specified, ready-made courses of action, they may find it necessary to plan in order to accomplish their goal.

Familiarity, Planning and Performance

Examinations of the relationship between familiarity and planning have produced contradictory findings. While Langer and Weinman (1981) found that planning increased the fluency of truthful statements about unfamiliar topics, Greene (Greene, Smith, Smith & Cashion, 1987) found that planning increased the fluency of truthful statements about familiar topics. Given that Langer and Weinman (1981) failed to assess whether or not their manipulations resulted in the desired conditions, it is uncertain whether their findings are attributable to planning's interaction with familiarity. Both studies conceptualized familiarity as how commonplace a topic was, rather than how frequently a speech act was deployed in response to a topic or situational exigence. Neither study examined familiarity and planning as they relate to deception. While these two studies produced contradictory findings, planning has been found to benefit performance of lies told in response to unfamiliar exigencies.

Plans and Performance

While overplanning may undermine performance (Berger et al., 1989), planning theorists note the benefit planning affords as a means to anticipate, monitor and adjust to the environment (Cambis, 1979; Faludi, 1973; Holloway, 1986; Sackman & Citerbaum, 1972). While access to numerous action alternatives may make for disfluent performance in shorter interactions, in longer interactions, plans incorporating more alternative routes of action may be to interactants' advantage (Berger, 1988; Berger et al., 1989). Indeed, planning for a work-related task correlates positively with ratings of performance (Earley, Wojnaroski & Prest, 1987). If planning enables the navigation of unfamiliar situations, then planning should benefit persons telling lies that are unfamiliar.

Research suggests that deceivers benefit from planning. Judges viewing videotapes of the lying and truthtelling of subjectsógiven until the next day to plan a truthful answer and plausible lie about their favorite television show or vacation spotóare less accurate in detecting planned lies than they are in detecting those not planned (Littlepage & Pineault, 1984). Ten minutes to plan a lie about one's favorite vacation spot seems to enable persons' inhibition of deception cues stemming from body movement (Greene, O'Hair, Cody & Yen, 1985). Assumedly because of the readiness and confidence planning affords, students' who take ten minute to plan lies about their grade point averages display shorter response latencies, more affirmative head nods, less smiling and make shorter statements than students' making truthful statements do (O'Hair et al., 1981). Finally, when persons lie about the content of pictures showing either landscapes or burn victims, high self monitors who plan lies are better at evading detection than low self monitors who have not rehearsed and those who have rehearsed (Miller et al., 1983). In contrast to these studies, DePaulo, Lanier and Davis (1983) find no effect for planning versus spontaneity on success in evading detection; yet, DePaulo (DePaulo et al., 1983) likens her study's questioning format to a conversation shifting unexpectedly to a topic one wishes to avoid and suggests more extensive planning may be needed for one to successfully navigate such situations. If planning benefits deception, it is worthwhile to ask how deceivers use planning.

Channel Effects

Research suggests that persons use planning to control nonverbal expressiveness rather than verbal communication (Littlepage, Tang & Pineault, 1986). Planned lies are more difficult to detect in visual and auditory formats than spontaneous lies are, but this effect does not hold for written transcripts. In conjunction with findings regarding the controllability of various nonverbal channels (DePaulo, Zuckerman & Rosethal, 1980; Ekman & Friesen, 1969, 1974; Zuckerman et al., 1981), the findings of Littlepage (Littlepage et al., 1986) suggest that persons plan the expressivity of those channels of which they are most aware, the face and the voice (Bond, Kahler & Paolicelli, 1985; Ekman, 1985; Kraut, 1978; Littlepage & Pineault, 1978, 1979; Zuckerman, Kernis, Driver & Koestner, 1984; Zuckerman, Larrance, Spiegel & Kloman, 1981). Moreover, since persons exercise little control over the voice as a channel (DePaulo, Lassiter & Stone, 1982; DePaulo et al., 1980; Zuckerman, DePaulo, Rosenthal, 1981; Zuckerman, Larrance, Spiegel, Klorman, 1981), facial expressivity is probably the larger recipient of planning.

While research has considered lies as spontaneous and planned, a number of relationships remain unexamined, specifically (a) deceiver's sex and ability to tell white lies, (b) the relationship between lie type, preparation type, sex and channel and (c) the relationship between planning and the familiarity or unfamiliarity of the lie, specifically whether certain lies are scripted.

Hypotheses

If white lies are routine, or ready-made, in contrast to other lies that must be created in response to an unfamiliar situation then

Hypothesis 1:. Lies that are planned and told in response to an unfamiliar situation will be perceived as more truthful than such lies delivered spontaneously; in contrast, the difference between the perceived truthfulness of lies that are planned and told in response to a familiar situation and such lies delivered spontaneously will be attenuated.

In addition, if women have more experience in telling white lies than men do, this experience should be manifested in their performance of such lies; consequently

Hypothesis 2:. Women telling white lies will be perceived as more truthful than men telling white lies.

Finally, if planning effort is directed toward nonverbal expressiveness to a greater degree than verbal content,

Hypothesis 3:. Deception communicated through an audiovisual channel will be easier to detect than deception communicated through a video-only channel.

Method

Creation of Stimulus Tapes

Subjects were undergraduates at a large midwestern university who received extra credit for their participation. Subjects brought a friend, ostensibly to be their partner in the experiment. Four males and four females were assigned to spontaneously deliver a lie in response to a familiar exigence and an unfamiliar exigence, and four males and four females were assigned to deliver a planned lie in response to a familiar exigence and an unfamiliar exigence. Depending on assigned condition, subjects received one of the following sets of written instructions (pronoun references matched sex of the confederate).

Instruction Set A

This is an experiment about the influence of cooperativeness. In a minute you and your friend will play a game. You and your friend will each be on a different team. Each of you will have a partner whose friend is also on the other team; we will introduce you to your partner shortly. In this particular game, the teams can cooperate and arrive at a mutually beneficial outcome, or they can compete, so that one team wins and the other team loses. We want you to be as cooperative with the other team as possible during the course of the game. We want to see what effect this has on the behavior of your teammate. We will be bringing your teammate to meet you in just a minute, and then we will take the two of you to the room where you and the other team, on which your friend is a member, will play the game. Since we will be watching for how your cooperativeness influences your teammate's behavior during the course of the game, it is essential that she or he does not know what we have told you.

Instruction Set B was the same as above but had the additional instructions:

Chances are good that since we told your teammate just to be her/himself during the course of the game, that s/he will ask you if we gave you special instructions. Also your teammate will be wearing this headset that has a microphone attached. S/he may be uncertain how s/he looks and may ask you for feedback. We need for her/him to wear the headset during the game so that we can tape her/his remarks.

Instruction set B was meant to cue subjects to plan lies both about the experiment (an unfamiliar exigence) and about the teammate's appearance (a familiar exigence). Instruction set A was meant to provide no such cue.

After subjects read the instructions, the experimenter took them to another room to meet the teammate. The room they were taken to, in actuality, was an audiovisual equipment storeroom. When the subject entered the room, the experimenter introduced the confederate, who was waiting there, as the teammate. Confederates' sex was counterbalanced across subjects' sex. The experimenter asked subjects to sit in the seat indicated. The experimenter apologized for putting the two in what was essentially a storage-area but claimed that the number of subjects present had resulted in a room shortage.

Of the cameras stored in the room, four were pointing in the subject's general direction. Of these, one recorded the seated subject from the waist up. The experimenter asked the two "teammates" to get acquainted and then left and closed the door.

The confederate subjects encountered wore a white vinyl headband with what looked like the bottom of a phone receiver attached to it by a curved metal bar; trailing from the headband was a thick brown cable. The awkward mouth piece, broad vinyl headband and trailing cord were meant to appear unsightly, cumbersome and somewhat ridiculous.

After the experimenter left, confederates probed subjects about what they had been told:

I heard that we were going to play a game with your friend and my friend. They told me to just be myself. Did they tell you anything special?

Confederates also probed subjects' thoughts regarding the confederate's appearance:

They said I have to wear this thing [referring to the headset] so that they can record what I say. My friend is going to see me wearing this, and your friend, someone I don't even know. Does it look stupid?

Order of probes was counterbalanced across subjects. In response to probes, all actors masked what they had been told regarding the experiment's purpose, and all actors expressed sentiments that the confederate did not look odd.

After subjects responded to the second probe, the experimenter reentered the room. He asked subjects to tell the truth about the instructions they had received and to state their true feelings about the confederate's appearance. Then subjects completed two manipulation checks. These manipulation checks revealed that no subject suspected the interaction's real purpose; moreover, all subjects, except for one male and one female, felt the confederate's appearance looked "stupid" to some degree; nonetheless, these two subjects, when asked earlier to voice their true feelings, expressed sentiments that the headset looked odd.

After subjects completed the manipulation check, the experimenter explained the interaction's true purpose and told subjects they had been videotaped. The experimenter asked if the videotape could be used for a later experiment and told subjects that the tape would be erased if they so chose, and that choosing to do so would not affect their extra credit. All subjects agreed to allow the tape to be used.

The above manipulations provided videotapes of persons who either planned and lied (cued for the probes) or spontaneously lied (not cued for the probes) in response to a familiar exigence (inquiry about personal appearance) and in response to an unfamiliar exigence (breaching the trust of a partner) and corresponding instances of truthtelling.

Instances of lying and truthtelling were randomized so that for any given person no instance of their lying or truthtelling directly followed a previous instance. The order of this stimulus tape was again randomized, maintaining the condition that no instance of any given person's behavior should directly follow a previous instance. Thus, there were two randomized versions of the stimulus tape. Truthtelling was included in the stimulus tape only to provide judges with a base for comparing deceptive responses.

Judgment of Lies

Forty students from the same midwestern university viewed one of the two versions of the stimulus tape; for their participation they received extra credit. Twenty students viewed an audio-visual channel and twenty students viewed a video-only channel. For each instance of behavior viewed, judges made a dichotomous decision as to whether the behavior constituted a lie or the truth. Additionally, subjects judged the degree to which the behavior was truthful or deceptive. This judgment was made using a Likert Scale with the terms "truthful" and "deceptive" at the beginning or end of the scale. Order of the words "truthful" and "deceptive" was randomized.

Design

Because there were multiple actors representing each condition, the scores of same-sex actors within a condition were combined to form indices representative of each given condition. Judges scores for these indices were the unit of analysis. These scores were entered into a 2 x 2 x 2 x 2 ANOVA (Lie Type x Preparation Type x Deceiver's Sex x Channel). Lie type, preparation type and sex of deceiver were within subjects factors. Channel was a between subjects factor.

Results

Analysis of the data in this manner revealed significant main effects for preparation type and deceiver's sex as well as a significant interaction between preparation type and deceiver's sex. In addition, there was a significant three-way interaction between channel, preparation type and deceiver's sex (F(1,38)=9.98, p < .003). As a consequence, the three way interaction was decomposed. Comparison of detectability scores between the audiovisual channel and the video-only channel revealed that while type of preparation had no significant effect on perceived truthfulness across channel for males telling planned lies and females telling planned and spontaneous lies, males spontaneously lying were perceived as less truthful in the audiovisual channel than in the video-only channel (F (1,38) =10.69, p <.002). Additionally, comparisons of means were made within channels. To control for type I error, alpha was set at .004. Comparisons within channel revealed that while in the audiovisual channel, males spontaneously lying were perceived as less truthful than males telling planned lies (t(19)=-5.16, p < .001), females telling planned lies (t(19)=-4.14, p. < .001), and females who told lies spontaneously (t(19)=-4.58, p < .001), in the video-only channel, there were no significant differences between males spontaneously lying and their counterparts.

Analysis of the data also revealed a significant interaction between channel and lie type, as well as a significant interaction between channel, lie type and deceiver's sex (F(1,38)=4.52, p < .040). Decomposition of the three-way interaction revealed that while type of lie across channels did not significantly affect the perceived truthfulness of females telling unfamiliar lies, or the perceived truthfulness of males telling familiar and unfamiliar lies, females telling familiar lies were perceived as less truthful in the audiovisual channel than they were in the video-only channel (F(1,38)=5.49, p < .024). Additionally, comparisons of means were made within channels. To control for type I error, alpha was set at .004. Comparisons of means within channels revealed that while there were no significant differences between the perceived truthfulness of females telling familiar lies in the audiovisual format and their counterparts in the same format, in the video-only format, females telling familiar lies were perceived as significantly more truthful than males telling familiar lies (t(19)=-3.18 p < .005). In addition, in the audiovisual channel, males telling unfamiliar lies were perceived as significantly less truthful than females telling unfamiliar lies (t(19)=-3.17, p < .005).

Means and standard deviations for each condition are displayed in Table 1. Because analyses were performed on an index representing a compilation of actors' scores within a condition, follow-up tests were conducted to determine if within conditions, collapsing channel, there were actor differences. These analyses revealed that there were significant differences between the perceived truthfulness of actors within all conditions except the condition in which females spontaneously told familiar lies and the condition in which males spontaneously told familiar lies. Because certain actors, regardless of their actual veracity are judged consistently as truthful or deceptive (Riggio, Tucker & Widaman, 1987), it was decided that further analyses should be conducted to determine if this demeanor bias (Riggio et. al, 1987) underlaid actor differences.

Table 1

Means for Lie Type by Condition, Sex and Channel

	Familiar						Unfamiliar
	Spontaneous			Planned			Spontaneours		Planned
	(M)	(F)	(M)		(F)	(M)		(F)		(M)	(F)
Channel
Audiovisual
M	11	15	17		14	12		18		16	15
SD	4.0	3.9	5.3		4.7	3.9		4.6		3.8	3.9
Video
M	14	18	16		17	15		15		15	16
SD	3.7	4.6	3.4		4.5	4.4		4.7		3.3	4.7
Combined
M	12.5	16.5	16.5		15.5	13.5		16.5		15.5	15.5

Note. All n's=20

Dichotomous judgments of whether the actor was lying or telling the truth were analyzed using oneway chi-square tests. These analyses revealed that six of the males and three of the females were judged consistently as lying or as consistently telling the truth, regardless of their actual behavior. Five of the males were consistently judged as lying, while one of the males was consistently judged as telling the truth; in addition, one of the females was consistently judged as lying, while two were consistently judged as telling the truth.

Actors for whom there was or was not a bias were found in all conditions. In addition, no condition consisted solely of actors who had a "liar's" demeanor or a "truthteller's" demeanor. Different types of actor demeanor were distributed across conditions; therefore, actor differences in themselves do not account for the effects found.

Study 2

One possible cause for the lack of a clearcut relationship between preparation type and lie type is that white lies are not in fact more familiar than lies involving violations of trust, even though theorists have suggested that white lies are "field-tested," and "occur with a very high frequency in everyday interaction" (Camden et al., 1984, p. 319). To examine this possibility 58 males and 58 females were asked to complete a two-item questionnaire contrasting their experience in telling white lies with that of telling lies that violate another's trust. Subjects were asked to indicate on a seven point scale which experience, as characterized by an example, they had encountered more frequently. The order of items was counterbalanced across subjects.

Responses were analyzed using a 2 X 2 ANOVA (Sex: male/female X Lie Type: tact/trust), with the last factor within subjects. Analysis of the data in this manner revealed a main effect for sex (F(1,114)=16.50, p < .001). There was also a main effect for lie type (F(1,114)=29.29, p < .001) and a sex by lie type interaction (F(1,114)=4.26, p < .041). Decomposition of the interaction revealed that for both males (F(1,57)=4.49 p < .039) and females (F(1,57)=37.26 p < .001) telling a lie for tact is a significantly more familiar experience than telling a lie involving breach of trust, and that for females in particular telling lies for tact is a significantly more familiar experience than it is for males (F(1,115)=24.36 p < .001). Means and standard deviations are arrayed in Table 2.

Table 2

Means Familiarity of Lie Types by Sex

	Sex
	Male	Female
Type
Tact
M	3.35	2.03
SD	1.6	1.2
Trust
M	4.09	3.69
SD	1.8	1.8

Note. All n's=58

Discussion

The results of study 2 indicate that telling white lies is a more familiar experience than telling lies that violate another person's trust. Thus study 2 provides support for the contention that white lies, by virtue of their frequency, may be stored and deployed differently from lies that are less frequently deployed.

While no clear-cut relationship between lie type and preparation type was found, the results of this study are informative regarding the relationship between lie type, preparation type, sex and channel. Results suggest that the female encoders may have been better at dissembling than the male encoders. Secondly, results suggest that for familiar lies, the women's advantage may have resided in their nonverbal expressivity. Finally, findings suggest that the male encoder's deceptive efforts benefited from planning regardless of whether the lie was familiar or unfamiliar, and that planning enables the verbal component of the lie.

The results of this study, suggest that the female encoders may have been better at conveying truthfulness than males. Whether their lies were planned or delivered spontaneously, when viewed in the audiovisual channel, women were perceived as more truthful than males lying spontaneously. Moreover, whether the lie was familiar or unfamiliar, women were perceived as more truthful than their male counterparts. This held true for unfamiliar lies viewed in the audiovisual channel and familiar lies viewed in the video-only channel. The perception of truthfulness garnered by women across preparation type and lie type suggests that in some way they had an advantage over males in dissembling.

For familiar lies this advantage may have resided in the female encoder's expressivity. It was in the video-only format that women telling familiar lies were perceived as more truthful than males telling familiar lies. Moreover, the familiar lies of females viewed in the video-only format were perceived as more truthful than the same lies viewed in the audio-visual format.

The difference in the perceived truthfulness of familiar lies told by women between the two channels suggests that the verbal content of the white lie may lessen its perceived sincerity. It may be that the verbal content contradicts the nonverbal expressiveness, or it may be that white lies sound patent. Since one way to lie is to exaggerate the expected response, another possibility is that the obvious content of the white lie may be compounded by exaggerated expressiveness. The effect of these two factors in conjunction may be to lessen perceived sincerity. This explanation would be congruent with DePaulo's (DePaulo et al., 1985) findings that ingratiating lies are more detectable than noningratiating lies and that planned lies are perceived as more deceptive than spontaneous lies (DePaulo et al., 1983). While DePaulo's (DePaulo et al., 1985) findings would suggest that the presence of visual nonverbal cues heightens such perceptions, the results of this study potentially suggest that at least for white lies, verbal content and vocalics are important determinants of the decoder's perceptions of the message's sincerity.

Finally, this study adds to the literature that suggests that planning benefits deceptive effort. In this study, planning benefited males, whether telling white lies or lies breaching trust, especially when the verbal content of the lie was available to the decoder. The fact that males spontaneously lying in the video-only format were perceived as more truthful than when they were lying in the audiovisual format suggests that nonverbal expressivity accompanying the spontaneously told lie was not, in and of itself, enough to signal deceptiveness. This finding would suggest that the verbal content of the lie and the vocalics that accompanied it reflected the spontaneity of the lie.

In sum, this study, while finding support for the familiarity of white lies in terms of what persons recognize as familiar, finds no evidence (in terms of efficiency and goal-success) for this familiarity in the performance of the lie. What the findings of this study do suggest is that the female encoders in this study seemed to have some advantage over the male encoders in the act of dissembling. Moreover, findings suggest that the nonverbal expressivity of females may have been to their advantage in telling white lies, especially if its effect was not counteracted by the patency of the lie's content. Finally, this study suggests that the lies of the male encoders, especially the verbal content of such lies, benefited from planning. Further research needs to examine what the prevalence of white lies means for their familiarity and to further question what familiarity with a lie, or any other speech act, means for the effortfulness of its deployment.

References

Anderson, J. R. (1974). Cognitive psychology and its implications. New York: Freeman.

Berger, C. R. (in press). A plan-based approach to strategic communication. In D. E. Hewes (Ed.), Cognitive bases of interpersonal communication. Hillsdale, NJ: Erlbaum.

Berger, C. R. (1988). Communication plans and communicative performance. Paper presented at the annual convention of the International Communication Association, New Orleans.

Berger, C. R., Karol, S. H., & Jordan, J. M. (1989). When a lot of knowledge is a dangerous thing: The debilitating effects of plan complexity on verbal fluency. Human Communication Research, 16, 91-119.

Bond, C. F., Kahler, K. N., & Paolicelli, L. M. (1985). The miscommunication of deception: an adaptive perspective. Journal of Experimental Social Psychology, 21, 331-345.

Bower, B. H., Black, J. B. & Turner, T. J. (1979). Scripts in memory for text. Cognitive Psychology, 11, 177-220.

Cambis, M. (1979). Planning theory and philosophy. London: Tavistock Publications.

Camden, C., Motley, M. T., & Wilson A. (1984). White lies in interpersonal communication: A taxonomy and preliminary investigation of social motivations. The Western Journal of Speech Communication, 48, 309-325.

Chiesi, H. L., Spilich, G. J., & Voss, J. F. (1979). Acquisition of domain related information in relation to high and low domain knowledge. Journal of Verbal Learning and Verbal Behavior, 18, 257-273.

DeLisi, R. (1987). A cognitive-development model of planning. In S. L. Friedman, E. K. Scholnick & R. R. Cocking (Eds.), Blueprints for thinking: The role of planning in cognitive development (pp. 79-109). Cambridge: Cambridge University.

DePaulo, B. M., Lanier, K., & Davis, T. (1983). Detecting deceit in the motivated liar. Journal of Personality and Social Psychology, 45, 1096-1103.

DePaulo, B. M., Lassiter, G. D., & Stone, J. I. (1982). Attentional determinants of success at detecting deception and truth. Personality and Social Psychology Bulletin, 8, 273-278.

DePaulo, B. M., Stone, J. I., & Lassiter, G. D. (1985). Telling ingratiating lies: Effects of target sex and target attractiveness on verbal and nonverbal deceptive success. Journal of Personality and Social Psychology, 48, 1191-1203.

DePaulo, B. M., Zuckerman, M., & Rosenthal, R. (1980). Detecting deception. In L. Wheeler (Ed.) Review of Personality and Social Psychology, 1 (pp. 125-162). Beverly Hills: Sage.

deTurck, M. A., & Miller, G. R. (1990). Training observers to detect deception: Effects of self-monitoring and rehearsal. Human Communication Research, 16, 603-620.

Earley, C. P, Wojnaroski, P. , & Prest, W. (1987). Task planning and energy expended: Exploration of how goals influence performance. Journal of Applied Psychology, 72, 107-114.

Ekman, P. (1985). Telling lies. New York: Berkley Books.

Ekman, P., & Friesen, W. V. (1974). Detecting deception from the body or face. Journal of Personality and Social Psychology, 29, 288-298.

Ekman, P., & Friesen, W. V. (1969). Nonverbal leakage and clues to deception. Psychiatry, 32, 88-106.

Faludi, A. (1973). Planning Theory. Oxford: Pargamon.

Feldman, R. S., Jenkins, L., & Popoola, O. (1979). Detection of deception in adult and children vi facial expressions. Child Development, 50, 350-355.

Fiske, S. T., Kinder, D. R., & Larter, M. W. (1983) The novice and the expert: Knowledge-Based strategies in political cognition. Journal of Experimental Social Psychology, 19, 381-400.

Grasesser, A. C., Woll, S. B., Kowalski, D. J., & Smith, D. A. (1980). Memory for typical and atypical actions in scripted activities. Journal of Experimental Psychology, 6, 503-515.

Greene, J. O. (1984). Speech preparation processes and verbal fluency. Human Communication Research, 11, 61-84.

Greene, J. O., & Lindsey, A. E. (1989). Encoding processes in the production of multiple-goal message. Human Communication Research, 16, 120-140.

Greene, J. O., O'Hair, H. D., Cody, M. J., & Yen, C. (1985). Planning and control of behavior during deception. Human Communication Research, 11, 335-364.

Greene, J. O., Smith, S. W., Smith, R. C., & Cashion, J. L. (1987). The sound of one mind working: Memory retrieval and response preparation as components of pausing and spontaneous speech. In McLaughlin, M. L. (Ed.), Communication Yearbook, 10, (pp. 241-258). Newbury Park: Sage.

Greenglass, E. (1972). Effects of age and prior help on "altruistic" lying. The Journal of Genetic Psychology, 121, 303-313.

Hammond, K. J. (1989). Case-based planning: Viewing planning as a memory task. New York: Academic.

Holloway, C., (1986). Strategic planning. Chicago: Nelson-Hall.

Holzman, P. S. & Rousey, C. (1966). The voices as a precept. Journal of Personality and Social Psychology, 4, 79-86.

Knapp, M. L., & Comadena, M. E. (1979). Telling it like it isn't: A review of theory and research on deceptive communications. Human Communication Research, 5, 270-285.

Kraut, R. E. (1978). Verbal and nonverbal cues in the perception of lying. Journal of Personality and Social Psychology, 36, 380-391.

Langer, E. J. (1985). Playing the middle against both ends: The usefulness of adult cognitive activity as a model for cognitive activity in childhood and old age. In S. Yessen (Ed.), The growth of reflection (pp. 267-285). New York: Academic.

Langer, E. J., & Weinman, C. (1981) When thinking disrupts intellectual performance: Mindfulness on an overlearned task. Personality and Social Psychology Bulletin, 7, 240-243.

Lippard, P. V. (1988). "Ask me no questions, I'll tell you no lies": Situational exigencies for interpersonal deception. Western Journal of Speech Communication, 52, 91-103.

Littlepage, G. E., & Pineault, M. A. (1978). Verbal, facial, and paralinguistic cues to the detection of truth and lying. Personality and Social Psychology Bulletin, 4, 461-464.

Littlepage, G. E., & Pineault, M. A. (1979). Detection of deceptive factual statements from the body and the face. Personality and Social Psychology Bulletin, 5, 325-328.

Littlepage, G. E., & Pineault, M. A. (1984). Detection of deception of planned and spontaneous communications. The Journal of Social Psychology, 125, 195-201.

Littlepage, G. E., Tang, D. W. & Pineault, M. A. (1986). Nonverbal and content factors in the detection of deception in planned and spontaneous communications. Journal of Social Behavior and Personality, 1, 439-450.

Metts, S. (1989). An exploratory investigation of deception in close relationships. Journal of Social and Personal Relationships, 6, 159-179.

McClintock, C. C., & Hunt, R. G. (1975). Nonverbal indicators of affect and deception in an interview setting. Journal of Applied Social Psychology, 5, 54-67.

Miller, G. R., deTurck, M. A., & Kalbfleisch, R. J. (1983). Self-monitoring, rehearsal, and deceptive communication. Human Communication Research, 10, 97-117.

O'Hair, H. D., Cody, M. J. & McLaughlin, M. L. (1981). Prepared lies, spontaneous lies, Machiavellianism, and nonverbal communication, Human Communication Research, 7, 325-339.

O'Keefe, B. J., & Shepard, G. J. (1987). The pursuit of multiple objectives in face-to-face-persuasive interactions: Effects of construct differentiation on message organization. Communication Monographs, 54, 396-417.

Parham, I. A., Feldman, R. S., Oster, G. D., & Popoola, O. (1981) Intergenerational differences in nonverbal disclosure of deception. The Journal of Social Psychology, 113, 261-269.

Pea, R. D., & Hawkins, J. (1987). Planning in a chore scheduling task. In S. L. Friedman, E. K. Scholnick, & R. R. Cocking (Eds.), Blueprints for thinking: The role of planning in cognitive development (pp. 273-302). Cambridge, England: Cambridge University.

Planalp, S. (1985). Relational schemata: A test of alternative forms of relational knowledge as guides to communication. Human Communication Research, 12, 3-29.

Planalp, S. & Hewes, D. (1982). A cognitive approach to communication theory: Cogito ergo dico. Communication Yearbook, 5, 49-62. Newbury Park: Sage.

Pryor, J. B., & Merluzzi, T. V. (1985). The role of expertise in processing social interaction scripts. Journal of Experimental Social Psychology, 21, 362-379.

Riggio, R. E., Tucker, J., & Widaman, K. F. (1987). Verbal and nonverbal cues as mediators of deception ability. Journal of Nonverbal Behavior, 11, 126-145.

Roloff M., Berger, C. R. (1982). Social cognition and communication: An introduction. In M. Roloff & C. R. Berger (Eds.), Social Cognition and Communication (pp. 9-32). Beverly Hills: Sage.

Sackman, H., & Citerbaum, R. (1972). Online planning: Towards creative problem solving. New Jersey: Prentice Hall.

Schank, R., & Abelson, R. P. (1977). Scripts, goals, plans and understanding. Hillsdale, NJ: Lawrence Erlbaum.

Scholnick, E. F., & Friedman, S. L. (1987). The planning construct in the psychological literature. In S. L. Friedman, E. K. Scholnick & R. R. Cocking (Eds.), Blueprints for thinking: The role of planning in cognitive development (pp. 1-33). Cambridge: Cambridge University.

Spilich, G. J., Vesonder, G. T. Chiesi, H. L. & Voss, J. F. (1979). Text processing of domain-related information for individuals with high and low domain knowledge. Journal of Verbal Learning and Verbal Behavior, 18, 275-290.

Streeter, L. A., Krauss, R. M. Geller, V., Olson, C., & Apple, W. (1977). Pitch changes during attempted deception. Journal of Personality and Social Psychology, 35, 345-350.

von Cranach, M., Kalbermatten, U., Indermuhle, K., & Gugler, B. (1982). Goal-directed action. London: Academic.

Wilensky, R. (1981). Metaplanning: Representing and using knowledge about planning in problem solving and natural language understanding. Cognitive Science, 5, 197-233.

Zuckerman, M., DePaulo, B. M., & Rosenthal, R. (1981). Verbal and nonverbal communication of deception. In L. Berkowitz (Ed.), Advances in experimental social psychology, 14, (pp. 61-106). New York: Academic.

Zuckerman, M., Kernis, M. R., Driver, R. Koestner, R. (1984). Segmentation of behavior: Effects of actual deception and expected deception. Journal of Personality and Social Psychology, 46, 1173-1182.

Zuckerman, M., Larrance, D. T., Spiegel, N. H., & Klorman, R. (1981). Controlling nonverbal displays: Facial expressions and tone of voice. Journal of Experimental Social Psychology, 17, 506-524.

Response from Leonard J. Shedletsky

While white lies are often accomplished by speaking, they are not, in and of themselves, a speech act, anymore than telling the truth is a speech act. Nevertheless, telling white lies does seem to be something we do relatively frequently, more frequently than telling lies that violate trust. It is assumed in this research that white lies are more routinized than lies violating trust. This research does not find a clear-cut relationship between lie type (white versus trust violation) and preparation type. When sex and channel are taken into account, lie type and preparation type matter. It would seem that this outcome ought to cast doubt on the theory that white lies represent routinized behavior and lies violating trust represent behavior that is less planned; moreover, we can doubt that planning theory accounts for the encoding and decoding carried out in this research.

Compliance-Gaining Behavior: Analysis of Cognitive Processes Underlying Message Production

Elizabeth M. Goering and Melissa M. Spirek

About the Authors: Elizabeth M. Goering (Ph.D., Purdue University) is an assistant professor in the Department of Communication Studies at Indiana University, Indianapolis, IN 46202/ (317) 274-0565. Melissa M. Spirek (Ph.D., Purdue University) is an assistant professor in the Department of Telecommunications at Bowling Green State University, Bowling Green, OH 43403-0235/ (419) 372-8641. The authors thank Teresa Ely Bowen for her assistance in collecting and coding the data.

Abstract: The current investigation was designed to explore the cognitive processes and structures which underlie communicative behavior in the context of interpersonal persuasion; specifically, it sought to test several of the processes implicit in a general cognitive model, action assembly theory. Initially, this investigation examined the effect of recency and frequency on compliance-gaining message production. In addition, this experiment explored several temporal indicators of cognitive activity involved in compliance-gaining message construction, particularly in the production of subsequent attempts to gain compliance.

Between 1977 and 1984, over 40 studies examined the topic of compliance-gaining behavior in interpersonal relationships (Seibold, Cantrill & Meyers, 1988). This trend has continued, for a surge of research on this topic has emerged since 1984 (Dillard, 1988; Roiger & Hellweg, 1986). The focus of the majority of these investigations has been on the production of compliance-gaining messages in interpersonal relations. More specifically, these studies have examined the likelihood with which individuals might use a particular solution (e.g., Marwell & Schmitt, 1967; Miller, Boster, Roloff & Seibold, 1977) or the differences in substantive content when a study's participants construct their own messages as opposed to adopting a checklist method (e.g., Applegate, 1982; Clark & Delia, 1976; Delia, Kline & Burleson, 1979; Neuliep & Mattson, 1990). In most cases, the researchers have been interested in establishing either individual difference (e.g., Boster & Levine, 1988; deTurck, 1987; Hunter, 1988; O'Hair, O'Hair, Southward, Morris & Krayer, 1987) or situational correlates (e.g., Neuliep, 1987; Yuda, 1982) of compliance-gaining strategy choice. Although this line of questioning has been extremely helpful in furthering the understanding of compliance-gaining, this concentration on content hinders the development of models of the cognitive structures and processes that underlie compliance-gaining behavioral production.

Developing various cognitive models of structures and processes is important for several reasons (Waldron & Cegala, 1992). First, as Greene (1984a) suggests, it is these elements that allow the production of communicative messages. Moreover, Greene (1984c) argues that a focus on structures and processes will enhance production and explanation of communicative behavior. In addition, Anderson (1985) asserts that an explicit understanding of how humans acquire, process and structure knowledge and of the cognitive demands required to do so, will enable researchers to improve training and performance.

The vast majority of studies in the compliance-gaining literature focus solely on initial attempts at persuasion. Few studies examine what happens when an initial compliance-gaining effort fails and an individual is forced to try other persuasive alternatives. In direct contrast, a number of researchers have investigated the quantity of compliance-gaining requests. One of many examples is a study by Richard and Dodge (1982) that explored the ways in which children's peer status impacted the number of solutions generated in response to two hypothetical situations. They discovered that "popular" children were able to generate significantly more solutions than the "aggressive" or "isolated" subjects. However, such research on subsequent compliance-gaining methods is limited in that many studies have only explored the compliance-gaining behavior of children (e.g., McQuillen, 1986) or teachers and students (Kearney & Plax, 1987; Kearney, Plax, Sorenson & Smith, 1988). As noted earlier, very few investigations explore the nature of the cognitive structures and processes which underlie subsequential attempts to gain compliance. Although studies have looked at the repertoire of strategies, these same investigations do not really provide an explanation of the cognitive processes.

In light of these concerns, the present paper briefly reviews existing models of cognitive processing which specifically focus on the construction of compliance-gaining messages. Following this, a general cognitive approach to human communication is reviewed and applied to an investigation of compliance-gaining behavior, particularly subsequent message production.

Cognitive Models of

Persuasive Message Construction

A substantial number of cognitive models of persuasion have been proffered in recent years. In fact, Eagly and Chaiken (1984) argue that most contemporary persuasive models are cognitive. However, the majority of these cognitive accounts of compliance-gaining are receiver-oriented; they focus on the ways in which the recipient of a persuasive message decodes and processes that message. Because the primary concern of this inquiry is on the production of persuasive messages, such receiver-oriented accounts lie beyond the scope of the current investigation.

Nonetheless, several researchers have forwarded cognitive explanations of persuasive message construction. Some have attempted to explain compliance-gaining message production in terms of situational factors. For example, Reardon's (1981) ACE theory suggests that persuasive messages are selected on the basis of their perceived appropriateness, consistency and effectiveness; while Clark and Delia (1979) claim that behavioral choices are made on the basis of three interactional goals: instrumental, relational and identity management. Sillars' (1980) research offers support for the notion that compliance-gaining messages are selected on the basis of a risks-benefit analysis, and Smith (1982, 1984) proposes a contingency rules or schemata model which argues that interpretive and behavioral rules govern behavioral choices in a compliance-gaining situation.

Yet another cognitive explanation of compliance-gaining message selection is suggested by Hunter and Boster (1978). These researchers propose that individuals have a range of potential strategies arranged on a continuum from positive to negative in terms of their emotional impact on the listener. Each individual has an "ethical threshold" that constrains and controls compliance-gaining message choices.

Perhaps one of the most detailed cognitive frameworks for analyzing persuasion is Schank and Abelson's "persuade package" (Schank & Abelson, 1977; Rule, Bisanz & Kohn, 1985). These theorists suggest that individuals store goals and methods for accomplishing these goals in persuasive schemata. They further assume that the number of goals in the persuasive scheme is a small, finite set, and that these goals are accompanied by a limited set of methods ordered so that when one method fails, the individual will select an alternative method further down the sequence.

Unfortunately, communication researchers have only begun to explicate both the underlying cognitive structures and the processes which operate through these structures. Some of the available cognitive explanations fail to specify the structural dimensions of compliance-gaining message storage; while others offer no account of the processes involved in assembling, creating, retrieving or implementing interpersonal persuasive messages. Consequently, some of these accounts are "inherently unfalsifiable," for as Greene (1984a) explains, "As long as only structure or process is specified [in a cognitive model], any observation can be explained" (p. 245). In a similar vein it is interesting to note that these same theories might be falsifiable if they made incorrect predictions about content.

In direct contrast, a limited number of studies have begun to explicate both the processes and structures which underlie compliance-gaining behavior (Rule, Bisanz & Kohn, 1985; Smith, 1982, 1984). Specifically, two studies that tested an associative network model of memory for understanding compliance-gaining messages were conducted by Wilson (1990) and Greene, Smith & Lindsey (1990). In addition, Wilson, Cruz and Kang (1992) examined intrapersonal variability in order to see if attributions made about a compliance-gaining interaction were related to the observer's perspective.

One cognitive model that specifies both the processes and the structures which underlie human behavior was developed by Greene (1984a, 1984c). This model, action assembly theory, is a general cognitive account of human behavior, which lends itself quite easily to application in the specific context of compliance-gaining (Cody, Greene, Marston, O'Hair, Baaske & Schneider, 1985).

According to action assembly theory, memory is a "repository of procedural knowledge whose units are records of action-outcome; relationships which have been acquired through past experiences" (Greene, 1984c, p. 292). In the context of compliance-gaining, these procedural records would include features of compliance-gaining behaviors at various levels of abstraction and the outcomes associated with those behaviors through experiences. When an individual is confronted with a particular situation, the action assembly model suggests that relevant procedural records are reinforced or activated, with speed of activation being a "function of the recency and frequency of activation of the element" (Greene, 1984b, p. 64). When an individual is confronted with a compliance-gaining situation, she or he retrieves the procedural records relevant to gaining compliance in that situation, and the behaviors likely to be most highly activated are those which most closely correspond to the current situation and have been associated with successful compliance-gaining efforts, assuming that the activating conditions for all procedural records are equally met. This activation process occurs in parallel and makes no demands on cognitive processing capacity as it is simply a more or less automatic retrieval from memory.

A second process involved in this model is that of assembly, which is the process whereby the activated procedural records are integrated into an "output representation." Unlike the activation process, the assembly process involves a considerable amount of cognitive processing effort as one must formulate the desired plan for action rather than merely automatically retrieving information from memory (Greene & Lindsey, 1989). The compliance-gaining "strategies" that an individual employs in a particular situation, then, are not necessarily stored as strategies in memory. Rather they are the result of the activation of relevant procedural records and the assembly of those procedural records that the individual deems most appropriate for the particular situation into a "strategy," or plan of action.

In summary, a number of the existing cognitive accounts of interpersonal persuasion are inadequate as cognitive theories because they lack precision in terms of structures and processes; however action assembly theory does provide the specificity needed to generate testable hypotheses and to allow falsification of the theory itself while permitting the authors the opportunity to specifically examine compliance-gaining behavior.

Pausal Indicators of Cognitive Processes

Greene (1987) identifies "temporal measure of speech production," or pausal indicators, as one method for investigating demand on processing capacity and thus, indirectly, for testing the cognitive processes implicit in any cognitive model. In other words, such nonverbal cues can be thought of as the "output" by which the "black box" of the human mind can be understood (Greene & Cappella, 1986).

Hesitation in speech is one nonverbal behavior which has been empirically linked with increased cognitive effort. Goldman-Eisler's (1961a, 1961b) classic studies comparing hesitation in the speech of participants who were instructed to interpret cartoons and those who were asked just to describe the same cartoon suggest that as the cognitive difficulty of a task increases, the amount of hesitation (operationalized as silent pauses) increases as well. Goldman-Eisler concluded, after replicating the findings of her original studies, "pauses in speech are the delay due to processes taking place in the brain whenever speech ceases to be the automatic vocalization of learned sequences" (1968, p. 58).

Studies by several other researchers have confirmed the conclusions of Goldman-Eisler (Dechert & Raupach, 1980). One study (Levin, Silverman & Ford, 1967) discovered that when children were asked to either describe or explain what happened when two colorless liquids were mixed together to form a colored solution, the more difficult task was linked to more speech hesitation. As expected, the children who were instructed to try and explain the phenomenon displayed more frequent and longer silent pauses than the children who were asked only to describe what they had seen.

In another early investigation, Reynolds and Paivio (1968) compared the frequency of silent and filled pauses in the speech of participants who were asked to define abstract terms and those who were to give definitions of concrete words. Again, both silent and filled pauses were more frequent in the more difficult taskóthe definition of the abstract concepts. Siegman (1979) replicated Goldman-Eisler's original cartoon experimental design and measured a variety of hesitation-related dependent variables such as "reaction time," or "the period between the presentation of a response card to the participant and the participant's first word in response to the card," and pause duration ratio, or the sum of silent pauses within a participant's response over the total duration of the response. Siegman found that more difficult tasks were consistently associated with a variety of these measures of speech hesitation; she concludes "cognitive decision-making is associated with hesitation in speech" (p. 175).

In 1987, Greene, Smith, Smith and Cashion utilized an analysis of pausal phenomena to directly investigate certain properties of the cognitive processes suggested by action assembly theory. Specifically, they sought to test the plausibility of the notion that memory retrieval and response planning are separate procedures in behavior production by placing participants in situations where the amount of time allotted for retrieving information and actually planning responses was controlled. A comparison of speech onset latency and pause rate provided evidence that the two processes are, indeed, separate procedures.

In summary, while many of the existing models of compliance-gaining message construction do not provide a basis for generating clear, falsifiable predictions concerning behavior, action assembly theory does, because it clearly specifies the cognitive processes and structures underlying human behavior (Booth-Butterfield, 1987). Moreover, by focusing on nonverbal cues, one can test the predictions generated by this theory.

Three hesitation variables are of particular interest in this investigation: response latency, or the amount of time between the stimulus statement or question and the onset of vocalization by the participant; filled pause rate, or the number of filled pauses (e.g., "ers," "ums," and "ahs") uttered by the individual divided by the person's total talk time, and silent pause rate, which will be operationalized as speech rate. This decision is justified by the high correlations between speech rate and the occurrence and length of silent pauses (r=.80-.90) reported in previous research (Goldman-Eisler, 1968; Sabin, Clemmer, O'Connell & Kowal, 1979). The use of these three pausal measures in this context is warranted because these speech phenomena have been shown to be useful indices of the amount of cognitive processing capacity required for a communicative task (Siegman, 1978; Siegman & Pope, 1965).

One prediction implied by action assembly theory is that the speed of activation process, is in part, dependent on the recency and frequency with which a particular procedural record has been employed. Thus, the following hypotheses are advanced:

Hypothesis 1: Subjects' first constructed responses will be rated as most recently used with decreasing monotonicity for subsequent responses, such that later responses will be rated as less recently used.

Hypothesis 2: Subjects' first constructed responses will be rated as most frequently used with decreasing monotonicity for subsequent responses, such that later responses will be rated as less frequently used.

In addition, action assembly theory distinguishes the activation process from another process termed assembly. Upon presentation of a stimulus situation, participants would have to engage in both the activation and assembly processes. While activation does not place demands on cognitive processing capacity, it does take time (as evidenced by Greene et. al., 1987). Therefore, initial response latency is expected to be relatively long. Subsequent message production, however, should entail shorter response latencies, because activation occurs in parallel, thus several relevant procedural records can be activated simultaneously, requiring time only for assembly. In addition, these reduced temporal demands should also be evidenced by a decrease in pausal behavior during the utterance of the message. However, once an individual's readily accessible procedural records have been exhausted, response latency should increase as the individual is required to search her or his memory for additional relevant behavioral alternatives. This increased demand on cognitive processing capacity should also be evidenced in an increase in pauses during speech. Furthermore one would expect this effect to be observed across situations because prior research in compliance-gaining (e.g., Cody et. al., 1985; Boster & Stiff, 1984; Clark, 1979) suggests that messages are situation specific. Therefore two additional hypotheses are advanced:

Hypothesis 3: A quadratic trend will be observed between response latency and response number (i.e., whether the response is the 1st, 2nd, 3rd, and so on, one produced by the participant) across both stimulus situations, with response latency being larger for the 1st response generated, decreasing for subsequent responses, and increasing for the last responses generated.

Hypothesis 4: A quadratic trend will be observed between pause rate and response number, with pause rate being larger for the 1st response generated, decreasing for subsequent responses, and increasing for the last responses generated.

The third pausal indicator examined in this study will be filled pauses. Inconsistencies have been found among studies utilizing this measure (Greene, 1984b; Greene et al., 1987; Siegman, 1978). Therefore, due to the exploratory nature of this particular measure for our study, a research question as opposed to a formal hypothesis is advanced:

Research Question 1: Will the sequence of the message produce a relationship for filled pause rate that is similar to that expected for the measures of response latency and speech rate?

Method

Participants

A total of 46 college students enrolled in undergraduate communication courses at a large midwestern university were recruited as participants. Of these 46, two failed to keep appointments for their interviews, two failed to complete all phases of the experiment, two interviews could not be transcribed because of the inaudible recordings of their interviews, and 12 were unable to generate multiple compliance-gaining messages and, consequently, could not be used in the repeated measure analysis employed in this study. Thus, 28 of those recruited served as subjects in the final analysis. Participants ranged in age from 17 to 26 years of age (M = 19.87 years). The sample consisted of 19 females and nine males. Each participant received extra credit points for participating in this investigation.

Procedures and Measures

General Procedures. Subjects were individually interviewed by one of two female experimenters. Participants were told that the interviewer was interested in how people interact with one another. Each subject was asked to state her or his name, age, major area of study and classification and then to respond to the two hypothetical situations involving the goal of gaining the compliance of another person. At no point prior to the presentation of the first situation or throughout the interview did either interviewer reveal that the focus of the situations was compliance-gaining. All responses were audio-taped, and verbatim transcripts of these responses were used for the analysis.

Following the response to these hypothetical tasks, the participant was asked to complete a set of scales for each strategy she or he had supplied. The questionnaire focused on the recency and frequency with which the participant employed the alternatives she or he suggested.

Compliance Gaining Strategy Elicitation and Coding. Each participant was presented with two randomly ordered hypothetical situations involving the goal of gaining compliance from another individual. The situations utilized in this study were the two situations rated by college students as being most believable and easiest to imagine of five situations (Burleson, Wilson, Waltman, Goering, Ely & Whaley, 1989) developed for compliance-gaining research [Applegate's (1982) office manager and roommate situations, Burke's (1979) group project and vacation situations, and Wiseman and Schenck-Hamlin's (1981) loud stereo situation].

The situations selected described: (a) a roommate whose entertaining was interfering with one's own life and (b) a classmate who failed to successfully complete her or his part of a group project, thereby endangering the grade for all of the group members. Following the presentation of each situation, the participant was asked, "What would you do if this happened to you?" After completing her or his first response, subsequent messages were elicited by asking, "If that failed, what else would you do or say?" This line of questioning was repeated until the participant had generated five responses or appeared to have exhausted her or his behavioral repertoire (i.e., usually by stating that she or he could not think of anything else).

Audio recordings of the participant's responses were timed to obtain the hesitation variables. Throughout the analysis, three pausal indicators were used: response latency, speech rate and filled pause rate. Initially, the recordings were timed to measure response latency and total talk time. Response latency was assessed by determining the length of time between the completion of the interviewer's description of the hypothetical situations and the first vocalization in the participant's first produced response. For subsequent responses, this was measured as the length of time between the completion of the probe and the first vocalization of each subsequent response. Total talk time was computed by subtracting response latency from total duration of talk on each response. Total duration of talk was measured as the length of time between the end of the interviewer's explanation of the situation and/or probe and the participant's last vocalization in each response. Speech rate was computed by dividing the total number of words, excluding filled pauses and speech disruptions, by total talk time. Filled pause rate was computed by dividing the total number of filled pauses ("ers," "uhms," and "ahs") by the total talk time. The number of words in each response and the number of filled pauses were determined from the verbatim transcripts of each subject.

Reliability for total duration of talk, response latency, and filled pauses was established by having two individuals independently time 20% of the protocols. The interrater reliability coefficient was .92. All remaining responses were timed by one of the coders.

Table 1

Cell Means and Standard Deviations for

Recency, Frequency, Response Latency, Speech Rate and Filled Pause Rate

Message Number

123

RMGPRMGPRMGP

Recency

Means4.794.400.423.243.503.18

Standard Deviations1.641.541.922.011.962.04

Frequency

Means5.274.883.723.853.63.78

Standard Deviations1.381.231.781.611.841.87

Response Latency

Means3.182.622.241.683.503.06

Standard Deviations1.652.481.382.172.143.24

Speech Rate*

Means3.263.053.263.234.293.30

Standard Deviations1.360.641.540.816.560.99

Fille Pause Rate**

Means0.090.080.110.090.080.10

Standard Deviations0.090.050.090.090.090.12

Note. RM is the abbreviation for the noisy roommate stimulus situation and GP is the abbreviation for the group project situation (n=28).

*Speech rate indicates number of words per minute.

**Filled pause rate indicates number of filled pauses per second.

Recency-Frequency Questionnaire. Following the interview task, participants were asked to complete a questionnaire designed to assess the recency and frequency of use for each strategy supplied during the interview. For each compliance-gaining response generated, each participant was asked to complete a set of six semantic differential scales: three were designed to measure frequency of strategy use ("I frequently use this" or "I do not frequently use this;" "I often use this" or "I do not often use this;" "I always use this" or "I never use this") and three were employed as assessments of recency of strategy use ("I have recently used this" or "I have not recently used this;" "I have used this lately" or "I have not used this lately;" "I used this not long ago" or "I have not used this for a long time"). Strategies selected for use with this questionnaire were individual thought units provided within the participants' responses during the interview. The thought unit, as a unit of analysis, refers to an utterance or set of utterances containing a single focus or referent (Hamilton & Hunter, 1985). Interviewers noted the substantive content of strategies during the interviews and reminded participants of these alternatives for completion of the recency-frequency questionnaire.

Reliability for this questionnaire was assessed by computing Cronbach alphas for the three scales tapping frequency of use and those measuring recency of use. This procedure produced coefficients of .93 for the frequency scales and .91 for those assessing recency.

Data Analysis and Results

Recency

Recency of use scores were computed for each response within each participant's monologues by averaging the three items measuring recency for each generated message on the recency or frequency questionnaire. Table 1 contains the means and standard deviations for recency, as well as for

Table 2

Results of Multiple Regressions Between Recency and Frequency of Use

and Message Number, Situation, and Order of Presentation

Recency of Use Frequency of Use

MultipleFMultipleF

RRatioRRatio

Message Number0.2720.09*0.3736.33*

Situation0.271.220.370.02

Order of Presentation0.292.84**0.370.57

*p _ .0001

**p _ .10

frequency and the temporal variables, in each of the stimulus situations.

In order to determine the extent to which variance in recency ratings could be explained by message number (whether the message was the first, second, or third one generated), a repeated-measures analysis of variance was performed. The responses of those participants who were able to generate three or more compliance-gaining messages in the stimulus situation were employed in this and all repeated measures analyses, and in order to maintain the equal cell sizes, the first three responses from each participant were included. The results of this analysis revealed a significant main effect for message number [F(2,54)=4.18, p< .0205, for the group project situation; F(2,54)=5.21, p < .0085, for the noisy roommate situation]. In order to pinpoint the nature of this effect, a repeated measures orthogonal trend analysis was computed. This contrast revealed a significant linear trend for the group project situation [F(1,27)=9.70, p < .0097].

A multiple regression analysis was conducted using the variables of message number, order of presentation and situation as predictors of recency. The total regression of these measures was significant [R=.29, R =0.09, adjusted r =0.07; F(3,231)=7.25, p < .0001]. However, a review of the summary of this analysis revealed that only message number significantly contributed to the prediction of the criterion (see Table 2). These analyses support, in part, the relationship predicted in the hypothesis 1 , which suggested that those strategies generated first should be the ones participants would rate as most recently used, with subsequent strategies receiving decreasing recency ratings.

Frequency

Frequency of use ratings were computed in the same manner as recency scores (i.e., as the average of the participants' frequency ratings for each timed sequence of speech). Once again, a repeated measures analysis of variance was computed to determine the extent to which message number could explain the variance in frequency scores. The effect was found to be significant [F(2,54)=3.93, p < .0255, for the group project situation; F(2,54)=8.35, p < .0007, for the noisy roommate situation]. Once again, a repeated-measures orthogonal trend analysis was computed, revealing, for the noisy roommate situation, a statistically significant linear trend in the data [F(1,27)=14.22, p < .0008]. For the group project situation, the linear trend contrast also proved to be significant [F(1,27)=6.31, p < .02].

Using message number, order of presentation and situation as predictors of frequency of use, a multiple regression analysis was conducted. The total regression was significant [R=.37, R =.14. adjusted r =.12, F(3,231)=12.11, p < .0001). But once again, only message number was found to significantly contribute to the prediction of the criterion (see Table 2). These results are consistent with hypothesis 2 which predicted that the compliance-gaining message generated first would be rated as the one that was used most often, with subsequent methods of persuasion receiving lower frequency of use ratings.

Table 3

Results of Multiple Regressions Between the Hesitation Variables

and Frequency of Use, Situation, Message Number and Frequency of Use

Hesitation Variables

ResponseSpeech RateFilled Pause

LatencyRate

MultipleFMultipleFMultipleF

RRatioRRatioRRatio

Frequency0.170.500.010.070.070.46

Situation0.180.640.040.360.080.79

Message

Number0.250.210.050.070.090.02

p _ .01

Response Latency

As mentioned earlier, response latency was operationalized as the length of time between the end of the interviewer's question or probe and the participant's verbalization. The initial analysis performed on these data was a repeated-measures analysis of variance of response latency by message number. This analysis yielded a significant main effect [F(2,540=7.72, p < .0011, for the group project situation; F(2,54)=4.83, p < .0017, for the noisy roommate situation]. Once again, a repeated-measures orthogonal trend analysis was conducted. For the group project situation, it revealed a statistically significant quadratic trend [F(1,27)=32.63, p < .0000]. For the noisy roommate situation, the quadratic contrast was found to be similarly significant [F(1,27)=14.58, p < .0007].

Including frequency of use, recency of use, situation and message number as predictors of response latency, a multiple regression analysis was conducted. The total regression was significant [R=.25, R2 = .06, adjusted r 2=.05; F(4,230)=3.91, p < .004). Once again, however, the only significant contributor to the prediction of the criterion was message number (see Table 3). These analyses suggest that initially response latencies are relatively long, but the response latencies of subsequent strategies decrease and then increase quite dramatically. These analyses support hypothesis 4 hypothesis which predicted a quadratic relationship between response latency and message number.

Speech Rate

Speech rate, or the number of words divided by the total amount of talk time, was employed as an indicator of the occurrence of silent pauses. A repeated measures analysis of variance produced no significant results [F(2,54)=.63, p < .53, for the group project situation; F(2,54)=.59, p < .56, for the noisy roommate situation]. In addition, employing frequency of use, recency of use, situation, and message number as predictors of speech rate in a multiple regression analysis revealed no significant main effects or interactions (see Table 3). Therefore, hypothesis 4 hypothesis , concerning speech rate, was not supported.

Filled Pause Rate

As earlier stated, a subsidiary question concerning the relationship between filled pause rate (i.e., the number of filled pauses divided by total talk time) and message number was advanced. In order to investigate this relationship, a multiple regression analysis was conducted using frequency of use, recency of use, situation and message number as predictors. The total regression was found to be not significant (see Table 3). A repeated-measures analysis of variance also revealed an effect that was not significant when looking specifically at message number [F(2,54)=.49, p < .62 for the group project situation; F(2,54)=1.60, p <.21, for the noisy roommate situation].

Discussion

This investigation was designed to explore the cognitive processes and structures which underlie communicative behavior in the context of interpersonal persuasion; specifically, it sought to test several of the processes implicit in a general cognitive model, action assembly theory. Initially, this investigation explored the effect of recency and frequency on compliance-gaining message production. In addition, this experiment explored several temporal indicators of cognitive activity involved in compliance-gaining message construction, particularly in the production of subsequent attempts to gain compliance.

Recency and Frequency

Action assembly theory identifies recency and frequency as two primary factors influencing the activation of relevant procedural records. The theory proposes that when an individual is confronted with a compliance-gaining situation, she or he retrieves the procedural records relevant to gaining compliance in that particular situation, and the behaviors likely to be most highly activated are those which more recently and more frequently have been associated with successful compliance-gaining efforts, assuming that activating conditions for all procedural records are equally met. This notion provided the basis for hypotheses one and two. It was predicted that the first compliance-gaining messages generated would be what the participants would later report as the ones they used most frequently and most recently. Subsequent messages would be rated lower on the recency or frequency questionnaire. The analysis of the recency of use data moderately supported hypothesis 1; similarly, hypothesis 2 was also partially supported.

One finding that appears to be somewhat inconsistent with the hypothesized predictions is that the mean recency of use rating in one of the situations (the noisy roommate situation) actually increases slightly from the second to the third message (see Table 2). Although the increase is quite small (.1), and not statistically significant, it would seem that the final message generated would be one that is not used frequently and/or has not been employed recently. A possible explanation of this anomalous result is that, to keep cell sizes equal, only the first three messages generated by each subject were employed in the repeated-measures analysis. Only three subjects were able to come up with a total of five compliance-gaining messages in each situation, but artificially ending these participants' responses at the third message may have artificially increased the observed recency of use rating slightly.

Pausal Indicators of Cognitive Processing

The second primary concern addressed in this investigation involved exploring several nonverbal hesitation cues as indicators of cognitive effort. It was suggested that response latency would be relatively high initially and would decrease with the generation of subsequent messages until the participants repertoire of possible compliance-gaining responses was depleted, at which time, response latency would once again increase. The basis of this argument was that participants would initially need time for both activation and assembly of relevant procedural records. Although activation does not place any demands on cognitive processing capacity, it does take time. However, because activation occurs in parallel and does not require cognitive effort, the amount of time required to produce subsequent messages should decrease, reflecting only the assembly demands on cognitive processing capacity. Nonetheless, once participants exhaust their repertoire of potential responses, it was suggested that response latency would once again increase as they activated and assembled additional alternatives.

This line of reasoning provided the basis for hypothesis 3, which suggested that a quadratic trend would be observed between response latency and message number. The results of this investigation support this prediction. Two situations were utilized in this investigation to insure that the results obtained were not unique to a particular situation. For both situations, response latency was significantly shorter in generating the second response than the first and significantly longer in generating the third response than the second.

Hypothesis 4 was developed along the same argument provided for hypothesis 3 . This prediction suggested that the relationship between speech rate and message number would also be quadratic in nature. Contrary to this argument, no significant effects were observed. Perhaps this finding can be understood by examining the correlations between response latency and speech rate. If response latency and speech rate are expected to have a similar relationship with message number, as was hypothesized in this investigation, one might expect the correlations between these two variables to be quite high. In fact, the opposite is true (rs range from -.20 to .05). However, this relationship, and thus the lack of support for hypothesis 4, may be explained on the basis of the intuitive incongruence between these two hesitation cues. If an individual takes time prior to uttering a response (i.e., response latency), one might expect that person to exhibit fewer silent pauses during the actual utterance, and vice versa. Further, in the case of rather routine compliance-gaining tactics, once interaction level choices have been carried out, other assembly operations may proceed very quickly.

Even though rather clear-cut hypotheses could be advanced with regard to response latency and speech rate on the basis of prior research, the relationship between filled pause rate and message number remained rather vague. Thus, research question 1 was proffered for exploratory purposes. Results of this inquiry revealed no effects for filled pause rate with respect to message number, situation, recency or frequency of use. This finding is not totally surprising, however, as previous research suggests that filled pauses occur more in dyadic interaction than in monologues, because filled pauses serve the function of preventing interruptions (e.g., Siegman & Pope, 1966; Greene et al., 1986). Because participants in this investigation were provided the floor for as long as they desired (i.e., the interviewers made no attempt to stop the flow of their responses), the task was essentially monologic rather than dyadic in nature. Consequently, the fact that no significant effects for filled pause rate were found is, perhaps, not unusual.

In conclusion, this investigation provides evidence that action assembly theory can be utilized as a basis for generating falsifiable hypotheses regarding compliance-gaining message production. The results of this inquiry suggest that the recency and frequency with which a particular procedural record is activated influences the production of compliance-gaining behavior, as one would expect from the processes implicit in action assembly theory. In addition, the inquiry provides support for the use of nonverbal cues such as hesitation behavior as indicators of the cognitive processing involved in compliance-gaining. Moreover, this approach can be useful in examining the cognitive structures and processes underlying subsequent message production, an area which has been overlooked in most compliance-gaining literature.

References

Anderson, J. R. (1985). Cognitive psychology and its implications. New York: W. H. Freeman and Company.

Applegate, J. L. (1982). The impact of construct system development on communication and impression formation in persuasive contexts. Communication Monographs, 49, 277-289.

Booth-Butterfield, S. R. (1987). Action-assembly theory and communication apprehension: A psychophysiological study. Human Communication Research, 13(3), 386-398.

Boster, F. J., & Levine, T. (1988). Individual differences and compliance-gaining message selection: The effects of verbal aggressiveness, argumentativeness, dogmatism, and negativism, Communication Research Report, 5(2), 114-119.

Boster, F. J., & Stiff, J. B. (1984). Compliance gaining message selection behavior. Human Communication Research, 10(4), 539-556.

Burke, J. A. (1979). The relationship of interpersonal cognitive development for the adaptation of persuasive strategies in adults. Paper presented at the annual conference of the Central States Speech Association, St. Louis, MO.

Burleson, B. R., Wilson, S. R., Waltman, M. S., Goering, E. M., Ely, T. K., & Whaley, B. B. (1989). Item desirability effects in compliance-gaining research: Seven studies documenting artifacts in the strategy selection procedure. Human Communication Research, 14(4), 429-486.

Clark, R. A. (1979). The impact of self interest and desire for liking on selection of communicative strategies. Communication Monographs, 46, 257-273.

Clark, R. A., & Delia, J. G. (1976). The development of functional persuasive skills in childhood and early adolescence. Child Development, 47, 1008-1014.

Clark, R. A., & Delia, J. G. (1979). Topoi and rhetorical competence. Quarterly Journal of Speech, 65, 187-206.

Cody, M. J., Greene, J. O., Marston, P. J., O'Hair, H. D., Baaske, K. T., & Schneider, M. J. (1985). Situation perception and message strategy selection. Paper presented at the annual convention of the International Communication Association, Honolulu, HI.

Dechert, H. W., & Raupach, M. (Eds.) (1980). Temporal variables in speech. New York: Mouton Publishers.

Delia, J. G., Kline, S. L., & Burleson, B. R. (1979). The development of persuasive communication strategies in kindergartners through twelfth-graders. Communication Monographs, 46, 241-256.

deTurck, M. A. (1987). When communication fails: Physical aggression as a compliance-gaining strategy. Communication Monographs, 54(1), 106-112.

Dillard, J. P. (1988). Compliance-gaining message-selection: What is our dependent variable? Communication Monographs, 55(2), 162-183.

Eagly, A. H., & Chaiken, S. (1984). Cognitive theories in persuasion. In L. Berkowitz (Ed.), Advances in experimental social psychology, vol. 17. New York: Academic.

Goldman-Eisler, F. (1961a). Hesitation and information in speech. In C. Cherry (Ed.), Information theory (pp. 162-174). London: Butterworth.

Goldman-Eisler, F. (1961b). A comparative study of two hesitation phenomena. Language and Speech, 4, 18-26.

Goldman-Eisler, F. (1968). Psycholinguistics: Experiments in spontaneous speech. New York: Academic.

Greene, J. O. (1984a). Evaluating cognitive explanations of communicative phenomena. Quarterly Journal of Speech, 70, 241-254.

Greene, J. O. (1984b). Speech preparation processes and verbal fluency. Human Communication Research, 11, 61-84.

Greene, J. O. (1984c). A cognitive approach to human communication: An action assembly theory. Communication Monographs, 51, 289-306.

Greene, J. O. (1987). Cognitive processes: Methods for probing the black box. In C. H. Tardy (Ed.), A Handbook for the study of human interaction (pp. 37-66). Norwood, NJ: Ablex.

Greene, J. O., & Cappella, J. N. (1986). Cognition and talk: The relationship of semantic units to temporal patterns of fluency in spontaneous speech. Language and Speech, 29(2), 141-157.

Greene, J. O., & Lindsey, A. E. (1989). Encoding processes in the production of multiple-goal messages. Human Communication Research, 16(1), 120-140.

Greene, J. O., Smith, S. W., & Lindsey, A. E. (1990). Memory representations of compliance-gaining strategies and tactics. Human Communication Research, 17(2), 195-231.

Greene, J. O., Smith, S. W., Smith, R. C., & Cashion, J. L. (1987). The sound of one mind working: Memory retrieval and response preparation as components of pausing in spontaneous speech. In M. L. McLaughlin (Ed.), Communication Yearbook 10 (pp. 241-258). Newbury Park, CA: Sage.

Hamilton, M. A., & Hunter, J. E. (1985). Analyzing utterances as the observational unit. Human Communication Research, 12(2), 285-294.

Hunter, J. E. (1988). Failure of the social desirability response set hypothesis . Human Communication Research, 15(1), 162-168.

Hunter, J. E., & Boster, F. J. (1978). Compliance gaining message selection experiments: Some evidence for an empathy model. Paper presented at the annual convention of the Speech Communication Association, Minneapolis, MN.

Kearney, P. , & Plax, T. G. (1987). Situational and individual determinants of teachers' reported use of behavior alteration techniques. Human Communication Research, 14(2), 145-166.

Kearney, P. , Plax, T. G., Sorenson, G. S., Smith, V. R. (1988). Experienced and prospective teachers' selections of compliance-gaining messages for "common" student misbehaviors. Communication Education, 37(2), 150-164.

Levin, H., Silverman, I., & Ford, B. L. (1967). Hesitation in children's speech during explanation and description. Journal of Verbal Learning and Verbal Behavior, 6, 560-564.

Marwell, G., & Schmidt, D. R. (1967). Dimensions of compliance-gaining behavior: An empirical analysis. Sociometry, 30, 350-364.

McQuillen, J. S. (1986). The development of listener-adapted compliance-resisting strategies. Human Communication Research, 12(3), 359-375.

Miller, G., Boster, F., Roloff, M., & Seibold, D. (1977). Compliance-gaining message strategies: A typology and some findings concerning the effects of situational differences. Communication Yearbook 2 (pp. 29-48). New Brunswick, NJ: Transaction.

Neuliep, J. W. (1987). The influence of Theory X and Theory Y management styles on the selection of compliance-gaining strategies. Communication Research Reports, 4(1), 14-19.

Neuliep, J. W., & Mattson, M. (1990). The use of deception as a compliance-gaining strategy. Human Communication Research, 16(3), 409-421.

O'Hair, D., O'Hair, M. J., Southward, G., Morris, J., Krayer, K. (1987). Physician communication and patient compliance. Journal of Compliance in Health Care, 2(2), 125-129.

Reardon, K. K. (1981). Persuasion: Theory and context. Beverly Hills, CA: Sage.

Reynolds, A., & Paivio, A. (1968). Cognitive and emotional determinants of speech. Canadian Journal of Psychology, 22, 164-175.

Richard, B. A., & Dodge, K. A. (1982). Social maladjustment and problem solving among school-age children. Journal of Consulting and Clinical Psychology, 50, 226-233.

Roiger, J. F., & Hellweg, S. A. (1986). Influence and compliance-gaining strategies in interpersonal settings and control and influence strategies in organizational settings: A synthesis of the literature. Paper presented at the annual conference of the Western Speech Communication Association, Tucson, AZ.

Rule, B. G., Bisanz, G. L., & Kohn, M. (1985). Anatomy of a persuasion schema: Targets, goals and strategies. Journal of Personality and Social Psychology, 48, 1127-1140.

Sabin, E. J., Clemmer, E. J., O'Connell, D. C., & Kowal, S. (1979). A pausological approach to speech development. In A. W. Siegman & S. F. Feldstein (Eds.), Of speech and time: Temporal patterns in interpersonal contexts (pp. 35-55). Hillsdale, NJ: Erlbaum.

Schank, R., & Abelson, R. (1977). Scripts, plans, goals and understanding. Hillsdale, NJ: Erlbaum.

Seibold, D. R., Cantrill, J. G., & Meyers, R. A. (1985). Communication and interpersonal influence. In M. L. Knapp & G. R. Miller (Eds.), Handbook of interpersonal communication. Beverly Hills, CA: Sage.

Siegman, A. W. (1978). The telltale voice: Nonverbal messages of verbal communication. In A. W. Siegman & S. Feldstein (Eds.), Nonverbal behavior and communication (pp. 183-243). Hillsdale, NJ: Erlbaum.

Siegman, A. W. (1979). Cognition and hesitation in speech. In A. W. Siegman & S. Feldstein (Eds.), Of speech and time (pp. 151-178). Hillsdale, NJ: Erlbaum.

Siegman, A. W., & Pope, B. (1965). Effects of question specificity and anxiety-producing messages on verbal fluency in the initial interview. Journal of Personality and Social Psychology, 2, 522-530.

Siegman, A. W., & Pope, B. (1966). Ambiguity and verbal fluency in the TAT. Journal of Consulting Psychology, 30, 239-245.

Sillars, A. L.(1980). The stranger and the spouse as target persons for compliance-gaining strategies: A subjective expected utility model. Human Communication Research, 6, 265-279.

Smith, M. J. (1982). Cognitive schemata and persuasive communication: Toward a contingency rules theory. In M. Burgoon (Ed.), Communication Yearbook 6 (pp. 330-362). Beverly Hills, CA: Sage.

Smith, M. J. (1984). Contingency rules theory, context, and compliance behaviors. Human Communication Research, 10(4), 461-488.

Waldron, V. R., & Cegala, D. J. (1992). Assessing conversational cognition levels of cognitive theory and associated methodological requirements. Human Communication Research, 18(4), 599-622.

Wilson, S. R. (1990). Development and test of a cognitive rules model of interaction goals. Communication Monographs, 57(2), 81-103.

Wilson, S. R., Cruz, M. G., & Kang, K. H. (1992). Is it always a matter of perspective? Construct differentiation and variability in attributions about compliance-gaining. Communication Monographs, 59(4), 350-367.

Wiseman, R. L., & Schenk-Hamlin, W. (1981). A multidimensional scaling validation of an inductively-derived set of compliance-gaining strategies. Communication Monographs, 48, 251-270.

Yuda, A. (1982). Toward a taxonomy of compliance-gaining situations: An application of multidimensional scaling. Journal of Child Development, 18, 16-26.

Response from Leonard J. Shedletsky

Is it contradictory to claim that activation does not place demands on cognitive processing, but that it does take time? Perhaps "cognitive processing" is being used in a special way here, possibly to mean effortful, conscious processing. In any event, if activation takes time and if assembly takes time, we do not know from this study just what underlying processes take the time. It is difficult to know how to interpret latencies, hesitations and message number. Perhaps more significantly, it is not clear that latencies, frequencies, recency, or message number point to procedural records. Finally, as always, can "what would you do if" data generalize to "what we do do?"

Homepages: Communication Resources (Aitken) | Communicating (Berko) | Intimate Communication | Intrapersonal | Public Speaker/Public Listener (Berko) | "Speech Communication Teacher" (NCA)