VOCAL SIGNS OF CONFIDENCE Charles E. Kimble Steven D. Seidel
ABSTRACT: This study examined whether two paralinguistic variables, vocal Ioud-
ness and response latency, were associated with confidence in answers to trivia questions. Audience presence and size were manipulated and subjects' assertiveness was measured. Subjects verbally responded to trivia questions by indicating their choice and how confident they were in each answer. Tapes of these responses were later analyzed for latency of response and Ioudness of speech. As expected, the more confident inclividuals were in their answers, the laster and Iouder they responded. Assertive subjects spoke Iouder. The presence of an audience had no effects on vocal responding. Apparently, the confidence of a speaker can be inferred from the speed and Ioudness of the speaker's responses. Much of the research on the portrayal or display of confidence has been done by presenting different types of speech to listeners and having them rate how confident the speaker appeared to be (e.g., Scherer, London, & Wolf, 1973). In this study, we were interested in whether vocal features of the response reflected the amount of confidence that a speaker had in a particular answer. In other words, we examined the encoding of confidence in a message rather than its decoding. There are several potential ways that a person can show confidence in what he or she is saying. One way would be through verbal presentation (e.g., 'I am positive t h a t . . . ," "1 am absolutely certain t h a t . . . " ) . Howerer, these overt indications of confidence are sometimes distrusted. One study (Erickson, Lind, Johnson, & O'Barr, 1978) conducted in a courtroom setting indicated that overuse of intensifiers (e.g., so, very, absolutely) and qualifiers made the speaker less convincing. The notion of emotional leakage (Ekman & Friesen, 1969), whereby a person's nonverbal actions are presumed to betray otherwise hidden emotions, suggests that observers are more likely to believe nonverbal signs of confidence. This study was a Master's thesis project conducted by the second author under the supervision of the first author. Steven Seidel is now at Miami University. Requests for reprints should be addressedto Charles E. Kimble, Department of Psychology, University of Dayton, Dayton, OH 45469-1430. Ioumal of Nonverbal Behavior 15(2], Summer 1991 © 1991 Human Sciences Press. Inc.
99
100 JOURNALOF NONVERBALBEHAVIOR What are the likely nonverbal signs of confidence? We contend that these signs are any actions that indicate composure or the lack of nervousness or anxiety. Some actions which might indicate confidence are few self-touches or self-adapters (LeCompte, 1981), few speech disturbances (Cook, 1969), Iower pitched tone of voice (Apple, Streeter, & Krauss, 1979; Streeter, Krauss, Geiler, Olson, & Apple, 1977), fastet speech rate (Miller, Maruyama, Beaber, & Valone, 1976), rauch eye contact while speaking (Exline, Ellyson, & Long, 1975; Hemsley & Doob, 1978), little blinking (DePaulo, Stone, & Lassiter, 1985), Iouder speech (Kimble & Musgrove, 1988), and more Duchenne smiles (Ekman, Davidson, & Friesen, 1990), an unconflicted type of smile. In Scherer's et al. (1973) decoding study, listeners associated Iouder speech, higher pitch, shorter pauses, and laster speech rate with texts read in a "confident" manner. Also, MacLachlan, Czepiel, and Laßarbera (1979) found that laster responses to questions about factual information were indicative of greater confidence in an individual's answer. In this study, we focused on the speed and Ioudness of vocal responses as indicators of confidence. Social facilitation research (Bond & Titus, 1983; Latane, 1981; Zajonc, 1965) led us to expect that the presence of others would also affect the paralinguistic aspects of subjects' responses. For example, drive theory (Zajonc, 1965) stares that for well-learned behavior, the presence of an audience increases the performer's drive or arousal level. This increase in arousal level energizes dominant responses, making them Iouder and faster. For confident answers, it was expected that the presence of an audience would make vocal responses Iouder and laster; for unconfident answers, an audience should have the opposite effect. So the presence of an audience should have magnified the differences between confident and unconfident responses, There is also evidence that assertive people talk rnore and Iouder when they are engaged in an argument (Kimble & Musgrove, 1988). So we investigated how trait assertiveness affects vocal features of speech and how it might interact with confidence and audience presence to affect these speech characteristics. Method Overview
Fifty male and 51 female subjects responded aloud to 10 trivia questions presented on a computer monitor in a small room. No, one, or two experimenters were present in the room for three groups of subjects. Their responses were recorded so that latency and Ioudness could be assessed
101 CHARLESE. KIMBLE,STEVEND. SEIDEL later. Subjects rated their confidence on each item on a I to 7 scale which appeared on the screen after each response. These confidence ratings were used to group questions into five high confidence and five low confidence questions for each subject in the later analyses. Subjects completed the Rathus (1973) Assertiveness Schedule after the quiz.
Procedure Upon arrival, each subject was greeted by the male experimenter and introduced to the female assistant in conditions where there were to be two experimenters observing. The experimenter led the subject to a sound-isolated cubicle and seated hirn/her in front of a computer. He explained that the study was on responses to questions on the computer screen and that it was necessary to record his/her responses on tape. The subject was asked to place a clip-on lavelier microphone (Shure model SM11-CN) on his/her shirt. The subject was informed that the single vocal response (A, B, or C) was important in the experiment and that the subject should refrain from "talking out" the answers, laughing, or making other noises. The experimenter gave instructions about the task and activated the Marantz model PMD201 tape recorder. Depending on the audience condition to which the subject was randomly assigned, the experimenter and the assistant either left the cubicle or stayed. When the experimenter was present, he stood behind and to the side of the subject and made "fake" markings in a notebook. The assistant stood to the other side and did the same thing in the two-person audience condition. The task of each subject was to give an oral answer to each of 10 trivia questions presented on the computer monitor. They were instructed to respond "A", "B", or "C" to indicate their answer to questions such as "Who developed the first polio vaccine?" with response alternatives "A. Salk B. Samuels C. Simpson." The questions were selected from a pool of 30 items pretested with 70 respondents to represent a wide range of difficulty. When each item appeared on the monitor, the computer generated a tone which served as a starting marker for measuring latency from the audiotape later. After answering each question, subjects indicated their confidence in their answer by responding to an explicit 7-point scale on the computer screen with 1 indicating no confidence to 7 indicating complete certainty. Subjects responded by depressing a number key from 1 to 7, and the computer recorded their response. After the trivia quiz, subjects completed the Rathus' (1973) Assertiveness Schedule. This survey asks subjects to indicate how assertively they characteristically act in 30 situations. The possible fange of scores is - 9 0 to +90 with a typical median of 10, which was used to divide subjects into
102 IOURNAL OF NONVERBAL BEHAVIOR
TABLE 1 Mean Response Times in Seconds and Transformed Latencies as a Function of Confidence and Audience Size
Audience Size
Confidence Low High Seconds Latencies S e c o n d s Latencies
No Experimenter
3.39
M = 2.95 SD= .27
2.55
M = 3.92 SD= .20
One Experimenter
3.03
M = 3.30 SD= .34
2.28
M = 4.38 SD= .26
Two Experimenters
3.11
M = 3.22 SD= .12
2.35
M = 4.25 SD= .31
Overall Ms
3.16
3.16"
2.39
4.19"
Note. The latencies reported hefe are reciprocal transformations. Therefore, higher latency values reflect laster responses. Cell Ns were 34 subjects each except the Two Experimenters conditions had only 33 subjects. *p < .0001
low and high assertive categories in this study. Finally, subjects were told the purpose of the study, thanked for their participation, and excused. Vocal loudness was measured by processing the tape-recorded responses through a Commodore 64 system and program with a special interface circuit which did an analog to digital conversion on the audio signal. The signal was integrated, amplified, and digitized to amplitude output values (ranging from 0 to 255) 42.5 times a second. Only output values greater than 10 on the 0-255 scale were cumulated for the amplitude averages so that system noise (which sometimes produced signals of I or 2) and silent periods would not affect the averages. These average loudness scale values are linearly related to signal amplitudes, not Iogarithmically related like decibel values. There were 10 loudness readings for each subject, one for each item. The volume level on the tape recorder and the parameters of the interface circuit were constant for all subjects. Results The confidence measures for each subject were sorted and the five Iowest ratings for each individual were classified as Low Confidence items and the
103 CHARLES E. KIMBLE, STEVEN D. SEIDEL
TABLE 2 Mean Loudness as a Function of Confidence and Audience Size
Low
Confidence High
No Experimenter
M = 57.68 SD= 5.38
M = 62.58 SD= 7.20
60.13
One Experimenter
M = 54.98 SD= 4.27
M = 57.83 SD= 3.40
56.90
Two Experimenters
M = 61.84 SD= 6.33
M = 65.32 SD= 4.91
63.58
Audience Size
Overall Ms
58.47* N=101
Overall Ms
61.87* N=101
Note. Larger scores indicate Iouder speech. Cell Ns for the three Audience Size conditions were 34, 34, and 33, the same as Table 1. *p < .0001.
five highest ratings were classified as High Confidence items. The average Ioudness and latency scores for the five Iow and high confidence items for each individual were used as dependent measures in the analyses. Assertiveness scores were also dichotomized so that we could examine the main effects and interactions involving assertiveness in the analyses of variance. A 2 by 2 by 3 one within-subjects, two between-subjects mixed design was used in the analyses. Confidence was the within-subjects variable, and Assertiveness and Audience Size (0, 1, 2) were the betweensubjects variables. Participants answered questions of which they were more confident faster than those of which they were doubtful, F(1, 98) = 171.51, p < ,0001. The latency analyses were conducted on reciprocal transformations which normalized the scores, but average response times in seconds are also presented in Table 1. They also responded more Ioudly on the high confidence items, F(1, 98) = 26.29, p < .0001. Audience presence had no effects. Overall, assertive individuals spoke Iouder than unassertive ones, Low Assertive M = 57.19, High Assertive M = 63.10, F(1, 95) = 4.91, p < .05. No interactions were significant, but the Assertiveness by Confidence interaction on the latency measure indicating that assertive
104 IOURNALOF NONVERBALBEHAVIOR individuals spoke fastest on the High Confidence items approached significance, Interaction F(1, 95) = 3.62, p < .06. Additional analyses revealed that there were no sex differences in latency; F(1, 95) = 1.13; Female transformed latency M = 3.64; Male transformed latency M = 3.70; or Ioudness; F(1, 95) = 1.90; Female Loudness M = 59.39; Male Loudness M = 61.12. Sex was not involved in any significant interactions either.
Discussion When people were confident of what they were saying, that confidence was reflected in Iouder speech and laster response times. This study is the most direct test of how people display confidence nonverbally. While other signs of composure such as rnuch eye contact or iower-pitched voice or an unaffected srnile may accompany feelings of confidence, none of them have been tested and verified so directly. Bold, assertive people spoke Iouder regardless of their confidence on a topic. In addition, assertive people tended to respond laster when they were confident of their answers. So trait assertiveness seems to be related to Ioudness and latency the same way that transitory confidence is. The presence of onlookers did not affect these vocal features of speech. It is possible that if the audience had been bigger or more centrally Iocated so that the subjects were saying their answers directly to them, the presence of the audience would have affected the subjects in ways implied by social facilitation theory. However, if the audience had been so visually prominent to the subjects, they may have distracted subjects and obscured the effects of confidence on their responses. Eren though audience presence did not have the effect predicted by Zajonc's drive theory of social facilitation, the main results are consistent with the theory. That is, confident responses seem to be dominant responses which are given more energetically (Iouder and laster). Perhaps the high confidence items were rated that way because subjects recognized that they had a strong dominant (habitual) response to those items. Whether subjects made an independent cognitive judgment of the correctness of their answers or inferred their confidence from their vocal responses, it is clear that confidence was strongly associated with laster, Iouder responses. While many people consider confidence to be broader than confidence about the correctness of a single answer, it is our position that confidence is most appropriately studied in relation to such specific situations.
105 CHARLES E. KIMBLE, STEVEN D. SEIDEL
In this study, confidence was studied in relation to audience presence as weil as to the specific questions. Audience presence did not affect the vocal signs of confidence as strongly as the specific items did. Assertive people may be "confident" people in the broader sense of confidence, but trait assertiveness was less related to the vocal signs of Ioudness and latency than the specific item confidence measures were. Since confidence is related positively to performance in many situations, the abitity of observers to read these signs of confidence may be useful in many everyday interactions. References Apple, W., Streeter, L. A., & Krauss, R. B. (1979). Effects of pitch and speech rate on personal attributions. ]ournal of Personality and Social Psychology, 37, 715-727. Bond, C. E., & Titus, L. J. (1983). Social facilitation: A meta-analysis of 241 studies. PsychoIogical Bulletin, 94, 265-292. Cook, M. (1969). Anxiety, speech disturbance, and speech rate. British ]ournal of Social and Clinical Psychology, 8, 13-21. DePaulo, B. M., Stone, J. I., & Lassiter, D. G. (1985). Telling ingratiating lies: Effectsof target sex and target attractiveness on verbal and nonverbal deceptive success. Iournal of Personality and Social Psychology, 48, 1191-1203. Ekman, P., Davidson, R. J., & Friesen, W. V. (1990). The Duchenne smile: Emotional expression and brain physiology II. ]oumal of Personality and Social Psychology, 58, 342-353. Ekman, P., & Friesen, W. V. (1969). Nonverbal leakage and cues to deception. Psychiatry, 32, 88-108. Erickson, B., Lind, E. A., Johnson, B. C., & O'Barr, W. M. (1978). Speech style and impression formation in a court setting: The effects of "powerful" and "powerless" speech. ]ournal of Experimental Social Psychology, 14, 266-279. Exline, R. V., Ellyson, S. L., & Long, B. D. (1975). Visual behavior as an aspect of power role relationships. In P. Pliner, L. Krames, & T. AIIoway (Eds.), Advances in the study of communication and affect (Vol. 2). New York: Plenum. Hemsley, G. D., & Doob, A. N. (1978). The effect of Iooking behavior on perceptions of a communicator's credibility. ]oumal of Applied Social Psychology, 8, 135-144. Kimble, C. E., & Musgrove, J. I. (1988). Dominance in arguing, mixed-sex dyads: Visual dominance pattems, talking time, and speech Ioudness. ]ournal of Research in Personality, 22, 1-16. Latane, B. (1981). The psychology of social impact. American Psychologist, 36, 343-356. LeCompte, W. A. (1981). The ecology of anxiety: Situational stressand rate of self-stimulation in Turkey. ]ourna/ of Personality and Social Psychology, 40, 712-721. MacLachlan, J., Czepiel, J., & LaBarbera, P. (1979). Implementation of response latency measures. ]ournal of Marketing Research, 16, 573-577. Miller, N., Maruyama, G., Beaber, R. J., & Vallone, K. (1976). Speed of speech and persuasion. ]oumal of Personality and Social Psychology, 34, 615-624. Rathus, S. (1973). A 30-itern schedule for assessingassertive behavior. Behavior Therapy, 4, 398-406. Scherer, K. R., London, H., & Wolf, J. I. (1973). The voice of confidence: Paralinguistic cues and audience evaluation. ]ournal of Research in Personality, 7, 31-44. Streeter, L A., Krauss, R. M., Geiler, V., Olson, C., & Apple, W. (1977). Pitch changes during attempted deception. ]ournal of Personality and Social Psychology, 35, 345-350. Zajonc, R. B. (1965). Social facilitation. Science, 149, 269-274~ , ' . . , r