Abstract
It has long been recognized in the nonverbal communication literature that individual differences in extraversion are expected to be most clearly evident in expressive forms of nonverbal behavior. Although several studies designed to examine the correlation between extraversion and nonverbal behavior have been conducted, most attempts to summarize these studies reflect the inconsistent nature of the research findings. Indeed, correlations range from −.36 to .73. The extant literature summaries are limited to narrative reviews, however, which tend to be highly selective. The purpose of this study was to meta‐analyze the relevant studies. Results indicated that the mean correlation coefficient weighted for sample size was .13 for the entire sample of studies. Although 63% of the variance was attributable to sampling error, the distribution of correlations was heterogeneous. The remaining variance was due to two methodological artifacts: sample size and the number of nonverbal behaviors coded. For example, hierarchical breakdowns by moderator indicated that the mean correlation was .50 for highly focused studies—those in which researchers concentrated on a few behaviors and limited coding to 40 or fewer subjects. A cognitive‐overload model for these findings is offered and corroborated by findings from other research domains. Implications for communication research and theory are discussed. In addition to clarifying the empirical relationship between extraversion and nonverbal behavior, the contributions of this study reside, in part, in its implications for large samples and large numbers of nonverbal behaviors coded in research and for meta‐analyses of such research literatures.
Notes
Betty H. La France (Ph.D., Michigan State University) and Alan D. Heisel (Ed.D., West Virginia University) are Assistant Professors and Michael J. Beatty (Ph.D., Ohio State University) is Professor and Chair, Department of Communication, University of Missouri – St. Louis. The authors would like to thank Frank Boster and the anonymous reviewers for their assistance with this manuscript. Correspondence to Betty H. LaFrance, Department of Communication, University of Missouri – St. Louis, 8001 Natural Bridge Rd., St. Louis, MO 63121‐4499, USA. Email: [email protected].
Authors of some studies (Riggio & Friedman, Citation1983, Citation1986; Riggio, Lippa, & Salinas, Citation1990; Simpson, Gangestad, & Biek, Citation1993) performed exploratory factor analysis to reduce the number of nonverbal behaviors coded into an abbreviated list of variables that were then correlated with personality variables such as extraversion. Unfortunately, this method of data reduction resulted in factors that combined many behaviors across nonverbal codes. For example, after performing a principal components exploratory factor analysis with oblique rotation, Riggio et al. (Citation1990) formed a factor labeled “self focused behaviors” that was comprised of “grooming behavior, self‐touching, and hand‐to‐head contacts” (p. 21). Similarly, Riggio and Friedman (Citation1986) combined “posture shifts, head movements, and the incidence of speech disturbances” to form “body emphasis” (p. 424). Nonverbal variables that were composite variables created by combining behaviors across nonverbal codes were not included in the current meta‐analysis because the effect of extraversion on a particular nonverbal behavior could not be calculated.
Readers may note that the sampling error variance is greater than the observed variance. Hunter and Schmidt (Citation1990) describe that this result occurs as a consequence of second‐order sampling error in meta‐analyses that include a small number of effects. They further note, “Negative estimated variances are not uncommon in statistical estimation” (p. 413).
The number of effects (nonverbal behaviors correlated with extraversion) as reported in each study (presented in ) and the number of nonverbal behaviors coded within each study are different variables. For example, Iizuka (Citation1992) only coded five nonverbal behaviors (gazing at the interviewer, gazing at the interviewer while listening, gazing at the interviewer while speaking, gazing during subjects' talk time, and gazing during interviewers' talk time), but six total effects are reported because an additional variable was calculated using these extant behaviors. Riggio and Friedman (Citation1986) had raters code many nonverbal cues. Although some cues only included one behavior (e.g., smiles) other cues were comprised of multiple behaviors (e.g., “body‐focused” movements were comprised of “hands coming into contact with any part of the body, clothing, or chair” (p. 423). In total, the number of behaviors raters coded was 27, but 12 effects were generated by correlating extraversion with specific nonverbal cues (but see Note 1).