Robert Rosenthal’s research while affiliated with Harvard University and other places

Teacher behaviour reflecting their differential expectations was investigated in a context-minimal method, where judges rate extremely brief (10-second) clips of videotaped teacher behaviour, separated into isolated non-verbal and verbal channels (face, body, speech content, tone of voice, etc.). Teachers were recorded when talking about and talking to high- and low-expectancy students. Contrary to recent claims that teacher expectancy effects are negligible and that teachers' differential behaviour is generally appropriate and reality-based, expectancy effects of substantial magnitude were found in this study, especially in affective and non-verbal behaviours. Teachers were rated as showing more negative affect in the non-verbal channels, and as more dogmatic in the non-verbal and transcript channels, when talking about low expectancy compared to high expectancy students. When talking to students and teaching them briefly, facially communicated expectancy differences were found in ratings of negative affect and active teaching behaviour. The findings supported a view of teachers as attempting to compensate low-expectancy students in controllable, direct teaching behaviours, at the same time transmitting (or “leaking”) negative affect in less controllable, mostly non-verbal channels. It was also found that teachers who were more susceptible to biasing information were more negative and showed more intense expectancy effects than unbiased teachers in certain verbal and non-verbal channels.

A bstract Subjects' facial expressions were videotaped without their knowledge while they watched two pleasant and two unpleasant videotaped scenes (spontaneous facial encoding). Later, subjects' voices were audiotaped while describing their reactions to the scenes (vocal encoding). Finally, subjects were videotaped with their knowledge while they posed appropriate facial expressions to the scenes (posed facial encoding). The videotaped expressions were presented for decoding to the same subjects. The vocal material, both the original version and an electronically filtered version, was rated by judges other than the original senders. Results were as follows: (a) accuracy of vocal encoding (measured by ratings of both the filtered and unfiltered versions) was positively related to accuracy of facial encoding; (b) posing increased the accuracy of facial communication, particularly for more pleasant affects and less intense affects; (c) encoding of posed cues was correlated with encoding of spontaneous cues and decoding of posed cues was correlated with decoding of spontaneous cues; (d) correlations, within encoding and decoding, of similar scenes were positive while those among dissimilar scenes were low or negative; (e) while correlations between total encoding and total decoding were positive and low, correlations between encoding and decoding of the same scene were negative; (f) there were sex differences in decoding ability and in the relationships of personality variables with encoding and decoding of facial cues.

A bstract The structure of skill at decoding nonverbal cues was examined for 150 high school students and 95 college students. An overall principal components analysis yielded four factors differing in the complexity of the message (pure versus mixed) and in the relative importance of the video versus the audio modality. Factor 1 (pure video) was defined by accuracy at face and body cues of ordinary (2 second) and very brief exposure length. Factor 2 (mixed video) was defined by accuracy at face and body cues with a “noisy” background. Factor 3 (mixed audio) was defined by accuracy at decoding discrepant cues and “noisy” audio cues. Factor 4 (pure audio) was defined by accuracy at pure tone of voice cues. The overall evidence suggested that despite a nontrivial degree of relationship among all measures of skill at decoding nonverbal cues (Armor's Theta = .62), it would increase our theoretical and empirical precision to conceptualize nonverbal decoding ability as made up of several relatively unrelated subskills.

The analysis of self-other data for the study of person perception accuracy is illustrated and discussed. Length of acquaintance, length of cohabitation, and trait empathy were investigated for their moderating effects on person perception accuracy, defined as the level of self-other agreement. Self-other agreement was computed four ways. A trait-by-trait analysis was performed twice, first using the moderator variable to form subgroups from which self-other correlations within each trait were computed and then using the moderator as a continuous variable in a series of moderated multiple regressions. Next, a profile analysis was performed that isolated two accuracy components, implicit profile accuracy and ideographic accuracy, which were conceptually similar to Cronbach's stereotype accuracy and differential accuracy components. The analyses, taken together, provided a componential and informative (if not comprehensive) analysis regarding accuracy as it is manifest in self-peer agreement data. Sex and cohabitation length moderated accuracy whereas acquaintance length and trait empathy did not.

The validity of a rating paradigm used to measure relative degrees of movement coordination occurring within social interaction (i.e., interactional synchrony) was examined. Untrained judges viewed video clips from 60 dyads, each recorded in two interaction contexts. Two types of video displays were generated. In one condition, judges rated interactional synchrony from standard video. In the mosaic display condition, all features and fine detail that normally would provide a rich array of socially informative data (e.g., facial expressions and twitches) were removed by digitalizing the video image into a mosaic of monocolored blocks. The patterns of synchrony results generated from the two display conditions were virtually identical. Whereas the ratings of synchrony from standard displays could have been influenced by expressed affect and interest, such confounds were unlikely to influence ratings made using the silent mosaic displays.

We asked active psychological researchers to answer a survey regarding the following data-analytic issues: (a) the effect of reliability on Type I and Type II errors, (b) the interpretation of interaction, (c) contrast analysis, and (d) the role of power and effect size in successful replications. Our 551 participants (a 60% response rate) answered 59% of the questions correctly; 46%) accuracy would be expected according to participants' response preferences alone. Accuracy was higher for respondents with higher academic ranks and for questions with “no” as the right answer. It is suggested that although experienced researchers are able to answer difficult but basic data-analytic questions at better than chance levels, there is also a high degree of misunderstanding of some fundamental issues of data analysis.

Students differing in ages and teachers differing in experience were exposed to extremely brief samples of teacher behavior when talking about, and talking to, students for which they held high or low expectations. Judgments of teacher characteristics as well as those of the unseen student with whom the teacher was involved were collected. An expectancy detection effect was found such that when teachers were involved with their high-expectancy student, raters judged the unseen student more positively than when teachers were involved with their low-expectancy student. This detection was facilitated differentially by the teachers’ verbal and nonverbal behavior. Ratings of teacher characteristics showed similar expectancy effects but only for older raters. Findings demonstrate the detectability of teachers’ expectancy-related behavior. We discuss the implications of young students’ detecting teacher expectancies from brief samples of behavior and the educational significance of the observed discrepancies between verbal and nonverbal communications.

This study tested the prediction that 3 variables—self-reported trait relevance, consistency, and observability—would moderate correlations between self-ratings and peer ratings. These predictions received considerable support when the 3 moderators were measured by ranking procedures (i.e., rank ordering traits in terms of their standing on each moderator) and very litle support when the 3 moderators were measured by rating scales (i.e., rating each trait in terms of its standing on each moderator). The advantage of the ranking measure may indicate an advantage for moderators that distinguish among traits across or within individuals (intertrait and intraindividual moderators) as opposed to moderators that distinguish among individuals. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

In this article we present a brief exposure method of assessing teachers' verbal and nonverbal behavior. Highly biased and unbiased teachers were videotaped addressing their classes, and judges rated randomized 10-s clips. Leakage, the transmission of more positive affect in controllable channels while negative affect is given away involuntarily in less controllable channels, was assessed by linear contrast analyses of three channels in a leakage hierachy: transcript of speech content, face, and body. As hypothesized, biased teachers demonstrated systematic and substantial leakage effects in affective variables (factor-based, composite scores reflecting dogmatic behavior and negative affect), whereas unbiased teachers showed no leakage. As predicted, no leakage was found for any group in active teaching behavior, a nonaffective composite variable. Biased and unbiased teachers did not differ in comparisons for each separate channel. These findings are consistent with previous findings on differences between biased and unbiased teachers. (PsycINFO Database Record (c) 2012 APA, all rights reserved)