Perspective Taking and Communication Satisfaction in Coworker Dyads

When individuals rate themselves on a perspective-taking scale, the rating represents self-reported tendency to understand the other person’s point of view. According to Davis (1994), “the items of this scale explicitly ask about the respondent’s likelihood of engaging in attempts to entertain the point of view of other people (process); they do not inquire about the purported outcome of that process (e.g., accuracy or social insight)” (p. 57). One way to assess the actual ability of perspective taking is to check empathic accuracy. A common procedure used in assessing children’s perspective-taking ability involves expert judges evaluating children’s performance on standard tasks in the spatial, cognitive, and affective domains, although conceptual and methodological problems and difficulties have been raised (Davis, 1994). Another way is to have either uninvolved observers or the actual interactants (e.g., conversation partners involved in interaction with the target individual) assess the target individual’s perspective-taking ability or tendency.

This research examines individuals’ self-rating of their perspective taking, their coworkers’ rating of the individuals’ perspective taking, and the discrepancy between the two. Since it is not our goal to assess how biased or accurate individuals are in their assessment of self and others, this research does not use expert judges’ evaluation of individuals’ performance on standard tasks in multiple domains.

This procedure of including about five participants from each kindergarten (even from a kindergarten with more than five teachers) and allowing the participants to choose whomever they wanted to evaluate was necessary because of the concerns for voluntary participation and anonymity issues.

Even when robust statistics were not used, the analysis yielded an acceptable fit (NFI = .985, CFI = 1.00, GFI = .991, AGFI = .973).

Within-group agreement (r _WG ) (James et al. 1984) was calculated for each individual because each individual received multiple ratings from her coworkers. The average r _WG was .73. For 74.5% of the individuals, r _WG was equal to or greater than .70. The intraclass correlation (Kashy and Kenny 2000) was .25, p < .001. The reliability of the individual mean calculated by Hierarchical Linear Modeling (HLM) was .50. In considering that individuals received only one to four ratings, coworker ratings of several individuals showing the low reliability estimated by HLM are not abnormal (for comments on reliability in HLM, see Pollack 1998).

The mean of r _WG across individuals was .55. For 55.1% of the individuals, r _WG was greater than .70. The intraclass correlation was .19, p < .001. The reliability of the individual mean calculated by HLM was .43.

Even when robust statistics were not used, CFAs showed an acceptable fit for a unidimensional solution with most of the fit indexes above .92.

For more detailed explanation on the role of the group-mean centering in disentangling within- and between-group variances in the Level-1 intercept and in disentangling within- and between-group effects in the Level-1 slope, see Hofmann et al. (2000), Park (2008), Raudenbush (1989a, b), and Raudenbush and Bryk (2002). These authors use models with individuals for the Level-1 and groups for the Level-2, while this study uses models with coworker ratings for the Level-1, individuals for the Level-2, and groups for the Level-3. Thus, this study uses group-mean centering to disentangle within-individual and between-individual variances in the Level-1 intercept and to disentangle within-individual and between-individual effects in the Level-1 slope.

Other group level variables such as interaction length of the work group, age variation among the group members, size of the group, and group average other-rated perspective taking were examined for this Level-3 model, but none were statistically significant.