Abstract
The two central constructs of family communication patterns theory—conversation orientation and conformity orientation—have been studied in their relation to a host of outcomes. In measuring these orientations, researchers often acknowledge that the empirical relationship between the two constructs tends to be negative. A meta-analytic review was conducted (k = 32, N = 8,171) to assess the magnitude of this relationship. In addition, respondent type (i.e., child versus parent respondents) was tested as a moderator. The weighted mean effect size was –.33 (95% CI [–.35, –.31]). The findings gave preliminary evidence that the relationship was weaker for parent respondents compared to child respondents. The findings hold implications for future work, such as by suggesting certain analytic decisions.
Notes
[1] Preliminary checks were conducted in order to determine whether any noticeable publication biases existed. First, the correlation between sample size and effect size was examined; this technique can be used to gauge whether there was a bias against studies with nonsignificant relationships among the variables of interest (Levine, Asada, & Carpenter, Citation2009). This correlation was not substantial, r = –.12, p = .507. Second, Duval and Tweedie’s (Citation2000) trim and fill method was implemented in order to assess whether the inclusion of (theoretically) missing studies would substantially alter the relationship between the variables of interest. The adjusted value for all studies was r = –.28, which suggests that the effect would not be meaningfully reduced after adjusting for missing studies. Third, the relationship between publication year and effect size was examined in order to see whether the strength of the relationship is decreasing over time. This relationship was neither negative nor significant, r = .26, p = .143. Fourth, potential author effects were examined after it was discovered that of the 28 studies found, eight (28.6%) were carried out by Prof. Andrew M. Ledbetter and Prof. Paul Schrodt, either individually or as coauthors. These works were dummy coded (Ledbetter and/or Schrodt = 1, all others = 0). The observed point-biserial correlation was negligible, r = .03, p = .857.