Immediacy and Learning Meta-Analysis 23
Limitations of the Study
There were a number of difficulties encountered in conducting a thorough and specific
meta-analysis. As noted above, some meaningful studies of immediacy and learning were
excluded from the analysis because the type or detail of statistical information they reported was
not sufficient for correlational comparison. However, the general findings of this group of
studies fall within the range of those included in the analysis, so their exclusion probably had
little effect on the overall results of the meta-analysis. In addition to the problems with
questionnaires noted above, the combination of verbal and nonverbal immediacy into a single
measure in some studies precluded specific analysis of separate immediacy types for those
studies. Problems persist in the valid measurement of immediacy and learning: the questionnaire
on verbal immediacy (Gorham, 1988) may have difficulties with content and construct validity
(Robinson & Richmond, 1995), and obtained reliability is seldom reported for the learning loss
measure. Results involving cognitive learning were relatively few and focused primarily on
lower-level learning. There were few experimental studies and heavy reliance on survey-
questionnaires, which minimized causal conclusions.
We averaged obtained effects and corrected for statistical artifacts and research design in
order to compute equivalent effect statistics across the wide variety of reporting styles and
research designs. However, the resulting r’ used in the analyses served only as a best estimate of
actual correlations. Computing accurate effects proved particularly problematic in a few
instances (e.g., Hackman & Walker, 1990) and may have served to underestimate actual
correlations in those few samples.
The most troublesome aspect of this meta-analysis pertained to the persistent finding of
heterogeneity within the samples, even in the 3 x 3 analysis. It is likely that adjusting for low
Convention