Abstract
Although previous research has linked students' expected grades to numerous pedagogical variables, this factor has been all but ignored by instructional communication scholars. In the present study, 315 undergraduates were presented with grading scenarios representing differing combinations of course rigor, teacher immediacy, and student self-efficacy. For each scenario, students estimated the numerical grades they would expect to receive on basic course speaking assignments. Componential analysis (Kenny, 1994) was used to decompose expected speech scores into perceiver, target, and the unique reactions of individual respondents, which respectively accounted for 20.4%, 60.2%, and 19.4% of the variable of interest. Within the target effect, student perceptions of how they had prepared and performed in classroom speaking situations represented the largest single variance subcomponent (31.4%), followed by perceptions of course rigor (23.3%) and teacher immediacy (5.5%). Recommendations for future research and instructional practice are advanced.
Notes
1. When using componential analysis, it is customary to combine participants into study groups of equal size before computing variance estimates. Subsequently, the variance attributable to the perceiver, target, and interaction effects are then averaged across these groups. This provides the opportunity to conduct significance testing on the model at each stage and to examine the range of variances for components. In the current study, a maximum of five components would be analyzed, namely the perceiver effect, three target effect subcomponents, and the uniqueness effect. In order to determine the minimum size for each study group, these components were treated as multiple predictors in a linear equation explaining a single dependent variable. It was further presumed that the overall effect size for this model would be large. Based upon these assumptions, a power analysis (Faul, Erdfelder, Lang, & Buchner, Citation2007) indicated the minimum size of each study group should be 63. Fortuitously, the total number of study participants (315) is a multiple of 63. Therefore, five equal-sized groups of that number were used in this study.
2. The following formulae were used in computing variance estimates for the model. In each case, the terms “n” and “r” refer to Perceivers and Targets, respectively.
The Mean Square for Perceivers (MSP) was estimated from:
The Mean Square for Targets (MST) was estimated from:
The Mean Square for the Perceiver × Target interaction (MSI) was estimated from:
Variance estimates for Perceivers, Targets, and the Interaction term were computed as random effects:
The formulae listed above and in the body of the paper are used strictly in componential analysis for half-block designs, i..e., when perceivers rate a common set of targets but not other perceivers. Computation of model components and variance estimates for block and round-robin designs are considerably more complex. For a more detailed discussion of these issues, consult Kenny (Citation1994; Kenny, Kashy, et al., Citation2006).
3. Statistical significance for the Perceiver and Target variances was based on the following tests:
A significant F-test for any component indicates it is needed as part of the model explaining the dependent variable. On the other hand, the proportion of variance attributable to any component indicated its relative importance.