THE KELLEY MODEL AS AN ANALYSIS OF VARIANCE ANALOGY - HOW FAR CAN IT BE TAKEN

Kelley's (1967, 1973) suggestion that the attribution process can be compared to the statistical procedure of analysis of variance (ANOVA) was investigated. More specifically, I examined whether person and entity attributions vary as a function of different degrees of covariation of the effect with the respective cause, and whether certainty about the causal judgments is influenced by the number of observations. Unlike other studies in this field, complete rather than incomplete covariation information relevant for the ANOVA was presented. In addition, the numerical "raw" data rather than "prepackaged" (summary) data were given. Also, different degrees of covariation of the effect with the possible causes rather than merely two levels of covariation (e.g., high vs low consensus) were presented to the subjects. The comparison of subjects' attributions with the statistical parameters of the ANOVA revealed that causes were rated as increasingly important for an effect inasmuch as the variation attributable to the respective cause increased and the variation due to the alternative cause decreased. This relationship closely paralleled the effect-size parameter (η2). In addition, person attributions were more resistant to new information than entity attributions. Certainty ratings were uninfluenced by the manipulations. © 1992.