Contrast and assimilation effects: A meta-analytic review
Rudolph, Amy Spence
Gaugler, Barbara B.
Doctor of Philosophy
The effects of contrast and assimilation in person and sensory perception tasks were reviewed and examined within and across psychosocial and psychophysical research domains. A wide range of effect sizes that varied in both magnitude and direction were found. Meta-analysis of 57 studies containing 172 effect sizes across the total sample revealed a mean corrected d and variance of $-.21$ and 1.06, respectively, indicating contrast. The mean corrected d for studies in the psychosocial domain was $-.22,$ and the mean corrected d for psychophysical studies showed little effect, $-.04.$ Effect sizes were corrected for sampling error and unreliability in the dependent measure, accounting for little variance in study outcome. Sufficient variance remained both within and across domains after correcting for statistical artifacts to justify the search for moderator variables. Across the total sample, effect size was moderated by type of rater and stimulus presentation order. Serial presentation of context and target stimuli resulted in contrast effects, and simultaneous presentation resulted in assimilation effects. Graduate students produced ratings with the greatest magnitude of contrast effects, followed by psychology undergraduates, and unspecified undergraduates. The ratings of nonprofessional adult subjects showed assimilation. Contrast effects resulted when studies were published in perceptual psychology journals, when stimuli were presented simultaneously, and when the degree of discrepancy was high. Type of rater did not moderate effect size within the psychophysical domain. Within the psychosocial area, contrast effects were seen when the study was unpublished or found in the education literature, when the context and target stimuli were presented in similar forms, when the research was conducted in an applied lab setting, when stimuli were presented serially, when subjects were instructed to form an impression or evaluate performance, and when subjects actively rated the contextual stimuli. Assimilation effects were found in this domain when nonprofessional, unspecified adults served as subjects, when subjects were familiar with the stimuli, and when subjects were not trained in the rating process. The degree of discrepancy between the context and target stimuli, the time span between observation and rating, the presence of a distracter task, and subjects' interaction with other did not moderate effect size within the psychosocial domain. The findings suggest that, although contrast and assimilation may be pervasive, many variables moderate the magnitude and direction of the effects. In addition, integration across psychophysical and psychosocial domains may not be appropriate. Limitations of meta-analysis, implications, and suggestions for future research are discussed.