The Automatic Social Categorization Test: Validating a New Measure

Abstract

Categorization is an essential process in making sense of the world. However, it simplifies perception in a way that distorts reality to some degree. For instance, the color spectrum is a continuum of wavelengths; but colors are grouped into linguistically defined categories (Berlin & Kay, 1969). Although people can tell the difference between different hues within a category, they are better at discriminating colors from different categories than colors from the same category, even if they differ by the same degree in terms of physical wavelength. Harnad (2003) describes this process as “'warping' perceived similarities and differences so as to compress some things into the same category and separate others into different categories.” This dissertation describes the Automatic Social Categorization Test (ASCAT), a new way to measure the extent to which this process occurs for social categories (e.g., gender, age, race) by examining confusion errors among stimuli that vary continuously between two categories. We operationally define automatic categorical perception as the degree to which confusion rates are higher for a pair of stimuli within the same subjective category than for a pair belonging to different subjective categories. We first present a series of validation studies with non-social stimuli, showing that automatic categorical perception is stronger for categorical stimuli (i.e., with a gap in the middle) than for continuous stimuli. Next, we extend our method to social stimuli (i.e., morphed faces varying in racial composition), demonstrating a categorical trend that is reduced when social information is removed (i.e., when the faces are scrambled). The degree of automatic social categorization also depends on which social category is presented; faces varying in gender and race are perceived more categorically than faces varying in age. Finally, we show that there are reliable individual differences in automatic race categorization that cannot be attributed to working memory ability or differential fatigue effects. Performance on the ASCAT is also relatively stable across multiple testing sessions. These results suggest that the ASCAT is a useful new tool for research on the correlates and consequences of automatic social categorization.