Abstract
This study examines the extent to which holism in visual perception can be revealed by way of the presence or absence of crowding. Martelli, Majaj, and Pelli (2005) used crowding to propose an operational definition for holism. Specifically, they posited that holistic perception of an object is implicated if that object can be identified when the entire object is presented within an isolation field (defined as an area proportional to one-half eccentricity). Conversely, parts-based processing is implicated if identification is impaired when the entire object is within an isolation field, with an attenuation or elimination of that impairment when each part of the object is isolated by critical spacing. Martelli et al. found evidence of crowding—increases in threshold for contrast as a function of eccentricity for faces and words—suggesting that foveally-presented objects are processed holistically, and peripherally-presented objects are processed by parts. We consider this operational definition from the perspective of general recognition theory (GRT, Ashby & Townsend, 1986). GRT provides theoretical characterizations of perceptual and decisional independence and separability, with violations of independence and separability allowing for multiple characterizations of holism. In this experiment, we use accuracy of identification responses to link Martelli et al.'s operational definition to the definitions of holism provided by GRT. A set of schematic faces were presented under conditions modeled on those used by Martelli et al. Results show that we are able to use these simple schematic figures to replicate the patterns documented by Martelli et al.: specifically, evidence for the benefit of a facial context in foveal presentation, and impairment in peripheral presentation. At the same time, however, our GRT analyses suggest important disparities between the current operational and theoretical definitions of holistic processing.