Abstract
The present research addresses the long-standing question of the similarity between the perception of individual “count” objects (“three cats”), and continuous “mass” objects (“some sand”). Such distinctions have been important in studies of higher cognition (e.g., linguistics) and basic visual representation (e.g., ensemble features). Can mass and count be represented in a shared format? A prominent theory by Chierchia (1998; Natural Language Semantics 6, 339-405) proposes that mass and count representations share a similar format as ensembles of discrete individuals. This view, in concert with certain theories of mid-level vision (Pylyshyn, 1989; Cognition 32, 65-97), takes objects to be the foundation upon which higher cognition is built. Our study tested whether number and area perception share a visual format through sets of individuals or through noisy continuous approximations. In Experiment 1, 20 adults were presented with two sets of dots for 150 ms on each trial and were asked to evaluate which set was greater in number (see Figure 1). An additional 20 subjects saw, for 150 ms, images of 2-D blobs made up of two colors, and had to judge which color had a greater area (see Figure 2). A classic psychometric curve derived from a Gaussian subtraction model was highly correlated with the observed data for each condition (R2 > 0.90), and suggests that the representations of number and area share a similar format through noisy approximations. Therefore, the underlying representation, for vision, of both count and mass objects is not a set of discrete individuals, but a continuous Gaussian variable (or a Bayesian probability distribution). These findings speak not only to the psychophysics of area and number discrimination in adults, but also contribute to long-standing debates in lexical semantics, as well as to the relationship between vision and language and the extraction of ensemble features.