Category-specific naming and the 'visual' characteristics of line drawn stimuli

It has been argued that greater intra-category structural similarity for living things may make them more difficult to recognize and name (e.g. Humphreys et al., 1988). Nevertheless, the precise meaning and quantification of 'structural similarity' remain unclear. We developed three new visual measures derived from the Snodgrass and Vanderwart (1980) corpus and examined their relationship with picture naming in a speeded presentation paradigm. The three measures were: the proportion of black pixels (PB); the degree of pixel overlap within subcategories using Euclidean Overlap (EO); and the degree of consistency in inter-pixel distribution across each picture (IPC). Within-category EO was greater for nonliving than living things, indicating less within-category visual overlap for living things. Finally, EO correlated significantly with error rates (PB and IPC did not). These findings contradict existing notions that line drawings of living things have greater visual similarity than nonliving things.