Abstract

We report on two experiments investigating the nature of conceptual tokening—whether concepts are accessed via the content of lexical labels representing whole objects or constituent conceptual properties—using a picture-word masked priming congruency task with brief exposures (50/60 ms or 190/200 ms). In Experiment 1, participants were presented with picture-word pairs and had to judge whether the stimuli were related to each other. In Experiment 2, participants completed the same task while wearing red-blue anaglyph glasses—allowing us to control for the potential overlap between retinal projections, and investigate the role of visual pathways and early posterior visual projections during object and word recognition. For each picture (e.g., a dog), one of four word probes was presented for congruency decision: the basic level category label of the picture (e.g., dog), a superordinate label (e.g., animal), a high-prototypical (e.g., bark), or low-prototypical feature (e.g., fur). Response times and accuracy to congruency decisions were analyzed through linear mixed effects models. Results showed that (a) at 50/60 ms, pictures paired with superordinate labels engendered faster and more accurate responses than those paired with high- and low-prototypical features—but no differences with basic level labels, and (b) at 190/200 ms, superordinate and basic level labels yielded faster and more accurate responses than high- and low-prototypical features. We suggest that object concepts are represented in the brain by non-decompositional abstract atomic symbols carrying information about their superordinate categories or information akin to their most generic lexical labels, not through their constituent or salient features.