ABSTRACT
Each concept consisted of a randomly generated pattern of vectors across the “visible” units and a subset of the semantic units. The distribution of visual and functional features for each living and nonliving concept was consistent with the assumptions of the sensory/functional hypothesis: Notably, representations of concepts in the two categories were asymmetrically distributed across the sensory and functional semantic units. Concepts belonging to the category of living things were represented by an average of 2.68 visual features and 0.35 functional features. In contrast, concepts belonging to the category of nonliving things were represented by an average of 1.57 visual features and 1.1 functional features. Thus, representations of living things contained a ratio of 7.7:1 visual to functional features, while representations of nonliving things contained a ratio of 1.4:1. In the network representations, consistent with these ratios, living things were represented by an average of 16.1 visual and 2.1 functional units, and nonliving things were represented by an average of 9.4 visual and 6.7 functional units. The representations of all concepts encoded in the network contained both visual and functional semantic memory units.
