Semantic Representations in 3D Perceptual Space

Abstract

Barsalou's (1999) perceptual theory of knowledge echoes the pre-20th century tradition of conceptualizing all knowledge as inherently perceptual. Hence conceptual space has an infinite number of dimensions and heavily relies on perceptual experience. Osgood's (1952) semantic differential technique was developed as a bridge between perception and semantics. We updated Osgood's methodology in order to investigate current issues in visual cognition by: (1) using a 2D rather than a 1D space to place the concepts, (2) having dimensions that were perceptual while the targets were conceptual, (3) coupling visual experience with another two perceptual domains (audition and touch), (4) analyzing the data using MDS (not factor analysis). In three experiments, subjects (N = 57) judged five concrete and five abstract words on seven bipolar scales in three perceptual modalities. The 2D space led to different patterns of response compared to the classic 1D space. MDS revealed that perceptual modalities are not equally informative for mapping word-meaning distances (Mantel min = −.23; Mantel max = .88). There was no reliable differences due to test administration modality (paper vs. computer), nor scale orientation. The present findings are consistent with multidimensionality of conceptual space, a perceptual basis for knowledge, and dynamic characteristics of concepts discussed in contemporary theories.