Predicting the Good Guy and the Bad Guy: Attitudes are Encoded in Language Statistics

Gabriel RecchiaInstitute for Intelligent Systems, University of Memphis
Alexandra SlaterDepartment of Psychology, University of Memphis
Max LouwerseTilburg Center for Cognition and Communication, Tilburg University

Abstract

Various studies have provided evidence that people activate introspective simulations when making valence judgments. Such evidence is in line with an embodied cognition account that argues that cognition is fundamentally embodied, with perceptual simulation rather than language statistics being the source of lexical semantics. Recently, demonstrations that conceptual knowledge is encoded in language have been used to argue that semantic processing involves both language statistics and perceptual simulation, with linguistic cues allowing meaning to be bootstrapped with minimal symbol grounding. Whether language also encodes attitudes towards concepts is unclear. In three studies, negative-valence words were found to be more closely associated in language with individuals commonly considered villains, and positive-valence words with heroes (both fictional and historical). These results suggest that attitudes toward persons can be inferred from lexical associations.

Files

Predicting the Good Guy and the Bad Guy: Attitudes are Encoded in Language Statistics (244 KB)



Back to Table of Contents