Shared meaning in representational and abstract visual art: an empirical study

Abstract

A longstanding and important question is how meaning is generated by visual art. One view is that abstract art uses a universal language whereas representational art is tied to specific knowledge. This view predicts that meaning for abstract is shared across viewers to a greater extent than for representational art. This contrasts with a view of greater shared meaning for representational than abstract art, because of shared associations for the entities depicted in representational art, as supported by recent empirical findings. This study examined the contrasting predictions derived from these two views. 49 nonexpert adult participants wrote brief descriptions of meanings that they attributed to 20 abstract and 20 representational artworks, generating a corpus of 1918 texts. Computational analyses (semantic textual similarity, latent semantic analysis) and linguistic analysis (type-token ratio) provided triangulated quantitative data. Frequentist and Bayesian statistical analyses showed that meanings were shared to a somewhat greater extent for representational art, but that meanings for abstract artworks were also shared above baseline. Triangulated human and machine analyses of the texts showed core shared meanings for both art types, derived from literal and metaphoric interpretations of visual elements. The findings support the view that representational art elicits higher levels of shared meaning than abstract art. The empirical findings can be used to enhance theoretical and computational models of aesthetic evaluation, and the rigorous new methodologies developed can be deployed in many other contexts.