Ad hoc creature: Lost and added in translation from description to depiction

Abstract

When asked to draw a “rabbit”, almost all people can draw a rabbit correctly. In contrast, is it possible that using only descriptions of appearance features of a rabbit could achieve the same task? We prepared a description of rabbit from a dictionary (230 words), and masked “rabbit” in it. The description contains 40 appearance features of rabbit. Then, subjects (n=41) were asked to draw a creature from the description. As the result, the average 15 features (35%; min 2%, max 53%) have appeared in each picture. This indicates that people did not use all of the description to draw the target. Moreover, some people tend to draw undescribed features: a tail (40%), whiskers (10%) and wings (4%). The result suggests that description-to-depiction translation might involve not only bottom-up processing from appearance features of a target, but also top-down processing from an ‘ad hoc creature’ image in mind.


Back to Table of Contents