Emotional Resemblance: Perception of Facial Emotion in Written English

Written language is comprised of simple line configurations (i.e., letters) that, in theory, elicit affect by virtue of the concepts they symbolize, rather than their physical features. However, we propose that the line configurations that comprise letters vary in their visual resemblance to canonical features of facial emotion and, through such emotional resemblance, influence affective responses to written language. We first describe our data-driven approach to indexing emotional resemblance in each letter according to its visual signature. This approach includes cross-cultural validation and neural-network modeling. Based on the resulting weights, we examine the extent to which emotional resemblance in Latin letters is incidentally processed in a flanker paradigm (Study 1), shapes unintentional affective responses to letters (Study 2), accounts for affective responses to orthographically controlled letter strings (Study 3), and shapes affective responses to real English words (Study 4). Results were supportive of hypotheses. We discuss mechanisms, limitations, and implications.