The Effects of Horizontal Bias Training on Face Identification

Abstract

Horizontally oriented structure drives the accuracy of face identification: Individuals who rely more on faces’ horizontal structure relative to vertical structure (horizontal bias) perform better on facial identity discrimination tasks (Pachai et al., 2013); superior identification performance for familiar faces is linked to stronger horizontal bias (Pachai et al., 2017); and training on inverted faces leads to increased horizontal bias (Pachai et al., 2019). Although it is now established that training on faces can enhance horizontal bias, here we ask whether targeted perceptual training of horizontal structure can enhance face identification, and the extent of transfer of any learning effects. This study utilized a ten-alternative forced choice task implemented over four consecutive days. On days one and four, participants underwent baseline testing of two face sets with varying horizontal and vertical bandwidth filters; while on days two and three, participants underwent training of horizontal structure of only one of the two selected face sets and only horizontal bandwidth filters. Further, participants were separated into two training groups: one with additional uninformative vertical context added to the horizontally filtered faces, and the other with only horizontally filtered information. Preliminary data showed that horizontal training led to a decreased face identification thresholds within the trained face set. Both context present and context absent conditions led to slight transfer of learning to the untrained horizontally filtered face set. However, transfer to vertically filtered faces was seen only in the training conditions in which uninformative vertical context was included. These results provide useful insights for the development of training programs to enhance face perception.