Allison B. Sekuler¹²³, Jamie G.E. Chocrane³, Runzhi (Nina) Yue¹², Konka Paul¹, Yifei Wang¹², Ali Hashemi³⁴, Eugenie Roudaia¹, Patrick J. Bennett³; ¹Rotman Research Institute, Baycrest Academy for Research & Education, Toronto, ON, Canada, ²University of Toronto, Toronto, ON, Canada, ³McMaster University, Hamilton, ON, Canada, ⁴Wilfred Laurier University, Milton, ON, Canada

The preferential use of horizontal facial structures, known as horizontal bias, positively correlates with face identification performance. Older observers, who have lower face identification abilities than younger observers, tend to show less horizontal bias; and, within each age group, individuals with greater horizontal bias tend to have better face identification. However, the extent of an individual's horizontal bias is not always fixed: Training younger adults to enhance horizontal bias improves face identification for both trained and novel faces, with the greatest degree of transfer observed when uninformative vertical context is included during testing. The current study investigates whether training horizontal bias, with or without uninformative vertical context, can enhance the horizontal bias and improve face identification in older adults, and whether these effects transfer to novel faces. Older adults completed a 10-alternative forced-choice task, identifying briefly presented faces filtered with horizontal or vertical orientation filters (alternating across blocks). Identification thresholds, which were estimated by varying the orientation filter bandwidth across trials, were measured pre- and post-training (days 1 and 4) in eight conditions that differed in face set (trained/novel), filter orientation (horizontal/vertical), and context (present/absent). Training occurred on days 2 and 3, with participants randomly assigned to one of two conditions: horizontal filters without context, or horizontal filters with uninformative vertical context. Both types of horizontal bias training improved face identification performance and reduced threshold bandwidths. Both training programs also showed partial transfer to novel faces with informative horizontal and vertical structure, with and without uninformative context. However, greater transfer was seen in the horizontally filtered conditions. These results suggest training horizontal bias can enhance face identification in older adults. Future research will examine whether this training improves face identification in other face-perception impaired groups, and determine how training affects the neural signatures of face perception.

Acknowledgements: Support from NSERC Discovery Grants (ABS and PJB), and Baycrest Foundation (ABS)