Abstract
Perceptual learning improves face recognition, and learning is highly specific and long-lasting (Hussain et al., Psych Sci 2011). Even inverted faces can benefit from learning (Hussain et al., Vis Res 2009). But what changes in our representations of faces with learning? Last year, we demonstrated that the preferential use of horizontal information ("horizontal tuning") is correlated with upright face identification accuracy and the size of the face inversion effect (Pachai et al., VSS 2011). In the current study, we asked whether perceptual learning for faces is associated with an increase in horizontal tuning. Specifically, we tested inverted faces in a 10AFC identification paradigm where stimuli were the average of 10 faces viewed through different filters. Information from the target face alone was visible only within orientation bandwidths ranging from 10 degrees to 180 degrees (full-face) in 10 degree steps, centred around horizontal or vertical. In the first session, observers completed 10 trials in each condition to measure initial horizontal tuning. In the following three sessions, observers completed 300 trials/session of full-face identification. The fifth session was identical to the first. Observers returned 3-5 days later to assess maintenance of learning and transfer of learning to a new face set. As expected, training significantly improved inverted full-face identification. Critically, training also improved accuracy for faces with narrow-band filters centred on horizontal, but not vertical, suggesting an increase in horizontal tuning. Tuning was maintained in the follow-up session, but did not transfer to novel faces. These results suggest that perceptual learning improves horizontal tuning for trained face stimuli while improving overall identification accuracy, further implicating the importance of horizontal information for accurate face identification regardless of picture-plane orientation, and suggesting that the relatively high efficiency of processing horizontal information for upright faces may be a result of learning across the lifespan.
Meeting abstract presented at VSS 2012