Re-learning face recognition: evidence for efficient strategies without holistic processing

In previous experiments, we showed an advantage for visual over haptic face recognition. Promoting serial encoding in vision using a novel gaze-restricted display, we found that these differences are due to modality-specific encoding strategies (holistic in vision vs serial in haptics), and that serial encoding leads to featural (vs holistic) processing. Here, we test how the observed encoding and processing differences might be affected by expertise. Participants were trained on five consecutive days on a set of 19 faces using an old/new recognition task for which three faces were learned with feedback, followed by four test-blocks. On days 1 (pre-test), 4 (post-test), and 5, the task consisted of two upright and two inverted test-blocks. On day 5, we tested participants on a different set of faces. Performance for upright faces was low on day 1, but improved significantly through training ( d1'=1.11, d4'=3.75). Importantly, this learning effect generalized to a new face-set on day 5 ( d5'=3.07). Although performance significantly improved through training, we found no inversion effect on any day ( d1'=1.70, d4'=3.54) indicating no change in processing strategies. Our results show that participants can develop efficient, generalizing strategies to compensate for encoding differences, and that these strategies do not require holistic encoding