Abstract
Perceptual learning is the ability to improve perception through practice. The hallmark of perceptual learning is its specificity for the trained visual features such as the stimulus orientation and location. For example, training in discriminating Gabors’ orientation improves performance. However, that training does not improve Gabor orientation discrimination at untrained locations. Perceptual learning is mostly studied without eye movements. Nevertheless, in everyday life, a given stimulus is actively explored through eye movements, resulting in successive projections of that stimulus at different retinal locations. Here, we studied perceptual learning of orientation change discrimination across saccades. In Experiment 1, observers trained to discriminate orientation changes of a Gabor grating presented in the periphery while fixating in the center. Before and after training (pre- and post-training tests), observers were tested with the same task but with an orthogonal orientation at the trained location, and the trained orientation at an untrained location. Expectedly, observers improved performance, and perceptual learning was orientation and location specific. In Experiment 2, instead of fixating, observers in training and pre- and post-training test trials saccaded to the peripheral grating, and discriminated the orientation change occurring during the saccade. Orientation and location conditions were the same as in Experiment 1. Interestingly, we found trans-saccadic perceptual learning for orientation change discrimination. This perceptual learning did not transfer to the untrained orientation. Surprisingly however, we found transfer to the untrained location. Additionally, after the trans-saccadic training, observers also improved their performance at discriminating orientation changes in the fixation condition such as described in Experiment 1. We propose that perceptual learning within an active perception framework might reflect different mechanisms than classic perceptual learning.
Acknowledgement: This work was supported by grants of the Deutsche Forschungsgemeinschaft to HD (DE336/6-1).