Abstract
Visual statistical learning allows the rapid acquisition of sequential regularities in the visual environment. However, the implications of this type of learning for subsequent visual perception are not well understood. We have recently shown that statistical learning of natural image sequences influences perceptual selection during binocular rivalry by increasing the likelihood of perceiving an unexpected image over an expected, learned image (VSS 2012). Here, we asked whether this tendency for statistical learning to enhance perception of the unexpected image improves performance on an objective target identification task. Subjects viewed a stream of three-item image sequences (triplets). The sequences of images in the triplets were consistent throughout the experiment in order to allow statistical learning of the triplets to take place over the course of the experimental session. The image stream was presented dichoptically, and on some image presentations, target face images were embedded within the stream by presenting them to one eye, while the other eye was presented with the appropriate next image in the stream. Therefore, target images and learned images were engaged in binocular rivalry. Subjects' task was to identify faces as male or female whenever they saw them. Comparing the first half to the second half of the experiment, we found improvements in target accuracy (percent correct) and discrimination (percent correct normalized by percent detected) that were specific to targets paired with the third image in the triplets. These results are consistent with the idea that learning the image sequence regularities selectively improved perception of targets that rivaled with the most strongly predicted images, compared to targets that rivaled with images less strongly predicted by the learned sequence. These results raise the possibility that one function of visual statistical learning may be to enhance perception of unexpected items relative to those that can be anticipated.
Meeting abstract presented at VSS 2014