Abstract
Several resent studies found that training amblyopes to detect contrast with/without flankers or identify positional offset can improve the visual acuity in the amblyopic eyes (Polat et al., 1996; Polat et al., 2004; Zhou et al., 2006). However, provided the general observation that perceptual learning is highly specific to the training set (e.g., grating spatial frequency), one might expect perceptual learning of limited use as a clinical tool. Generalizability of learning is a necessary condition for perceptual learning to be an effective treatment. We provide the first systematic evaluation. We measured the magnitude and generalizability of perceptual learning in adults with anisomotropic amblyopia and their controls. Ten amblyopes (Group I) and 15 normal adults (Group II) were trained in the amblyopic eyes (amblyopes) or the non-dominant eyes (normals) in a contrast detection task at their cutoff spatial frequencies. Another seven normal adults (Group III) were trained at the average cutoff spatial frequency of Group I. Contrast sensitivity functions were measured in the trained and untrained eyes before and after training. We found that the amblyopic subjects improved their contrast sensitivity more than normal adults (10.66 vs 4.88 dB) at their respective cutoff spatial frequencies; Group III did not show any significant learning. Learning effects generalized to a broader range of spatial frequencies in the amblyopic (5.02±0.70 octaves) group than the normal group (2.03±0.34 octaves). Moreover, learning transferred to the untrained eyes in Group I, but was specific to the trained eyes in Group II. Our results suggest that the amblyopic visual system is more plastic — learning can be generalized to a wider range of conditions. Improvements in amblyopic eyes are not at the expense of the fellow eyes. Both observations may provide important theoretical basis for using perceptual learning to treat amblyopia.