In amblyopes, one of the eyes has poorer vision than the other, usually because the eye was neglected (e.g. due to squint) in early childhood. It is unclear whether the poor vision in complex visual tasks can be explained by lack of sensitivity to simple stimuli, or whether information from the amblyopic eye is processed differently in the brain (Levi et al. 2002).

We performed experiments on five amblyopic women (aged 18-30) with Ethical Committee approval. We measured thresholds in both eyes of each observer for two or four sets of natural scene stimuli: series of slightly different pictures made by morphing one picture into another (Parraga et al. 2000). For instance, pictures might show slight changes in facial expression. A total of 14 such experiments was performed. For comparison, we also measured the observer’s contrast detection thresholds for sinusoidal gratings of different spatial frequencies. Thresholds were measured by presenting pictures or gratings (2 deg square) on a CRT monitor, following a modified two-alternative forced-choice protocol (Parraga & Tolhurst, 2000).

In all experiments with complex natural-scene stimuli, the thresholds in the amblyopic eye were 1.83 times those in the fellow eye (mean = 1.83, n = 14, S.D. = 0.44). Often, the contrast thresholds for the gratings were also higher in the amblyopic eye, raising the question whether poor performance in complex discriminations results just from a failure to see the Fourier components of the complex scenes as well.

We have previously developed a simplistic model of discrimination for normal observers (Parraga et al. 2000), in which we presume that the visual system breaks complex pictures down into separate spatial frequency bands; discrimination depends on detecting differences in contrast in those bands. The parameters of the model include the oberver’s thresholds for sinusoidal gratings. When we ran such models on the amblyopic eyes, we found that the poor sensitivity to gratings did predict poor performance on the complex discrimination task. However, the actual performance in the complex task was worse even than predicted by the model; the actual thresholds were on average 1.4 times the predicted thresholds (S.D. = 0.41).

There are many assumptions in such a simplistic model, so it is interesting to note that our two best-studied observers (4 experiments each) actually had similar grating thresholds in their two eyes but performed the complex discrimination worse in their amblyopic eye than their fellow eye; the amblyopic threshold was 1.57 times the threshold in the good eye (n = 8; S.D. = 0.27).

CAP was a Fight for Sight Fellow.