Confidence and feedback in visual perceptual learning

The classical view that perceptual learning is highly specific to the stimulus properties used during training is being challenged by accumulating evidence of transfer under specific conditions. Perceptual learning in the adult brain may result from an element of cortical reorganisation at a perceptual level or a reorganisation of decision weights at a higher level. This thesis focuses on individual differences in perception, and the mechanisms of perceptual learning. Since the role of feedback is not usually the primary area of interest for perceptual learning research, this was one of the main focus points of the methodology used. Perception and perceptual learning were assessed for a range of tasks at local and global levels, including depth perception in random dot stereograms, and global motion and form coherence. The results of these experiments established that transfer occurred when the task matched the tuning of the global processing area, even when the untrained task was a locally processed task. The results also provided evidence of robust learning in local and global tasks with or without feedback, as long as easy and difficult trials were interleaved within the same task. Furthermore, in some conditions internal feedback provided better learning at sub-threshold stimulus levels than did explicit external feedback. These results suggest that, for low stimulus intensities, external feedback may reduce observers' confidence in their own perceptual decisions. Confidence in perceptual decisions was a key factor throughout all the studies, and where this was measured it was found to be highly correlated with performance. Confidence in perceptual decision and accuracy was low for a depth perception study using anti-correlated random-dot stereogram (ACRDS). The binocular energy model of neural responses predicts that depth from binocular disparity may be perceived in the reversed-direction when the contrast of dots are anti-correlated. Depth was mostly perceived in the correct direction for ACRDS conditions with some inconsistencies. These differences are likely to reflect the inconsistent depth signals, across scale and across first- and second-order channels, elicited by anticorrelated stimuli. To ensure that the rich variability embedded within the individual differences in heterogeneous data sets were not excluded from analyses, mixed effects models were employed throughout the thesis. This technique considers between-individual variation as a random factor which made it possible to investigate behavioural differences between individuals with migraine and control groups. Specifically we evaluated the spatial extent of excitatory and inhibitory interactions using a classic lateral masking task. Overall, contrast thresholds in the baseline condition for the migraine group were lower than those in the control group. There was no difference in the degree of lateral interaction in the migraine group. The results suggest the that impaired performance in perceptual tasks in individuals with migraine may not be as a results of altered local mechanisms.