Abstract
Accumulating evidence suggests that multiple levels of processing and representation may be involved in perceptual learning, and that tasks assumed to reflect low-level changes in perceptual sensitivity may be subject to a variety of high-level influences. For example, organizing patterns of contrast as meaningful stimuli (e.g., faces vs. textures) supports higher levels of performance than with equivalent levels of contrast in non-meaningful stimuli. With respect to representation, the possibility of multiple sources of influence raises at least two questions: can perceptual learning be accomplished with multiple features, and if so, what are the informational relationships among the learned features? That is, to what extent are these acquired representations dependent or separable, either perceptually or decisionally? The present study investigates these questions using stimuli (sized to be completely foveated) containing 0, 1 or 2 contrast-defined target pattern features, drawn from the 1865 drawing of the Cheshire Cat in Alice's Adventures in Wonderland. Participants began by performing a complete identification task (CID), in which stimuli contained 0, 1, or 2 of the components of the target image, presented at either 50% or 10% contrast. They then practiced with all possible stimuli for 10-15 days, using an adaptive staircase procedure to track thresholds. Finally, they again performed the CID, with stimuli presented at 50%, 10% and threshold contrast levels. EEG data were collected during both pre- and post-practice performance of the CID. Reliable reductions in thresholds were accompanied by reliable shifts in decisional criteria, the emergence of violations of separability, and increases in the amplitude of early ERP components and power in gamma-band activity. These results suggest that perceptual learning can be obtained with multiple features and that learning can involve interacting sources of information at both perceptual and decisional levels.
Meeting abstract presented at VSS 2014