The Interplay between Feature-Saliency and Feedback Information in Visual Category Learning Tasks

What is the role of feedback information in different visual category learning (VCL) scenarios? To address this question we tested participants' performance in VCL tasks in which stimuli varied in three feature dimensions, one of which was relevant for the task and the other two were irrelevant. The relevant feature could be identified based on trial-by-trial feedback. In one condition the task relevant and irrelevant features were highly-salient. In the second condition all features had low-visual-saliency. Feedback information was also manipulated: In the high-information condition the task relevant feature could be identified by the information provided in each trial whereas in the mid-information condition the feedback was ambiguous and information from several learning trials was required in order to confidently identify the relevant feature. Surprisingly, our data shows that mid- and high-information feedback are similarly effective in high-saliency VCL tasks. In contrast, in low-saliency VCL tasks, mid-information feedback impairs learning. We suggest that VCL can be done effectively either when feedback is ambiguous or in low-saliency conditions, but not in scenarios when both challenges occur concurrently.