Title
Learning in the visual brain: Generalization versus sharpening
Processing capabilities for many low-level visual features are experientially malleable, aiding sighted organisms in adapting to dynamic environments. In this presentation, we discuss how visuocortical responses change as human observers learned to associate exemplars drawn from a given feature dimension with aversive outcomes. Using classical aversive conditioning while recording dense-array EEG and pupillometry, we tested the pre-registered hypotheses of either sharpening or generalization for a range of feature dimensions, including orientation, motion direction, object category, and spatial location. Models of gaussian (generalization) or difference-of-gaussian (sharpening) changes after, compared to pre-conditioning were directly compared in a Bayesian framework. We found that visuocortical responses were selectively heightened when viewing aversively paired features for all feature dimensions. In the case of orientation, motion direction, and spatial location, effects displayed a non-linear, difference-of-gaussian profile across neighboring exemplars on a feature gradient, consistent with suppressive surround modulation of non-prioritized features. Measures of alpha band (8 – 12.8 Hz) activity and pupil diameter showed evidence of generalization. These results indicate that aversive conditioning of low-level visual prompts sharpened tuning in visual cortex. By contrast, aversive conditioning of higher-level features such as object category prompts linearly graded (generalization) modulation in visual cortex. These effects mirror the effects seen for top-down influences indexed by alpha power reduction and autonomic responses, also showing generalization. We summarize these changes in a computational model of adaptive population tuning as a function of experience.
Biography
The long-term goal of my laboratory’s research is to characterize how human beings adapt to perceived environmental challenges. Much of our work has focused on how healthy individuals and patients with psychiatric or neurological disorders process sensory events and how experience changes the neurophysiological systems underlying constructs such as perception, emotion, and motivation. To address these questions, we use a multi-method framework which includes self-report, electrophysiological, neuroimaging, and somato-visceral measures. My work has been continuously funded by NIH since 2009. We also have led efforts to establish robust, reliable, and valid indices of sensation and perception that are based in human electrophysiology, and that are objective and continuous in nature.