Correlates of perceptual learning in extrastriate visual cortex

Gregor Rainer

Max Planck Institute

Experience with objects can allow us to recognize them more accurately and efficiently. To study the neural basis of these perceptual improvements, we have developed parametric sets of stimuli consisting of natural scenes at different levels of coherence corresponding to perturbations of Fourier phase spectra. Practice with a particular set of natural scenes allowed monkeys to identify them at lower coherence levels compared to novel natural scenes. We recorded local field potentials (LFPs) and single unit activity (SUA) from extrastriate visual area V4. Several distinct modes of processing were evident in the LFP recordings. While LFPs showed little learning related changes during the transient visual response (~50 to 90ms after onset), large robust changes were observed later in the response. The systematic changes in mean LFPs reflecting inputs to V4 were correlated with changes in SUA representing V4 output. In particular, while neurons communicated similar amounts of information about novel and familiar objects early during the response, information was amplified for familiar objects at intermediate coherence levels during later portions of the response. Many V4 neurons showed higher activity during difficult discriminations particularly for familiar objects, consistent with participation in recurrent processing. Our results reveal that while V4 neural activity is determined by sensory stimulus properties early in the response, the later response reflects recurrent processing probably involving feedback from higher areas.