Content area
Visual perception depends critically on the brain’s ability to integrate information across spatial context. In this thesis, I examine the circuit mechanisms by which primary visual cortex (V1) implements flexible normalization to modulate local feature representations based on surrounding stimuli. I address three interrelated questions: (1) What are the cortical origins of orientation-dependent and figure-ground contextual signals in V1? (2) How do network oscillations reflect and constrain mechanistic models of contextual modulation? (3) Can a biologically grounded model of E–I circuitry reproduce empirical signatures of orientation selectivity in V1?
In Chapter 2, I employ ultra–high-field laminar fMRI in human V1 to disentangle feedforward and feedback contributions to contextual modulation, revealing depth-specific BOLD profiles for distinct forms of surround influence. Chapter 3 analyzes the orientation tilt illusion and neural oscillations measured from LFP, demonstrating that contextual effects leading to altered perceptions of orientation and leading to changes in narrow-band gamma oscillations can both be explained with a biologically plausible implementation of divisive normalization. In Chapter 4, I use a stabilized supralinear network (SSN) to explore mechanisms of orientation selectivity in ferret V1 by directly comparing model predictions to single-cell measurements from two-photon calcium imaging.
Together, this work bridges multiple spatial scales from synaptic circuits through population dynamics to cortical interactions across the visual hierarchy and advances our understanding of how recurrent and feedforward interactions implement context-dependent computations in visual cortex.