UC Berkeley

Visual information processing in the central visual pathway is mediated by three main types of inter-cellular connections. These are feedforward, feedback, and horizontal connections. Most neurophysiological studies have been conducted in a perspective of feedforward connections assuming a hierarchical model. They revealed that convergent feedforward inputs determine size and structure of classical receptive fields (CRFs) of recipient cells in a higher level area. Non-feedforward connections are known to integrate visual information from outside the CRF and use this information to modulate the CRF activity. But relative roles of feedback and horizontal connections are not clear. My research interest is to figure out different functions for three major pathways in vision. In order to do this, I measure single unit activity in cat visual cortex while specific visual or magnetic stimulation is being applied to provide weighted activity from feedforward, feedback, and horizontal connections.

In Chapter 1, we introduce visual stimuli specially designed to maximize relative differential involvements of three main types of neural connections. The approach is based on well-established anatomical and physiological features of neural projections in the central visual pathway. We show clear segregation of fast and slow components of surround modulation. Then, we present supporting evidences that they are primarily mediated by feedback and horizontal connections, respectively.

Chapter 2 quantitatively describes transcranial magnetic stimulation (TMS) effects on neural activity. We focus specific stimulation parameters with limited visual stimuli and examine how TMS affects response selectivity (orientation, spatial frequency & contrast tuning) of visual neuron. Our findings suggest that TMS interrupts the existing balance between sub-cortical and intra-cortical inputs for a relatively extended time period.

According to anatomical studies, proportions of sub-cortical and intra-cortical inputs to a single cell in the visual cortex are clearly different depending on cortical layers. In Chapter 3, we examine how intra-cortical inputs are differently contributing to direction selectivity of visual neurons by comparing various features of direction selectivity among multiple layers. Consistent with the classical view, our results show that linear estimation of direction selectivity (only sub-cortical inputs are considered) is quite accurate in input layers but substantially smaller in non-input layers, suggesting a considerable non-linear contribution of intra-cortical connections in these layers.

Lastly, Chapter 4 describes issues related to trial-to-trial variability. Spiking response of a V1 single neuron to repeated presentations of the same visual stimulus is too variable to explain excellent behavioral performance in discrimination task. From the viewpoint of experimenter who controls the parameters of visual stimuli systematically, this trial-to-trial variability can be regarded as unpredictable noise which may reduce potential performance of neural computation. The source and advantage of trial-to-trial variability is not yet understood. Here, I introduce several response variability related issues that we already know and need to know.