Investigating invariant motion encoding in mouse visual cortex

PROJECT SUMMARY A fundamental feature of perception is the ability to recognize objects or object features despite changes in presentation, such as size, position, or context. Elucidating the mechanisms by which cortical populations and circuits build representations that are invariant to these manipulations will significantly advance our

understanding of the biological underpinnings of perception. Based on limited biological work examining invariant tuning properties of neurons in visual cortex, studies in the field of object recognition hypothesize that invariance, or tolerance to identity-preserving transformations, increases gradually as signals ascend the visual processing

hierarchy. In contrast, studies investigating how invariance is computed for perception of motion across different object types have hypothesized that different types of invariances are computed in discrete, subsequent steps. For example, humans and nonhuman animals alike perceive the pattern direction of motion of drifting plaids

despite the fact that plaids are composed of two overlaid gratings that drift in divergent directions. This perception requires direction invariance (tolerance to the component grating directions of motion) and spatial invariance (tolerance to differences in spatial features across different types of plaids). Previously published work proposes

that direction and spatial invariance are computed in series. First, spatial invariance is achieved by complex, direction selective cells in V1 and next, direction invariance is computed upon integration of the signals encoding component directions of motion in higher order areas of cortex. However, data from other studies and my own

preliminary data do not agree with this two-stage model. We find direction invariance in V1 prior to the representations achieving spatial invariance. These data suggest that the computation for spatial invariance for pattern motion encoding may occur gradually, as is proposed in object recognition pathways. Therefore, my

central hypothesis is that motion encoding circuits compute spatial invariance gradually over multiple transformations to build an invariant percept of pattern motion. Using multiphoton imaging and optogenetic

manipulations in mouse visual cortex, I will test this hypothesis with two specific aims. In Aim 1, I will test if spatial invariance increases gradually throughout the processing hierarchy and, if so, whether the increase is inherited by receiving biased inputs that come from the most invariant neurons in the preceding population, or if spatial

invariance is computed de novo upon integration of heterogenous inputs. In Aim 2, I will investigate how single cell and population level pattern motion representations inform pattern motion perception by training mice in a direction discrimination task. Overall, the significance of this work will be to advance our understanding of how

invariance is built for pattern motion perception, resolving current controversies in the field and revealing fundamental principles of cortical computations that increase invariance for perception.