UC Berkeley

Perceiving objects' positions is one of the fundamental purposes of vision and is crucial to day-to-day life. Much work has been devoted to finding and characterizing spatial maps in the brain, but we still know strikingly little about how these maps support perceptual localization. The experiments in this thesis employ a combination of functional brain imaging and visual psychophysics, drawing on a body of behavioral literature on spatial perception and visual attention, to ask how position coding in the brain adapts to support the demands of the task at hand.

The experiments in Chapter 2 ask how and where perceived object positions are represented in the brain. Perceived position depends on many factors such as attention, frame of reference, adaptation, and motion. At what stage of visual processing is this information integrated into the brain's retinotopic maps? By measuring variability in perceived position alongside variability in the multivariate pattern of neural responses in a host of visual areas, we identify a percept-centered reference frame in high-level object-, face-, scene-, and motion-selective regions.

The experiments in Chapter 3 ask how the visual system achieves improved spatial resolution with focused attention. Attention is known to boost the amplitude of neural responses, but it might also sharpen position tuning at the neural population level, making the activations produced by adjacent objects more distinct within the cortex. Employing a combined imaging and modeling approach, we find that attention narrows the spatial spread of the fMRI BOLD response in early visual cortical areas, including V1. This narrowed population position tuning is an efficient means of achieving visual resolution improvements.

The experiments in Chapter 4 ask how the brain suppresses distracting visual input in order to limit its negative impact on performance. During a selective attention task, we measure the BOLD response in the pulvinar nucleus of the thalamus, an area whose damage appears to lead to distractor filtering deficits. We find that while attended items are represented with high spatial and featural precision in the pulvinar, ignored items are gated out, leaving no discernable trace in the pattern of response in the pulvinar. These results suggest that the spatial maps in the pulvinar may serve as an interface in which distracting visual input is filtered out.

Collectively, the findings presented here speak to a remarkable flexibility in the way the brain represents spatial information. The brain's visual maps adapt on a moment-by-moment basis to support the demands of the task at hand.