Funded Grants


Linking attentional modulations in human visual cortex with perception and behavior

Our experience of perception is immediate and effortless. This impression, however, is misleading. Faced with the infinite complexity of the world, sensory systems must be selective -- fluidly shifting to focus attention on the objects and features that are important to action, while leaving the majority of the world largely unanalyzed. Understanding how selective attention shapes sensory signals into the perceptual representations that guide behavior is thus fundamental to understanding cognition. However, attention poses a significant scientific problem: even the most sophisticated computational models of attention fall far short of matching human capabilities, and disruptions to attention are associated with a spectrum of clinical disorders ranging from impaired focus in the classroom to debilitating inabilities to perceive objects across large regions of space. To understand selective attention, my work uses a multidisciplinary combination of theoretical and novel methodological approaches.

Perception is based on large populations of feature?selective sensory neurons that pool their output to provide stable representations of visual features in the environment. Therefore, examining how attention shapes these response profiles is necessary to understand how relevant stimuli are efficiently processed. However, predicting and measuring modulations at the population level is challenging from both theoretical and empirical points of view.

From a theoretical perspective, there is not a single 'optimal' pattern of neural activity that should be universally adopted when processing a visual stimulus. For instance, most models suggest that attention acts like a volume knob that turns up the gain of sensory evoked responses. However, our work suggests that the ideal activity pattern is highly dependent on the specific perceptual task confronting an observer (e.g. trying to detect your friend wearing a red jacket versus trying to discriminate a subtle anomaly from normal tissue on a brain scan). Therefore, we employ computational models and information theoretic metrics to predict how attention should ideally modulate population response profiles in the context of specific behavioral goals.

From an empirical perspective, measuring feature?selective response profiles in human visual cortex is daunting. Even though non?invasive techniques like functional magnetic resonance imaging (fMRI) are capable of indexing activation changes across large populations of sensory neurons, traditional analysis techniques largely discard feature?selective information. To overcome this limitation, we have developed a set of sophisticated analytic tools to measure changes in feature?selective response profiles using fMRI signals. In contrast to increasingly popular 'multivoxel pattern analysis' techniques, our approach maps explicit models of neural tuning functions onto observed changes in fMRI signals. Thus, we can evaluate theory?driven questions about how attention modulates population responses in human visual cortex with a high degree of precision.

By bridging these theoretical and empirical approaches, research in my lab critically evaluates novel ideas about how attention facilitates information processing at the level of population codes in visual cortex. Moreover, by establishing a complete link between theory, data, and behavior, our approach will support new insights into the functional properties of attentional modulations, which is a key element in building models that can account for deficits that arise when selective attention goes awry.