The visual system must constantly extract behaviorally relevant stimulus information from an abundance of irrelevant inputs from the environment, using cognitive phenomena such as attention and learning to guide this continuously adapting process. Understanding the mechanisms by which task-relevant information is extracted from the high-dimensional activity of neuronal populations will be vital to understanding the complex etiology of many neurological diseases, such as disorders of attention. A longstanding assumption has been that this process is optimized for each specific visual task, maximizing the amount of information extracted from the activity of neuronal populations. While this may be possible in highly reductionist lab settings with simple stimuli, such specific optimization would be virtually impossible in the face of the abundant and rapidly changing stimuli and task goals encountered in the natural world. Our recent work suggests a new hypothesis: the extraction of information from neuronal population activity is optimized not for each specific visual task, but generally for the wide variety of stimuli and tasks encountered in realistic environments. In each of our Aims, we will use feature-rich, realistic visual stimuli, precise psychophysical measurements of perceptual performance, simultaneous recordings from populations of visual neurons, and cutting edge data analysis techniques to test one prediction of our central hypothesis.
In Aim 1, we will test the prediction that in realistic environments with changes to both task-relevant and -irrelevant visual features, neuronal information extraction is optimized generally for all of the encountered feature changes, instead of just for the task-relevant changes.
In Aims 2 and 3, we will test the extent to which our central hypothesis is true across different time frames.
In Aim 2, we will test the prediction that the neuronal information extraction process is optimized to be flexible on short time scales, in the face of rapidly changing task goals.
In Aim 3, we will test the prediction that information extraction can also be flexibly optimized on long time scales, explaining gradual and highly specific improvements in perceptual ability due to perceptual learning. The results of these studies will have broad implications both for biophysical models of visual perception and for our understanding of how neuronal mechanisms in general are able to flexibly adapt to our constantly changing natural environment. The proposed project will not only further our understanding of how neuronal activity guides behavior in the context of realistic visual environments, but will provide me with the necessary technical and analytical skills to launch my career as an independent investigator. By receiving expert training to create and operate complex, feature-rich visual stimuli, to collect precise psychophysical measurements by parametrically varying specific aspects of those stimuli, and to apply advanced computational techniques to analyze complementary behavioral and neuronal datasets, I will be fully prepared to independently pursue questions of how neuronal activity guides perception and behavior in my own research program.
Unraveling the complex etiology of many neurological diseases will require understanding how the coordinated responses of populations of neurons guide behavior in natural environments. The projects in this proposal will shed light on our fundamental understanding of the neuronal mechanisms that underlie behavior in complex and feature-rich realistic settings, which require allocating attention to relevant stimuli and behavioral goals. Furthering our understanding of these mechanisms will allow us to better address neurological disorders such as attention deficit hyperactivity disorder (ADHD), a disease thought to affect as many as 10% of the children in the United States.
