CAREER: Studying the effects of task goals on brain representations of visual motion

People do not always notice the same things when they look at the same scene, because perception depends on the goal. At the neural level, many perceptual regions in the brain do not necessarily respond the same way to the same stimuli-activity depends on both the stimulus and on task goals. The ability to focus on task-relevant features and thereby change activity in the brain is called top-down attention. Attention is affected in many neurological disorders, including ADHD and Schizophrenia, and varies across the healthy population. Despite its importance, brain mechanisms of attention are not well understood. One hypothesis holds that a network spanning frontal and parietal cortex supports task-general functions including attention and accumulating or weighing evidence. However, parts of this network respond to visual features of stimuli including motion and shape, and the dual influence of task and stimulus are seldom studied in the same experiments. The goal of this project is to integrate these findings into a coherent computational model. Instead of trying to attribute one function to each parcel of the brain, the researchers aim to quantify the degree to which brain responses across the brain vary with the stimulus, with the cognitive components of the task, and/or with task difficulty. The central aim of the research is to determine the degree to which different tasks change the pattern of activity within and across brain regions. The broader impacts include plans to widen the opportunities for undergraduates to develop computer and data science skills through course development and communication. There are also plans to recruit students from historically underrepresented groups to participate in the research. The project seeks answers to these questions in a cumulative series of behavioral and functional magnetic resonance imaging (fMRI) experiments. Participants view moving stimuli on which they perform different visual tasks while their brains are measured with fMRI. For example, they judge whether one object will collide with another object or whether the participant (proxied by the point of view of the camera) will collide with a static object. These tasks serve as an experimental proxy for many natural tasks. For example, driving, walking, and many sports involve estimating one’s own motion relative to static obstacles and moving objects. Importantly, among stimulus-based factors, self-motion is a particularly strong influence on responses in many areas also driven by task-related factors. The project aims to analyze the data using cutting-edge encoding models based on labels for task conditions and motion parameters derived from deep neural networks. Models are compared with variance partitioning, which assesses the relative effect of each factor on brain responses. The models developed in this project can provide a detailed quantitative baseline that enables sensitive measurement of individual differences in attention and task processing in future studies. Finally, since the analytic approach of this project is computationally intensive, another aim is to improve data science education at University of Nevada Reno by developing and teaching an introductory applied research computing course. This course is also intended to teach the “hidden curriculum” of research computing, including use of the command line, version control, and dependency management. The aim is to help build a community of practice in data science at both graduate and undergraduate levels. This project is jointly funded by Cognitive Neuroscience and the Established Program to Stimulate Competitive Research (EPSCoR). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.