Dissertation/Thesis Abstract

Measures and Models of Covert Visual Attention in Neurotypical Function and ADHD
by Mihali, Andra, Ph.D., New York University, 2018, 222; 10751244
Abstract (Summary)

Covert attention allows us to prioritize relevant objects from the visual environment without directing our gaze towards them. Attention affects the quality of perceptual representations, quality which can be quantified with precision (or its inverse, variability) parameters in simple psychophysical models that capture the relationship between stimulus strength and an observer’s behavior. Two main types of attention, divided and selective, have been studied in the recent decades with two corresponding classic paradigms, visual search and visual spatial orienting.

In this thesis, we developed variants of these tasks to address questions related to visual attention, in neurotypicals and ADHD. In addition to precision-related parameters derived from behavior, we measured the observers’ fixational eye movements, developed a new algorithm to detect microsaccades and explored their possible role as an oculomotor correlate of precision.

In a first investigation, we built upon a paradigm designed to increase the chances of probing divided attention. Specifically, we extended a visual search task with heterogenous distractors and explored the effects on performance of set size, task—detection and localization, time (perception and memory) and space. An optimal observer model with a variable precision encoding stage and an optimal decision rule was able to capture behavior in a task more naturalistic than target detection, namely target localization. Performance decreased with the set size of the search array for both detection and localization; so did precision. As expected, precision was higher in the perception condition relative to the memory condition. We found the same pattern of results with visual search arrays with reduced stimulus spacing; observers achieved comparable precision parameters, albeit with increased reaction times.

The nature of the attentional impairment in ADHD has been elusive. By using a new task that combines visuo-spatial orienting with feature dimension switch between orientation and color, we found an increased perceptual variability parameter in the ADHD group, which was correlated with an executive control metric. A classifier based on perceptual variability yielded high diagnosis accuracy. These results suggest that using basic psychophysical paradigms to capture encoding precision of low-level features deserves further study in ADHD, especially in conjunction with attention and executive function.

Measures of covert attention have included aspects of fixational eye movements, especially microsaccades. Inferences about the roles of microsaccades in perception and cognition depend on accurate detection algorithms. By using a new hidden semi-Markov model to capture sequences of microsaccades amongst drift and an inference algorithm based on this model, we found that microsaccades were more robustly detected under high measurement noise from the eye tracker. Applying this algorithm to the eye movement traces of ADHD and Control participants, we found a correlation between post-stimulus microsaccade rate and the perceptual variability parameter, suggesting a potential oculomotor mechanism for the less precise perceptual encoding in ADHD.

We conclude that by using and developing variants of visual attention paradigms, psychophysical models and oculomotor measurements, we can enhance our understanding about the brain processes in health and disease.

Indexing (document details)
Advisor: Ma, Wei Ji
Commitee: Carrasco, Marisa, Kiani, Roozbeh, Martinez-Conde, Susana, Simoncelli, Eero
School: New York University
Department: Neural Science
School Location: United States -- New York
Source: DAI-B 79/12(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Neurosciences, Cognitive psychology
Keywords: ADHD, Attention, Bayesian models, Microsaccades, Visual perception
Publication Number: 10751244
ISBN: 9780438171565
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest