Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 2010. / Cataloged from PDF version of thesis. / Includes bibliographical references. / Ecologically relevant search typically requires making rapid and strategic eye movements in complex, cluttered environments. Attention allocation is known to be influenced by low level image features, visual scene context, and top down task constraints. Scene specific context develops when observers repeatedly search the same environment (e.g. one's workplace or home) and this often leads to faster search performance. How does prior experience influence the deployment of eye movements when searching a familiar scene? One challenge lies in distinguishing between the roles of scene specific experience and general scene knowledge. Chapter 1 investigates eye guidance in novel scenes by comparing how well several models of search guidance predict fixation locations, and establishes a benchmark for inter-observer fixation agreement. Chapters 2 and 3 explore spatial and temporal characteristics of eye guidance from scene specific location priors. Chapter 2 describes comparative map analysis, a novel technique for analyzing spatial patterns in eye movement data, and reveals that past history influences fixation selection in three search experiments. In Chapter 3, two experiments use a response-deadline approach to investigate the time course of memory-based search guidance. Altogether, these results describe how using long-term memory of scene specific representations can effectively guide the eyes to informative regions when searching a familiar scene. / by Barbara Hidalgo-Sotelo. / Ph.D.

Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, February 1999. / Includes bibliographical references (p. 120-125). / This dissertation considers ambiguity resolution in regard to two issues. First, it investigates factors that lead the human parser to favour some types of interpretations over others when faced with some types of ambiguous input. Second it examines the reanalysis process that takes place when initial biases lead to incorrect interpretations. The first part of the dissertation (Chapter 1) proposes that reanalysis is a process that requires the maximal satisfaction of constraints (similar to first pass parsing as in Gibson & Pearlmutter, 1998; MacDonald, Pearlmutter & Seidenberg, 1994) rather than the minimization of the number of operations involved as has been suggested previously (Frazier, 1994; Fodor & A. Inoue, 1994). Three experiments in Japanese using main/embedded clause ambiguities are reported in support of this claim. The second part of the dissertation (Chapters 2, 3 and 4) uses a well-established generalization as a starting point, namely, that a modifying phrase is preferentially attached to the closest available site. A recent question in the literature has been to determine ways of parameterizing this principle in order to account for cross-linguistic variations observed in the attachment of relative clauses (Cuetos & Mitchell, 1988). In Chapter 2, the potential parametrizations that may explain the phenomenon at hand are restricted based on data from Brazilian Portuguese. In Chapter 3, an experiment in Japanese investigates the locality preference in this head-final language. Finally, in Chapter 4, it will be suggested that ambiguities in relative clause attachment are not only well suited to investigate cross-linguistic phenomena but also various properties of the human parser that lie beyond the realm of grammatical well-formedness, and two on-going projects are briefly described. / by Edson T. Miyamoto. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 2009. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student submitted PDF version of thesis. / Includes bibliographical references (p. 123-132). / Over the past two decades several hierarchical learning models have been developed and applied to a diverse range of practical tasks with much success. Little is known, however, as to why such models work as well as they do. Indeed, most are difficult to analyze, and cannot be easily characterized using the established tools from statistical learning theory. In this thesis, we study hierarchical learning architectures from two complementary perspectives: one theoretical and the other empirical. The theoretical component of the thesis centers on a mathematical framework describing a general family of hierarchical learning architectures. The primary object of interest is a recursively defined feature map, and its associated kernel. The class of models we consider exploit the fact that data in a wide variety of problems satisfy a decomposability property. Paralleling the primate visual cortex, hierarchies are assembled from alternating filtering and pooling stages that build progressively invariant representations which are simultaneously selective for increasingly complex stimuli. A goal of central importance in the study of hierarchical architectures and the cortex alike, is that of understanding quantitatively the tradeoff between invariance and selectivity, and how invariance and selectivity contribute towards providing an improved representation useful for learning from data. A reasonable expectation is that an unsupervised hierarchical representation will positively impact the sample complexity of a corresponding supervised learning task. / (cont.) We therefore analyze invariance and discrimination properties that emerge in particular instances of layered models described within our framework. A group-theoretic analysis leads to a concise set of conditions which must be met to establish invariance, as well as a constructive prescription for meeting those conditions. An information-theoretic analysis is then undertaken and seen as a means by which to characterize a model's discrimination properties. The empirical component of the thesis experimentally evaluates key assumptions built into the mathematical framework. In the case of images, we present simulations which support the hypothesis that layered architectures can reduce the sample complexity of a non-trivial learning problem. In the domain of speech, we describe a 3 localized analysis technique that leads to a noise-robust representation. The resulting biologically-motivated features are found to outperform traditional methods on a standard phonetic classification task in both clean and noisy conditions. / by Jacob V. Bouvrie. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1996. / Includes bibliographical references (p. 139-150). / by Emanuela Bricolo. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 2007. / Includes bibliographical references (leaves 114-123). / To what extent can experience shape perception? In what ways does perception vary across people or even within the same person at different times? This thesis presents three lines of research examining the role of experience on perception. The first section presents evidence from synesthesia suggesting that learning can influence letter-synesthesia pairings and that associative learning can affect relatively early visual processing. The second section examines the role of linguistic categorization in color judgments, finding that language can play an online role even in a relatively simple color discrimination task. The final section examines how perception adjusts over relatively short time scales using face adaptation. The adaptation experiments show that adaptation to faces can improve recognition performance on famous faces. The results further demonstrate that these effects can be obtained without extensive training and that contrary to proposals from experiments using face spaces, that identity based adaptation effects can be found on trajectories which do not pass through the average face. / by Nathan Witthoft. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 2012. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student-submitted PDF version of thesis. / Includes bibliographical references (p. 100-105). / In the following thesis, I describe the investigation of two problems related to the organization and structural analysis of visual information: perceptual grouping and silhouette analysis and representation. For the problem of perceptual grouping, an intuitive model framework was developed which operates on raw images and locates relevant groupings utilizing a higher dimensional space that contains not only the two spatial dimensions of the image but one or more dimension corresponding to relevant image features such as luminance, hue, or orientation. A psychophysical experiment was run to measure how human visual observers perform perceptual grouping across a variety of spatial scales and luminance differences. These results were compared with the predictions of our grouping model, and the model was able to capture much of the grouping behavior of the human subjects. A second experiment was run in which the perception of groups was disrupted by the presence of noise or shifts in brightness. Though the experiments showed only small effects resulting from these disruptions on the behavior of human subjects, the model was still able to successfully capture much of the image-to-image variability. For the question of silhouette representation and analysis, I suggest that human silhouette representation may be inextricably tied to 3D interpretation of 2D shapes. To support this, I propose a novel algorithm for 2D silhouette inflation called Puffball, which closely matches human intuition for a variety of simple shapes and can be run on almost any input. Using this algorithm, a new model of human part segmentation was derived using 2D-to-3D inflation; this model was evaluated against human-generated part segmentations and two competing part segmentation algorithms. Across a variety of different analyses, Puffball part segmentation performed as well or better than its competitors, suggesting a potential role for 2D-to-3D inflation in the segmentation of silhouette parts. Finally, I suggest several avenues of research which may further illuminate the role of inflation in the human representation and analysis of 2D and 3D shape. / by Nathaniel R. Twarog. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Brain and Cognitive Sciences, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references. / How animals establish and maintain the sizes of myriad tissues and organs in tight proportion to one another is a fundamental question of developmental biology. Planarian flatworms regenerate from diverse injuries, in each case precisely restoring body parts to their appropriate proportions. Underlying this ability is a pluripotent population of dividing cells called neoblasts, which are required for homeostatic maintenance and regeneration of all planarian tissues. Whether neoblasts restore proportion by sensing and responding to the presence or absence of specific tissues during regeneration is unknown. We used the planarian eye lineage to address this problem. Following decapitation, neoblasts normally give rise to a large number of eye progenitors, facilitating eye regeneration. Remarkably, we found that eye absence alone was not sufficient to induce this response. Tissue-specific eye regeneration was achieved by homeostatic eye progenitor production accompanied by a decreased rate of cell death specifically in the regenerating eye. Conversely, large head wounds were sufficient to increase eye progenitors, even in the presence of intact eyes. Therefore, eye absence is not sufficient or necessary for neoblasts to increase eye progenitor production. Our findings suggest a "target-blind" model for planarian regeneration in which progenitor production by neoblasts does not depend on feedback from the presence or absence of specific target tissues to be regenerated. / by Samuel Alexander LoCascio. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, 1996. / Includes bibliographical references. / by Bhavin R. Sheth. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, February 2007. / Includes bibliographical references. / The origins and adaptive significance of music, long an elusive target, are now active topics of empirical study. I argue that empirical results can constrain discussions of the adaptive significance of music by testing whether various musical traits are innate, uniquely human, and specific to music. This thesis extends the body of empirical research along these lines, with a focus on comparative experiments in nonhuman animals. A series of studies in nonhuman primates explores whether they have preferences for sounds that might be related to music perception in humans. A second set of studies explores whether preferences for music can be instantiated in nonhuman animals by exposure to music. One study examines pet dogs, who receive extensive exposure to music courtesy of their owners. Another examines the effect of artificial music exposure on a colony of laboratory monkeys. Although there are a few potential homologies between the human response to music and that of nonhuman animals, the bulk of our results suggest dramatic differences between humans and other species. This leaves open the possibility of uniquely human music-specific capacities that might constitute an adaptation for music. / by Joshua H. McDermott. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Brain and Cognitive Sciences, September 2011. / Cataloged from PDF version of thesis. "September 2011." / Includes bibliographical references (p. 125-156). / Parkinson's disease (PD) is an age-related degenerative disease of the brain, characterized by motor, cognitive, and psychiatric symptoms. Neurologists and neuroscientists now understand that several symptoms of the disease, including hallucinations and impulse control behaviors, stem from the dopaminergic medications used to control the motor aspects of PD. Not all patients experience these nonmotor symptoms and tools that can predict a priori which patients are likely to have an adverse response to medication do not exist. This thesis begins to fill this gap by elucidating the mechanisms underlying the adverse effects of dopaminergic medications. Converging evidence from animals and humans shows that individual differences in particular genes that affect the dopamine system may alter the response of PD patients to dopaminergic medication. We examined the hypothesis that patients taking dopamine replacement therapy who carry candidate alleles that increase dopamine signaling experience a dopamine overdose, causing unwanted psychiatric symptoms. / by Paymon Ashourian. / Ph.D.