Neural signatures of compositionality in the human brain

Thesis (Ph. D.)--University of Rochester. Department of Brain & Cognitive Sciences, 2019. Chapter 2 was written with Bradford Mahon and published as Shay, Chen, Garcea, & Mahon (2019) in Cognitive Neuroscience. Chapter 3 was written with Bradford Mahon and is in preparation as Shay, Kersey, Mahon, & Cantlon (in preparation). A version of Chapter 5 was written with Rajeev Raizada and Scott Grimm and published as Shay, Grimm, & Raizada (2017). Chapters 1, 4, and 6 were written by Elizabeth Shay under the guidance of Bradford Mahon and Jessica Cantlon.Since the start of the cognitive revolution, researchers have thought about how our cognitive system handles the large number of stimuli we experience every day. In order for our system to be flexible and able to process and respond to novel stimuli, it is reasonable to suppose that our brains utilize composition. Composition is the process of combining small, simple representations using a composition function to get to more complex representations. For example, within the language domain, semantic composition is the process of combining the meanings of individual words into the meanings of phrases. To understand the complete composition process, we must understand the simple components that act as inputs to the composition process, the composition function, and the outputs of the composition process. Cognitive neuroscience researchers have made some progress on these areas, although mainly on the simple components or the end-state of the composition process – complex representations. The present work aims to add to this understanding of the potential components utilized by the composition function and provide a path for making progress on the broader process of composition. First, I demonstrate the process of finding potential components utilized in composition, as well as the regions important for the neural representation of these components. I explore two domains for this work: motor actions (Chapter 2) and number processing (Chapter 3). I then review and analyze the research performed thus far related to composition in the language, motor, and number domains (Chapter 4). Finally, I discuss some fruitful ways to approach composition through an interdisciplinary approach to research, focusing on combining expertise from the areas of cognitive neuroscience and linguistics (Chapter 5). This approach will allow the field of cognitive neuroscience to capitalize on progress made in composition within other fields. Taken together, this work provides some potential simple components that could be utilized in composition, as well as important regions of the brain that may be involved in representing these components and/or their composition. These findings, as well as the outline for making progress in research by working in interdisciplinary teams, provides a framework for research on the composition of neural representations