Shades of Meaning: Capturing Meaningful Context-Based Variations in Neural Patterns

When cognitive psychologists and psycholinguists consider the variability that arises during the retrieval of conceptual information, this variability it is often understood to arise from the dynamic interactions between concepts and contexts. When cognitive neuroscientists and neurolinguists think about this variability, it is typically treated as noise and discarded from the analyses. In this dissertation, we bridge these two traditions by asking: can the variability in neural patterns evoked by word meanings reflect the contextual variation that occurs during conceptual processing? We employ functional magnetic resonance imaging (fMRI) to measure, quantify, and predict brain activity during context-dependent retrieval of word meanings. Across three experiments, we test the ways in which word-evoked neural variability is influenced by the sentence context in which the word appears (Chapter 2); the current set of task demands (Chapter 3); or even undirected thoughts about other concepts (Chapter 4). Our findings indicate that not only do the neural patterns evoked by the same stimulus word vary over time, but we can predict the degree to which these patterns vary using meaningful, theoretically motivated variables. These results demonstrate that cross-context, within-concept variations in neural responses are not exclusively due to statistical noise or measurement error. Rather, the degree of a concept’s neural variability varies in a manner that accords with a context-dependent view of semantic representation. In addition, we present preliminary evidence that prefrontally-mediated cognitive control processes are involved in expression of context-appropriate neural patterns. In sum, these studies provide a novel perspective on the flexibility of word meanings and the variable brain activity patterns associated with them.