Cognitive modes detectable by functional magnetic resonance imaging during working memory

Task-based functional magnetic resonance imaging (fMRI) can be used to identify distinct spatiotemporal configurations of blood-oxygen-level-dependent (BOLD) activity associated with specific cognitive functions. However, separating cognitive processes that coincide and occur sequentially in time remains a challenge. This is particularly relevant for working memory (WM) tasks, where encoding, maintenance and response/probe phases must follow a fixed order. In this thesis, I investigated two approaches for improving the separation of spatiotemporal BOLD configurations underlying WM during the Sternberg Item Recognition Paradigm (SIRP). First, I combined fMRI data from a newly collected SIRP task with data from a previously published Thought Generation (TG) task. In this SIRP version, encoding, maintenance and probe phases were presented sequentially in a fixed order. Both tasks required the maintenance of attention to internal representations, but the TG task does not require a motor response, allowing temporal separation of maintenance and probe phases. Six cognitive modes emerged using fMRI-CPCA: Initiation (INIT), Peripheral versus Central visual processing (PvC), Maintaining Internal Attention (MAIN), Sensorimotor (SM), Default Mode B (DMB), and Auditory Perception (AUD). Separation of MAIN and SM was successful, but some overlap remained. Secondly, I analyzed a publicly available dataset from the functional Biomedical Informatics Research Network (fBIRN). In this SIRP version, multiple probe presentations separated by jittered delays followed each encoding period, providing another approach for separating temporally overlapping maintenance and probe phases. Five cognitive modes emerged: INIT, MAIN, SM, DMB, and Somatosensory Integration (SSI). Separation of MAIN and SM was successful, but SM did not emerge robustly. Together, these analyses demonstrate two distinct approaches helpful for separating maintenance from probe phases in the SIRP, but one could not be considered a more optimal method than the other. INIT, MAIN, SM and DMB replicated across both analyses, and replicated configurations observed in past SIRP and non-SIRP research. These findings demonstrate methods for separating task phases for fMRI, support the interpretation of cognitive modes as task-general processes, confirm replication of cognitive modes across WM and other tasks, and demonstrate how fMRI-detectable cognitive processes involve the temporally organized and coordinated engagement of multiple large-scale brain systems.