OBJECTIVE: The functional strategic mechanisms in the brain during performing visuospatial working memory tasks, especially tasks with heavy load, are controversial. We conducted the functional magnetic resonance imaging (fMRI) while sixteen subjects were performing face- and location-matching n-back tasks to examine causal relations within the frontoparietal networks. METHODS: We applied a sophisticated method, the structural equation modeling (SEM), to the fMRI data. The imaging data were analyzed by extracting the task-related eigenseries using the principal component analysis (PCA) and then by applying a form of data-driven model called the automated search method. RESULTS: The SEM analyses revealed a functional shift of network connectivity from the right to the left hemisphere with increasing load in the face-matching n-back tasks while the location-matching tasks required bilateral activation. In the locating matching n-back tasks, a pattern of parallel processing was observed in the left phonological loop and the right inferior parietal regions. Furthermore, object working memory-related activities in the left hemisphere reliably contributed to performance of both the face- and location-matching 2-back tasks. CONCLUSION: Our results are consistent with previous reports in terms of demonstrating parallel and distributed information processing during performing working memory tasks with heavy loads. Our results additionally suggest a dynamic shift between the fast imagery circuit (right hemisphere) and the stable verbal circuit (left hemisphere), depending on task load.
Automatic Data Processing ; Brain ; Magnetic Resonance Imaging ; Memory, Short-Term ; Principal Component Analysis