Abstract
The growing literature of task-based functional magnetic resonance imaging (task-fMRI) has increased calls for an adequate organizing ontology, or taxonomy, of task fMRI experiments. Researchers differ over what should be the dominant features of such an ontology: should it be concrete/observable dimensions, such as task paradigm (e.g. n-back vs. flanker task), or latent/theoretical dimensions, such as cognitive domains (e.g. working-memory vs. bottom-up attention)? This dissertation attempts to address what is important for a task-fMRI ontology in a quantifiable manner. We use a simple quantitative criterion for categories/dimensions of a task-fMRI ontology: the ability to explain observed variations in task-fMRI activation patterns Using meta-analysis tools and multivariate statistical methods, we identify those dimensions and categories of the task-fMRI environment that explain observed variations in task-fMRI activation patterns. In study one, we observed that a preliminary ontology of four or seven latent cognitive categories provides a simplified description of the observed differences in whole-brain blood-oxygen-level dependent (BOLD) activation patterns. In study two, we observed that while these categories may provide an adequate description of BOLD activation patterns at the population level, inter-subject variability restrains inferences from these population level distinctions to the subject level. In conclusion, results from both studies suggest that a data-driven task-fMRI ontology is a viable project for cognitive neuroscience.