Flexible Reconfigurations of Brain Networks During Decisions With Predefined Versus Self-Generated Options.

How do large-scale brain networks dynamically reorganize to support different types of decision-making with distinct yet overlapping cognitive demands? While we often make decisions by evaluating and choosing from a set of externally defined choice options, we can also generate options internally from our existing knowledge store. Such flexibility suggests the ability for decision-related brain networks to reconfigure in response to the need to recruit sensory processing and semantic retrieval modules to evaluate externally and internally generated options, respectively. Here we sought to test this hypothesis by applying graph-theoretic tools to functional neuroimaging data obtained for (i) decisions with externally provided options (external-menu choices/EMC); (ii) decisions with self-generated options (internal-menu choices/IMC); and (ii) a semantic fluency condition in which individuals generated but were not required to evaluate options (semantic fluency; SF). Using categorical multi-slice community detection, we found that variations in cognitive demands across the tasks were associated with distinct reconfigurations of hierarchically organized modular brain networks. Specifically, global network organization that differed primarily along the dimension of external visual/sensory input distinguished EMC from both of the internally oriented tasks (IMC and SF). At submodular levels, IMC was distinguished from SF by stronger interactions between presumptive semantic retrieval and valuation networks hypothesized to support the generation and evaluation of choice options. These findings are consistent with a hierarchical architecture in which modules at multiple levels interact to support adaptive decision-making.