Research & Projects
Research & Projects
Previous work on hierarchical representation structures has suggested a rostral to caudal gradient for hierarchical representations where more concrete levels are encoded more caudally and less concrete levels (e.g., context) are encoded more rostrally. However, much of this previous work has used distinct cues/tasks to create a hierarchical task representation structure.
Therefore, we utilized representational similarity analysis (RSA) with a hierarchical control task where the hierarchical representation structure was embedded in a single cue object. We then investigated if distinct hierarchical levels showed distinct representational similarity patterns in distinct regions of interest (ROIs).
Distinct hierarchical levels were associated with distinct ROIs showing greater representational similarity when either context, subordinate rule, task, or response decision matched.
Everyday behavior requires the integration of many diverse and ambiguous/volatile sources of information in order to engage in successful goal-directed behavior. Despite this fact, most studies on this topic focus on the integration of non-ambiguous & observable perceptual inputs.
Therefore, the primary objectives of this project were to leverage probabilistic (i.e., ambiguous) inputs that were both observable and non-observable to more fully characterize the integration process underlying the formation of task representations that support goal-directed behavior.
Our primary finding was that frontoparietal hub regions utilize integrated uncertainty (entropy) to drive cognitive flexibility (task output and updating) required for goal-directed behavior.
Cognitive flexibility relies heavily on contextual representations that can be utilized to determine what aspects of the environment are currently goal-relevant. Distinct contexts are often associated with distinct perceptual cues. However, previous work investigating hierarchical representation structures incorporating contextual information have largely ignored interactions between underlying hierarchical representation structures and perceptual systems.
Therefore, we sought to address this gap by considering perceptual influences on trial-to-trial neural reconfiguration estimates. More specifically, we investigated whether the reconfiguration of distinct hierarchical levels (i.e., context versus subordinate rules) was associated with more or less interference from perceptual factors associated with task cues.
Briefly, the reconfiguration of subordinate rules showed susceptibility to perceptual interference whereas contextual reconfiguration appeared largely shielded from perceptual interference.