UC Berkeley

Parietal cortex has long been considered essential for complex executive functions necessary during human decision-making (Mesulam, 1998). Following an introduction, Chapter 2 of this dissertation investigates one particularly pertinent region of parietal cortex called the intraparietal sulcus (IPS) in sensory and abstract decision-making processes. The IPS has been implicated in numerous functions that range from representation of visual stimuli to action planning, but its role in abstract decision-making has been unclear, in part because functions in sensorimotor processing often act as confounds. This thesis addresses this problem using a novel task developed to dissociate abstract decision-making from sensory salience, attentional control, and motor output. Functional MRI data were collected from healthy female and male subjects while they performed a policy abstraction task requiring use of a more abstract (second-order) rule to select between two less abstract (first-order) rules that informed the motor response. By identifying IPS subdivisions with preferential connectivity to prefrontal regions that are differentially responsive to task abstraction, we found that a caudal IPS (cIPS) subregion with strongest connectivity to the pre-premotor cortex (pre-PMd) was preferentially active for second-order cues, while a rostral IPS subregion (rIPS) with strongest connectivity to the dorsal premotor cortex (PMd) was active for attentional control over first-order cues. These effects for abstraction were seen in addition to cIPS activity that was specific to sensory salience, and rIPS activity that was specific to motor output. Notably, topographic responses to the second-order cue were detected along the caudal-rostral axis of IPS, mirroring the broader organization seen in lateral prefrontal cortex . Together these data demonstrate that subregions within IPS exhibit activity responsive to policy abstraction, and they suggest that IPS may be organized into fronto-parietal subnetworks that support hierarchical cognitive control.Chapter 3 next assessed whether parietal areas involved in decision-making vary as a function of strategy. One can perform the same goal-oriented task using different strategies (Morrison, Rosenbaum, Fair, & Chein, 2016). Studies have shown dissimilarities in neural activity for strategies that depend on unique sensory or mnemonic systems (Iaria, Petrides, Dagher, Pike, & Bohbot, 2003), but few have investigated if different strategies in performing a cognitive control task produce activation differences within the fronto-parietal control network if those strategies differ in the form of abstraction. Half of all trials in the same abstract decision-making task developed for this thesis required use of a hierarchically-defined “policy strategy”, and the other half permitted a heterarchical match/nonmatch “comparison strategy” (Badre, 2008). This unique design feature was next used to investigate whether frontal and parietal control regions implicated in policy abstraction differentially represent strategy when it employs different forms of abstraction. Previously collected fMRI data from healthy female and male subjects was re-analyzed. Accuracy was higher for trials allowing the comparison strategy, and a significant behavioral interaction was observed between sensory coherence of the second-order cue and strategy trial type. Across the brain, known decision-making regions for hierarchical policy abstraction tasks had increased activity for the policy strategy, including frontal regions pre-PMd and PMd. A region in the superior frontal sulcus covaried activity with subject strategy use. However, parietal activity did not strongly differentiate between the strategies, suggesting that IPS subregions specifically important for policy abstraction decision information may not preferentially represent control of hierarchical policy abstraction over other non-hierarchical types of task abstraction. Finally, Chapter 4 explores if retained versus removed working memory items are reflected in parietal activity during decision-making. Every discrete decision requires selection from at least two unique options, and in the absence of consistent external representations, those options are held in working memory. As a result, relevant working memory decision items are retained for use in a future motor response while irrelevant items must be removed to reduce retrieval interference (Lewis-Peacock, Drysdale, Oberauer, & Postle, 2012). Parietal cortex and IPS specifically are associated with working memory item representation and control (Chatham, Frank, & Badre, 2014; Cowan et al., 2011; M. D’Esposito et al., 1998), implicating it in possible retention and removal processes during decision-making. To test whether an active process of removal and/or retention exists in parietal cortex, fMRI data collected from the same abstract decision-making task developed for this thesis work were re-analyzed. A subset of the data were selected where second-order cues were presented in position 2 within a trial, effectively acting as a retro-cue for the relevance or irrelevance of the previously presented cue in position 1. Investigating activity at this timepoint would reveal working memory item removal if the previous cue was irrelevant, and retention if the cue was relevant. Hypothesis 1 predicted that parietal cortex would reflect an active process of content unbinding during irrelevant working memory item removal, but results instead suggested that removal may be a latent process that is undetectable via univariate imaging. Hypothesis 2 investigated working memory item retention and predicted that rIPS would increase activity for retaining response-relevant first-order contents while cIPS would not change activity for working memory control of first-order information. Contrary to these hypotheses, results revealed that rIPS did not significantly change activity for working memory item retention of relevant first-order information, while cIPS increased activity at this timepoint. This finding supports an alternative hypothesis that cIPS and rIPS should instead be considered to be primarily engaged with decision-making aspects of the policy abstraction task, and not necessarily implicated in aspects of working memory item retention. Continued research is necessary to further expose the neural underpinnings of working memory item removal and retention.