ORCID

https://orcid.org/0000-0002-7956-2609

Date of Award

Winter 12-15-2016

Author's School

Graduate School of Arts and Sciences

Author's Department

Psychology

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

What patterns of brain activity reflect engagement with highly demanding cognitive tasks? How do these patterns relate to subjective, phenomenal effort? Answering these questions is critical to understanding what causes some people to experience cognitive tasks as more effortful than others. Subjective experience, in turn, is vital, with trait tendencies to exert effort having been linked to career and academic success. High subjective effort, as in schizophrenia and depression, can thus be extremely problematic. And yet, poor operational definitions have constrained research into basic questions about what neural dynamics track subjective effort. Here, a powerful, new behavioral economic operationalization is employed, in combination with fMRI, to investigate brain dynamics corresponding to subjectively costly cognitive effort. Brain regions varying in activity by working memory load and cognitive control demands are strong candidates for tracking subjective effort (Westbrook & Braver, 2015). To identify such regions, I examined BOLD data, collected while participants performed a well established working memory task (the N-back; Kirchner, 1958) that is both subjectively effortful, and for which subjective effort varies as a monotonic function of load (Westbrook et al., 2013). I focused my search within independently-defined networks of nodes that co-vary (within-network) across a wide range of brain states. Specifically, I examined a subset of a priori "task-positive" networks, as identified by Power et al. (2011), which typically show increasing, and a "task-negative" network which typically shows decreasing activity with greater load. Importantly, variation was examined over N-back loads for which data has never been published, thus the present study reveals novel insights about activity-load functions in independently-defined functional networks from very low (N = 1) to very high loads (N = 6). As expected, all task-positive networks showed robustly greater activity during the N-back. However, patterns of variation by load differed by network. While the task positive fronto-parietal (FP), dorsal attention (DorAtt), and salience (Sal) networks showed inverted-U functions, peaking mid-range (at the 2- or 3-back) and decreasing after, the cingulo-opercular network (CO) showed robust activity that did not further vary by load. Rather than encoding load per se, the CO simply encoded that a participant was performing the N-back. The task-negative default mode network (DMN) was robustly and increasingly de activated across all load levels examined. Given that both subjective effort (Westbrook et al., 2013) and DMN deactivation are approximately monotonic functions of load, the DMN is a strong candidate for tracking variation in subjective effort with load. By contrast, inverted-U functions in the FP, Sal, and DorAtt networks do not straightforwardly map to monotonically increasing effort. Performance measures instead suggest that inverted-U functions tracked individual differences in adaptive strategy shifting. Namely, when participants were divided by 3-back performance, better performers showed a pronounced inverted-U (over N = 1—3) while worse performers did not. Interestingly, a similar pattern was found when dividing participants according subjective effort, providing tentative support to a hypothesis that subjective effort acts as a cue to shift strategies adaptively under excessive demands. In any case, surprisingly, in none of the networks did load-specific changes in brain activity predict load-specific changes in subjective effort. Critically, although load-specific patterns of brain activity did not predict subjective effort, load-independent brain activity predicted individual differences in subjective effort. Namely, higher average brain activity in any of the task-positive networks predicted greater subjective effort. At the sub-network level, this was notably true for two key regions that have been implicated as core components of a cognitive control system, and also hypothesized to track effort costs: the dorsal anterior cingulate cortex (dACC) and the dorsolateral prefrontal cortex (dlPFC) (McGuire et al., 2010). Importantly, after controlling for performance, the dACC remained a reliable predictor of subjective effort, while the dlPFC did not, supporting that the dACC tracks cognitive effort apart from task difficulty (while the dlPFC may not). This is consistent with strong prior theory implicating the dACC in regulating the intensity of cognitive control in response to flagging performance and in proportion to the expected value of doing so (Shenhav et al., 2013). The present results begin to answer basic questions about how the brain tracks subjective effort. They also lay the foundation for future work addressing why subjective effort can be so much greater for some individuals, like those with schizophrenia or depression, and also future work developing interventions for promoting desirable effort expenditure.

Language

English (en)

Chair and Committee

Todd S. Braver

Committee Members

Cynthia Cryder, Ian Dobbins, Joel Myerson, Jeffrey Zacks

Comments

Permanent URL: https://doi.org/10.7936/K7SF2TMM

Share

COinS