Lesion mapping of cognitive control and value-based decision making in the prefrontal cortex

Abstract

A considerable body of previous research on the prefrontal cortex (PFC) has helped characterize the regional specificity of various cognitive functions, such as cognitive control and decision making. Here we provide definitive findings on this topic, using a neuropsychological approach that takes advantage of a unique dataset accrued over several decades. We applied voxel-based lesion-symptom mapping in 344 individuals with focal lesions (165 involving the PFC) who had been tested on a comprehensive battery of neuropsychological tasks. Two distinct functional-anatomical networks were revealed within the PFC: one associated with cognitive control (response inhibition, conflict monitoring, and switching), which included the dorsolateral prefrontal cortex and anterior cingulate cortex and a second associated with value-based decision-making, which included the orbitofrontal, ventromedial, and frontopolar cortex. Furthermore, cognitive control tasks shared a common performance factor related to set shifting that was linked to the rostral anterior cingulate cortex. By contrast, regions in the ventral PFC were required for decision-making. These findings provide detailed causal evidence for a remarkable functional-anatomical specificity in the human PFC.

Fig. 1.

Results from the lesion overlap analysis of different tests of cognitive control and value-based decision making. VLSM results are thresholded at P < 0.05 (FDR) and projected on a template brain in neurological convention (R = right; all midsaggital hemispheric views are left hemisphere). (A) Results for the TMT, STROOP, WCST, and COWA. (B) Results for the IGT. The graphs at Right show the mean test score (error bars = SEM) for the participants with lesions in the local maximum of each particular test compared with participants with lesions elsewhere. The respective local maximum is highlighted by the white and black circles on the overlay in the middle column. Grayed out areas show regions without sufficient statistical power to detect a lesion deficit effect (see Fig. S2 for complete power maps).

Fig. 2.

Results for all VLSM analyses projected onto a template brain in neurological convention (R = right). (A) All results are thresholded at P < 0.05 (FDR) and coded in different colors. (B) Overlap ratio [(Number of significant voxels in Test A and Test B) / (Number of significant voxels in either A or B)]. This ratio quantifies the volumetric overlap in those regions significant for one task relative to another task (minimum = 0, maximum = 1). Because of the different number of significant voxels in each test (base rate), the overlap matrix is asymmetrical. (C) Scatter plot (r = −0.37, P = 0.0001) of IGT performance and the extent to which individual lesions overlap with the vmPFC. Highlighted are patients who participated in other decision making tasks along with the IGT.

Fig. 3.

(A) Results of a VLSM analysis of the factor scores of a cognitive control factor. The loadings were computed by extracting a single factor using common factor analysis. (B) The cognitive control factor correlates with a region in the ACC that was significant for TMT and WCST (magnification of the saggital slice). (C) Difference image between IGT and the cognitive control factor scores. The images map out the mean differences between both z-scored variables highlighting a valuation network (i.e., patients with lower IGT scores) in blue and a cognitive control network (i.e., patients with lower executive factor scores) in red. (Top) Whole-brain reconstructions (with part of dorsal PFC cut away on the right to visualize internal details). (Middle and Bottom) Slices as indicated.