Dissociable contributions of ventromedial prefrontal and posterior parietal cortex to value-guided choice

Abstract

Two long-standing traditions have highlighted cortical decision mechanisms in the parietal and prefrontal cortices of primates, but it has not been clear how these processes differ, or when each cortical region may influence behaviour. Recent data from ventromedial prefrontal cortex (vmPFC) and posterior parietal cortex (PPC) have suggested one possible axis on which the two decision processes might be delineated. Fast decisions may be resolved primarily by parietal mechanisms, whereas decisions made without time pressure may rely on prefrontal mechanisms. Here, we report direct evidence for such dissociation. During decisions under time pressure, a value comparison process was evident in PPC, but not in vmPFC. Value-related activity was still found in vmPFC under time pressure. However, vmPFC represented overall input value rather than compared output value. In contrast, when decisions were made without time pressure, vmPFC transitioned to encode a value comparison while value-related parameters were entirely absent from PPC. Furthermore, under time pressure, decision performance was primarily governed by PPC, while it was dominated by vmPFC at longer decision times. These data demonstrate that parallel cortical mechanisms may resolve the same choices in differing circumstances, and offer an explanation of the diverse neural signals reported in vmPFC and PPC during value-guided choice.

Keywords: Decision making; Parietal cortex; Reward; Ventromedial prefrontal cortex; fMRI.

Fig. 1

Task schematic. Short, middle and long trials were grouped in alternating blocks of 10 trials. Every 10 trials, a precue signalled the condition for the next 10 trials. In the short and middle condition, subjects could respond as soon as the options were onscreen. In the long condition, there was a fixed viewing period of 3 s before the central question mark appeared, prompting them to respond within 1 s.

Fig. 2

Behavioural results for the three conditions. A) Percent choices of the higher value option. B) Parameters from prospect theory for subjective weighting of reward magnitude (α) and probability (γ), and softmax temperature (τ). C) Plotting the objective versus subjective magnitudes using the weighting parameter displayed in (B) shows underweighting of reward magnitudes in the middle and long, but not short condition. D–F) Regression of experimental parameters against subjects' choice. Positive values indicate that the parameter increases the probability to select the option. RP: reward probability, RM: reward magnitude, pO: outcome on previous trial, pC: choice on previous trial, RP × RT: interaction between reward probability and reaction time, RM × RT: interaction between reward magnitude and reaction time.

Fig. 3

Results for the behavioural control experiment. Bars represent regression coefficients (mean ± SEM) obtained from a regression of experimental parameters against subjects' choices. Positive values indicate that the parameter increases the probability to select the option. RP: reward probability, RM: reward magnitude, pO: outcome on previous trial, pC: choice on previous trial, RP × RT: interaction between reward probability and reaction time, RM × RT: interaction between reward magnitude and reaction time.

Fig. 4

Whole-brain results and masks used for ROI analyses. A) Effect of value difference at p < 0.001 in the vmPFC and posterior cingulate (left) and, at lower threshold (p < 0.01) in the pSPL (right). B) Regions of interest selected from a previous study showing activity related to value difference in the vmPFC (green), midcingulate cortex (yellow), posterior cingulate cortex (blue) and pSPL (red).

Fig. 5

Timecourse of value-related effects in the ventromedial prefrontal cortex (vmPFC) over the course of a trial in the three conditions. Left column: effects of value difference and value sum on BOLD activity in vmPFC. Right column: Effects of chosen and unchosen option value on vmPFC BOLD activity. Solid lines represent mean effect sizes across participants, and shaded areas are standard error of the mean.

Fig. 6

Timecourse of value-related effects in the posterior superior parietal lobule (pSPL) over the course of a trial in the three conditions. Left column: effects of value difference and value sum on BOLD activity in pSPL. Right column: Effects of chosen and unchosen option values on pSPL BOLD activity. Solid lines represent mean effect sizes across participants, and shaded areas are standard error of the mean.