Frontoparietal representations of task context support the flexible control of goal-directed cognition

Abstract

Cognitive control allows stimulus-response processing to be aligned with internal goals and is thus central to intelligent, purposeful behavior. Control is thought to depend in part on the active representation of task information in prefrontal cortex (PFC), which provides a source of contextual bias on perception, decision making, and action. In the present study, we investigated the organization, influences, and consequences of context representation as human subjects performed a cued sorting task that required them to flexibly judge the relationship between pairs of multivalent stimuli. Using a connectivity-based parcellation of PFC and multivariate decoding analyses, we determined that context is specifically and transiently represented in a region spanning the inferior frontal sulcus during context-dependent decision making. We also found strong evidence that decision context is represented within the intraparietal sulcus, an area previously shown to be functionally networked with the inferior frontal sulcus at rest and during task performance. Rule-guided allocation of attention to different stimulus dimensions produced discriminable patterns of activation in visual cortex, providing a signature of top-down bias over perception. Furthermore, demands on cognitive control arising from the task structure modulated context representation, which was found to be strongest after a shift in task rules. When context representation in frontoparietal areas increased in strength, as measured by the discriminability of high-dimensional activation patterns, the bias on attended stimulus features was enhanced. These results provide novel evidence that illuminates the mechanisms by which humans flexibly guide behavior in complex environments.

Keywords: attention; cognitive control; decision making; prefrontal cortex.

Figure 1.

Experimental design. The task was organized into miniblocks of three consecutive trials with the same active rules, which were cued at the start of the block. The stimuli in this figure demonstrate the two possible feature values along the dimensions of shape, color, and pattern.

Figure 2.

Main behavioral results. Mean within-subject median reaction time (top row) and mean within-subject proportion correct responses (bottom row) sorted by the two rule sets (A, B), by the decision rule and whether the attended features matched or differed (C, D), and by the number of trials since the rule switch plotted separately with respect to the dimension and decision rules (E, F). The time courses in E and F are dashed between trials 3 and 4 to indicate passage through a miniblock boundary. Error bars on all facets indicate bootstrapped SEs of the aggregate values across subjects. RTs are plotted only for trials included in the imaging analyses.

Figure 3.

Targeted regions for decoding analyses. A, Source labels defining the ROIs on the Freesurfer average cortical surface. Please see the original reference (Yeo et al., 2011) for more information and additional visual perspectives. B, Schematic diagram of the procedure used to label voxels in native functional space for decoding analyses. The second panel shows contours of the template curvature plotted in the spherical coordinate system over the curvature from a representative subject. Darker areas indicate sulci. Cortical surfaces were visualized using PySurfer (http://pysurfer.github.io/).

Figure 4.

Decoding results for the dimension rules in prefrontal cortex. A, Time-resolved decoding results. Solid lines indicate mean decoding accuracy within each time bin and error bands denote bootstrapped SE across subjects. The horizontal dashed line shows the empirical measure of chance performance derived from the permutation analysis. Vertical dashed line is placed at the onset of the first stimulus. B, Height of each bar shows the percentile in the shuffled null distribution corresponding to the observed accuracy for each subject and region. Horizontal dashed line demarcates the criterion of significance at a (corrected) α = 0.05. Bars are sorted by height within region. C, D, Decoding accuracies fit to data averaged across the time points at 3 and 5 s after stimulus onset. Points and error bars represent the mean and bootstrapped SE across subjects, respectively.

Figure 5.

Searchlight analysis of the dimension rules in prefrontal cortex. A, Map of group-average searchlight accuracy after surface-based normalization and smoothing. The map is thresholded at p < 0.005 (uncorrected) from a group t test against expected chance. Voxels falling outside of the lateral PFC mask are dimmed and the IFS region used in ROI-based analysis is outlined in gray. B, Example slice through the native functional volume showing the search space. This region was defined by combining the masks for each individual PFC ROI and dilating the result by three voxels. Decoding models were fit within spheres of 10 mm radius and then the resulting accuracy maps were projected onto the surface for group testing. C, Distribution of searchlight accuracies within the PFC mask after group averaging.

Figure 6.

Decoding results for the decision rules in prefrontal cortex. Plot conventions are identical to those in Figure 4. The y-axis is scaled to span similar binomial probabilities as in Figure 4.

Figure 7.

Decoding results for the dimension rules in posterior neocortex. A, Points and error bars show the mean and bootstrapped SE, respectively, for decoding accuracy in the original time bins. The solid traces show the gamma PDF models used to derive temporal information averaged across subjects. Plot conventions are otherwise as in Figure 4A. B, Plot conventions are as in Figure 4C. C, Boxplots showing the distribution of model coefficient autocorrelation across subjects sorted by region. D, Boxplots showing the distribution of relative differences in the time of peak decoding accuracy for the IFS and IPS models relative to the OTC models. Negative numbers indicate later peaks in the OTC.

Figure 8.

Relationship between frontoparietal classifier evidence and visual classifier performance. A, Points show the mean OTC classifier accuracy sorted by binned IFS and IPS classifier evidence (the logit-transformed probability of the target class). Error bars represent the bootstrapped SE across subjects. Solid traces show the predictions of a logistic regression model fit to all data points. Horizontal dashed line shows chance performance. B, C, Histograms showing the distribution of classifier evidence in IFS and IPS for all trials and subjects; the x-axis and bins correspond across all three panels.

Figure 9.

Relationship between control demands and dimension rule decoding. Points and error bars show means and bootstrapped SEs across subjects, respectively. Solid traces show the predictions of a log-linear fit to all data points. Traces are dashed between trials 3 and 4 to indicate passage through a miniblock boundary. Horizontal dashed line shows chance performance.

Figure 10.

Schematic diagram of the relationship between control demands and context representation. Points illustrate high-dimensional patterns of activity across a population of voxels (or neurons) on each trial sorted into different contexts by color. The plane implements a linear classifier, which portrays both our analytical approach and the hypothesized readout mechanism of downstream areas (Rigotti et al., 2013). As control demands increase, patterns associated with different contexts grow more separated in the neural state space, which leads to increased classification accuracy and greater top-down control.