Francisco Barceló

Cognitive control enables flexible adaptation of behavior, but it has associated costs: task rules must be held in memory and updated when they change. It remains unclear whether these costs arise from a unitary control bottleneck or from the interaction of multiple computational processes, and how such processes are implemented in the brain. Here we recast rule maintenance and switching within a computational model based on probabilistic (active) inference, which we apply to trial-by-trial behavioral and EEG data from a cued task-switching paradigm. Behavior showed a speed–accuracy dissociation, with slower responses during rule repetition and reduced accuracy during rule switching. We identified three separable model parameters as necessary to explain behavior: context stability, cue sensitivity and action precision. These parameters were dynamically reweighted across task contexts, with context stability predominating during rule maintenance and cue sensitivity during rule updating. They also influenced behavior through dissociable routes: higher action precision selectively improved accuracy without slowing responses. Using hierarchical Parametric Empirical Bayes with out-of-sample cross-validation, we related these latent parameters to electrophysiological signals, revealing both context-invariant and context-specific mappings: parietal P3b latency tracked cue sensitivity across conditions, whereas frontal beta latency and the frontal P3a indexed context stability in a context-dependent way. The neural implementation of context stability reorganized with task demands, shifting from frontal beta during switching to frontal theta during maintenance. Together, these findings move beyond the bottleneck view, suggesting that cognitive control is best understood as an adaptive computational architecture that integrates endogenous and exogenous information while reconfiguring its neural implementation according to task demands.

Cognitive flexibility, the ability to adjust behavior to changing task demands, is critically dependent on distinct control processes supported by a multiple demand frontoparietal cortical network. While behavioral deficits following focal frontal lesions are well documented, most past studies relied on global indices of accuracy or reaction time, without a detailed analysis of specific error types. Here, we investigated task-switching behavior in patients with focal frontal (n = 18) and parietal (n = 8) brain lesions—and in their age-matched healthy controls—by employing a fine-grained taxonomy of error subtypes. Group differences were tested separately for each of the first three target trials following switch and repeat cues using Kruskal–Wallis tests, followed by Dunn’s post-hoc pairwise comparisons when appropriate.Distinct error profiles emerged, with subtle differences across clinical groups: perseverative errors denoted failures in task-setting or rule updating mostly in frontal patients, distraction errors reflected deficits in monitoring or working memory maintenance mainly in parietal patients, whereas omission errors were linked to impaired energization or alerting in both clinical groups. Crucially, patients’ correct responses to target trials were paradoxically preserved most of the time, underscoring that their impairments reflect time-varying fluctuations in the efficiency of cognitive control rather than permanent cognitive deficits or global motor slowing. Complementary machine learning analyses confirmed that multivariate patterns of post-cue errors provided sensitive markers of brain injury, although they did not neatly distinguish between frontal and parietal damage. Together, these findings support evidence for a novel mechanistic account of dysexecutive deficits, suggesting that both frontal and parietal nodes of the frontoparietal network jointly contribute to task-setting, monitoring, and energization subcomponents of cognitive control, although selective damage to different frontoparietal nodes gives rise to subtly distinct behavioral error profiles.

For decades, a common assumption in cognitive neuroscience has been that prefrontal executive control is mainly engaged during target detection [Posner, M. I., & Petersen, S. E. The attention system of the human brain. Annual Review of Neuroscience, 13, 25–42, 1990]. More recently, predictive processing theories of frontal function under the Bayesian brain hypothesis emphasize a key role of proactive control for anticipatory action selection (i.e., planning as active inference). Here, we review evidence of fast and widespread EEG and magnetoencephalographic fronto-temporo-parietal cortical activations elicited by feedback cues and target cards in the Wisconsin Card Sorting Test. This evidence is best interpreted when considering negative and positive feedback as predictive cues (i.e., sensory outcomes) for proactively updating beliefs about unknown perceptual categories. Such predictive cues inform posterior beliefs about high-level hidden categories governing subsequent response selection at target onset. Quite remarkably, these new views concur with Don Stuss' early findings concerning two broad classes of P300 cortical responses evoked by feedback cues and target cards in a computerized Wisconsin Card Sorting Test analogue. Stuss' discussion of those P300 responses—in terms of the resolution of uncertainty about response (policy) selection as well as the participants' expectancies for future perceptual or motor activities and their timing—was prescient of current predictive processing and active (Bayesian) inference theories. From these new premises, a domain-general frontoparietal cortical network is rapidly engaged during two temporarily distinct stages of inference and learning of perceptual categories that underwrite goal-directed card sorting behavior, and they each engage prefrontal executive functions in fundamentally distinct ways.

For decades a common assumption in Cognitive Neuroscience has been that prefrontal executive control is mainly engaged during target detection (Posner & Petersen, 1990, 3: 25–42, Ann Rev Neurosci.). More recently, predictive processing theories of frontal function under the Bayesian brain hypothesis emphasize a key role of proactive control for anticipatory action selection (i.e., planning as active inference). Here, we review evidence of fast and widespread electroencephalographic (EEG) and magnetoencephalographic (MEG) fronto-temporo-parietal cortical activations elicited by feedback cues and target cards in the Wisconsin Card Sorting Test (WCST). This evidence is best interpreted when considering negative and positive feedback as predictive cues (i.e., sensory outcomes) for proactively updating beliefs about unknown perceptual categories. Such predictive cues inform posterior beliefs about high-level hidden categories governing subsequent response selection at target onset. Quite remarkably, these new views concur with Don Stuss' early findings concerning two broad classes of P300 cortical responses evoked by feedback cues and target cards in a computerized WCST analogue. Stuss' discussion of those P300 responses —in terms of the resolution of uncertainty about response (policy) selection, and the subject's expectancies for future perceptual or motor activities and their timing— were prescient of current predictive processing and active (Bayesian) inference theories. From these new premises, a domain-general frontoparietal cortical network is rapidly engaged during two temporarily distinct stages of inference and learning of perceptual categories that underwrite goal-directed card sorting behavior, and they each engage prefrontal executive functions in fundamentally distinct ways.

Quantifying Contextual Information For Cognitive Control

Cognitive control warrants efficient task performance in dynamic and changing environments through adjustments in executive attention, stimulus and response selection. The well-known P300 component of the human event-related potential (ERP) has long been proposed to index “context-updating”—critical for cognitive control—in simple target detection tasks. However, task switching ERP studies have revealed both target P3 (300–350 ms) and later sustained P3-like potentials (400–1,200 ms) to first targets ensuing transition cues, although it remains unclear whether these target P3-like potentials also reflect context updating operations. To address this question, we applied novel single-trial EEG analyses—residue iteration decomposition (RIDE)—in order to disentangle target P3 sub-components in a sample of 22 young adults while they either repeated or switched (updated) task rules. The rationale was to revise the context updating hypothesis of P300 elicitation in the light of new evidence suggesting that “the context” consists of not only the sensory units of stimulation, but also associated motor units, and intermediate low- and high-order sensorimotor units, all of which may need to be dynamically updated on a trial by trial basis. The results showed functionally distinct target P3-like potentials in stimulus-locked, response-locked, and intermediate RIDE component clusters overlying parietal and frontal regions, implying multiple functionally distinct, though temporarily overlapping context updating operations. These findings support a reformulated version of the context updating hypothesis, and reveal a rich family of distinct target P3-like sub-components during the reactive control of target detection in task-switching, plausibly indexing the complex and dynamic workings of frontoparietal cortical networks subserving cognitive control.

One common assumption has been that prefrontal executive control is mostly required for target detection (Posner and Petersen in Ann Rev Neurosci 13:25–42, 1990). Alternatively, cognitive control has also been related to anticipatory updating of task-set (contextual) information, a view that highlights proactive control processes. Frontoparietal cortical networks contribute to both proactive control and reactive target detection, although their fast dynamics are still largely unexplored. To examine this, we analyzed rapid magnetoencephalographic (MEG) source activations elicited by task cues and target cards in a task-cueing analogue of the Wisconsin Card Sorting Test. A single-task (color sorting) condition with equivalent perceptual and motor demands was used as a control. Our results revealed fast, transient and largely switch-specific MEG activations across frontoparietal and cingulo-opercular regions in anticipation of target cards, including (1) early (100–200 ms) cue-locked MEG signals at visual, temporo-parietal and prefrontal cortices of the right hemisphere (i.e., calcarine sulcus, precuneus, inferior frontal gyrus, anterior insula and supramarginal gyrus); and (2) later cue-locked MEG signals at the right anterior and posterior insula (200–300 ms) and the left temporo-parietal junction (300–500 ms). In all cases larger MEG signal intensity was observed in switch relative to repeat cueing conditions. Finally, behavioral restart costs and test scores of working memory capacity (forward digit span) correlated with cue-locked MEG activations at key nodes of the frontoparietal network. Together, our findings suggest that proactive cognitive control of task rule updating can be fast and transiently implemented within less than a second and in anticipation of target detection.

ERP research on task switching has revealed distinct transient and sustained positive waveforms (latency circa 300–900 ms) while shifting task rules or stimulus‐response (S‐R) mappings. However, it remains unclear whether such switch‐related positivities show similar scalp topography and index context‐updating mechanisms akin to those posed for domain‐general (i.e., classic P300) positivities in many task domains. To examine this question, ERPs were recorded from 31 young adults (18–30 years) while they were intermittently cued to switch or repeat their perceptual categorization of Gabor gratings varying in color and thickness (switch task), or else they performed two visually identical control tasks (go/no‐go and oddball). Our task cueing paradigm examined two temporarily distinct stages of proactive rule updating and reactive rule execution. A simple information theory model helped us gauge cognitive demands under distinct temporal and task contexts in terms of low‐level S‐R pathways and higher‐order rule updating operations. Task demands modulated domain‐general (indexed by classic oddball P3) and switch positivities—indexed by both a cue‐locked late positive complex and a sustained positivity ensuing task transitions. Topographic scalp analyses confirmed subtle yet significant split‐second changes in the configuration of neural sources for both domain‐general P3s and switch positivities as a function of both the temporal and task context. These findings partly meet predictions from information estimates, and are compatible with a family of P3‐like potentials indexing functionally distinct neural operations within a common frontoparietal “multiple demand” system during the preparation and execution of simple task rules.