Learning Cognitive Flexibility: Neural Substrates of Adapting Switch-Readiness to Time-varying Demands.

An individual's readiness to switch tasks (cognitive flexibility) varies over time, in part, as the result of reinforcement learning based on the statistical structure of the world around them. Consequently, the behavioral cost associated with task-switching is smaller in contexts where switching is frequent than where it is rare, but the underlying brain mechanisms of this adaptation in cognitive flexibility are not well understood. Here, we manipulated the likelihood of switches across blocks of trials in a classic cued task-switching paradigm while participants underwent fMRI. As anticipated, behavioral switch costs decreased as the probability of switching increased, and neural switch costs were observed in lateral and medial frontoparietal cortex. To study moment-by-moment adjustments in cognitive flexibility at the neural level, we first fitted the behavioral RT data with reinforcement learning algorithms and then used the resulting trial-wise prediction error estimate as a regressor in a model-based fMRI analysis. The results revealed that lateral frontal and parietal cortex activity scaled positively with unsigned switch prediction error and that there were no brain regions encoding signed (i.e., switch- or repeat-specific) prediction error. Taken together, this study documents that adjustments in cognitive flexibility to time-varying switch demands are mediated by frontoparietal cortex tracking the likelihood of forthcoming task switches.

Context-independent scaling of neural responses to task difficulty in the multiple-demand network.

The multiple-demand (MD) network is sensitive to many aspects of cognitive demand, showing increased activation with more difficult tasks. However, it is currently unknown whether the MD network is modulated by the context in which task difficulty is experienced. Using functional magnetic resonance imaging, we examined MD network responses to low, medium, and high difficulty arithmetic problems within 2 cued contexts, an easy versus a hard set. The results showed that MD activity varied reliably with the absolute difficulty of a problem, independent of the context in which the problem was presented. Similarly, MD activity during task execution was independent of the difficulty of the previous trial. Representational similarity analysis further supported that representational distances in the MD network were consistent with a context-independent code. Finally, we identified several regions outside the MD network that showed context-dependent coding, including the inferior parietal lobule, paracentral lobule, posterior insula, and large areas of the visual cortex. In sum, a cognitive effort is processed by the MD network in a context-independent manner. We suggest that this absolute coding of cognitive demand in the MD network reflects the limited range of task difficulty that can be supported by the cognitive apparatus.

One-shot stimulus-control associations generalize over different stimulus viewpoints and exemplars.

Cognitive control processes are central to adaptive behavior, but how control is applied in a context-appropriate manner is not fully understood. One way to produce context-sensitive control is by mnemonically linking particular control settings to specific stimuli that demanded those settings in a prior encounter. In support of this episodic reinstatement of control hypothesis, recent studies have produced evidence for the formation of stimulus-control associations in one-shot, prime-probe learning paradigms. However, since those studies employed perceptually identical stimuli across prime and probe presentations, it is not yet known how generalizable one-shot stimulus-control associations are. In the current study, we therefore probed whether associations formed between a prime object and the control process of task-switching would generalize to probe objects seen from a different viewpoint (Experiment 1), to different exemplars of the same object type (Experiment 2), and to different members of the object category (Experiment 3). We replicated prior findings of one-shot control associations for identical prime/probe stimuli. Importantly, we additionally found that these episodic control effects are expressed regardless of changes in viewpoint and exemplar, but do not seem to generalize to other category members. These findings elucidate the scope of generalization of the episodic reinstatement of cognitive control.

The Neural Correlates of Updating and Gating in Procedural Working Memory.

Goal-directed behavior relies on maintaining relevant goals in working memory (WM) and updating them when required. Computational modeling, behavioral, and neuroimaging work has previously identified the processes and brain regions involved in selecting, updating, and maintaining declarative information, such as letters and pictures. However, the neural substrates that underlie the analogous processes that operate on procedural information, namely, task goals, are currently unknown. Forty-three participants were therefore scanned with fMRI while performing a procedural version of the reference-back paradigm that allowed for the decomposition of WM updating processes into gate-opening, gate-closing, task switching, and task cue conflict components. Significant behavioral costs were observed for each of these components, with interactions indicating facilitation between gate-opening and task switching, and a modulation of cue conflict by gate state. In neural terms, opening the gate to procedural WM was associated with activity in medial pFC, posterior parietal cortex (PPC), the basal ganglia (BG), thalamus, and midbrain, but only when the task set needed to be updated. Closing the gate to procedural WM was associated with frontoparietal and BG activity specifically in conditions where conflicting task cues had to be ignored. Task switching was associated with activity in the medial pFC/ACC, PPC, and BG, whereas cue conflict was associated with PPC and BG activity during gate closing but was abolished when the gate was already closed. These results are discussed in relation to declarative WM and to gating models of WM.

Transfer of Learned Cognitive Flexibility to Novel Stimuli and Task Sets.

Adaptive behavior requires learning about the structure of one's environment to derive optimal action policies, and previous studies have documented transfer of such structural knowledge to bias choices in new environments. Here, we asked whether people could also acquire and transfer more abstract knowledge across different task environments, specifically expectations about cognitive control demands. Over three experiments, participants (Amazon Mechanical Turk workers; N = ~80 adults per group) performed a probabilistic card-sorting task in environments of either a low or high volatility of task rule changes (requiring low or high cognitive flexibility, respectively) before transitioning to a medium-volatility environment. Using reinforcement-learning modeling, we consistently found that previous exposure to high task rule volatilities led to faster adaptation to rule changes in the subsequent transfer phase. These transfers of expectations about cognitive flexibility demands were both task independent (Experiment 2) and stimulus independent (Experiment 3), thus demonstrating the formation and generalization of environmental structure knowledge to guide cognitive control.

Ventromedial Prefrontal Cortex Drives the Prioritization of Self-Associated Stimuli in Working Memory.

Humans show a pervasive bias for processing self- over other-related information, including in working memory (WM), where people prioritize the maintenance of self- (over other-) associated cues. To elucidate the neural mechanisms underlying this self-bias, we paired a self- versus other-associated spatial WM task with fMRI and transcranial direct current stimulation (tDCS) of human participants of both sexes. Maintaining self- (over other-) associated cues resulted in enhanced activity in classic WM regions (frontoparietal cortex), and in superior multivoxel pattern decoding of the cue locations from visual cortex. Moreover, ventromedial PFC (VMPFC) displayed enhanced functional connectivity with WM regions during maintenance of self-associated cues, which predicted individuals' behavioral self-prioritization effects. In a follow-up tDCS experiment, we targeted VMPFC with excitatory (anodal), inhibitory (cathodal), or sham tDCS. Cathodal tDCS eliminated the self-prioritization effect. These findings provide strong converging evidence for a causal role of VMPFC in driving self-prioritization effects in WM and provide a unique window into the interaction between social, self-referential processing and high-level cognitive control processes.SIGNIFICANCE STATEMENT People have a strong tendency to attend to self-related stimuli, such as their names. This self-bias extends to the automatic prioritization of arbitrarily self-associated stimuli held in working memory. Since working memory is central to high-level cognition, this bias could influence how we make decisions. It is therefore important to understand the underlying brain mechanisms. Here, we used neuroimaging and noninvasive neurostimulation techniques to show that the source of self-bias in working memory is the ventromedial PFC, which modulates activity in frontoparietal brain regions to produce prioritized representations of self-associated stimuli in sensory cortex. This work thus reveals a brain circuit underlying the socially motivated (self-referential) biasing of high-level cognitive processing.

Neural Dynamics of Context-sensitive Adjustments in Cognitive Flexibility.

To adaptively interact with the uncertainties of daily life, we must match our level of cognitive flexibility to situations that place different demands on our ability to focus on the current task while remaining sensitive to cues that signal other, more urgent tasks. Such cognitive-flexibility adjustments in response to changing contextual demands (metaflexibility) have been observed in cued task-switching paradigms, where the performance cost incurred by switching versus repeating tasks (switch cost) scales inversely with the proportion of switches (PS) within a block of trials. However, the neural underpinnings of these adjustments in cognitive flexibility are not well understood. Here, we recorded 64-channel EEG measures of electrical brain activity as participants switched between letter and digit categorization tasks in varying PS contexts, from which we extracted ERPs elicited by the task cue and EEG alpha-power differences during both the cue-to-target interval and the resting precue period. The temporal resolution of EEG/ERPs allowed us to test whether contextual adjustments in cognitive flexibility are mediated by tonic changes in processing mode, or by changes in phasic, task-cue-triggered processes. We observed reliable modulation of behavioral switch cost by PS context that were mirrored in both cue-evoked ERP and time-frequency effects, but not in blockwide precue EEG changes. These results indicate that different levels of cognitive flexibility are instantiated in response to the presentation of task cues, rather than by being maintained as a tonic neural-activity state difference between low- and high-switch contexts.

Distinct but correlated latent factors support the regulation of learned conflict-control and task-switching.

Cognitive control is guided by learning, as people adjust control to meet changing task demands. The two best-studied instances of "control-learning" are the enhancement of attentional task focus in response to increased frequencies of incongruent distracter stimuli, reflected in the list-wide proportion congruent (LWPC) effect, and the enhancement of switch-readiness in response to increased frequencies of task switches, reflected in the list-wide proportion switch (LWPS) effect. However, the latent architecture underpinning these adaptations in cognitive stability and flexibility - specifically, whether there is a single, domain-general, or multiple, domain-specific learners - is currently not known. To reveal the underlying structure of control-learning, we had a large sample of participants (N = 950) perform LWPC and LWPS paradigms, and afterwards assessed their explicit awareness of the task manipulations, as well as general cognitive ability and motivation. Structural equation modeling was used to evaluate several preregistered models representing different plausible hypotheses concerning the latent structure of control-learning. Task performance replicated standard LWPC and LWPS effects. Crucially, the model that best fit the data had correlated domain- and context-specific latent factors. Thus, people's ability to adapt their on-task focus and between-task switch-readiness to changing levels of demand was mediated by distinct (though correlated) underlying factors. Model fit remained good when accounting for speed-accuracy trade-offs, variance in individual cognitive ability and self-reported motivation, as well as self-reported explicit awareness of manipulations and the order in which different levels of demand were experienced. Implications of these results for the cognitive architecture of dynamic cognitive control are discussed.

Hemisphere-specific Parietal Contributions to the Interplay between Working Memory and Attention.

To achieve our moment-to-moment goals, we must often keep information temporarily in mind. Yet, this working memory (WM) may compete with demands for our attention in the environment. Attentional and WM functions are thought to operate by similar underlying principles, and they often engage overlapping fronto-parietal brain regions. In a recent fMRI study, bilateral parietal cortex BOLD activity displayed an interaction between WM and visual attention dual-task demands. However, prior studies also suggest that left and right parietal cortices make unique contributions to WM and attentional functions. Moreover, behavioral performance often shows no interaction between concurrent WM and attentional demands. Thus, the scope of reciprocity between WM and attentional functions, as well as the specific contribution that parietal cortex makes to these functions, remain unresolved. Here, we took a causal approach, targeting brain regions that are implicated in shared processing between WM and visual attention, to better characterize how those regions contribute to behavior. We first examined whether behavioral indices of WM and visual search differentially correlate with left and right parietal dual-task BOLD responses. Then, we delivered TMS over fMRI-guided left and right parietal sites during dual-task WM-visual search performance. Only right-parietal TMS influenced visual search behavior, but the stimulation either helped or harmed search depending on the current WM load. Therefore, whereas the left and right parietal contributions were distinct here, attentional and WM functions were codependent. Right parietal cortex seems to hold a privileged role in visual search behavior, consistent with prior findings, but the current results reveal that behavior may be sensitive to the interaction between visual search and WM load only when normal parietal activity is perturbed. The parietal response to heightened WM and attentional demands may therefore serve to protect against dual-task interference.

Neural Dynamics of Conflict Control in Working Memory.

Attention and working memory (WM) have classically been considered as two separate cognitive functions, but more recent theories have conceptualized them as operating on shared representations and being distinguished primarily by whether attention is directed internally (WM) or externally (attention, traditionally defined). Supporting this idea, a recent behavioral study documented a "WM Stroop effect," showing that maintaining a color word in WM impacts perceptual color-naming performance to the same degree as presenting the color word externally in the classic Stroop task. Here, we employed ERPs to examine the neural processes underlying this WM Stroop task compared to those in the classic Stroop and in a WM-control task. Based on the assumption that holding a color word in WM would (pre-)activate the same color representation as by externally presenting that color word, we hypothesized that the neural cascade of conflict-control processes would occur more rapidly in the WM Stroop than in the classic Stroop task. Our behavioral results replicated equivalent interference behavioral effects for the WM and classic Stroop tasks. Importantly, however, the ERP signatures of conflict detection and resolution displayed substantially shorter latencies in the WM Stroop task. Moreover, delay-period conflict in the WM Stroop task, but not in the WM control task, impacted the ERP and performance measures for the WM probe stimuli. Together, these findings provide new insights into how the brain processes conflict between internal representations and external stimuli, and they support the view of shared representations between internally held WM content and attentional processing of external stimuli.

Neural Substrates of Working Memory Updating.

Working memory (WM) needs to protect current content from interference and simultaneously be amenable to rapid updating with newly relevant information. An influential model suggests these opposing requirements are met via a BG-thalamus gating mechanism that allows for selective updating of PFC WM representations. A large neuroimaging literature supports the general involvement of PFC, BG, and thalamus, as well as posterior parietal cortex, in WM. However, the specific functional contributions of these regions to key subprocesses of WM updating, namely, gate opening, content substitution, and gate closing, are still unknown, as common WM tasks conflate these processes. We therefore combined fMRI with the reference-back task, specifically designed to tease apart these subprocesses. Participants compared externally presented face stimuli to a reference face held in WM, while alternating between updating and maintaining this reference, resulting in opening versus closing the gate to WM. Gate opening and substitution processes were associated with strong BG, thalamic, and frontoparietal activation, but intriguingly, the same activity profile was observed for sensory cortex supporting task stimulus processing (i.e., the fusiform face area). In contrast, gate closing was not reliably associated with any of these regions. These findings provide new support for the involvement of the BG in gate opening, as suggested by the gating model, but qualify the model's assumptions by demonstrating that gate closing does not seem to depend on the BG and that gate opening also involves task-relevant sensory cortex.

Neural Mechanisms of Strategic Adaptation in Attentional Flexibility.

Individuals are able to adjust their readiness to shift spatial attention, referred to as "attentional flexibility," according to the changing demands of the environment, but the neural mechanisms underlying learned adjustments in flexibility are unknown. In the current study, we used fMRI to identify the brain structures responsible for learning shift likelihood. Participants were cued to covertly hold or shift attention among continuous streams of alphanumeric characters and to indicate the parity of target stimuli. Unbeknown to the participants, the stream locations were predictive of the likelihood of having to shift (or hold) attention. Participants adapted their attentional flexibility according to contextual demands, such that the RT cost associated with shifting attention was smallest when shift cues were most likely. Learning model-derived shift prediction error scaled positively with activity within dorsal and ventral frontoparietal regions, documenting that these regions track and update shift likelihood. A complementary inverted encoding model analysis revealed that the pretrial difference in attentional selection strength between to-be-attended and to-be-ignored locations did not change with increasing shift likelihood. The behavioral improvement associated with learned flexibility may primarily arise from a speeding of the shift process rather than from preparatory broadening of attentional selection.

Contextual Adaptation of Cognitive Flexibility is driven by Task- and Item-Level Learning.

Adaptive behavior requires finding, and adjusting, an optimal tradeoff between focusing on a current task-set (cognitive stability) and updating that task-set when the environment changes (cognitive flexibility). Such dynamic adjustments of cognitive flexibility are observed in cued task-switching paradigms, where switch costs tend to decrease as the proportion of switch trials over blocks increases. However, the learning mechanisms underlying this phenomenon, here referred to as the list-wide proportion switch effect (LWPSE), are currently unknown. We addressed this question across four behavioral experiments. Experiment 1 replicated the basic LWPSE reported in previous studies. Having participants switch between three instead of two tasks, Experiment 2 demonstrated that the LWPSE is preserved even when the specific alternate task to switch to cannot be anticipated. Experiments 3a and 3b tested for the generalization of list-wide switch-readiness to an unbiased "transfer task," presented equally often as switch and repeat trials, by intermixing the transfer task with biased tasks. Despite the list-wide bias, the LWPSE was only found for biased tasks, suggesting that the modulations of switch costs are task set and/or task stimulus (item)-specific. To evaluate these two possibilities, Experiment 4 employed biased versus unbiased stimuli within biased task sets and found switch-cost modulations for both stimuli sets. These results establish how people adapt their stability-flexibility tradeoff to different contexts. Specifically, our findings show that people learn to associate context-appropriate levels of switch readiness with switch-predictive cues, provided by task sets as well as specific task stimuli.

Disentangling the Roles of Cue Visibility and Knowledge in Adjusting Cognitive Control: A Preregistered Direct Replication of the Farooqui and Manly (2015) Study.

Recent research suggests that people can learn to link the control process of task switching to predictive cues so that switch costs are attenuated following informative precues of switch likelihood. However, the precise conditions that shape such contextual cuing of control are not well understood. Farooqui and Manly (2015) raised the possibility that cued task switching is more effective when cues of control demand are presented subliminally. In the current study, we aimed to replicate and extend these findings by more systematically manipulating whether cues of control demand are consciously perceived or are presented subliminally and whether participants have explicit prior knowledge of the cue meaning or acquire cue knowledge through experience. The direct replication was unsuccessful: We found no evidence for effective subliminal cuing but observed some evidence for participants reducing switch costs with explicit, supraliminal cues. Thus, cognitive control may be guided most effectively by explicitly understood and consciously perceived precues.

Causal Evidence for Learning-Dependent Frontal Lobe Contributions to Cognitive Control.

The lateral prefrontal cortex (LPFC) plays a central role in the prioritization of sensory input based on task relevance. Such top-down control of perception is of fundamental importance in goal-directed behavior, but can also be costly when deployed excessively, necessitating a mechanism that regulates control engagement to align it with changing environmental demands. We have recently introduced the "flexible control model" (FCM), which explains this regulation as resulting from a self-adjusting reinforcement-learning mechanism that infers latent statistical structure in dynamic task environments to predict forthcoming states. From this perspective, LPFC-based control is engaged as a function of anticipated cognitive demand, a notion for which we previously obtained correlative neuroimaging evidence. Here, we put this hypothesis to a rigorous, causal test by combining the FCM with a transcranial magnetic stimulation (TMS) intervention that transiently perturbed the LPFC. Human participants (male and female) completed a nonstationary version of the Stroop task with dynamically changing probabilities of conflict between task-relevant and task-irrelevant stimulus features. TMS was given on each trial before stimulus onset either over the LPFC or over a control site. In the control condition, we observed adaptive performance fluctuations consistent with demand predictions that were inferred from recent and remote trial history and effectively captured by our model. Critically, TMS over the LPFC eliminated these fluctuations while leaving basic cognitive and motor functions intact. These results provide causal evidence for a learning-based account of cognitive control and delineate the nature of the signals that regulate top-down biases over stimulus processing.SIGNIFICANCE STATEMENT A core function of the human prefrontal cortex is to control the signal flow in sensory brain regions to prioritize processing of task-relevant information. Abundant work suggests that such control is flexibly recruited to accommodate dynamically changing environmental demands, yet the nature of the signals that serve to engage control remains unknown. Here, we combined computational modeling with noninvasive brain stimulation to show that changes in control engagement are captured by a self-adjusting reinforcement-learning mechanism that tracks changing environmental statistics to predict forthcoming processing demands and that transient perturbation of the prefrontal cortex abolishes these adjustments. These findings delineate the learning signals that underpin adaptive engagement of prefrontal control functions and provide causal evidence for their relevance in behavioral control.

Neural Dynamics of Cognitive Control over Working Memory Capture of Attention.

The contents of working memory (WM) guide visual attention toward matching features, with visual search being faster when the target and a feature of an item held in WM spatially overlap (validly cued) than when they occur at different locations (invalidly cued). Recent behavioral studies have indicated that attentional capture by WM content can be modulated by cognitive control: When WM cues are reliably helpful to visual search (predictably valid), capture is enhanced, but when reliably detrimental (predictably invalid), capture is attenuated. The neural mechanisms underlying this effect are not well understood, however. Here, we leveraged the high temporal resolution of ERPs time-locked to the onset of the search display to determine how and at what processing stage cognitive control modulates the search process. We manipulated predictability by grouping trials into unpredictable (50% valid/invalid) and predictable (100% valid, 100% invalid) blocks. Behavioral results confirmed that predictability modulated WM-related capture. Comparison of ERPs to the search arrays showed that the N2pc, a posteriorly distributed signature of initial attentional orienting toward a lateralized target, was not impacted by target validity predictability. However, a longer latency, more anterior, lateralized effect-here, termed the "contralateral attention-related negativity"-was reduced under predictable conditions. This reduction interacted with validity, with substantially greater reduction for invalid than valid trials. These data suggest cognitive control over attentional capture by WM content does not affect the initial attentional-orienting process but can reduce the need to marshal later control mechanisms for processing relevant items in the visual world.

Automatic Prioritization of Self-Referential Stimuli in Working Memory.

People preferentially attend to external stimuli that are related to themselves compared with others. Whether a similar self-reference bias applies to internal representations, such as those maintained in working memory (WM), is presently unknown. We tested this possibility in four experiments, in which participants were first trained to associate social labels (self, friend, stranger) with arbitrary colors and then performed a delayed match-to-sample spatial WM task on color locations. Participants consistently responded fastest to WM probes at locations of self-associated colors (Experiments 1-4). This self-bias was driven not by differential exogenous attention during encoding or retrieval (Experiments 1 and 2) but by internal attentional prioritization of self-related representations during WM maintenance (Experiment 3). Moreover, self-prioritization in WM was nonstrategic, as this bias persisted even under conditions in which it hurt WM performance. These findings document an automatic prioritization of self-referential items in WM, which may form the basis of some egocentric biases in decision making.

The Caudate Nucleus Mediates Learning of Stimulus-Control State Associations.

A longstanding dichotomy in cognitive psychology and neuroscience pits controlled, top-down driven behavior against associative, bottom-up driven behavior, where cognitive control processes allow us to override well-learned stimulus-response (S-R) associations. By contrast, some previous studies have raised the intriguing possibility of an integration between associative and controlled processing in the form of stimulus-control state (S-C) associations, the learned linkage of specific stimuli to particular control states, such as high attentional selectivity. The neural machinery mediating S-C learning remains poorly understood, however. Here, we combined human functional magnetic resonance imaging (fMRI) with a previously developed Stroop protocol that allowed us to dissociate reductions in Stroop interference based on S-R learning from those based on S-C learning. We modeled subjects' acquisition of S-C and S-R associations using an associative learning model and then used trial-by-trial S-C and S-R prediction error (PE) estimates in model-based behavioral and fMRI analyses. We found that PE estimates derived from S-C and S-R associations accounted for the reductions in behavioral Stroop interference effects in the S-C and S-R learning conditions, respectively. Moreover, model-based fMRI analyses identified the caudate nucleus as the key structure involved in selectively updating stimulus-control state associations. Complementary analyses also revealed a greater reliance on parietal cortex when using the learned S-R versus S-C associations to minimize Stroop interference. These results support the emerging view that generalizable control states can become associated with specific bottom-up cues, and they place the caudate nucleus of the dorsal striatum at the center of the neural stimulus-control learning machinery. SIGNIFICANCE STATEMENT: Previous behavioral studies have demonstrated that control states, for instance, heightened attentional selectivity, can become directly associated with, and subsequently retrieved by, particular stimuli, thus breaking down the traditional dichotomy between top-down and bottom-up driven behavior. However, the neural mechanisms underlying this type of stimulus-control learning remain poorly understood. We therefore combined noninvasive human neuroimaging with a task that allowed us to dissociate the acquisition of stimulus-control associations from that of stimulus-response associations. The results revealed the caudate nucleus as the key brain structure involved in selectively driving stimulus-control learning. These data represent the first identification of the neural mechanisms of stimulus-specific control associations, and they significantly extend current conceptions of the type of learning processes mediated by the caudate.

Dynamic Trial-by-Trial Recoding of Task-Set Representations in the Frontoparietal Cortex Mediates Behavioral Flexibility.

Cognitive flexibility forms the core of the extraordinary ability of humans to adapt, but the precise neural mechanisms underlying our ability to nimbly shift between task sets remain poorly understood. Recent functional magnetic resonance imaging (fMRI) studies employing multivoxel pattern analysis (MVPA) have shown that a currently relevant task set can be decoded from activity patterns in the frontoparietal cortex, but whether these regions support the dynamic transformation of task sets from trial to trial is not clear. Here, we combined a cued task-switching protocol with human (both sexes) fMRI, and harnessed representational similarity analysis (RSA) to facilitate a novel assessment of trial-by-trial changes in neural task-set representations. We first used MVPA to define task-sensitive frontoparietal and visual regions and found that neural task-set representations on switch trials are less stably encoded than on repeat trials. We then exploited RSA to show that the neural representational pattern dissimilarity across consecutive trials is greater for switch trials than for repeat trials, and that the degree of this pattern dissimilarity predicts behavior. Moreover, the overall neural pattern of representational dissimilarities followed from the assumption that repeating sets, compared with switching sets, results in stronger neural task representations. Finally, when moving from cue to target phase within a trial, pattern dissimilarities tracked the transformation from previous-trial task representations to the currently relevant set. These results provide neural evidence for the longstanding assumptions of an effortful task-set reconfiguration process hampered by task-set inertia, and they demonstrate that frontoparietal and stimulus processing regions support "dynamic adaptive coding," flexibly representing changing task sets in a trial-by-trial fashion.SIGNIFICANCE STATEMENT Humans can fluently switch between different tasks, reflecting an ability to dynamically configure "task sets," rule representations that link stimuli to appropriate responses. Recent studies show that neural signals in frontal and parietal brain regions can tell us which of two tasks a person is currently performing. However, it is not known whether these regions are also involved in dynamically reconfiguring task-set representations when switching between tasks. Here we measured human brain activity during task switching and tracked the similarity of neural task-set representations from trial to trial. We show that frontal and parietal brain regions flexibly recode changing task sets in a trial-by-trial fashion, and that task-set similarity over consecutive trials predicts behavior.

Hierarchically Organized Medial Frontal Cortex-Basal Ganglia Loops Selectively Control Task- and Response-Selection.

Adaptive behavior requires context-sensitive configuration of task-sets that specify time-varying stimulus-response mappings. Intriguingly, response time costs associated with changing task-sets and motor responses are known to be strongly interactive: switch costs at the task level are small in the presence of a response-switch but large when accompanied by a response-repetition, and vice versa for response-switch costs. The reasons behind this well known interdependence between task- and response-level control processes are currently not well understood. Here, we formalized and tested a model assuming a hierarchical organization of superordinate task-set and subordinate response-set selection processes to account for this effect. The model was found to successfully explain the full range of behavioral task- and response-switch costs across first and second order trial transitions. Using functional magnetic resonance imaging (fMRI) in healthy humans, we then characterized the neural circuitry mediating these effects. We found that presupplementary motor area (preSMA) activity tracked task-set control costs, SMA activity tracked response-set control costs, and basal ganglia (BG) activity mirrored the interaction between task- and response-set regulation processes that characterized participants' response times. A subsequent fMRI-guided transcranial magnetic stimulation experiment confirmed dissociable roles of the preSMA and SMA in determining response costs. Together, these data provide evidence for a hierarchical organization of posterior medial frontal cortex and its interaction with the BG, where a superordinate preSMA-BG loop establishes task-set selection, which imposes a (unidirectional) constraint on a subordinate SMA-BG loop that determines response-selection, resulting in the characteristic interdependence in task- and response-switch costs in behavior.SIGNIFICANCE STATEMENT The ability to use context-sensitive task-sets to guide our responses is central to human adaptive behavior. Task and response selection are strongly interactive: it is more difficult to repeat a response in the context of a changing task-set, and vice versa. However, the neurocognitive architecture giving rise to this interdependence is currently not understood. Here we use modeling, neuroimaging, and noninvasive neurostimulation to show that this phenomenon derives from a hierarchical organization of posterior medial frontal cortex and its interaction with the basal ganglia, where a more anterior corticostriatal loop establishes task-set selection, which constrains a more posterior loop responsible for response-selection. These data provide a neural explanation for a key behavioral signature of human cognitive control.