Cognitive effort encompasses the mental exertion invested to achieve a goal as well as the experienced cost associated with that exertion. The dopaminergic midbrain regions (i.e., ventral tegmental area/substantia nigra), ventral striatum, and the anterior cingulate cortex have all been broadly implicated in effort regulation. However, less is known about the role of these regions in dissociable subprocesses such as effort evaluation versus effort preparation. In this preregistered functional MRI study with concurrent pupil-tracking, we disentangled the evaluation and preparation phase of reward- and effort-based motivation. In the evaluation phase, the typical valuation regions encoded high reward and low effort demand as a motivational factor. In the preparation phase, the anterior cingulate cortex displayed increased activity for demanding high reward trials, while the ventromedial prefrontal cortex responded to demanding trials in general. The dopaminergic midbrain exhibited a positive reward effect only during task execution. The results suggest that when separating evaluation from preparation, regions traditionally thought of as primarily reward-attuned can assign a motivating signal to effort, while the anterior cingulate cortex integrates effort and reward prospect to subserve preparatory control processes.

Background Modern accounts of cognitive effort posit that it is actively controlled based on a cost-benefit analysis, integrating (subjective) costs and possible rewards, which is furthermore moderated by efficacy (i.e., whether there is a clear relationship between successful task performance and reward). Although there is evidence that reward and efficacy interact, this relationship may not be stable over time, given that reward attenuates time-on-task effects while efficacy’s relationship with time-on-task is still unclear. Methods Thirty participants performed a Stroop task with a block design, with block-wise manipulation of efficacy (possible rewards were either contingent upon correct task performance or determined without regard to task performance) and reward (high or low). Behavioral measures (response time and accuracy), pupil data (tonic and phasic pupil size), and blink measures (rate and duration) were recorded and analyzed using a mixed-effects modelling approach, with each block split into two halves to assess time-on-task effects. Results The behavioral data showed strong effects for efficacy, a pattern that was also largely mirrored in phasic pupil size, suggesting substantial effort allocation. In contrast, we found no clear impact of time-on-task or reward on behavioral data or phasic pupil size. However, tonic pupil size and blink metrics (rate and duration) did reflect time-on-task effects, with reward and efficacy attenuating these in some cases. Taken together, the present results suggest a strong and stable control mode for high-efficacy/high-reward blocks, with low-efficacy/low-reward blocks leading to disengagement, whereas the mixed block types seemed to be tackled with suboptimal levels of cognitive effort while still showing signs of increasing fatigue. Conclusions Thereby, this study contributes corroborating evidence in favor of the account that information regarding efficacy is robustly integrated into the cost-benefit analysis underlying the allocation of cognitive control resources, and highlights the complex dynamics of cognitive effort over time.

The relationship between acute physical exercise and decision-making remains debated. Various evidence leans towards an inverted U-shaped curve where decision-making peaks during moderate acute physical exertion and deteriorates under heavy exercise. This study examined behavioural and computational markers of decision-making during a conflict task across three increasing cycling intensities, up to the heavy domain. Thirty-one participants completed the Eriksen Flanker task at light, moderate, and heavy exercise intensities. Throughout the experiment, heart rate, oxygen uptake, and rates of perceived exertion were measured and confirmed the prescribed exercise intensities. Results indicated that reaction times decreased, while accuracy remained stable. Diffusion Model for Conflict (DMC) analyses showed that drift rate followed the hypothesised inverted U-shaped pattern, peaking at moderate intensity, suggesting optimal evidence accumulation at this workload. Decision boundaries decreased linearly with intensity, indicating less cautious responding under higher exertion. This partially supports arousal-based models such as the Yerkes-Dodson law, highlighting moderate intensity as a potential “sweet spot” for decision-making. The integration of computational modelling provides novel insight into the interplay between intensity levels of acute exercise and cognitive processes. This has implications for sports performance, training strategies, and high-stakes professions requiring rapid decisions under physical stress.

People often need to maintain multiple task sets at the same time. This ability is thought to rely more on parallel than serial task-maintenance strategies, which have been linked to a flexibility-stability trade-off. However, it remains unclear whether people can voluntarily regulate these strategies through explicit instructions. In four between-subject experiments (total N = 650), we asked participants to adopt either a parallel (flexible) or serial (stable) strategy. These instructions had no effects on task-switch costs or task-rule congruency effects, suggesting people were unable to alter strategy use. In contrast, manipulating proportion congruency had a direct effect on conflict processing alone, indicating that targeted, experience-based adaptations are possible. Notably, we observed a flexibility-stability trade-off in the first block, reflected in a negative correlation between switch costs and congruency effects, which, interestingly, gradually dissipates over time. This finding was replicated in a meta-analysis across five independent datasets (total N = 1764). Together, these results suggest that while a flexibility-stability trade-off may be inherent, people can adapt to it through learning and practice, but not through explicit instructions.

People integrate social information asymmetrically, favoring agreement over disagreement. We propose that social disconfirmation functions as a mismatch signal analogous to cognitive conflict. We tested this hypothesis in two online experiments using perceptual tasks with controlled social feedback attributed to a previous participant. In Experiment 1 (N = 105), participants chose between environments offering high (70%) versus low (30%) agreement rates; preference progressively shifted toward high-agreement environments, and disagreement elicited post-disconfirmation slowing that was amplified when disagreement was infrequent – mirroring conflict-frequency effects. In Experiment 2 (N = 91), an Affect Misattribution Procedure revealed that neutral stimuli were rated as less pleasant following disagreement than agreement, particularly among participants who reliably encoded feedback valence. Across both experiments, disagreement was actively avoided, carried negative affective value, and triggered context-sensitive behavioral adjustments. These converging findings suggest that social disconfirmation operates as a mismatch signal, triggering hallmark signatures of cognitive conflict.

Modern accounts of cognitive effort posit that it is actively controlled based on a cost-benefit analysis, integrating (subjective) costs and possible rewards, which is furthermore moderated by efficacy (i.e., whether there is a clear relationship between successful task performance and reward). Although there is evidence that reward and efficacy interact, this relationship may not be stable over time, given that reward attenuates time-on-task effects while efficacy’s relationship with time-on-task is still unclear. Thirty participants performed a Stroop task with a block design, with block-wise manipulation of efficacy (possible rewards were either contingent upon correct task performance or determined without regard to task performance) and reward (high or low). Behavioral measures (response time and accuracy), pupil data (tonic and phasic pupil size), and blink measures (rate and duration) were recorded and analyzed using a mixed-effects modelling approach, with each block split into two halves to assess time-on-task effects. The behavioral data showed clear effects for efficacy, a pattern that was also largely mirrored in phasic pupil size, suggesting substantial effort allocation. In contrast, we found no clear impact of time-on-task or reward on behavioral data or phasic pupil size. Tonic pupil size and blink metrics (rate and duration) did suggest time-on-task effects, with reward and efficacy attenuating these in some cases. However, because these time-on-task effects diverged from behavioral outcomes and yielded theoretically ambiguous interaction patterns, they warrant cautious interpretation. These findings support the role of efficacy in effort-related cost-benefit calculations, even when efficacy information is sustained across relatively long task blocks. More broadly, the present study highlights the interpretive complexity of ocular measures for studying cognitive effort over time.

A long-standing debate in cognitive science is whether cognitive control recruits domain-general or domain-specific processes. A useful paradigm to study this is the congruency sequence effect, which reflects the reduced impact of congruency after an incongruent trial. Many studies examined the congruency sequence effect across task in order to investigate domain generality. However, differences in how task dissimilarity was manipulated led to different conclusions about the nature of adaptive control. Therefore, we conducted a systematic, preregistered meta-analysis of 174 experiments (229 effect sizes). On average, the cross-task congruency sequence effect (Hedges’ gz = 0.252) decreased as task dissimilarity increased, with some factors influencing adaptive control more than others. Especially when tasks used a different response modality or conflict arose from different irrelevant stimulus dimensions, we observed a decreased cross-task congruency sequence effect, suggesting adaptive control is both action-oriented and involves distractor suppression. Notably, however, when tasks shared the same conflict type, larger congruency sequence effects were observed when stimulus and response sets were distinct rather than partially overlapping. Together, these findings refine current theories by pointing to a central role for the different motor modalities, distractor suppression, and shared versus separate task representations, moving beyond a simple domain-general versus domain-specific distinction. Keywords: meta analysis, conflict adaptation, congruency sequence effect, domain generality, cognitive control

This study examines the role of internal motor signals and visual information in the detection of and confidence in Partial-errors (PEs), subtle endogenous motor corrections. Using electromyographic (EMG) recordings, we captured motor activations during a conflict task in which participants reported the presence of PEs and rated their confidence. Two experiments were conducted: Experiment 1 provided visible visual conflict through supraliminal primes, while Experiment 2 reduced visual feedback using subliminal primes. In both Experiments, participants demonstrated limited PE detection and above-chance metacognitive efficiency. Notably, when participants reported the absence of a PE, confidence was lower when a PE was actually present (unaware PE) compared to when there was no PE (correct rejection), suggesting implicit sensitivity to motor activation. Detection and confidence were systematically influenced by motor signals, with larger PE amplitudes and longer correction times leading to higher detection rates and confidence levels. However, a metacognitive bias emerged: confidence was paradoxically lower for detected PEs than for undetected ones, despite strong motor evidence. Visual information modulated the reliance on motor signals. In Experiment 2, where subliminal priming reduced visual feedback, motor signals had a more pronounced influence on both detection and confidence. These findings highlight the complementary roles of internal motor signals and external visual information in shaping sensorimotor confidence.

Human task performance elicits diverse subjective metacognitive experiences, such as boredom, effort, fatigue, and frustration, which are considered to play important roles in the monitoring and regulation of cognitive processes. Yet, their specific contributions to task performance remain poorly understood. Therefore, we investigated the temporal dynamics underlying these metacognitive experiences and latent cognitive processes supporting task performance. We used a time-on-task design using a conflict task and analyzed the data using a comprehensive approach encompassing behavioral, model-based, subjective, and neural measures (N = 111). Our results show that changes in cognitive processes can be understood as a rational attempt to optimize task performance and that distinct metacognitive experiences are related to different aspects of this rational endeavor. These findings suggest that metacognitive experiences act as tools for humans to gain insights into the optimality of their cognitive performance. Using computational modeling and EEG, the study suggests several metacognitive states correlate with rational adjustments in behavior; this might benefit steering behavior towards optimality

Error monitoring, a crucial aspect of behavioral regulation, has been studied extensively in cognitive psychology and neuroscience. Lab studies often observe that people become more cautious after both own and others' errors. Moreover, social error monitoring research suggests that cooperation elicits stronger adaptations to others' errors than competition. However, these insights primarily stem from controlled reaction time tasks, limiting generalization to real-world social settings. To address this, we investigated whether basketball players become more cautious after missing compared to making shots, and whether these adaptations differ when the miss was by themselves, teammates or opponents. Meta-analyses of four NBA-seasons revealed increased caution in terms of shot distance and angle following missed compared to made shots, with the most pronounced effect after own, next teammates' and last opponents' errors. While controlling for rebound shots reduced the effects, the general pattern remained. Additional analyses also revealed a positive correlation between post-error caution and accuracy, suggesting post-error caution may serve to improve shooting accuracy. These findings not only extend cognitive theories from controlled settings to a real-word sport setting, but also provide a crucial understanding of athletes' social error responses thereby paving the way for research aimed at optimizing this aspect of performance.

The subjective experience of mental effort is critical for adaptive cognitive control, yet its neural and computational underpinnings remain elusive. Here, we combine hierarchical drift-diffusion modeling and EEG to investigate how trial-by-trial fluctuations in both preparatory and task-related neural activity shape subjective effort ratings. Participants performed an arithmetic task of variable difficulty, choosing task difficulty in advance, which allowed us to isolate neural signatures of preparation (contingent negative variation) and task engagement (P3 amplitude). Computational modeling revealed that participants adjusted decision boundaries based on anticipated difficulty, reflecting heightened caution. Critically, subjective effort ratings tracked this increased caution, likely reflecting the cost of additional accumulation. EEG analyses showed that while subjective effort was sensitive to the P3 amplitude, indicating exerted effort during task performance, it was insensitive to preparatory CNV activity. Our findings offer novel insights into the computations underlying subjective effort, proposing a selective role for exerted, but not preparatory activity.

Pupil size is a well-established marker of cognitive effort, with greater efforts leading to larger pupils. This is particularly true for pupil size during task performance, whereas findings on anticipatory effort triggered by a cue stimulus are less consistent. For example, a recent report by Frömer et al. found that in a cued-Stroop task, behavioral performance and electrophysiological markers of preparatory effort allocation were modulated by cued reward and ‘efficacy’ (the degree to which rewards depended on good performance), but pupil size did not show a comparable pattern. Here, we conceptually replicated this study, employing an alternative approach to the pupillometry analyses. In line with previous findings, we found no modulation of absolute pupil size in the cue-to-target interval. Instead, we observed a significant difference in the rate of pupil dilation in anticipation of the target: pupils dilated more rapidly for high-reward trials in which rewards depended on good performance. This was followed by a significant difference in absolute pupil size within the first hundreds of milliseconds following Stroop stimulus onset, likely reflecting a lagging effect of anticipatory effort allocation. Finally, the slope of pupil dilation was significantly correlated with behavioral response times, and this association was strongest for the high-reward, high-efficacy trials, further supporting that the rate of anticipatory pupil dilation reflects anticipatory effort. We conclude that pupil size is modulated by anticipatory effort, but in a highly temporally-specific manner, which is best reflected by the rate of dilation in the moments just prior to stimulus onset.

Adaptive control refers to flexible adjustments in control settings in response to conflicting situations. There has been a long-standing debate as to whether this adaptation relies on a domain-general or domain-specific process. Recent models predict a U-shaped relation where only highly similar or highly dissimilar tasks show adaptation across tasks, because only those tasks can be represented or activated in parallel. While there has been an abundance of evidence for adaptation within and across highly similar tasks, only some recent studies have reported adaptation across highly dissimilar tasks, with some failures to replicate. In order to further investigate this, we interleaved two very different conflict tasks, a manual multi-source interference task and a vocal picture-word interference task. We ran this experiment in Dutch (Experiment 1) and Mandarin (Experiment 2). Across the two experiments, results show no cross-task conflict adaptation. These results do not fit with suggestion of domain-general adaptive processes nor with the hypothesis of a U-shaped model. Instead, our results are most compatible with a task-specific view on the mechanisms behind adaptive control.

ABSTRACTPupil size is a well‐established marker of cognitive effort, with greater efforts leading to larger pupils. This is particularly true for pupil size during task performance, whereas findings on anticipatory effort triggered by a cue stimulus are less consistent. For example, a recent report by Frömer et al. found that in a cued‐Stroop task, behavioral performance and electrophysiological markers of preparatory effort allocation were modulated by cued reward and ‘efficacy’ (the degree to which rewards depended on good performance), but pupil size did not show a comparable pattern. Here, we conceptually replicated this study, employing an alternative approach to the pupillometry analyses. In line with previous findings, we found no modulation of absolute pupil size in the cue‐to‐target interval. Instead, we observed a significant difference in the rate of pupil dilation in anticipation of the target: pupils dilated more rapidly for high‐reward trials in which rewards depended on good performance. This was followed by a significant difference in absolute pupil size within the first hundreds of milliseconds following Stroop stimulus onset, likely reflecting a lagging effect of anticipatory effort allocation. Finally, the slope of pupil dilation was significantly correlated with behavioral response times, and this association was strongest for the high‐reward, high‐efficacy trials, further supporting that the rate of anticipatory pupil dilation reflects anticipatory effort. We conclude that pupil size is modulated by anticipatory effort, but in a highly temporally‐specific manner, which is best reflected by the rate of dilation in the moments just prior to stimulus onset.

Many theories on cognitive effort start from the assumption that cognitive effort can be expended at will, and flexibly up- or down-regulated depending on expected task demand and rewards. However, while effort regulation has been investigated across a wide range of incentive conditions, few investigated the cost of effort regulation itself. Across four experiments, we studied the effects of reward expectancy and task difficulty on effort expenditure in a perceptual decision-making task (random-dot-motion) and a cognitive control task (colour-naming Stroop), and within each task comparted cues between short (cueing the next trial) and long (cueing the next six trials) prediction horizons. We found that participants used the cue information only when it was valid for multiple trials in a row. In the random-dot-motion task, a high reward expectancy resulted in better accuracy, especially in easy trials, but only with long prediction horizon. Similarly, in the Stroop task, the reward facilitation of reaction time was only observed after reward cues with a long prediction horizon. Together, our results indicate that people experience a cost to effort regulation, and that lower adjustment frequency can compensate for this cost.

A long-standing debate in cognitive science is whether cognitive control recruits domain-general or domain-specific processes. A useful paradigm to study this is the congruency sequence effect, which reflects the reduced impact of congruency after an incongruent trial. Many studies examined the congruency sequence effect across task in order to investigate domain generality. However, differences in how task dissimilarity was manipulated led to different conclusions about the nature of adaptive control. Therefore, we conducted a systematic, preregistered meta-analysis of 169 experiments (229 effect sizes). On average, the cross-task congruency sequence effect (Hedges’ gz = 0.252) decreased as task dissimilarity increased, with some factors influencing adaptive control more than others. Especially when tasks used a different response modality or conflict arose from different irrelevant stimulus dimensions, we observed a decreased cross-task congruency sequence effect, suggesting adaptive control is both action-oriented and involves distractor suppression. Notably, however, when tasks shared the same conflict type, larger congruency sequence effects were observed when stimulus and response sets were distinct rather than partially overlapping. Together, these findings refine current theories by pointing to a central role for the different motor modalities, distractor suppression, and shared versus separate task representations, moving beyond a simple domain-general versus domain-specific distinction. Keywords: meta analysis, conflict adaptation, congruency sequence effect, domain generality, cognitive control

According to Fechner’s law, perceived sensations increase logarithmically with stimulus intensity. In the context of exercise, rising physical effort is therefore accompanied by increasingly salient bodily signals such as heart rate. Interoception, the sensing of internal bodily states, supports adaptive responses to these changing demands, for instance, by guiding the regulation of exertion. Beyond physical regulation, interoceptive signals have also been proposed to inform higher-order monitoring processes, such as error detection. Yet it remains unclear whether acute exercise sharpens cardiac interoception and whether enhanced access to bodily signals translates into greater awareness of errors. In a within-subject design, 36 healthy adults completed two tasks while cycling at low versus moderate intensity: a Heart Rate Discrimination Task assessing cardiac interoceptive accuracy and insight, and a Go/No-Go task measuring error awareness through post-error reports. Moderate effort selectively improved interoceptive accuracy (threshold closer to true heart rate), without changes in precision of lapse rate. Physical effort also made participants more confident in their interoceptive judgments, but this confidence increase was unrelated to actual accuracy. In the Go/No-Go task, commission errors were more frequent during cycling, yet the proportion of consciously detected errors was unaffected. Interoceptive performance did not reliably predict error awareness. These results indicate that acute exercise recalibrates cardiac interoception, likely by enhancing the salience of afferent signals, but does not strengthen higher-order monitoring of action. Error-related cardiac responses may thus reflect affective or orienting processes rather than a causal basis for awareness.

Previous research suggests that the mere presence of a smartphone can detrimentally affect performance. However, other studies failed to observe such detrimental effects. A limitation of existing studies is that no indexes of (potentially compensating) effort were included. Further, time-on-task effects have been unexplored. Here, we address these limitations by investigating the mere-presence effect of a smartphone on performance in two continuous-performance experiments (Experiment 1 using an n-back and a number judgement task at two difficulty levels, and Experiment 2 using a pure, challenging n-back task), measuring pupil size to assess invested effort, and taking into account time-on-task effects. Finally, contrary to previous studies that predominantly used between-subject designs, we utilized within-subject designs in both experiments. Contrary to expectations, Experiment 1 largely yielded no significant effects of smartphone presence on performance. Nonetheless, the presence of a smartphone triggered larger tonic pupil size in the more difficult task, and a more rapid decrease over time. Experiment 2 similarly failed to demonstrate smartphone effects on performance, but replicated the finding of larger tonic pupil size in the presence of a smartphone. In addition, tonic pupil size showed a slower decrease over time when a smartphone was present. In Experiment 2, we could furthermore look at phasic pupil size, which decreased over time in the absence of a phone but not in its presence. These findings suggest a complex relationship between smartphone presence, effort, and time-on-task, which does not necessarily express itself behaviorally, highlighting in particular the need to also explore potential contributions of (compensatory) effort.

Current theories propose that mental effort is invested only when the anticipated benefits, such as rewards, outweigh the associated costs, like task difficulty. Yet, it remains unclear whether this motivational and mitigating aspect of reward processing is reflected in the evaluation of reward/difficulty cues as such, and to what extent it depends on task experience. In a pre-registered experiment (N=84), we used the affect misattribution procedure (AMP) to gauge affective evaluations of nonword cues predicting reward and task difficulty levels. Contrary to previous studies, the AMP was administered at the outset, after cue instructions, and after the cues were used in a random dot motion (RDM) task. Compared to baseline, cues predicting a larger reward were evaluated more positively after RDM task experience, and most importantly, already after cue instructions, with no difference between the two phases. This evaluative effect manifested in increased performance after larger reward cues in the RDM task. Our results suggest that AMP effects may generally capture performance expectations which are independent of task experience. Importantly, these instructed expectations of reward and difficulty play a crucial role in the evaluation and subsequent investment of mental effort.

Cognitive conflict is typically experienced as negative, which has been argued to drive adaptive behavior following a conflict. We tried to change the negative value of conflict using evaluative conditioning, and measured changes in conflict adaptation in a subsequent Stroop task (N = 416 Prolific participants, English native speakers from different countries). We did not find evidence for decreased conflict adaptation following positive evaluative conditioning of conflict. However, we also did not find evidence for the change of conflict evaluation measured with the affect misattribution procedure in the follow-up experiment (N = 70). Interestingly, the exploratory follow-up analysis showed that people with low goal motivation (as measured through BAS Drive) did show the expected effect. A memory test for the evaluative conditioning pairings and the follow-up experiment suggest that, although the affective value of conflict was difficult to change, people with low goal motivation experienced less difficulty remembering the association between conflict stimuli and positive pictures. Our findings show additional evidence that conflicts are inherently negative, however, there is no clear support for, or against, the affective signaling hypothesis, that is the idea that conflict negativity drives control adaptations.