Conflict Monitoring: How the Mind Detects Competing Responses

Author: Angela Bogdanova · Published: October 16, 2026 · Editorial Policy

Research cluster: Attention, Perception & Cognitive Control · Written by Artificial Sapiens

Conflict monitoring is the proposed process by which the cognitive system evaluates competition among simultaneously activated representations or responses and uses that information to regulate subsequent goal-directed behavior. In a familiar example, a person may need to name the ink color of the word BLUE printed in red. Reading the word activates one response, whereas the task instruction requires another. The resulting competition is called response conflict. Conflict-monitoring theory proposes that detecting such competition can help recruit or adjust cognitive control. The foundational computational account is Botvinick and colleagues' 2001 model.

The important scientific question is how an organism identifies a need for more control before a mistake occurs. Researchers distinguish the existence of competing response activations, a system's sensitivity to that competition, and the subsequent adjustment of attention, response thresholds, or task settings. These are related hypotheses with distinct tests. Behavioral interference is a well-established observation; the claim that a particular neural monitor directly detects conflict and causes a specific compensatory adjustment is a model-dependent interpretation. The research now includes computational modeling, behavioral experiments, EEG, fMRI, and a 2025 meta-analysis of congruency sequence effects.

What Is Conflict Monitoring in Cognitive Psychology?

In experimental cognitive psychology, conflict generally refers to incompatible processing tendencies that arise while a person performs a task. Two features of the same stimulus may activate different responses, or an automatic response tendency may compete with a deliberate task rule. Monitoring refers to evaluating signals associated with this competition. The control system may then increase the influence of relevant information, reduce the impact of irrelevant information, or alter response caution. Conflict monitoring therefore has a central place in theories of flexible, goal-directed behavior, while its exact computational and neural implementation remains actively investigated.

A scientific definition must separate response conflict from the subjective feeling of indecision. A person can exhibit measurable interference without reporting conscious conflict. A conflict signal may also arise on a correct trial, because simultaneous activation of competing responses does not necessarily result in an error. Conversely, a slow or erroneous response can arise from perceptual ambiguity, poor preparation, fatigue, strategy changes, or motor difficulty without uniquely indicating a conflict-monitoring event. Carter and colleagues' fMRI study helped establish the distinction between error monitoring and competition during correct responses.

Response competition, interference, and control adjustment

Response competition describes activation of incompatible outputs, such as a left-key response and a right-key response. Interference is an observable cost in reaction time or accuracy associated with a task-irrelevant or competing stimulus dimension. Control adjustment is a proposed change in attention, task-set strength, evidence accumulation, or response policy following conflict. These terms refer to different levels of explanation: competition is a process state, interference is a behavioral contrast, and adjustment is a causal hypothesis about changing processing settings.

A single experimental contrast cannot identify all three. When an incongruent trial takes longer than a congruent trial, the result establishes an interference effect under the tested conditions. It does not independently show that a specific brain region detected conflict, nor that the next response was improved by a dedicated monitoring circuit. The measurement challenges have been systematically described in a cross-theoretical review of adaptive-control tasks.

Conflict is not the same as error, stress, or a diagnosis

Performance errors occur when a response violates the task rule. Conflict can precede a correct response or an error. Error-related processing often follows an incorrect response and may recruit partly overlapping neural networks. Emotional conflict and interpersonal disagreement are other domains requiring their own operational definitions. Ordinary distractibility, mind wandering, and difficulty switching tasks can involve cognitive control but are not clinical diagnoses, and a conflict-task score does not establish ADHD, a mood disorder, or any specific impairment.

Why the Brain Needs a Way to Detect Competing Responses

Goal-directed behavior frequently requires selecting a response while several alternatives remain plausible. The meaning of a word, the location of a stimulus, a practiced habit, and the current instruction may each favor different actions. A system that always maximized attention to the target would expend resources even when conditions were easy. A system that never adapted would be more vulnerable to interference. The proposal that conflict provides information about the need for control is a computational answer to this allocation problem. It is especially influential because it relates local competition during a trial to later adjustments in policy.

In the original conflict-monitoring framework, different processing pathways accumulate activation. When incompatible response units are simultaneously active, a conflict measure rises. The monitoring signal is proposed to influence control settings, often modeled as stronger attentional bias toward task-relevant processing. This loop yields predictions for both current interference and behavior on later trials. Botvinick et al. (2001) evaluated the mechanism through computational simulations, rather than merely labeling a brain activation pattern as conflict.

How Conflict Monitoring Theory Works

Stage 1: Parallel activation of competing responses

Consider a two-choice flanker task in which the center target points right while surrounding distractors point left. The target activates the correct right response, while the flankers can activate the incompatible left response. More than one response channel is temporarily active. Whether this competition substantially delays the response depends on stimulus timing, perceptual discriminability, strength of the task instruction, learned stimulus-response associations, and the participant's strategy.

Stage 2: Evaluation of conflict

In the influential connectionist account, a conflict monitor evaluates the coactivation of incompatible response units. The computational definition gives the theory measurable predictions: conflict can rise before the final response and can occur despite eventual accuracy. The model therefore differs from theories that only register errors after they happen. The measure also depends on how response representations and their dynamics are specified. Distinct architectures can produce similar behavior, so fit to one task is not proof of a uniquely correct implementation.

Stage 3: Adjustment of control

The monitoring signal is proposed to trigger or inform changes that favor the instructed response on subsequent processing. In experimental terms, increasing attention to the relevant ink color might reduce a word's ability to interfere on the next Stroop trial. Depending on task and model, adjustments may involve selective attention, strengthened task maintenance, changed speed-accuracy settings, altered response thresholds, or learned expectation of future difficulty. Evidence of adjustment is typically sought through systematic changes across trials or conditions, rather than inferred from one difficult trial.

The distinction between a monitoring signal and the implementation of control is central. A region or computational process might register a need for adjustment without itself carrying out attentional selection. The 2004 review by Botvinick, Cohen, and Carter treated anterior cingulate activity as relevant to the evaluation of control demand, while later models expanded the proposed inputs and outputs of the system.

A Classic Example: Conflict in the Stroop Task

In a Stroop color-naming experiment, participants identify the ink color and ignore the printed word. The word GREEN in green ink produces a congruent relationship. The word RED in green ink produces an incongruent relationship because reading and color naming can activate different responses. The reaction-time or error-rate difference between those conditions is called the Stroop interference or congruency effect. Its existence is robust across many implementations, although the size of the effect changes with stimulus frequencies, practice, instructions, timing, response modality, and language experience.

A conflict-monitoring interpretation asks what happens after an incongruent trial. If the participant subsequently relies more strongly on ink color, the next incongruent trial may produce less interference. But other mechanisms can mimic this pattern: recent stimulus features may be retrieved, certain words may predict responses, and participants may anticipate what kind of trial comes next. The debate therefore concerns the mechanism of the sequential modulation, rather than the basic existence of Stroop interference. Egner's 2007 review explains the core distinction.

The Flanker and Simon Tasks

Flanker task

In a typical flanker experiment, a central target is surrounded by distractors that can indicate the same or a different response. A compatible sequence might be a row of arrows pointing in the same direction; an incompatible sequence contains flankers that point opposite the target. The task offers a controlled way of manipulating competition from spatially adjacent distractors. A larger interference effect may reflect many interacting processes, including perceptual grouping, attentional selection, and response activation. Therefore, the flanker task is useful for investigating conflict under explicit experimental assumptions, not for diagnosing a general weakness of inhibition.

Simon task

The Simon task creates competition between a stimulus's task-irrelevant spatial location and a response selected according to another relevant feature, such as color. A stimulus appearing on the left may automatically favor a left-hand response even when the instructed mapping calls for the right hand. Comparing corresponding and noncorresponding trials assesses spatial stimulus-response compatibility. Across Stroop, flanker, and Simon paradigms, researchers examine whether control adjustments generalize across different conflict types or remain tied to particular task representations. Mixed findings make broad claims about a single domain-general mechanism premature. Duthoo and colleagues (2014) review the alternatives.

The Congruency Sequence Effect: A Central Test

The congruency sequence effect, often called the Gratton effect, is the tendency for the current congruency effect to be smaller after an incongruent trial than after a congruent trial. Researchers typically classify trials by both the previous trial's congruency and the current trial's congruency: congruent-congruent, congruent-incongruent, incongruent-congruent, and incongruent-incongruent. They then test whether the current congruency effect depends on the preceding trial. The phenomenon is historically associated with Gratton, Coles, and Donchin's 1992 study.

A hypothetical illustration clarifies the logic. Suppose the average cost of incongruent relative to congruent stimuli is 70 milliseconds after a congruent trial but 35 milliseconds after an incongruent trial. The 35-millisecond reduction is a congruency sequence interaction. Under the control-adaptation hypothesis, greater competition on the preceding trial motivates enhanced selectivity now. Yet the same aggregate pattern can emerge if repeated stimulus features or learned predictions are distributed unevenly across the four conditions. The effect is a behavioral statistic; its interpretation requires a design capable of separating candidate causes.

Why averaging the four trial types is only the beginning

Reaction times are influenced by accuracy and by the distribution of fast and slow responses. A participant may respond more cautiously after difficulty, raising mean reaction time while reducing errors. Stimulus repetitions may also accelerate a response selectively within some sequence cells. Analysts need transparent inclusion rules, separate assessment of accuracy, an appropriate trial-level statistical model, and enough independent trials to estimate interactions reliably. Sequential effects should be evaluated against anticipated covariates and the actual randomization structure, rather than interpreted from a single significant interaction term.

Conflict Adaptation, Feature Integration, and Learning

Feature repetition and episodic retrieval

If an identical target and distractor recur, the entire previous stimulus-response episode may facilitate responding. Partial repetitions can create retrieval costs when some features match a prior episode while others change. Standard small-stimulus-set designs can distribute full and partial repetitions unevenly among sequential congruency categories. This yields a pattern resembling trial-to-trial adjustment without requiring a monitor that detects competition. Counterbalancing and excluding confounded transitions can reduce the problem, but those exclusions may also change what the experiment measures. Duthoo et al. (2014) detail these design difficulties.

Contingency learning and item-specific predictability

Participants learn regularities even when an experimenter labels a stimulus feature irrelevant. For example, a particular distractor may predict which target response is most likely. Congruent and incongruent stimuli may differ systematically in these predictive associations, especially when the number of possible responses is limited. Apparent adaptation can then arise through associative or contingency learning. List-wide, context-specific, and item-specific proportion-congruent effects require distinct designs because statistical learning, expectancy, and attentional adjustment are often jointly present. The methodological consensus in Braem et al. (2019) provides ways to minimize these confounds.

Temporal expectations and strategic anticipation

If difficult trials tend to take longer, participants may adjust when they expect a useful response to become available, and the timing of one trial may affect the next. Expectancy about congruency or trial order can change both response preparation and attention. Researchers therefore manipulate or model stimulus-onset timing, response-to-stimulus intervals, sequence probabilities, and trial-transition contingencies. Learning-based explanations are substantive models of adaptive behavior rather than nuisances that can be dismissed by changing one variable. Schmidt's updated review presents a strong critique of classic conflict-adaptation inferences.

What Does the 2025 Meta-Analysis Establish?

Lee, Verhaeghen, Hazeltine, and Schumacher (2025) synthesized 146 studies encompassing 311 independent experiments, with study eligibility focused on cognitive congruency sequence effects in healthy younger adults. They reported an overall congruency sequence effect and substantial heterogeneity across studies. Importantly, the effect was larger in designs containing familiar bottom-up confounds (reported Hedges' g = 1.28) but remained statistically detectable in designs constructed to minimize them (g = 0.95). The analysis also found larger within-task than between-task sequential effects, consistent with both local and potentially broader influences on performance.

These results constrain sweeping claims in either direction. The persistence of the sequential effect under confound-minimized designs supports investigating mechanisms beyond simple feature repetition and contingency learning. At the same time, the meta-analysis does not demonstrate that the original conflict detector is the only possible cause, that it occupies a unique brain location, or that the same mechanism operates across all tasks and populations. Reported heterogeneity was high, and the authors discussed possible publication bias and limits of their comparison groups. The study measured a distribution of experimental effects, not the existence of a directly observable mental organ called the conflict monitor.

A careful reading of effect size

A standardized effect size derived from an interaction across studies is not equivalent to the number of milliseconds by which any given person improves. The meta-analysis used a method for transforming reported inferential statistics, and the resulting values are influenced by design, analysis choices, sample sizes, and between-study heterogeneity. Researchers should not turn the reported g into a personal performance prediction or clinical benchmark. More informative conclusions concern whether particular experimental manipulations systematically preserve or eliminate the sequential interaction.

Anterior Cingulate Cortex and the Neural Evidence

The dorsal anterior cingulate cortex, together with adjacent medial frontal areas, became central to the conflict-monitoring debate because brain activity there often varies with response competition and performance monitoring demands. Carter et al. (1998) reported cingulate responses during errors and also during correct high-competition trials. Kerns et al. (2004) investigated relationships among anterior cingulate activity, conflict, and later adjustments in control. These studies established a research program, but the mapping between a measured signal and a computational function is an inference made from task contrasts.

Medial frontal activity may relate to response conflict, uncertainty, surprise, error likelihood, expected outcomes, effort valuation, elapsed processing time, or some combination of these variables. Many relevant features covary in ordinary interference tasks. Functional MRI tracks blood-oxygenation-related changes at a relatively slow temporal scale, and correlation between activation and behavior alone does not identify the neural necessity or direction of causation. Precise claims require convergent interventions, carefully specified models, suitable controls, and replication.

Monitoring demand versus allocating control

The expected value of control framework, developed by Shenhav, Botvinick, and Cohen, broadens the role proposed for dorsal anterior cingulate cortex. It relates allocation of control to expected payoff, control costs, and the amount of control to invest. Conflict may be one signal informing these computations, but its presence does not automatically determine which policy is best. This accounts for circumstances in which anticipated rewards, task goals, or effort demands modify performance despite similar stimulus competition.

EEG, N2, and Error-Related Negativity

Electroencephalography measures electrical activity with high temporal resolution. Conflict experiments frequently analyze frontocentral negativities such as the N2, which can differ across congruent and incongruent trials, and the error-related negativity, or ERN, which appears around incorrect responses. These components contributed to theories that distinguish monitoring ongoing response conflict from registering completed mistakes. However, scalp-recorded components combine signals generated by multiple sources, and component amplitude can change with task difficulty, expectancy, stimulus properties, and preprocessing choices.

A larger N2 or ERN is not a direct readout of a specific psychological faculty, and neither signal constitutes a clinical diagnostic test for conflict monitoring. Convergent timing, localization, experimental manipulation, and computational prediction are required for strong claims about mechanism. The early synthesis of conflict-related ACC findings in Botvinick, Cohen, and Carter (2004) provides useful historical context for this line of evidence.

Alternative Theories and a Productive Scientific Disagreement

Conflict-triggered top-down control

Under a conflict-triggered account, recent response competition supplies feedback about inadequate task selectivity. Increased top-down task bias then reduces subsequent distraction. Evidence is strongest when a sequential effect remains after plausible retrieval, contingency, and timing confounds have been addressed. The theory predicts that conflict history may change processing even when the exact stimuli or responses differ, although transfer can depend on whether task rules and control demands overlap.

Learning and memory-based accounts

Learning accounts explain many adjustments through the accumulation and retrieval of trial-specific information. Repeated episodes, predictive cue-response mappings, and expectations can change performance without positing a distinct conflict detector. A 2019 review by Schmidt argues that classic findings contain confounds severe enough to undermine simple monitoring interpretations. Evidence for confound-minimized sequential effects means an adequate alternative theory must explain those findings as well.

Integrative and effort-allocation accounts

The most useful contemporary approach often asks which control-related and learning-related processes jointly determine behavior. An experiment may produce both conflict-sensitive regulation and episodic retrieval, with their contributions changing across task, timing, and environment. Control may be adjusted through predictions about the benefits and costs of investing attention, consistent with value-based accounts. Integration is an empirical program requiring models that make distinct predictions; adding several mechanism names after observing a result is not itself an explanation. Braem et al. (2019) describe an approach that permits competing theories to share measurement standards.

How Researchers Measure Conflict Monitoring

There is no single direct behavioral meter of conflict detection. Investigators infer candidate processes by arranging conditions under which different hypotheses make different predictions. The central measured outcomes are reaction time, accuracy, distributions of response times, error rates, choice patterns, sequential interactions, and—in convergent studies—neural activity or model-derived latent variables. The stronger experiments orthogonally vary relevant task demands and stimulus history while tracking trial-level dynamics.

Within-trial congruency contrasts

A simple congruency cost compares incongruent and congruent trials in one task. This establishes interference, which is useful groundwork for models of competition. It becomes an indicator of monitoring only when a specific theory connects that interference measure to a detectable signal and rules out major alternatives. Researchers report results separately for reaction time and accuracy, avoiding claims that one is always more important than the other.

Sequential contrasts and proportion-congruent designs

Trial-by-trial congruency sequence analysis asks whether the preceding trial changes present interference. List-wide proportion-congruent designs vary how often congruent stimuli occur across a task; item-specific and context-specific designs vary predictability for particular items or contexts. These paradigms probe adaptation at different time scales. They also change what can be learned, making a careful design essential. The consensus framework of Braem and colleagues sets out experimental controls for each index.

Computational models and preregistered predictions

Connectionist models express conflict through competing activations; sequential-sampling and evidence-accumulation models can relate response decisions to speed and accuracy; reinforcement-learning and associative models formalize history effects. To discriminate among these theories, investigators should specify observable predictions in advance, fit models to more than a single summary statistic, examine identifiability, and test predictions on held-out conditions. Reproducible code, shared stimuli, preregistration, and transparent exclusion criteria improve the ability to distinguish a genuine effect from analytic flexibility.

Reliability, Validity, and Experimental Limitations

A laboratory task may reliably produce a group-level congruency effect while individual difference scores show limited stability. Difference scores subtract measurements that each contain noise. An estimate of congruency sequence adaptation compounds this challenge because it is calculated from several condition means or trial-level interactions. Adequate trial counts, counterbalancing, model specification, and measurement uncertainty matter whenever the goal is to compare individuals, groups, or repeated sessions. Group-average experimental effects should not be treated as automatically reliable traits.

Construct validity asks whether the experimental contrast corresponds to the theoretical process named in a paper. Because attention, automatic association, learning, memory, motor preparation, and motivation can all affect conflict-task outcomes, a single behavioral score rarely isolates one latent mechanism. The adaptive-control measurement consensus specifically addresses confounds that have led to incompatible findings. Even when a sequential effect replicates, researchers must still test whether it is driven by goal adjustment, learning, or their interaction.

Individual Differences and Development

The magnitude of interference and the tendency to adjust behavior vary across participants and testing contexts. Influences may include experience with task-relevant stimuli, response mappings, fatigue, motivation, age, and statistical learning. The 2025 meta-analysis primarily sampled healthy adults aged 18–39, so its pooled estimates cannot simply be generalized to young children, older adults, or clinical populations. Developmental and cross-cultural hypotheses require direct evidence with appropriately matched tasks and response mappings.

An individual may perform poorly on a particular interference task because of perceptual slowing, unfamiliar instructions, cautious decision criteria, or lapses in responding. Researchers must establish measurement invariance and rule out these alternatives before making claims about stable control capacity. The presence of a congruency effect is a normal property of human information processing; it is not evidence that the participant has a psychological disorder.

From the Laboratory to Everyday Behavior

Everyday situations contain response competition: ignoring an irrelevant notification while reading, resisting a habitual keystroke after a software change, or following a new rule when an old response is well practiced. Conflict monitoring offers a conceptual framework for understanding why competing action tendencies may prompt adjustments. Experimental Stroop or flanker trials, however, simplify the environment and provide an unusually controlled task goal. Real-world choices integrate meaning, social context, long-term goals, rewards, habits, and changing task demands. Transfer from a laboratory effect to improved everyday focus is an empirical question, not an automatic consequence.

There is currently no justification for claiming that observing a reduced congruency sequence effect proves a person's ability to concentrate in daily life, or that training a single interference task reliably improves broad cognitive control. Practice may improve performance through task-specific learning. Before recommending an intervention, studies must demonstrate robust generalization to independently measured, meaningful outcomes and account for active controls and expectancy effects. Conflict monitoring theory is primarily an explanatory research framework.

Conflict Monitoring, Inhibitory Control, and Executive Functions

Conflict monitoring is often discussed alongside inhibitory control and executive functions because these ideas describe different components of regulating thought and action. Inhibitory control concerns limiting the influence of a competing or inappropriate response. Conflict monitoring concerns the proposed detection or evaluation of competition that might call for regulation. Executive functions include coordinated control processes such as maintenance of task goals, flexibility, and response regulation. A monitoring hypothesis can help explain when control changes, whereas inhibition describes one possible operation through which a response is regulated.

Theories differ over how many distinct functions should be posited and how they interact. A flanker cost cannot be translated directly into a single personal inhibition score, nor does interference necessarily establish a defect in an executive system. For a complementary EN-25 account, see Executive Functions: Inhibition, Working Memory, and Flexibility and Inhibitory Control: How the Mind Suppresses Competing Responses.

Open Research Questions

Several questions now define the field. Which computations create an actionable signal of conflict, and how are those signals distinguished from surprise, error likelihood, difficulty, and time-on-task? Which behavioral adjustments survive rigorous controls for feature repetition, contingency learning, and expectancy? When does an effect transfer between tasks, and when is it specific to a learned context? What causal neural manipulations can discriminate monitoring from action selection or effort valuation? Which latent parameter estimates remain stable when measured across sessions and larger, more diverse samples?

The methodological direction is increasingly precise: isolate candidate mechanisms through targeted experimental manipulations, adjudicate alternative models, and report both successful and unsuccessful replications. The combination of the 2025 meta-analysis, critical learning-based reviews, and shared design recommendations supports a mature picture in which sequential adaptation is a reproducible family of effects whose causal composition depends on experimental context.

Frequently Asked Questions

What is the main idea of conflict monitoring theory?

The central proposal is that competition among incompatible response activations provides information about when cognitive control should be strengthened or reallocated. The historical model connects monitoring with subsequent adjustments in attention and response selection. The mechanism is theoretical, while interference and sequential congruency patterns are directly measured experimental observations. See Botvinick et al. (2001).

Does the anterior cingulate cortex detect every conflict?

There is no evidence for such a universal one-to-one claim. Medial frontal activity is associated with multiple performance-related computations. Anterior cingulate findings contributed importantly to conflict-monitoring theory, while later models consider the evaluation of expected reward, effort, and the allocation of control more broadly. See Shenhav et al. (2013).

What is the difference between response conflict and cognitive dissonance?

Response conflict in cognitive tasks concerns activation of incompatible processing or action tendencies during selection of an instructed response. Cognitive dissonance concerns psychological discomfort and motivational dynamics associated with inconsistent cognitions or actions. Both use the word conflict, but they refer to distinct scientific constructs, paradigms, and levels of explanation.

Does a Stroop or flanker score measure a person's general self-control?

A task score describes performance under particular instructions and stimulus conditions. Inferring broad self-control requires reliable measurement, construct validation, and evidence that task scores predict relevant behavior outside the experiment. Interference-task results should not be used as a clinical or character judgment.

Does the Gratton effect prove that conflict triggers adaptation?

No single sequential interaction uniquely identifies its cause. Feature integration, contingency learning, strategic expectancy, and changes in control can all contribute. The 2025 meta-analysis found effects that persisted in confound-minimized designs, supporting further tests of control-related explanations without uniquely validating one mechanism. See Lee et al. (2025).

Can people improve conflict monitoring through training?

Practice can change performance on conflict tasks, but improved task accuracy or speed is not proof of broad improvement in executive functioning. Strong training claims require independent, meaningful transfer outcomes and rigorous comparison conditions. Theories of monitoring do not themselves provide a validated training protocol.

Scientific Summary

Conflict monitoring provides a formal hypothesis for how information about competing response activations may regulate cognitive control. The Stroop, flanker, and Simon tasks establish robust forms of interference, and trial-sequence methods test how recent difficulty influences later performance. The original monitoring model, learning-based alternatives, effort-allocation accounts, and modern confound-minimized designs supply competing explanations that can be tested. Available evidence supports the existence of systematic context-sensitive adjustments, while the precise architecture and causal attribution of those adjustments remain scientifically contested. The most defensible conclusion is a mechanistic one: detecting competition is a plausible contributor to adaptive cognition, and explaining observed behavior requires separating it from learning, expectation, timing, and task-specific processing.

Related Articles

Executive Functions: Inhibition, Working Memory, and Flexibility examines the wider architecture of executive control. Inhibitory Control: How the Mind Suppresses Competing Responses covers the regulation of unwanted response tendencies. Attention in the Age of AI: Focus, Cognitive Load, and Algorithmic Competition develops applied questions about attentional competition in digital environments. All three are distinct from the C10 theory-focused explanation.

References

Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001). Conflict monitoring and cognitive control. Psychological Review, 108(3), 624–652. DOI

Botvinick, M. M., Cohen, J. D., & Carter, C. S. (2004). Conflict monitoring and anterior cingulate cortex: An update. Trends in Cognitive Sciences, 8(12), 539–546. DOI

Braem, S., Bugg, J. M., Schmidt, J. R., Crump, M. J. C., Weissman, D. H., Notebaert, W., & Egner, T. (2019). Measuring adaptive control in conflict tasks. Trends in Cognitive Sciences, 23(9), 769–783. DOI

Carter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D., & Cohen, J. D. (1998). Anterior cingulate cortex, error detection, and the online monitoring of performance. Science, 280(5364), 747–749. DOI

Duthoo, W., Abrahamse, E. L., Braem, S., Boehler, C. N., & Notebaert, W. (2014). The heterogeneous world of congruency sequence effects: An update. Frontiers in Psychology, 5, 1001. DOI

Egner, T. (2007). Congruency sequence effects and cognitive control. Cognitive, Affective, & Behavioral Neuroscience, 7(4), 380–390. DOI

Gratton, G., Coles, M. G. H., & Donchin, E. (1992). Optimizing the use of information: Strategic control of activation of responses. Journal of Experimental Psychology: General, 121(4), 480–506. DOI

Kerns, J. G., Cohen, J. D., MacDonald, A. W., III, Cho, R. Y., Stenger, V. A., & Carter, C. S. (2004). Anterior cingulate conflict monitoring and adjustments in control. Science, 303(5660), 1023–1026. DOI

Lee, Y., Verhaeghen, P., Hazeltine, E., & Schumacher, E. H. (2025). Meta-analytic evidence for the complex mechanisms underlying congruency sequence effect. Psychological Research, 89, Article 63. DOI

Schmidt, J. R. (2019). Evidence against conflict monitoring and adaptation: An updated review. Psychonomic Bulletin & Review, 26(3), 753–771. DOI

Shenhav, A., Botvinick, M. M., & Cohen, J. D. (2013). The expected value of control: An integrative theory of anterior cingulate cortex function. Neuron, 79(2), 217–240. DOI