RESEARCH DOSSIER
Self-Control, Desire Conflict, and Ego-Depletion Evidence
Abstract. Self-control is the organization of behavior when a presently attractive desire conflicts with another goal, standard, or longer-term commitment. This fundamental research monograph examines the conceptual history of volition and delayed gratification; real-world measurements of desires, resistance, and enactment; the original strength-resource account of ego depletion; the 2010 meta-analytic literature and its critiques; preregistered multilaboratory tests published in 2016 and 2021; and a 2025 time-on-task study. It critically compares evidence for motivational shifts, subjective effort costs, value integration, executive control, planning, habit formation, and proactive situational regulation. The synthesis distinguishes a documented everyday conflict from the stronger and inadequately established claim that acts of self-control consume a common finite resource. Experimental effect sizes are assessed with their task boundaries, analytic status, and uncertainty. Research priorities concern temporal mechanisms, cross-task transfer, ecological generalization, construct validity, and the organization of choice environments during the Artificial Era. This is a critical narrative synthesis of existing scholarship, not a new empirical study or protocol-based systematic review.
Keywords: self-control; self-regulation; desire conflict; temptation; willpower; ego depletion; strength model; registered replication; preregistration; experience sampling; effort cost; motivation; executive control; value-based choice; habit; implementation intentions; mental fatigue; Artificial Era
The Scientific Problem of Self-Control and Competing Desires
Self-control concerns the regulation of behavior when a currently attractive response competes with a goal, standard, or commitment that has greater priority in a broader time frame. It can involve interrupting an ongoing action, withholding an impulse, changing the salience of alternatives, avoiding a tempting setting, forming a useful habit, or making a future-oriented choice before a conflict emerges. The subject of this monograph is the relation among experienced desire, desire–goal conflict, the psychological processes mobilized to manage conflict, and the disputed claim that earlier acts of control temporarily consume a general resource. Successful self-control must be defined with respect to the actor's actual goal and the situation, rather than inferred from conformity to an observer's preferred behavior. A refusal can be goal-consistent, and so can a carefully chosen indulgence. The relevant evidence derives from experience sampling, intervention experiments, individual-difference studies, sequential-task paradigms, preregistered replications, and formal models of control allocation. Fujita (2011) Kotabe & Hofmann (2015)
The central explanatory questions are how desires arise and become salient; when desires count as temptations; why conflict recruits resistance in some episodes but not others; what determines whether resistance succeeds; and whether effortful regulation changes subsequent performance through a transferable depletion mechanism. These questions have different observational units. A desire episode unfolds over seconds or minutes, a habit develops across repeated contexts, a laboratory task assesses performance under specified rules, and a person's self-control score summarizes heterogeneous tendencies. Evidence at one level cannot automatically establish a mechanism at another. The dossier therefore treats ego depletion as a contested empirical prediction within a larger field of reliable observations about conflict, value, attention, and behavior. Hofmann et al. (2012a) Hagger et al. (2016) Vohs et al. (2021)
This work is a critical narrative research synthesis of verified scholarly publications and the studies described in their accessible original records. No original participants were recruited, no new effect size was calculated, and no systematic-review search protocol or meta-analysis was undertaken. Where numerical findings are reported, they are drawn from identified sources. Theory, measured association, randomized task effect, and generalization beyond the tested procedure remain separate analytical categories. The aim is a coherent evidence-based account of self-control that remains informative even where the evidence for any single depletion hypothesis is equivocal. Friese et al. (2019) de Ridder et al. (2012)
Conceptual Boundaries: Self-Control, Self-Regulation, and Inhibition
Self-regulation is the broader organization of psychological activity around goals and standards, including setting goals, monitoring discrepancies, adjusting strategies, and revising aims. Self-control denotes a subset of regulatory problems in which competing motivational tendencies must be reconciled in favor of some prioritized commitment. Response inhibition is a family of cognitive operations measured by tasks such as withholding or stopping a prepared response. Inhibition can contribute to self-control, but performance on a simple stop or interference task is only one indicator of one component. A participant may suppress a motor response successfully while finding it difficult to plan around environmental temptations, or may regulate behavior successfully by removing cues without engaging in visible inhibition. Fujita (2011) Hofmann et al. (2012b)
Desire means a subjectively experienced wanting directed at an activity, object, or state. Temptation is a relational description: a desire is tempting when its enactment threatens another salient goal. Conflict refers to the perceived or functionally detectable incompatibility between motivational options. Resistance is an attempt to oppose enactment; enactment is the resulting behavior, which may follow strong desire, failed resistance, or an intentional permission to pursue that desire. These terms should be assigned different variables in an empirical design. Treating desire, conflict, effort, and unsuccessful behavior as interchangeable creates circular accounts in which an outcome is both explained by and used to define diminished control. Hofmann et al. (2012a) Kotabe & Hofmann (2015)
Trait self-control is a relatively stable between-person construct usually assessed by a questionnaire or aggregated behavior. State self-control refers to context-dependent regulatory processing. Subjective depletion refers to the experience of tiredness or reduced willingness to exert control; the classical ego-depletion effect refers specifically to a decrement in a later task after a prior regulatory task. The distinction is decisive: an association between self-reported tiredness and temptation does not identify a consumed metabolic or psychological resource. Likewise, a stable trait correlation with grades or health is compatible with different causal mechanisms, including habit, planning, favorable environments, and social advantage. Tangney et al. (2004) Galla & Duckworth (2015) Milyavskaya & Inzlicht (2017)
Historical Foundations: Delay, Volition, and Resource Hypotheses
Experimental inquiry into delayed gratification supplied early evidence that the configuration of attention changes decisions under temptation. Work by Mischel, Ebbesen, and Zeiss in 1972 investigated cognitive and attentional mechanisms in situations requiring children to wait for a preferred reward. The subsequent theoretical account by Mischel, Shoda, and Rodriguez in 1989 positioned the delay task within a broader developmental problem: how immediate incentives, attention, and representations influence deferred outcomes. These experiments established tractable laboratory paradigms for studying delay, although their interpretation does not entail that all self-control operates as a single capacity or that delay scores are independent of social conditions. Mischel et al. (1972) Mischel et al. (1989)
In the 1990s, Muraven, Tice, and Baumeister and, separately, Baumeister, Bratslavsky, Muraven, and Tice developed a strength model in which diverse acts of volition draw on a finite shared capacity. The 1998 studies included emotional regulation, choice, persistence, and the now familiar comparison between eating radishes and resisting chocolates before working on difficult puzzles. Their conceptual achievement was a falsifiable cross-task claim: an initial requirement for regulation should impair later regulation even when the tasks differ. The experiments, however, also combine motivational, perceptual, and performance demands that allow competing explanations. The phrase ego depletion named the proposed aftereffect, while an underlying resource was posited to explain it. Muraven et al. (1998) Baumeister et al. (1998)
The strength account gained influence through a 2010 meta-analysis by Hagger and colleagues covering 83 studies. It also attracted growing methodological scrutiny about publication selection, small-study effects, task heterogeneity, and the flexibility with which many experimental activities could be labeled depleting. Carter and McCullough raised publication-bias concerns in 2014; Carter and colleagues performed additional meta-analytic tests in 2015. Large preregistered collaborations subsequently supplied more constraining tests, including the 2016 registered replication and multisite studies published in 2021. The history therefore involves a succession of theoretical proposals and increasingly demanding tests, rather than a simple replacement of one theory by another. Hagger et al. (2010) Carter & McCullough (2014) Carter et al. (2015) Hagger et al. (2016)
Desire as a Psychological Episode
Desires vary in object, strength, urgency, anticipated reward, and perceived legitimacy. An intense urge can be aligned with a person's long-term commitments; a weaker urge can become a temptation because its consequences conflict with a valued standard. The strength of wanting is thus different from the degree of conflict. The major naturalistic study by Hofmann, Baumeister, Förster, and Vohs asked 205 adults to report desire episodes over a week and obtained 7,827 reports. Its distinctive contribution was to observe desire intensity, conflict, resistance, and enactment as related but separate moments of real-world motivation. Many desires appeared with little or no relevant conflict. Desire frequency and consequences differed across domains and contexts. Hofmann et al. (2012a)
The companion analysis by Hofmann, Vohs, and Baumeister summarized substantial variation across types of desire and the probability of resisting them. Such data reveal how often a person experiences competing motives and how this experience varies across daily activities. They do not isolate a causal mechanism in the way that random assignment to a manipulated desire cue or task demand can. Reports collected during everyday life can be influenced by whether participants notice an impulse, what they count as a desire, the timing of prompts, willingness to disclose sensitive behavior, and changes in behavior caused by monitoring. Despite these constraints, experience sampling provides evidence about ecological occurrence that laboratory tasks cannot supply on their own. Hofmann et al. (2012c) Hofmann et al. (2012a)
The temporal ordering matters. A cue can activate wanting before a person explicitly articulates a goal; awareness of a conflicting goal can then elicit discomfort or resistance. Once a response begins, habits and situational constraints may alter the possibility of stopping. A person can also manage likely desire encounters prospectively by arranging an environment. A scientific model should specify the time at which each variable is measured because retrospectively asking whether an unsuccessful action was tempting can create an explanatory artifact. Direct behavioral records, time-stamped reports, and experimentally varied cue exposure help distinguish an initial desire from subsequent evaluation of that desire. Kotabe & Hofmann (2015) Duckworth et al. (2016)
The Architecture of Desire–Goal Conflict
A desire–goal conflict arises when enactment of a currently valued option is perceived to undermine another goal. Fujita's dual-motive account defines self-control through the advancement of more distal motivations over more proximal ones when they compete. Kotabe and Hofmann proposed a component model relating desire, higher-order goals, conflict, control motivation, control capacity, effort, and enactment constraints. Its value for research is in separating the motivational condition that elicits control from the resources or strategies that might be mobilized in response. The same action can change status from temptation to chosen priority when a person's goals change, making goal measurement essential to interpretation. Fujita (2011) Kotabe & Hofmann (2015)
Conflict involves more than the simple magnitude of an impulse. A weak but normatively charged desire may generate strong guilt; a strong desire aligned with a chosen goal may produce no resistance. Goal importance, temporal distance, self-relevance, expectations, perceived autonomy, and available alternatives can all modify whether a desire counts as threatening. Laboratory studies that infer conflict merely from the attractiveness of a stimulus risk misclassifying cases in which participants have little reason to suppress it. Conversely, strong laboratory task instructions can create compliance pressure independently of participants' preexisting goals. A methodologically sound design elicits or measures the relevant counter-goal and distinguishes willingness to comply with instructions from self-initiated regulation. Hofmann et al. (2012a) Berkman et al. (2017)
Conflict can also reactivate the very goals it threatens. Fishbach, Friedman, and Kruglanski reported experimental evidence that temptation-related cues may activate overriding goals, especially where these goals are important and accessible. Thus exposure to temptation can sometimes prime restraint rather than merely increase pressure toward gratification. This dependence on goal organization makes static accounts of temptation incomplete. Repeated exposure, learned cue–goal associations, institutional environments, and social expectations may alter the direction of the effect. The decisive experimental comparison is between conditions with matched stimulus intensity but different goal accessibility or consequence structures. Fishbach et al. (2003) Wood & Neal (2007)
From Conflict to Action: Resistance, Strategies, and Outcomes
Resistance is a behaviorally and psychologically heterogeneous category. It may consist of looking away, reframing a stimulus, remembering a commitment, suppressing a motor response, postponing a decision, recruiting another person's assistance, or entering a different environment. The strength model emphasized direct effortful inhibition, whereas situational approaches emphasize choosing or modifying circumstances before temptation becomes acute. Duckworth, Gendler, and Gross's analysis of situational strategies explicitly describes means of selecting and changing contexts so that goal-incompatible impulses are less likely to dominate. Regulatory success can therefore be achieved by decreasing exposure, changing meaning, increasing goal salience, or automating an appropriate response. Duckworth et al. (2016) Fujita (2011)
A fundamental interpretive problem is that visible struggle may be inversely related to effective self-control. People who encounter fewer temptations, experience less conflict, or possess useful habits may exert less conscious resistance yet show better goal attainment. In six studies totaling 2,274 participants, Galla and Duckworth found evidence that beneficial habits mediated links between trait self-control and desirable outcomes across activities such as studying, sleep, exercise, and eating. Mediation in these designs supports a plausible pathway but does not by itself prove that a questionnaire trait caused habit formation. Nevertheless, the findings require a conception of self-control that includes the organization of behavior before a crisis of inhibition occurs. Galla & Duckworth (2015)
A further challenge is the meaning of success. Resistance followed by nonenactment may represent a desired result, but resisting a legitimate need can be maladaptive. Persistence may be beneficial until information changes the desirability of the goal. Research needs participant-defined priorities, independently observed behavior, and an explicit outcome horizon. Immediate refusal of a reward is not always better than taking it, and short laboratory persistence is not always equivalent to long-term achievement. Empirical claims are most informative when success is measured against a prespecified, substantively justified goal rather than assumed from conformity with a task instruction. Berkman et al. (2017) Watts et al. (2018)
Measurement and Identification of Regulatory Mechanisms
Experience sampling improves ecological coverage by repeatedly probing whether a desire occurred, how strong it was, whether it conflicted with another goal, and whether it was enacted. Yet reports are conditional on availability, memory over the reporting interval, and the ability to label internal states. The original everyday-desire protocol offers a model for temporally close measurement, while its observational design prevents a causal conclusion that conflict alone produces resistance or that resistance depletes capacity. Research can extend it with passive time stamps, situational covariates, preregistered coding, and within-person models separating deviations from a person's usual level from between-person differences. Hofmann et al. (2012a) Hofmann et al. (2012c)
Trait inventories such as the Tangney–Baumeister–Boone Self-Control Scale summarize tendencies across many domains. Correlations with grades, adjustment, and interpersonal functioning show predictive relevance, but the questionnaire combines self-description with normative judgments and can share method variance with the criterion. In a meta-analysis, de Ridder and colleagues assembled 102 studies with a total sample of 32,648 participants and demonstrated the wide range of outcomes related to trait self-control. Prediction from trait scores should be tested against conscientiousness, opportunity, socioeconomic conditions, habit, and domain-specific skill when the intended inference is causal or mechanistic. Tangney et al. (2004) de Ridder et al. (2012)
A sequential-task experiment randomly assigns participants to a purportedly demanding versus control first task and measures a second outcome such as persistence or response inhibition. The critical causal estimand is the effect of assigned first-task condition on second-task performance, holding instructions and measurement conditions fixed. Task difficulty, baseline performance, boredom, frustration, expectancy, experimenter demand, time-on-task, motivation to continue, and speed–accuracy strategy may all produce changes in the criterion. A manipulation check for feeling tired does not establish that a distinct general resource was consumed. Replication demands specified tasks, outcome scoring, exclusions, preregistration, and sufficient power for an effect compatible with recent large-sample estimates. Hagger et al. (2016) Vohs et al. (2021) De Luca et al. (2025)
The Strength Model and Its Testable Predictions
The classical strength model makes a distinctive causal claim: diverse acts of self-control draw on a finite, domain-general resource whose availability declines following exertion and can recover after an interval. Under the strongest interpretation, the resource produces a sequential-task decrement beyond ordinary task-specific adaptation, practice, motivational change, or changes in attention. This explanatory structure generated substantial interest because it linked regulation of eating, emotion, choice, and persistence under one mechanism. The underlying resource itself, however, required independent identification. A behavioral difference between conditions can support the effect's existence without demonstrating a specific resource, and a failed behavioral replication can challenge a proposed paradigm without logically excluding every version of a more restricted mechanism. Baumeister et al. (1998) Muraven et al. (1998)
A useful hypothesis must identify which acts consume the resource, what quantity is reduced, how depletion is measured independently of the outcome, what replenishes the resource, and what functional relation links initial demand to later impairment. If any difficult task may be reclassified retrospectively as not sufficiently depleting, the theory risks losing falsifiability. Conversely, a design that lacks a plausible initial control demand may test an uninformative boundary condition. Research can improve discrimination by measuring first-task engagement, defining an externally validated dose, and comparing resource-specific predictions against motivated task switching, fatigue, and expectancy explanations. Inzlicht & Berkman (2015) Friese et al. (2019)
The Original Sequential-Task Experiments
Baumeister and colleagues' 1998 experiments manipulated initial actions and measured subsequent persistence, problem solving, or passivity. In a particularly well-known condition, participants constrained their eating to radishes when chocolates were present and subsequently gave up sooner on unsolvable puzzles. Other experiments examined personal choice and emotional suppression. The results motivated a cross-domain proposal, because the first task and the second were not identical. A direct effect of instructions on subsequent persistence, however, may also involve differences in frustration, perceived task meaning, appetite, mood, and willingness to cooperate. Persistence on an impossible problem is especially sensitive to beliefs about attainability and to the perceived value of additional attempts. Baumeister et al. (1998)
Muraven, Tice, and Baumeister's contemporaneous work likewise examined the possibility that regulation in one domain impaired control in another. Together these articles established the sequential-task paradigm as a research program. Their historical importance does not substitute for independent corroboration of effect size and scope. The common metaphor of a muscle or reservoir is not an observed experimental variable. It is a theoretical analogy that requires the same scrutiny as any other mechanistic explanation. A contemporary evaluation therefore separates successful manipulation, measured criterion difference, and evidence that this difference transfers through a general capacity, rather than treating these as one inferential step. Muraven et al. (1998) Baumeister et al. (1998)
Meta-Analytic Accumulation and Its Vulnerabilities
Hagger, Wood, Stiff, and Chatzisarantis published a 2010 synthesis of 83 studies that initially supported a substantial sequential-task phenomenon. The meta-analytic approach increased the apparent evidential base by combining many task pairs. Yet heterogeneous tasks differ in their likely cognitive mechanisms and in their susceptibility to expectancy, demand, fatigue, and affect. Pooling them becomes theoretically meaningful only if the common effect is sufficiently well defined. The size and consistency of estimates can be distorted if small studies with weak or unfavorable outcomes are less likely to be published. A meta-analysis of selected published outcomes cannot correct such selection without assumptions about the missing evidence. Hagger et al. (2010)
Carter and McCullough's 2014 critique examined publication bias and the possibility that the limited-strength evidence had been overestimated. Carter, Kofler, Forster, and McCullough's 2015 series of meta-analytic tests further challenged the inference that the accumulated record demonstrated consumption of a general resource. These analyses were methodologically important because they confronted the relation between statistical patterns and study-selection processes. Their conclusions also depend on the behavior of bias-correction methods, which can respond differently to heterogeneity and selection. Divergent corrections must be interpreted as sensitivity analyses rather than as transparent readings of a single true effect. Carter & McCullough (2014) Carter et al. (2015)
Dang's 2018 updated meta-analysis presented a different reading of the depletion literature and contributed to continued disagreement over effect size and design characteristics. Friese and colleagues' 2019 assessment of arguments from both proponents and critics made clear that neither a broad endorsement nor a blanket dismissal adequately specifies which phenomena the experiments identify. High-quality evidence must distinguish a general average effect from a task-specific result, and a tiny effect from the previously proposed substantial resource reduction. The direction of a meta-analytic estimate is only one part of that judgment: precision, design selection, preregistration, and explanatory specificity matter equally. Dang (2018) Friese et al. (2019)
The 2016 Registered Replication Report
A coordinated registered replication led by Hagger and colleagues used a standardized sequential-task protocol across 23 laboratories with a combined sample of 2,141 participants. Its meta-analytic estimate of the ego-depletion effect was d = 0.04, with a 95% confidence interval extending from −0.07 to 0.15. The interval included zero and was incompatible with many earlier expectations of a substantial average effect in that protocol. The design's preregistration and distributed implementation reduced the opportunities for selective outcome reporting and provided an explicit benchmark against which the original strength claim could be judged. Hagger et al. (2016)
This result is a strong constraint on claims of a robust, easily reproduced decrement under the tested conditions. Its inference is also properly bounded. A standardized paradigm captures a specified family of control demands, so failure to observe a substantial effect there is not a proof that every possible sequential-task design has a zero effect. The scientifically productive conclusion is to narrow proposed mechanisms and preregister the circumstances in which a new, independently meaningful effect is expected. Without this narrowing, post hoc explanations of null results generate additional flexibility without increasing predictive success. Hagger et al. (2016) Friese et al. (2019)
The 2021 Multisite Paradigmatic Replication
Vohs and collaborators published a preregistered paradigmatic replication across 36 laboratories and 3,531 participants. Laboratories implemented one of two initial control procedures before a second self-control measure. The confirmatory analysis yielded a nonsignificant effect of d = 0.06. An informed-prior Bayesian analysis using a predicted effect centered at 0.30 reported data about four times more likely under the null model than under the specified alternative. Exploratory analyses using the full sample irrespective of prespecified exclusions yielded d = 0.08, with a much less decisive Bayesian comparison. These are distinct analyses with different evidential status; the exploratory result must not replace the preregistered primary test. Vohs et al. (2021)
The project is particularly informative because researchers attempted paradigmatic replication rather than repeating only one earlier task pair. This addresses the criticism that a single laboratory script could miss the broader theory. The results still failed to support the larger confirmatory effect. Exploratory variation by self-reported fatigue invites more specific hypotheses, but fatigue reported after a difficult task may index subjective cost or perceived difficulty rather than a resource with independent physical identity. A moderator that emerges outside the confirmatory plan requires a future direct test and a prespecified interaction estimand. Vohs et al. (2021)
A methodological consequence follows from the comparison between 2016 and 2021. Large collaborations sharply reduce uncertainty about the average effect expected from their protocols. They do not establish that every kind of self-control operates independently of prior work, because self-control is not a single observed variable. The relevant question shifts from whether one broad metaphor can accommodate any result to which narrowly defined activities, tasks, populations, and outcome measures show credible changes after demanding work. This shift transforms the research program into testable conditional predictions. Hagger et al. (2016) Vohs et al. (2021)
A Small Positive Multilaboratory Effect and What It Establishes
A preregistered multilaboratory study led by Dang, published in 2021, tested a Stroop-to-antisaccade sequence across 12 laboratories and 1,775 participants. The pooled effect was small but statistically significant, d = 0.10; a specified analysis excluding participants who may have responded randomly to the outcome task yielded d = 0.16. This finding is important precisely because it does not fit an account in which all preregistered evidence is uniformly null. The effect is compatible with some transfer between tasks under this protocol and demonstrates the value of studying particular task combinations rather than collapsing all evidence into an undifferentiated verdict. Dang et al. (2021)
A small observed sequential difference remains mechanistically ambiguous. Stroop performance may alter attentional set, arousal, motivational expectations, or response strategies that affect subsequent antisaccade trials. A positive experimental estimate does not by itself show that the same general substance is consumed by emotion regulation, choosing food, inhibiting eye movements, and solving puzzles. The scientific interpretation is an empirically bounded task-sequence effect, whose generality and mediation must be tested separately. Across large collaborations, the most defensible conclusion is that the evidence does not establish the substantial, general, invariant depletion effect traditionally attributed to a finite strength resource, even though small context-sensitive aftereffects warrant continued investigation. Dang et al. (2021) Hagger et al. (2016) Vohs et al. (2021)
The 2025 Time-on-Task Prediction Test
De Luca, Zhang, Verschoor, and Hommel reported a 2025 online experiment examining a direct temporal prediction of depletion theory. Participants completed numerical Stroop blocks with different proportions of incongruent trials and a global–local task designed to measure later control. A depletion process with cumulative demand would predict that performance costs systematically increase with time spent under the more demanding condition. Although the experiments reproduced familiar interference effects, the authors found no evidence for the predicted progressive ego-depletion outcome in their task design. The work strengthens the argument that a theory of depletion must specify how demand accumulates and which outcome should track that accumulation. De Luca et al. (2025)
The study should be understood as an investigation of a particular temporal implication. It cannot settle every version of fatigue, all task sequences, or every possible motivational boundary condition. Its distinct contribution is methodological: a mechanism framed as gradual expenditure should make quantitative predictions about the dose of prior demand and the slope of subsequent change. A model that predicts only some decrement at an unspecified time can be sustained by many interpretations. A model that predicts a graded relation between controlled exposure and later performance is much more tightly constrained by data, including negative results. De Luca et al. (2025) Inzlicht & Berkman (2015)
Comparative Assessment of the Depletion Evidence
The evidential record contains early influential positive studies, meta-analytic disagreement, large near-zero preregistered estimates, one smaller but positive multilaboratory effect, and a newer negative time-on-task test. It supports treating broad resource depletion as an unconfirmed explanatory mechanism and encourages specific investigations of sequential-task interactions. This conclusion applies to the mechanism and effect family as historically framed; it does not negate the everyday fact that sustained effort can feel costly, that task performance sometimes declines, or that fatigue can influence behavior. Fatigue, boredom, motivation, task-specific learning, and opportunity costs require their own measurements and experimental discriminations. Baumeister et al. (1998) Carter et al. (2015) Hagger et al. (2016) Dang et al. (2021) Vohs et al. (2021) De Luca et al. (2025)
Three inferential layers should remain separate in subsequent scholarship. First, a manipulation may change a downstream score. Second, a genuine common process may explain the change across tasks. Third, the process may be a finite consumable resource. Each layer requires evidence beyond the preceding one. The third claim needs a measurable resource, quantitative depletion and recovery predictions, and comparisons that eliminate motivational and strategic accounts. In a critical synthesis, uncertainty should be located exactly at these inferential boundaries. This preserves observable regulatory phenomena while preventing metaphor from functioning as a substitute for a tested mechanism. Inzlicht et al. (2014) Friese et al. (2019)
Motivation and Attention as Alternatives to Resource Depletion
The process account proposed by Inzlicht, Schmeichel, and Macrae identifies changes in motivational priorities and attention as possible sources of declining willingness to continue demanding tasks. After an extended period of externally imposed control, a participant may increasingly prioritize activities that are intrinsically rewarding or less effortful. This account proposes a shift in relative priority between have-to and want-to activities rather than a literal consumption of a generic psychological fuel. It naturally predicts sensitivity to reward, interest, autonomy, and anticipated alternatives, while leaving open whether fatigue and task-specific capacity changes also contribute. Its explanatory promise lies in hypotheses about observable shifts in incentives and attention over time. Inzlicht et al. (2014)
An experiment contrasting resource and motivation accounts must measure more than total performance. It can manipulate the value of the second task, give participants an opportunity to switch to a reward, vary perceived control over continuing, or assess attentional orientation to alternative activities. If participants continue performing well when incentives are altered after an initial effortful task, a strict inability-to-perform interpretation is weakened. However, preserved performance under high incentives does not prove that all limitations are purely voluntary, because incentives can recruit compensatory processes. The strongest tests estimate the relation among effort, reward, performance, subjective fatigue, and choice within the same participants. Inzlicht et al. (2014) Inzlicht & Berkman (2015)
Subjective Effort and Expected Value of Control
Effort is a psychological cost that must be measured according to the task, decision, and time horizon at issue. Westbrook, Kester, and Braver used economic preference methods to quantify willingness to undertake different cognitive loads, showing that cognitive effort can be treated as a subjective cost in choice. Shenhav, Botvinick, and Cohen proposed the expected-value-of-control framework, in which deployment of control depends on anticipated benefits and costs. These approaches explain variation in control allocation without requiring a globally depleted reservoir. Their neural hypotheses and economic parameters nonetheless need empirical tests rather than automatic identification with introspective experiences of effort. Westbrook et al. (2013) Shenhav et al. (2013)
The expected-value framework changes what scientists consider the dependent variable. A person who declines another demanding task may be selecting the higher-valued option under prevailing conditions rather than exhibiting inability. Distinguishing inability, reluctance, altered reward sensitivity, and opportunity cost requires choice alternatives and calibrated incentives. A sequential-task experiment with only a mandatory second task hides the option structure that may produce apparent deficits. A design that offers choices, estimates subjective values, and models trialwise allocation can reveal the extent to which the first task changes motivational valuation rather than performance capability. This distinction also matters in education and employment, where external rewards and perceived autonomy can substantially alter engagement. Shenhav et al. (2013) Berkman et al. (2017)
Self-Control as Value-Based Choice
Berkman, Hutcherson, Livingston, Kahn, and Inzlicht presented self-control as a process of dynamic value integration. Immediate consumption, future well-being, identity commitments, social consequences, and effort can all contribute to the subjective valuation of available actions. The experienced conflict between impulse and restraint may emerge from competition between weighted attributes rather than from separate systems with one inherently capable of overriding the other. Control can work by changing attention to relevant attributes, changing the accessibility of future outcomes, or selecting situations in which the preferred action has greater value. Such a model invites quantitative predictions linking changes in stimulus features to choice. Berkman et al. (2017)
Value-based explanations should be discriminated from circular descriptions. If every outcome is explained retrospectively by saying its option had the greatest subjective value, the framework predicts little. Empirically useful models require values estimated independently, time-sensitive response measures, out-of-sample choice prediction, and manipulations of specific attributes. They should identify conditions in which a value change occurs without altered response inhibition and cases where inhibitory performance shifts without a change in expressed goals. By separating the components, a common theory may integrate motivational conflict with cognitive control while avoiding the reduction of all regulation to a single unmeasured disposition. Berkman et al. (2017) Kotabe & Hofmann (2015)
Willpower Beliefs and Metacognitive Expectations
Job, Dweck, and Walton's 2010 studies tested lay theories about whether willpower is limited. Experimental and longitudinal findings suggested that beliefs can moderate the consequences of demanding activities in some contexts. These studies introduced a psychologically substantive explanatory candidate: expectations about one's capacity may shape task interpretation, readiness to persist, and the experience of fatigue. They did not establish that all previous depletion findings are expectancy effects or that believing in unlimited willpower protects every person under every demand. Beliefs can operate alongside fatigue, motivation, available opportunities, and task-specific skill. Job et al. (2010)
The measurement challenge is separating belief as a stable individual difference from belief manipulated during a study, demand characteristics from actual engagement, and self-report from objective behavior. Participants explicitly told that willpower is unlimited may also infer that the experiment expects prolonged persistence. A rigorous intervention should use credible comparison conditions, assess manipulation uptake, measure performance unobtrusively, and determine how long any effect lasts. The large paradigmatic replication of 2021 did not find the expected moderation by willpower beliefs in its preregistered context, an important constraint on any unrestricted belief-based account. Job et al. (2010) Vohs et al. (2021)
Executive Functions and Their Relation to Motivational Control
Working-memory updating, behavioral inhibition, and task switching provide important tools for goal-directed regulation. Hofmann, Schmeichel, and Baddeley reviewed their possible contributions to self-regulation and described mechanisms by which temporary demands could impair goal-directed conduct. Yet executive performance is task-dependent and multifactorial. A Stroop interference score, a go/no-go error rate, a stop-signal measure, and real-world self-reported discipline are not identical observations of one common capacity. Their correlations depend on measurement reliability and on the tasks' specific cognitive requirements. Executive resources and motivational priorities may interact, so a change in one laboratory score should not be treated as a complete measurement of the ability to pursue personal goals. Hofmann et al. (2012b)
Strong causal identification requires predefining which executive component is proposed to mediate a desire–goal outcome. For example, prepotent-response suppression may matter when a salient cue directly elicits a practiced action; working-memory maintenance may matter when a counter-goal has to remain available despite distraction. Measuring both under repeated conditions allows a within-person test of whether change in the purported mediator precedes change in enactment. Failure of one proxy measure to relate to real-life control cannot falsify every executive contribution, but broad claims that inhibition universally determines restraint need consistent cross-task and longitudinal evidence. Hofmann et al. (2012b) Kotabe & Hofmann (2015)
Habits as Preventive Self-Control
Habits are associations between situational contexts and repeatedly performed responses. Wood and Neal's habit–goal framework explains how learned cue–response links can persist without renewed deliberation whenever the context recurs. Habits can assist valued aims or compete with them; their implications depend on how the behavior was learned and what goals currently matter. Galla and Duckworth's investigations suggest that people with higher trait self-control often rely on beneficial routines that reduce conflict and the subjective requirement for effortful inhibition. The capacity to establish and inhabit a supportive routine is therefore a plausible component of stable regulatory success. Wood & Neal (2007) Galla & Duckworth (2015)
Habit-based accounts yield a different intervention target from replenishing a depleted resource. Changing cues, stabilizing contexts for desired actions, and reducing access to recurrent triggers may alter behavior without increasing momentary capacity to say no. Evidence must still distinguish habit from deliberate planning and environmental constraint. A person who keeps tempting food out of reach may be expressing stable priorities or responding to circumstances rather than demonstrating an automatic association. Designs combining repeated behavior, context stability, self-reported automaticity, and experimental cue changes can test whether habit formation actually mediates lasting goal attainment. Wood & Neal (2007) Duckworth et al. (2016)
Temporal Control: Planning, Delay, and Reconsideration
Implementation intentions link a recognizable future situation to a specified response. Gollwitzer's 1999 program showed how if–then planning can facilitate initiation and persistence by making an appropriate response accessible when a cue appears. The strategy moves part of regulatory work to an earlier decision point. In desire conflict, selecting the response before a temptation becomes acute may reduce the need for effortful deliberation at the critical moment. The mechanism is prospective cue–response organization, which is conceptually distinct from acquiring a permanent habit and from suppressing a response already under way. Gollwitzer (1999)
Delay-of-gratification tasks also involve temporal valuation, attention, and the credibility of future rewards. Early experiments emphasized attentional strategies; later analyses explored the associations between childhood delay and later outcomes. Watts, Duncan, and Quan's 2018 conceptual replication showed that these associations were substantially attenuated when family background and early cognitive and environmental conditions were taken into account. This demonstrates how an apparently stable individual-difference predictor can partly reflect structured opportunities, learned expectations, and baseline characteristics. It does not make delay tasks useless: it clarifies which inferences require adjustment and independent measurement. Mischel et al. (1972) Mischel et al. (1989) Watts et al. (2018)
Psychological Fatigue and Regulation Over Time
Mental fatigue is a phenomenological and behavioral phenomenon involving experienced cost after sustained engagement and, in some settings, changing accuracy, speed, or motivation. It cannot be identified solely with ego depletion. A person may be fatigued while retaining the capacity to perform under altered incentives, or may exhibit task-specific decline without reporting substantial tiredness. The process model treats fatigue as potentially informative about the relative value of continued cognitive labor. Resource models treat it more directly as a possible sign of expenditure. These alternatives imply different temporal relations among task exposure, subjective reports, performance, and subsequent choice. Inzlicht et al. (2014) Inzlicht & Berkman (2015)
The distinction is practically consequential. Generalizations from short laboratory tasks to sleep loss, prolonged clinical distress, occupational exhaustion, or repeated everyday decisions require additional evidence. These situations differ in physiological state, incentive structure, personal control, and exposure duration. The evidence for a small or absent average sequential-task ego-depletion effect does not license claims about sleep or physical recovery. Similarly, feelings of depletion after a temptation-rich day can be genuine while their cause remains uncertain. The preferred research strategy is repeated measurement of effort, fatigue, desire exposure, and behavior with designs that distinguish temporal order from stable individual differences. Milyavskaya & Inzlicht (2017) De Luca et al. (2025)
Research Designs for Everyday Desire and Self-Control
Ecological momentary assessment samples psychological experiences while participants carry out everyday activities. An effective design asks about desire presence, strength, conflicting goals, resistance, affect, context, and actual enactment, with items differentiated so that self-control is not tautologically inferred from an unsuccessful outcome. Within-person models can estimate whether the same individual is more likely to resist when a goal is salient, while between-person models can examine whether stable tendencies predict exposure to conflict. These are different estimands and should be specified before analysis. The 2012 everyday-temptation studies illustrate the value of retaining each component of a desire episode as its own variable. Hofmann et al. (2012a) Hofmann et al. (2012c)
Longitudinal research can investigate whether repeated conflict predicts exhaustion, habit change, or eventual goal attainment. Yet reciprocal causation is likely: people who fall behind their goals may encounter more reminders of failure and report greater conflict; those with established routines may encounter fewer temptations. Dense sampling, prospective measurement of goals, lagged outcomes, and controlled situational interventions help distinguish these directions. Such research should report participant burden, missing prompts, attrition, sampling windows, domain coding, and clustering of observations. Without these details, many thousands of episodes from a small number of people can misleadingly appear to constitute thousands of independent participants. Milyavskaya & Inzlicht (2017) Hofmann et al. (2012a)
Experimental Manipulations and Causal Inference
A randomized sequential-task design provides an estimate of whether assignment to a first-task condition changes later performance. It must isolate control demand from other features such as unpleasantness, total duration, incentives, comprehension, and task familiarity. In a convincing factorial design, the difficulty of the initial activity, the participant's freedom to choose it, and the incentive for the second task are varied separately. Baseline cognitive performance can be measured independently; any exclusions must be defined before looking at outcome differences. Where multiple outcomes are plausible, primary and secondary endpoints should be distinguished to prevent selectively emphasizing whichever task happens to produce a significant result. Hagger et al. (2016) Vohs et al. (2021)
The resource hypothesis has stronger requirements than showing that performance sometimes drops after exertion. A candidate resource must be operationalized independently, a change in it must precede performance changes, and interventions that restore it should have specific effects relative to plausible alternatives. A task may cause frustration or make leisure more attractive without producing a decrement in a transferable capacity. For this reason, mediation of an assignment effect by post-task fatigue or subjective effort cannot automatically be interpreted causally: the mediator is influenced by the same task that also changes expectations and task engagement. Converging manipulation, measurement, and computational predictions are needed. Inzlicht & Berkman (2015) Shenhav et al. (2013)
Statistical Precision, Heterogeneity, and Replication
Statistical significance asks whether a particular test statistic is surprising under a model, whereas scientific importance depends on the effect's magnitude, precision, mechanism, and consequences. In the large 2016 registered replication, the pooled estimate was d = 0.04 with a 95% interval of −0.07 to 0.15; in Vohs and colleagues' 2021 confirmatory assessment it was d = 0.06. These estimates challenge large, stable, transferable claims under the tested protocols. Dang and collaborators' 2021 small significant estimate of d = 0.10 shows that task selection and protocol characteristics deserve investigation. None of these values should be interpreted as the percentage of people depleted, the proportion of mental energy lost, or a direct prediction about a specific individual. Hagger et al. (2016) Vohs et al. (2021) Dang et al. (2021)
A replication must specify which aspect of the original hypothesis it reproduces: exact task materials, theoretical manipulation, behavioral estimand, or underlying mechanism. A direct replication of a task script may improve the estimate for that script; a paradigmatic replication probes broader generality but often introduces heterogeneity. Multisite analysis should distinguish a common mean effect from variation among sites and procedures, account for clustering, and report deviations and exclusion rules. Unsuccessful and successful replications together constrain theory more effectively than any selectively quoted study. A cumulative evidence base also needs public protocols, code, and sufficiently informative null results to reduce selective survival of favorable findings. Friese et al. (2019) Hagger et al. (2016) Vohs et al. (2021)
Construct Validity and the Problem of Task Impurity
The same observed error or delay can arise through several processes. Stroop interference may reflect selective attention, response conflict, strategy, reading automaticity, or speed–accuracy adjustments; antisaccade errors require attention to stimulus detection, motor preparation, and task comprehension. A decrement in one task score cannot simply be relabeled loss of self-control without an independently justified mapping. Task impurity becomes especially serious when a first task is selected because it produces an effect and the second is selected because that effect is statistically detectable. Both must be validated against the theoretical constructs they are supposed to measure. Hofmann et al. (2012b) Dang et al. (2021)
Researchers should use multiple indicators for each component, assess reliability, and explicitly test whether a shared latent process explains covariation among measures. Repeated-measures tasks can separate stable differences from changes within individuals, but carry-over and practice effects must be modeled. Experiments using many researcher degrees of freedom should make analytic choices auditable. A useful report records not only that the manipulation was challenging but how participants performed, how their responses changed across blocks, and whether the effect persists after accounting for engagement and task order. Such measurements make competing explanations empirically distinguishable. De Luca et al. (2025) Inzlicht & Berkman (2015)
Individual Differences, Development, and Social Context
Self-control as a trait is related to multiple life outcomes, but these relationships are compatible with stable habits, social opportunities, personal beliefs, learned routines, and differences in goal conflict exposure. Tangney, Baumeister, and Boone reported associations between their self-control scale and educational, interpersonal, and adjustment measures. De Ridder and colleagues integrated evidence across numerous behaviors. The positive associations are important, while causal interpretation remains limited by measurement overlap and confounding. Individuals can display different levels of regulatory success in eating, studying, social interaction, and emotional expression because the relevant cues and skills differ. Tangney et al. (2004) de Ridder et al. (2012)
Developmental work offers an additional caution. Delay-of-gratification experiments can reveal how attention and future rewards are represented, while later outcomes depend on family resources, cognitive development, and environmental reliability. The 2018 conceptual replication by Watts and colleagues reduced long-term associations when early context and ability were statistically adjusted. Cross-cultural comparisons should not presume that the same behavioral response has identical social meaning or goal relevance across populations. In settings where immediate reward may be uncertain or access to future reward is insecure, accepting a current opportunity can be reasonable. Culturally and economically informed designs should define the participant's available choices and observed constraints. Mischel et al. (1989) Watts et al. (2018)
Applied Implications: Studying, Health Habits, and Work
The applied implication best supported by converging research is that regulation can be improved by changing the structure of situations, routines, and choices rather than relying exclusively on willful suppression at a moment of maximum temptation. Gollwitzer's implementation-intention work suggests specifying a goal-consistent response to a foreseeable cue. Duckworth and colleagues describe controlling exposure to tempting environments; Galla and Duckworth provide evidence linking beneficial habits to favorable outcomes. These approaches act on the conditions under which a conflict occurs. Their usefulness should still be evaluated by domain, participant, actual behavior, maintenance interval, and equity of access to resources. Gollwitzer (1999) Duckworth et al. (2016) Galla & Duckworth (2015)
In education, prospective scheduling and removing avoidable distractions may help because students repeatedly encounter similar choice points. In health-related behavior, environmental availability, social support, and personally valued aims can influence whether an impulse competes with a commitment. In employment, sustained effort depends on workload, autonomy, incentives, fatigue, and recovery opportunities. The ego-depletion debate provides no warrant for a universal assertion that every decision consumes a measurable fixed quantity of willpower. Interventions should target observed mechanisms and measure real outcomes, with attention to adverse effects of excessive constraint or rigid persistence. Inzlicht et al. (2014) Berkman et al. (2017) Milyavskaya & Inzlicht (2017)
The Artificial Era and Designed Choice Environments
Artificial systems increasingly organize human exposure to options, cues, reminders, and default actions. Recommender systems, digital assistants, and algorithmically ordered interfaces may change the frequency and intensity of desire–goal conflicts without altering an individual's underlying executive capacity. An application designed to present highly salient, immediately rewarding alternatives can create repeated choice points; an application that supports scheduling or reduces unnecessary interruptions can change the same environment in the opposite direction. These are design-level hypotheses derived from documented relationships among cue exposure, value, habit, and control. Their specific magnitudes require direct intervention studies rather than extrapolation from ego-depletion estimates. Wood & Neal (2007) Duckworth et al. (2016) Berkman et al. (2017)
In the intellectual horizon of the Artificial Era, the psychological object of interest includes how human and artificial agents jointly arrange opportunities, costs, and goal representations. A digital assistant might externalize reminders and precommitments, but its benefit depends on accuracy, user autonomy, and the actual allocation of responsibility. Anthropomorphic presentation is unnecessary to study these outcomes. Researchers can assess whether a system changes the number of temptation encounters, the latency of goal-consistent action, perceived autonomy, and the durability of acquired routines. This is a program of testable human–artificial interaction questions, not evidence that artificial systems possess human desires or experience depletion. Gollwitzer (1999) Shenhav et al. (2013)
Scientific Ethics and the Limits of Normative Interpretation
Self-control research frequently encodes an implicit judgment about which behavior is desirable. Abstaining from consumption, maximizing task persistence, following instructions, and delaying reward can serve some goals while conflicting with others. Scientific analysis should identify the source of the standard and measure participants' priorities. In a restrictive or unsafe environment, overriding an externally imposed demand may be adaptive. Consequently, high self-control cannot be equated with unqualified obedience, and low self-control cannot be inferred from choosing immediate relief without examining the costs of alternatives. Fujita (2011) Berkman et al. (2017)
Applications should also avoid converting unstable estimates of depletion into moral judgments about effort or responsibility. Describing a participant as resource-depleted may obscure genuine fatigue, competing obligations, limited access to supportive contexts, or a rational decision to allocate effort elsewhere. The appropriate intervention can be a change in context or in the goal itself. Ethical research connects behavioral measures with autonomy, diverse values, and transparent analysis of risk. It is especially important to distinguish descriptive evidence about average task effects from normative conclusions about character, worth, treatment, or educational policy. Watts et al. (2018) Milyavskaya & Inzlicht (2017)
Conflicting Findings and Evidential Hierarchies
The scientific record offers unusually clear evidence of why self-control must be studied at several explanatory levels. Experience-sampling research confirms that people repeatedly experience wants, sometimes report conflicts with broader commitments, and differ in attempts to resist. Trait research confirms that questionnaire self-control has broad predictive associations. Laboratory research establishes that initial tasks can sometimes change later task behavior. These observations do not collectively entail the existence of one exhaustible strength. The inference to a common resource demands independent evidence of a conserved, transferable entity and predictable depletion and recovery dynamics. The strongest preregistered tests have provided particularly weak support for large generalized aftereffects. Hofmann et al. (2012a) de Ridder et al. (2012) Hagger et al. (2016) Vohs et al. (2021)
An evidence hierarchy for this subject must fit each question rather than rank entire research genres indiscriminately. Large, preregistered, randomized multicenter experiments are particularly probative for average sequential-task effects. Intensive longitudinal studies are better suited to observing real-world fluctuations in desire and conflict. Independent predictive tests of formal models are necessary for claims about how cost and control are computed. Studies using interventions and follow-up outcomes are needed for applied claims. A null experimental result cannot erase observational experience, and a large observational association cannot confirm a causally specific neurocognitive resource model. The appropriate hierarchy is question-sensitive and construct-sensitive. Shenhav et al. (2013) Milyavskaya & Inzlicht (2017) De Luca et al. (2025)
Causal Ambiguities in the Experience of Depletion
Subjective tiredness after resisting temptation may follow the amount of effort expended, the frequency of conflict, perceived lack of progress, or the anticipation of further demands. Milyavskaya and Inzlicht's 2017 work used experience sampling, diary, and prospective measures to examine temptation, experienced depletion, and goal achievement. Their reported pattern placed temptation exposure and feelings of depletion in important predictive roles, while effortful self-control showed little consistent relationship with goal achievement in their analyses. This finding challenges the idea that successful regulation is primarily an accumulation of strenuous acts of resistance. It does not establish a single causal chain because several measures remain observational. Milyavskaya & Inzlicht (2017)
A rigorous mechanism study would randomize an environmental source of conflict, hold task demands comparable, track perceived depletion before and after exposure, and observe goal outcomes later. It would assess whether the manipulation changed the intensity of desire, the number of conflict episodes, actual effort, or the interpretation of fatigue. If participants report feeling depleted without a corresponding deficit in incentivized performance, the data would support an account of subjective cost distinct from inability. If performance declines independently of choice incentives and expectancy, a capacity-based account would retain explanatory value. The hypotheses can coexist within a broader system when they predict different patterns across conditions. Inzlicht et al. (2014) Westbrook et al. (2013) Berkman et al. (2017)
An Integrated Account of Regulatory Organization
The available evidence supports organizing self-control as a sequence of interrelated functions. Environmental cues and physiological or learned states make particular desires available. Personal commitments and expectations determine whether a desire conflicts with another goal. Conflict can activate a higher-order goal, evoke negative affect, prompt active resistance, or motivate a change in circumstances. Subsequent behavior emerges from perceived values, attention, available skills, habitual responses, constraints, and opportunity costs. Repeated experience can reshape cue–response associations and the anticipated burden of future regulation. This organization is a synthesis of explanatory components documented across existing models, offered as a comparative map rather than as a newly validated theory. Kotabe & Hofmann (2015) Wood & Neal (2007) Berkman et al. (2017)
The integrated account also clarifies levels of causation. Situational redesign affects which options are encountered. Cognitive appraisal affects the perceived value of an option. Working-memory and inhibitory operations affect the processing of competing responses. Habit affects the automaticity of behavior. Fatigue and motivational state may alter the allocation of attention and effort over time. Social expectations and access to resources determine which options are realistically attainable. The same observable self-control success can therefore have multiple routes, and the same observed failure can arise from distinct combinations of mechanisms. Research must identify which pathway is operative before choosing a targeted intervention. Hofmann et al. (2012b) Duckworth et al. (2016) Shenhav et al. (2013)
Unresolved Scientific Questions
The first unresolved question is the degree of cross-domain generality in sequential-task aftereffects. Future studies should identify which exact combinations of first and second tasks yield reproducible effects and whether shared operations explain transfer. The second concerns the time course: if a resource is consumed, performance should change according to a measurable exposure–response function. The third concerns heterogeneity: preregistered studies must distinguish stable effects of individual differences from differences in tasks, sites, or incentives. The small positive estimate in one 2021 collaboration and near-zero estimates in others motivate precisely these comparisons. Dang et al. (2021) Vohs et al. (2021) De Luca et al. (2025)
Further questions concern the role of desire frequency versus desire intensity, how goal accessibility changes apparent conflict, when proactive environmental strategies outperform reactive inhibition, and whether self-control inventories measure the same underlying constructs across cultures and age groups. Designs should include ethically meaningful goal definitions, repeated measures, reliable behavioral outcomes, and cross-validated predictions. Rather than seeking a universal amount of willpower, a productive research agenda can estimate the conditions under which a specific behavior becomes likely and the interventions that shift it. Strong evidence would arise from convergent observational, experimental, longitudinal, and computational predictions that all refer to the same operational constructs. Hofmann et al. (2012a) Galla & Duckworth (2015) Watts et al. (2018)
Research Priorities and Falsifiable Predictions
The most discriminating next experiments should preregister contrasting quantitative predictions. A strict consumable-resource model predicts exposure-dependent cross-task losses that remain when shifts in interest, value, and expectancy are accounted for. A motivational-priority model predicts that task costs and available alternative rewards modulate continued engagement and its performance consequences. A value-integration model predicts that changing the salience of delayed consequences or of action costs changes choice in measurable directions. A habit-based account predicts that context cues exert increasing control over repeated behavior even when momentary intention is experimentally manipulated. These predictions require validated procedures, adequately powered designs, and reports of effect sizes and uncertainty. Inzlicht et al. (2014) Berkman et al. (2017) Wood & Neal (2007)
The second priority is convergence between laboratory and everyday measures. Individuals completing cognitive-control tasks could also provide intensive reports of desire conflict and objective measures of goal-directed behavior. Researchers should distinguish whether laboratory performance predicts conflict frequency, resistance frequency, enactment, or longer-term attainment. A small laboratory effect may matter little in daily life if its relevant context is rare; a small change in exposure to temptation may accumulate across frequent daily episodes. This is an empirical question about frequency, persistence, and consequence, not a reason to extrapolate standardized laboratory effects directly into population-level health or productivity claims. Hofmann et al. (2012a) Milyavskaya & Inzlicht (2017)
Conclusion
Self-control is best investigated as the organization of competing motivations, attention, choice, and action under changing personal and environmental conditions. Everyday research shows that desires are common and variably conflictual; trait research reveals broad predictive associations; experimental work identifies multiple means of arranging behavior before temptation becomes acute. These bodies of evidence support a broad science of regulatory organization. A singular inference from felt effort to consumption of a finite general resource is considerably less secure. The historically influential strength model advanced a distinctive empirical prediction, and large preregistered tests have not established the substantial general ego-depletion effect initially anticipated. A smaller positive collaboration and continuing theoretical debate keep narrowly specified sequential-task mechanisms open to direct study. Hofmann et al. (2012a) Galla & Duckworth (2015) Hagger et al. (2016) Dang et al. (2021) Vohs et al. (2021)
The strongest scientific synthesis is therefore conditional and constructive. Researchers can investigate desire and conflict as separate states, identify whether regulation takes the form of resistance or preventive design, measure effort and fatigue independently, and compare mechanistic models under explicit causal contrasts. Sustainable regulatory success frequently depends on opportunity structures, useful habits, accessible goals, and the ability to revise action as circumstances change. Future work in the Artificial Era should extend this precision to digitally organized choice environments and human–artificial coordination, preserving empirical accountability for every proposed mechanism. The enduring question concerns the processes that transform competing desires into behavior; explaining that transformation requires experimentally discriminable models and a close fit between evidence and claim. Fujita (2011) Kotabe & Hofmann (2015) Duckworth et al. (2016) Shenhav et al. (2013)
References
- Baumeister, R. F., Bratslavsky, E., Muraven, M., & Tice, D. M. (1998). Ego depletion: Is the active self a limited resource? Journal of Personality and Social Psychology, 74(5), 1252–1265. https://doi.org/10.1037/0022-3514.74.5.1252
- Berkman, E. T., Hutcherson, C. A., Livingston, J. L., Kahn, L. E., & Inzlicht, M. (2017). Self-control as value-based choice. Current Directions in Psychological Science, 26(5), 422–428. https://doi.org/10.1177/0963721417704394
- Carter, E. C., & McCullough, M. E. (2014). Publication bias and the limited strength model of self-control: Has the evidence for ego depletion been overestimated? Frontiers in Psychology, 5, Article 823. https://doi.org/10.3389/fpsyg.2014.00823
- Carter, E. C., Kofler, L. M., Forster, D. E., & McCullough, M. E. (2015). A series of meta-analytic tests of the depletion effect: Self-control does not seem to rely on a limited resource. Journal of Experimental Psychology: General, 144(4), 796–815. https://doi.org/10.1037/xge0000083
- Dang, J. (2018). An updated meta-analysis of the ego depletion effect. Psychological Research, 82(4), 645–651. https://doi.org/10.1007/s00426-017-0862-x
- Dang, J., Barker, P., Baumert, A., Bentvelzen, M., Berkman, E. T., Buchholz, N., Buczny, J., Chen, Z., De Cristofaro, V., de Vries, L., Dewitte, S., Giacomantonio, M., Gong, R., Homan, M., Imhoff, R., Ismail, I., Jia, L., Kubiak, T., Lange, F., . . . Zinkernagel, A. (2021). A multilab replication of the ego depletion effect. Social Psychological and Personality Science, 12(1), 14–24. https://doi.org/10.1177/1948550619887702
- De Luca, A., Zhang, C., Verschoor, S., & Hommel, B. (2025). Ego does not deplete over time. Experimental Psychology, 72(2), 100–113. https://doi.org/10.1027/1618-3169/a000644
- de Ridder, D. T. D., Lensvelt-Mulders, G., Finkenauer, C., Stok, F. M., & Baumeister, R. F. (2012). Taking stock of self-control: A meta-analysis of how trait self-control relates to a wide range of behaviors. Personality and Social Psychology Review, 16(1), 76–99. https://doi.org/10.1177/1088868311418749
- Duckworth, A. L., Gendler, T. S., & Gross, J. J. (2016). Situational strategies for self-control. Perspectives on Psychological Science, 11(1), 35–55. https://doi.org/10.1177/1745691615623247
- Fishbach, A., Friedman, R. S., & Kruglanski, A. W. (2003). Leading us not unto temptation: Momentary allurements elicit overriding goal activation. Journal of Personality and Social Psychology, 84(2), 296–309. https://doi.org/10.1037/0022-3514.84.2.296
- Friese, M., Loschelder, D. D., Gieseler, K., Frankenbach, J., & Inzlicht, M. (2019). Is ego depletion real? An analysis of arguments. Personality and Social Psychology Review, 23(2), 107–131. https://doi.org/10.1177/1088868318762183
- Fujita, K. (2011). On conceptualizing self-control as more than the effortful inhibition of impulses. Personality and Social Psychology Review, 15(4), 352–366. https://doi.org/10.1177/1088868311411165
- Galla, B. M., & Duckworth, A. L. (2015). More than resisting temptation: Beneficial habits mediate the relationship between self-control and positive life outcomes. Journal of Personality and Social Psychology, 109(3), 508–525. https://doi.org/10.1037/pspp0000026
- Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54(7), 493–503. https://doi.org/10.1037/0003-066X.54.7.493
- Hagger, M. S., Chatzisarantis, N. L. D., Alberts, H., Anggono, C. O., Batailler, C., Birt, A. R., Brand, R., Brandt, M. J., Brewer, G., Bruyneel, S., Calvillo, D. P., Campbell, W. K., Cannon, P. R., Carlucci, M., Carruth, N. P., Cheung, T., Crowell, A., De Ridder, D. T. D., Dewitte, S., . . . Zwienenberg, M. (2016). A multilab preregistered replication of the ego-depletion effect. Perspectives on Psychological Science, 11(4), 546–573. https://doi.org/10.1177/1745691616652873
- Hagger, M. S., Wood, C., Stiff, C., & Chatzisarantis, N. L. D. (2010). Ego depletion and the strength model of self-control: A meta-analysis. Psychological Bulletin, 136(4), 495–525. https://doi.org/10.1037/a0019486
- Hofmann, W., Baumeister, R. F., Förster, G., & Vohs, K. D. (2012). Everyday temptations: An experience sampling study of desire, conflict, and self-control. Journal of Personality and Social Psychology, 102(6), 1318–1335. https://doi.org/10.1037/a0026545
- Hofmann, W., Schmeichel, B. J., & Baddeley, A. D. (2012). Executive functions and self-regulation. Trends in Cognitive Sciences, 16(3), 174–180. https://doi.org/10.1016/j.tics.2012.01.006
- Hofmann, W., Vohs, K. D., & Baumeister, R. F. (2012). What people desire, feel conflicted about, and try to resist in everyday life. Psychological Science, 23(6), 582–588. https://doi.org/10.1177/0956797612437426
- Inzlicht, M., & Berkman, E. T. (2015). Six questions for the resource model of control (and some answers). Social and Personality Psychology Compass, 9(10), 511–524. https://doi.org/10.1111/spc3.12200
- Inzlicht, M., Schmeichel, B. J., & Macrae, C. N. (2014). Why self-control seems (but may not be) limited. Trends in Cognitive Sciences, 18(3), 127–133. https://doi.org/10.1016/j.tics.2013.12.009
- Job, V., Dweck, C. S., & Walton, G. M. (2010). Ego depletion—Is it all in your head? Implicit theories about willpower affect self-regulation. Psychological Science, 21(11), 1686–1693. https://doi.org/10.1177/0956797610384745
- Kotabe, H. P., & Hofmann, W. (2015). On integrating the components of self-control. Perspectives on Psychological Science, 10(5), 618–638. https://doi.org/10.1177/1745691615593382
- Milyavskaya, M., & Inzlicht, M. (2017). What’s so great about self-control? Examining the importance of effortful self-control and temptation in predicting real-life depletion and goal attainment. Social Psychological and Personality Science, 8(6), 603–611. https://doi.org/10.1177/1948550616679237
- Mischel, W., Ebbesen, E. B., & Zeiss, A. R. (1972). Cognitive and attentional mechanisms in delay of gratification. Journal of Personality and Social Psychology, 21(2), 204–218. https://doi.org/10.1037/h0032198
- Mischel, W., Shoda, Y., & Rodriguez, M. I. (1989). Delay of gratification in children. Science, 244(4907), 933–938. https://doi.org/10.1126/science.2658056
- Muraven, M., Tice, D. M., & Baumeister, R. F. (1998). Self-control as limited resource: Regulatory depletion patterns. Journal of Personality and Social Psychology, 74(3), 774–789. https://doi.org/10.1037/0022-3514.74.3.774
- 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. https://doi.org/10.1016/j.neuron.2013.07.007
- Tangney, J. P., Baumeister, R. F., & Boone, A. L. (2004). High self-control predicts good adjustment, less pathology, better grades, and interpersonal success. Journal of Personality, 72(2), 271–324. https://doi.org/10.1111/j.0022-3506.2004.00263.x
- Vohs, K. D., Schmeichel, B. J., Lohmann, S., Gronau, Q. F., Finley, A. J., Ainsworth, S. E., Alquist, J. L., Baker, M. D., Brizi, A., Bunyi, A., Butschek, G., Campbell, C., Capaldi, J., Cau, C., Chambers, H., Chatzisarantis, N. L. D., Christensen, W. J., Clay, S. L., Curtis, J., . . . Albarracín, D. (2021). A multisite preregistered paradigmatic test of the ego-depletion effect. Psychological Science, 32(10), 1566–1581. https://doi.org/10.1177/0956797621989733
- Watts, T. W., Duncan, G. J., & Quan, H. (2018). Revisiting the marshmallow test: A conceptual replication investigating links between early delay of gratification and later outcomes. Psychological Science, 29(7), 1159–1177. https://doi.org/10.1177/0956797618761661
- Westbrook, A., Kester, D., & Braver, T. S. (2013). What is the subjective cost of cognitive effort? Load, trait, and aging effects revealed by economic preference. PLOS ONE, 8(7), e68210. https://doi.org/10.1371/journal.pone.0068210
- Wood, W., & Neal, D. T. (2007). A new look at habits and the habit–goal interface. Psychological Review, 114(4), 843–863. https://doi.org/10.1037/0033-295X.114.4.843