Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Craving, Cue Reactivity, and Incentive Sensitization in Addiction

Abstract. Craving, cue reactivity, and incentive sensitization explain interconnected but distinct aspects of addiction-related motivation. This fundamental research monograph examines how substance-associated stimuli acquire predictive and motivational significance, how drug history and context influence craving and approach, and how consciously reported desire relates to subsequent use. A critical narrative synthesis integrates historical cognitive and conditioning models, incentive-sensitization theory, animal reinstatement and incubation experiments, laboratory and ecological momentary assessment, prospective meta-analyses, clinical neuroimaging, and intervention research. Evidence establishes cue-elicited responses across multiple methods and a prospective association of cue and craving indicators with subsequent drug use, while significant uncertainty remains about causal mediation, individual prediction, and cross-species translation. The work evaluates substantive alternatives based on habits, negative reinforcement, and cognitive control, identifies methodological confounds, and specifies testable questions for longitudinal and treatment research. It is a source-based narrative scientific synthesis, not a newly conducted experiment, registered systematic review, independent meta-analysis, or clinical treatment guideline.

Keywords: drug craving; cue reactivity; incentive sensitization; incentive salience; wanting and liking; addiction psychology; substance use disorders; associative learning; conditioned cues; Pavlovian-to-instrumental interactions; attentional bias; mesolimbic dopamine; cue-induced craving; incubation of craving; ecological momentary assessment; relapse prediction; reward motivation; extinction learning; cognitive regulation; cognitive bias modification; virtual reality interventions; Artificial Era

The Research Object: Craving as Situated Motivational Organization

Craving in addiction denotes a subjective episode of desire, urge, or compelling motivation directed toward using a substance, often accompanied by attention to associated stimuli, anticipated consequences, and changes in bodily and emotional state. It varies in duration, intensity, object, and context. A report that a person wants a drug is a useful measurement of an experience, but it is not itself a measurement of dopamine sensitization, automatic drug seeking, or imminent consumption. The scientific object of this dossier is the system of relationships among craving, cue-elicited responding, incentive salience, patterns of exposure, and subsequent drug-related behavior. The central question is how learned environmental information becomes a momentary source of motivation and under what conditions that motivation produces action. Sayette (2016) Vafaie & Kober (2022)

Cue reactivity describes measurable responses elicited by substance-associated signals relative to a defined comparison condition. Responses may involve self-reported urge, attentional allocation, physiological change, overt approach, decision-making, or activity within distributed brain networks. These outcomes can covary without being interchangeable. Incentive sensitization is a theoretical account of persistent amplification in systems that confer motivational attraction upon drug-related rewards and their cues in susceptible individuals. The theory proposes that drug-associated cues can acquire excessive ability to elicit wanting despite declining pleasure from actual use. A complete explanation must therefore distinguish the cue as informational predictor, the cue as motivational attractor, a conscious appraisal of desire, and the opportunity to undertake a learned action. Robinson & Berridge (1993) Berridge & Robinson (2016) Robinson & Berridge (2025)

This monograph centers on substance-related addiction rather than treating every frequent behavior as evidence of an addictive disorder. Its scope encompasses human laboratory studies, naturalistic and prospective observations, clinical findings, and animal models of conditioned approach, reinstatement, and incubation. The methods are combined through a critical narrative synthesis of identifiable primary studies and reviews; no new data, systematic search protocol, independent meta-analysis, or clinical treatment trial is represented as having been performed. Evidence for a particular construct is assessed at the level of the actual operation used to produce it. The analysis follows the causal sequence from learning and motivation through measurement and predictive validity to treatment implications and translational limits. Shaham et al. (2003) Sayette (2016) Bergeria et al. (2024)

Terminology, Levels of Explanation, and Testable Claims

An urge is a consciously reportable propensity to engage in drug use; its rating depends on the question asked, response scale, anticipation of opportunity, and immediate goals. Cue-induced craving refers specifically to the difference in reported urge associated with exposure to drug-related stimuli under a specified experimental or naturalistic comparison. Cue reactivity is broader and may include physiological and neural signals even when a participant reports little urge. Incentive salience refers to motivational properties attributed to a predictive stimulus, which can be inferred in some animal paradigms from approach, conditioned reinforcement, or the potentiation of instrumental behavior. These definitions imply separate measurement models and guard against interpreting a single questionnaire item as an assay of all relevant motivational mechanisms. Carter & Tiffany (1999) Sayette (2016) Robinson et al. (2014)

Learning and sensitization designate different operations. Associative learning establishes relations among events, contexts, actions, and outcomes; sensitization denotes enduring changes in responsiveness after relevant histories of exposure. A cue must ordinarily acquire predictive significance, yet the motivational impact of an already learned cue may fluctuate with deprivation, stress, drug history, opportunity, and internal state. Incentive sensitization explains a proposed susceptibility-dependent change in the strength of cue-elicited wanting, while memory, expectancy, and executive control govern other components of the same episode. The existence of conditioning alone therefore does not demonstrate sensitization, and higher cue-related brain activity on a single scan does not establish a lasting neural adaptation. Robinson & Berridge (2008) Robinson & Berridge (2025) Fascher et al. (2026)

The explanatory levels are linked by a chain of hypotheses: repeated exposure can alter reward-related circuitry; learned cues can recruit that circuitry; the resulting state can orient attention and action; the presence of alternatives and constraints can change what is ultimately done. Each arrow needs its own evidence. Neuroimaging identifies correlates of task responses under specified contrasts; pharmacological and circuit manipulations can inform causal mechanism, especially in animals; prospective studies evaluate whether earlier measures predict later behavior. Treatment studies are needed to determine whether changing a candidate process improves clinical outcomes. The ambition of an integrated account is to make these transitions testable while retaining the distinct meanings of desire, neural activation, and drug consumption. Volkow et al. (2006) Vafaie & Kober (2022) Evohr et al. (2025)

Historical Development: Conditioned Responses, Automaticity, and Incentive Motivation

Earlier theories of addiction emphasized pharmacological reward, tolerance, withdrawal, and conditioning. Clinical observation made it evident that familiar surroundings and drug-related objects could evoke intense reactions even when the drug was absent. Experimental cue paradigms formalized this observation by exposing participants to substance-related sights, smells, paraphernalia, imagery, or contexts while recording subjective and physiological changes. The resulting program brought associative learning into contact with clinical relapse questions. Carter and Tiffany's 1999 meta-analysis synthesized 41 cue-reactivity experiments and documented more consistent effects for reported craving than for some autonomic measures. Their findings established the experimental phenomenon while also indicating that different response systems should be analyzed separately. Carter & Tiffany (1999)

Tiffany's 1990 cognitive model challenged a simple sequence in which felt urge must precede every episode of drug use. The model treated well-practiced consumption routines as potentially controlled by automatized action schemata, whereas conscious urges could arise when automatic execution was supported, blocked, or contested. This interpretation remains useful for understanding why an individual may report substantial craving without using, or perform a habitual behavior without a prominent preceding experience of urge. It also makes predictions about task demands and conflict: craving reports can change when anticipated access, inhibitory effort, or the need to articulate desire changes. The alternative cognitive account expanded the field beyond an exclusively withdrawal-based picture of motivation. Tiffany (1990) Sayette (2016)

Robinson and Berridge's original 1993 incentive-sensitization formulation introduced a long-term motivational account grounded in neural plasticity and the distinction between wanting and pleasure. Their 2008 reformulation examined how learning, context, and human addiction evidence interact with sensitization; their later work separated cue-triggered incentive motivation from hedonic liking and considered heterogeneity of vulnerability. The 2025 thirty-year review broadened the comparison to habit, negative reinforcement, and prefrontal-control accounts. This history should be understood as the evolution of competing and partly complementary explanations, rather than a single settled trajectory from conditioning to a universally confirmed mechanism. Robinson & Berridge (1993) Robinson & Berridge (2008) Berridge & Robinson (2016) Robinson & Berridge (2025)

The Incentive-Sensitization Hypothesis and Its Scientific Burden

The theory's organizing proposition is that repeated exposure to certain drugs may leave susceptible reward-related systems especially responsive to drug-associated cues and opportunities. This change increases the capacity of cues to recruit momentary incentive motivation and can outlast acute intoxication, withdrawal, and the positive enjoyment of taking the substance. The proposal is powerful because it makes sense of continued seeking in circumstances where the person expects diminishing pleasure or substantial costs. Its central empirical burden is more demanding: investigators must establish enduring differential motivation attributable to sensitization rather than to baseline preference, prior learning, generalized arousal, or changes in expected reward. Robinson & Berridge (1993) Robinson & Berridge (2025)

The distinction between wanting and liking is a distinction among experimentally investigated processes. Hedonic reactions index the pleasurable effect of reward consumption, whereas incentive motivation indexes attraction, pursuit, and willingness to act. Mesolimbic dopamine is implicated prominently in the latter in the theory's framework, although dopamine function also participates in learning, movement, and multiple forms of behavioral adaptation. A finding that a drug cue recruits striatal or mesolimbic activity does not uniquely establish pathological wanting; a finding of strong desire does not prove the absence of liking. Interpretation requires designs that separately assess value, behavior, subjective ratings, and exposure history. Berridge & Robinson (2016) Robinson & Berridge (2025) Volkow et al. (2006)

The model also generates specific predictions about state and circumstance. The same cue should acquire different motivational power depending on the individual's history, current physiological state, and the opportunity to act. Its effect may be amplified in a drug-associated environment and moderated by attentional or executive demands. People may differ before drug exposure in their tendency to attribute incentive value to predictors, an issue illuminated by sign-tracking research. These predictions make context dependence and between-person diversity central rather than inconvenient residuals. At the same time, a clinically meaningful theory must predict behavior and outcome at the individual level, which requires longitudinal designs and intervention tests beyond retrospective consistency with the model. Robinson et al. (2014) Flagel et al. (2011) Robinson & Berridge (2025)

Alternative Accounts: Habits, Withdrawal Relief, and Control

Habit theories describe increasingly stimulus-driven action selection, with some forms of drug seeking becoming less sensitive to changes in the expected value of the outcome. Everitt and Robbins reviewed evidence for transitions among goal-directed, habitual, and compulsive patterns, emphasizing striatal and prefrontal contributions. Habitual execution and incentive attraction can coexist: a salient cue may trigger a learned response, motivate pursuit of a valued outcome, or do both. Experimental outcome devaluation and contingency degradation identify particular aspects of action control, whereas self-reported wanting addresses a distinct level of behavior. A comparative theory must evaluate whether an apparent urge–action connection is mediated by action habits, reward expectations, or both. Everitt & Robbins (2016) Tiffany (1990)

Negative reinforcement accounts place emphasis on using substances to reduce distress, aversive withdrawal states, or stress. Such processes predict urgency especially when relief from negative states is expected. Stress and craving often covary in everyday reports, and the two can jointly influence the perceived value of an available drug. The coexistence of positive cue attraction and anticipated symptom relief means that an episode can involve several motivational mechanisms at once. Cross-sectional correlations cannot apportion their causal contributions; designs must vary cues, stressors, anticipated availability, and outcomes rather than classifying all urge as a single mechanism. Preston & Epstein (2011) Sinha (2011) Robinson & Berridge (2025)

Executive-control models examine inhibition, working memory, conflict monitoring, future-oriented appraisal, and regulation of emotion or craving. Experimental reappraisal can alter subjective desire and engagement of prefrontal–striatal networks; this is evidence that urge expression can be regulated and is not a fixed direct readout of cue strength. The full account therefore requires interaction among motivational attribution, learned action, affective state, availability, and control. Competing theories are most productively tested by predictions that differ across these dimensions, and by independent replication of those predictions in clinical populations rather than by assigning every observed finding to a preferred neural system. Kober et al. (2010) Everitt & Robbins (2016) Sayette (2016)

Pavlovian Prediction, Instrumental Action, and Contextual Control

A substance-associated cue begins with a history of correlations: a bottle, dealer location, cigarette smell, phone notification, or internal sensation has repeatedly preceded substance availability or a familiar action sequence. Pavlovian learning can give such a stimulus predictive significance even when no drug is present. Instrumental learning separately establishes relationships among actions, outcomes, and contingent consequences. Their interaction supplies a route from recognizing an opportunity to preparing or executing a response. An experimental cue effect therefore depends on what was learned, how strongly the stimulus discriminates availability, and whether behavior has been reinforced in a relevant context. Animal reinstatement procedures distinguish cue-, drug-, and stress-triggered resumption of seeking after extinction, thereby isolating some of these routes experimentally. Shaham et al. (2003) Everitt & Robbins (2016)

Conditioned stimuli can have multiple psychological functions. They may signal what outcome is expected, direct attention toward a reward opportunity, support a conditioned response, or become targets of approach in their own right. The latter function motivates the distinction between sign tracking and goal tracking. In sign tracking, an animal approaches the discrete cue predicting reward; in goal tracking, it approaches the place where reward is delivered. Both patterns can reflect learning, yet studies indicate differences in motivational attraction to specific stimuli and in associated dopamine signals. This work demonstrates that accurate prediction of reward and incentive attraction toward its predictor are not equivalent. It also cautions against generalizing a cue's effectiveness across all individuals and contexts. Flagel et al. (2011) Robinson et al. (2014)

Environmental contexts influence whether the meaning of a drug cue is expressed. A cue encountered in a laboratory, at work, during abstinent treatment, or in a formerly familiar place may generate different expectations and motivational effects. Contexts also alter which behavior is possible: someone can desire a substance strongly and choose a competing activity, while another person in an available-use setting may take it after little reflective deliberation. A useful research design manipulates or records place, time, cue density, social company, stress, access, and competing rewards. These are scientifically relevant variables rather than background noise to be removed indiscriminately. The immediate context can modify both cue-triggered wanting and the translation from wanting to action. Robinson & Berridge (2008) Sayette (2016) Vafaie & Kober (2022)

Incentive Salience, Attentional Capture, and Approach

Incentive salience is proposed to make a stimulus motivationally significant beyond its predictive content. Its effects can be operationalized through the capacity of cues to attract approach, serve as conditioned reinforcers, or invigorate reward-seeking behavior in appropriately controlled animal paradigms. In humans, approach–avoidance tasks, visual attention measures, subjective ratings, and effort-based choices provide related but imperfect proxies. A person can recognize a drug picture without being motivated by it; another may direct attention toward a drug cue yet judge the substance undesirable. Explicit awareness of a stimulus and motivational significance must be measured independently. The theory is strongest when its predictions are expressed as differences among these operations rather than inferred from a single increase in reaction time or brain activation. Robinson et al. (2014) Robinson & Berridge (2025)

Attentional capture can be both a consequence and a contributor to an episode of craving. Once a stimulus becomes salient, it may be detected quickly, held in working memory, and repeatedly retrieved; attention in turn affords additional opportunities to rehearse reward expectations. However, tasks that measure attentional bias differ in reliability, stimulus complexity, and sensitivity to current state. In the absence of sufficient controls, an attentional effect could index familiarity, visual features, emotional significance, or response strategy. The 2026 neural cue-reactivity meta-analysis documented systematic influences from stimulus matching and task design, including confounds associated with visually unmatched conditions. A mature mechanistic account should compare eye movements, behavioral choices, subjective urge, and carefully controlled stimulus properties rather than privileging any one index. Fascher et al. (2026) Sayette (2016)

Approach tendencies are especially useful because they connect cue valuation to potential action. A manipulation that shifts approach bias can establish that an experimentally defined response tendency is modifiable, while subsequent clinical outcomes test whether the change matters outside the task. Wiers and colleagues randomized 214 alcohol-dependent inpatients to experimental and control conditions in a brief approach-bias modification study and reported favorable subsequent outcomes in the trained groups, although the unadjusted intention-to-treat relapse comparison of 46% versus 59% at one year was marginal (p = .057). The result supports investigation of action tendencies as candidate intervention targets but does not establish that subjective craving, incentive sensitization, and relapse all share a single mechanism. Later work requires independent replication, reliable mediation analyses, and careful distinctions between task-specific learning and durable behavior change. Wiers et al. (2011)

Mesolimbic Dopamine and Distributed Neural Systems

Incentive-sensitization theory assigns a prominent role to mesolimbic dopamine pathways, including projections from the ventral tegmental area to striatal and related regions. Its mechanistic claim concerns the attribution and activation of incentive motivation, not a universal equation between dopamine and pleasure. Experiments with reward-predictive cues and individual differences in rats show that dopaminergic signaling can vary with the motivational role of a cue, even when predictive information is learned. The inference depends on species, task, temporal measurements, and manipulations. Changes after repeated drug exposure may interact with prior vulnerability, the drug class, and context, making a single scalar 'dopamine level' an inadequate theory of drug craving. Flagel et al. (2011) Berridge & Robinson (2016) Robinson & Berridge (2025)

Human neuroimaging complements the animal evidence. Childress and colleagues reported limbic activation during exposure to cocaine-related videos in individuals with cocaine dependence, providing early evidence that drug cues engage distributed motivational and emotional processing systems. Volkow and colleagues used positron emission tomography to investigate cocaine cues, dorsal-striatal dopamine changes, and craving, supporting an association between cue exposure and dopaminergic processes. Neither fMRI blood-oxygen responses nor regional PET estimates can independently identify all synaptic mechanisms mediating a conscious urge. Moreover, activation in the insula, cingulate, striatum, or prefrontal cortex has multiple potential functions across tasks; inference must attend to the complete contrast, population, and behavior. Childress et al. (1999) Volkow et al. (2006)

Amygdala pathways contribute to learning about emotionally and motivationally significant signals, while insular systems can participate in bodily-state representation and interoceptive awareness. Prefrontal and cingulate networks contribute to value comparison, monitoring, and regulation; striatal circuitry supports selection, vigor, and learning of action patterns. These systems operate as interacting circuits rather than isolated centers for one mental faculty. Janak and Tye's synthesis emphasized how distinct amygdala circuits can support different behaviors, and a later meta-analysis of clinical cue-reactivity imaging identified consistent but heterogeneous regional responses. Whole-network claims require the resolution of imaging method and anatomical interpretation to be matched to the question actually tested. Janak & Tye (2015) Evohr et al. (2025) Fascher et al. (2026)

The Brain Response as Predictor: What Neuroimaging Can Establish

A neural cue-reactivity signal becomes clinically useful only when it predicts a relevant outcome under a defined protocol and improves on practical alternatives. In a 2025 meta-analysis, Evohr and colleagues synthesized 51 functional neuroimaging cue-reactivity studies with 1,787 participants. Across selected treatment studies, insular and cingulate responses were associated with subsequent outcomes, and some regions showed posttreatment change. Such findings provide candidate biomarkers, but predictive association does not establish that changing the signal directly causes recovery. Imaging also poses problems of small analytic subgroups, heterogeneous endpoints, study selection, and differences in contrasts or recruitment. The study identified a smaller subset of 14 predictive-biomarker investigations involving 456 participants, underscoring the distance between a pooled neural map and a prospectively validated individual clinical decision tool. Evohr et al. (2025)

Fascher and colleagues' 2026 activation-likelihood meta-analysis included 92 fMRI studies with 3,647 participants and found that clinical characteristics and task design systematically moderated neural cue reactivity. In particular, visually mismatched drug and control stimuli created interpretable confounding signals. This matters because even a highly replicable activation peak can reflect differences in picture content rather than addiction-specific incentive processes. A stronger imaging paradigm preregisters contrasts, matches stimulus characteristics, reports acquisition and preprocessing decisions, and examines whether neural signals relate to behavior after accounting for the subjective craving response. Between-study consistency alone does not remove method-related bias. Fascher et al. (2026)

Neural responses also fluctuate with circumstances, abstinence history, current medication, stress, expectation, and treatment engagement. A scan is a measurement taken during one controlled period; it cannot summarize every episode of motivated behavior in a person's environment. Multi-session reliability testing, longitudinal follow-up, and out-of-sample validation would be necessary to establish a stable vulnerability signature. In the clinical domain, a useful predictive biomarker should distinguish the probability of outcome under specific treatment choices, while a response biomarker should reflect a meaningful change linked to improved functioning. The current evidence is promising for research stratification and mechanistic hypothesis testing, with clinical deployment requiring stronger validation. Evohr et al. (2025) Robinson & Berridge (2025)

Time, Sensitization, and the Paradox of Persistent Vulnerability

Incentive sensitization proposes an enduring susceptibility that is expressed episodically. A person may remain abstinent for an extended period and experience little average craving, yet encounter a cue that abruptly raises desire in a relevant setting. Distinguishing average levels of craving from heightened reactivity to a particular trigger resolves some apparent temporal contradictions. The underlying theory predicts that neural adaptations can persist while the overt expression of motivation remains dependent on cues, states, and opportunity. Thus, absence of frequent self-reported craving is compatible with occasional strong cue responses, but it cannot be treated as direct evidence of sensitization without independently demonstrating the implicated adaptation. Robinson & Berridge (2008) Robinson & Berridge (2025)

The duration of behavioral vulnerability cannot be inferred from a single laboratory exposure. Longitudinal studies should separately follow substance use, abstinence duration, remembered craving, cue-reactivity response, affective changes, and environmental exposure. If a person changes living circumstances or increases competing activities, the observed frequency of craving may decline despite an unchanged learned history. Conversely, frequent contact with a drug-associated social context may sustain reported desire without a progressive change in neural responsiveness. Only designs that measure these components repeatedly can identify whether change in cue impact is due to sensitization, new learning, altered access, or a state-dependent interaction. Sayette (2016) Vafaie & Kober (2022)

The clinical implication is that risk can be intermittent and situated. A drug-use episode is an outcome of interacting processes over time, whereas addiction-related vulnerability is an evolving distribution of possible responses to environments and internal states. This formulation encourages measurement of within-person variation instead of treating a group-average cue response as a fixed trait. It also underscores the scientific value of incorporating access, alternative rewards, and social support in prediction models, so that substance use is understood as behavior within an ecology of decisions rather than the automatic output of an isolated neural event. Epstein et al. (2010) Preston & Epstein (2011) Robinson & Berridge (2025)

Laboratory Cue Reactivity: Designs, Controls, and Response Systems

Controlled cue-reactivity experiments compare responses to substance-associated materials against a matched neutral or alternative condition. Typical methods present pictures, video, sounds, odors, actual paraphernalia, imagery, or environments previously associated with drug use. Investigators may ask for an immediate urge rating, collect skin conductance or heart-rate measures, monitor gaze, measure behavioral approach, or record neural activity. The control condition matters as much as the drug cue: if the drug image contains people, movement, emotional expression, or unusually salient objects absent from neutral images, a response contrast can reflect perceptual differences. Within-person counterbalancing and careful separation of trial phases improve interpretability. In their classic synthesis, Carter and Tiffany reviewed 41 cue-reactivity studies and found robust self-reported craving responses, alongside response-system heterogeneity that prohibits assigning the same meaning to every physiological signal. Carter & Tiffany (1999) Fascher et al. (2026)

Imaginal exposure and in vivo presentations are psychologically different procedures. Images invite recall and autobiographical elaboration; actual objects can supply multisensory information and expectations of immediate access. In a laboratory smoking-urge experiment, Drobes and Tiffany investigated the induction of craving through imagery and in vivo procedures and compared physiological and self-report manifestations. Such experiments illustrate how manipulating the mode of exposure changes the interpretation of the result. A comprehensive protocol should report substance-use history, abstinence interval, the instructions participants receive about possible consumption, the order of conditions, expectancy effects, ethical safeguards, and the presence of a postexposure recovery period. Claims about clinical relapse require an outcome measurement beyond an immediate laboratory urge. Drobes & Tiffany (1997) Sayette (2016)

Reactivity is best treated as a vector of outcomes rather than as one latent construct assumed to produce identical responses across methods. A study can detect increased urge without a reliable change in heart rate, or increased orienting without a strong consciously articulated desire. This discrepancy may be informative: awareness, strategic control, and physiological regulation can have distinct temporal profiles. Yet measurement error, low statistical power, inadequate controls, and highly variable baselines can also produce apparent dissociations. Multimethod designs should preregister their primary endpoint and test convergent and discriminant validity instead of post hoc selecting whichever signal increased most. A sufficiently powerful design also separates effects attributable to cue familiarity from those attributable to drug-related motivation. Carter & Tiffany (1999) Sayette (2016) Fascher et al. (2026)

Measuring Craving: Scales, Demand, and Construct Validity

Self-report instruments range from a single visual analog urge item to multidimensional questionnaires distinguishing desire, anticipated positive outcomes, negative reinforcement, and intention. A single question can be sensitive to a rapid within-person change; a multi-item scale may estimate a more stable or differentiated construct. The choice must match the theory's prediction. Asking 'How much do you want to use now?' is different from asking whether a participant expects to use later, believes use will relieve distress, or would spend resources to obtain a substance. Ratings are affected by language, scale anchors, social desirability, momentary opportunity, and expectations about how an experiment or treatment will unfold. These problems do not invalidate subjective experience as evidence; they require carefully defined target constructs and validated response processes. Sayette (2016) Tiffany (1990)

Temporal sampling is crucial. The same individual can experience a rapid peak of craving and then return to baseline between scheduled assessments. A daily retrospective rating cannot reveal the onset of a ten-minute urge, and a peak rating does not characterize a day of ordinary life. Repeated measures should distinguish tonic background craving, event-linked increases, and cumulative exposure to high-risk settings. Ecological momentary assessment is especially useful when researchers wish to observe the relation of urges to real-world settings. It allows random prompts and event-contingent records, but reporting itself can be influenced by compliance, participant awareness, and missed episodes. The operational question is whether one seeks to estimate an average state, the probability of an episode, or the slope of change surrounding drug use. Epstein et al. (2010) Preston & Epstein (2011)

Construct validity also requires distinguishing the intensity of an urge from the person's effort to resist it or the eventual decision to use. An increase in effortful control may accompany a particularly strong craving, yet a participant who successfully abstains can report high urge and low consumption. A treatment could reduce perceived interference or increase coping without immediately eliminating cue responsiveness. Accordingly, evaluations should retain separate measures of frequency, intensity, duration, distress, perceived control, actual drug use, and functional outcomes. This multidimensional approach is compatible with a mechanistic science that recognizes conscious reports as important data while testing whether they carry incremental predictive information over previous use, stress, exposure, and available alternatives. Tiffany (1990) Sayette (2016) Vafaie & Kober (2022)

Ecological Momentary Assessment and the Everyday Risk Process

Ecological momentary assessment, or EMA, uses repeated reports obtained close to everyday experiences. In addiction research, participants can record urges, drug use, mood, stress, locations, company, and exposure to substance-associated objects. The design reduces some distortions of long retrospective recall and permits analyses of within-person changes. Epstein and colleagues studied tobacco, cocaine, and heroin craving and use in daily life among methadone-maintained outpatients, showing the value of a design that distinguishes multiple substances and separates wanting from action. The causal interpretation remains limited by self-selection into settings and the fact that environmental cues, available drugs, stress, and social relationships are correlated. Nonetheless, EMA reveals the situational organization that a single laboratory session can miss. Epstein et al. (2010)

Stress is an especially important contextual covariate because it can influence emotional state, anticipated relief, attentional narrowing, and the value of drug use. In everyday-life research on cocaine and heroin users, Preston and Epstein assessed associations among stress, mood, craving, relapse triggers, and cocaine use. The work demonstrates why the temporal sequence matters: a stress report preceding craving is more informative for prediction than a single cross-sectional observation, but even close temporal precedence does not establish that stress causes a specific urge or subsequent use. Time-varying confounding is substantial; for example, the same social setting may increase stress, cue exposure, and drug access. Within-person modeling, timestamped events, and negative-control comparisons can narrow the range of plausible explanations. Preston & Epstein (2011)

Naturalistic studies introduce distinctive missing-data mechanisms. Participants may be least likely to respond during intense intoxication, interpersonal conflict, or rapid drug use, precisely when researchers most need data. Frequent prompts can themselves change attention to craving, while event-contingent entries may be delayed or selectively completed. Analyses should report adherence, prompt density, the proportion of events captured, handling of unequal measurement intervals, and sensitivity to missing-not-at-random episodes. A useful multi-level model distinguishes stable between-person differences from momentary deviations relative to each person's baseline and accounts for repeated observations nested within participants. Generalization requires checking whether samples drawn from treatment settings represent those who are not receiving care. Epstein et al. (2010) Preston & Epstein (2011) Sayette (2016)

Prospective Prediction, Effect Size, and Limits of Causal Inference

Vafaie and Kober's systematic review and meta-analysis pooled evidence from 237 prospective studies, representing 51,788 participants and 656 effect estimates, to evaluate whether cue responses and craving were associated with later drug use or relapse. The combined association favored higher subsequent use when drug cues or craving were more prominent, with a pooled odds ratio of 2.05 and a confidence interval of 1.94 to 2.15 in their principal analysis. The study also identified publication-bias concerns; a trim-and-fill adjustment yielded an odds ratio of 1.31 with a confidence interval of 1.25 to 1.38. These estimates pertain to defined study contrasts and pooled outcomes, rather than an automatically transferable prediction for an individual patient's probability of recurrence. The reduction under adjustment reinforces the need to examine selective publication and design heterogeneity. Vafaie & Kober (2022)

Prospective association is stronger evidence for predictive relevance than a retrospective description of episodes already completed. It still does not prove that craving independently causes use. A cue response can be a marker for a history of frequent use, environmental availability, stress sensitivity, or other factors that influence subsequent behavior. It can also be a partial mediator of a broader causal process. Causal claims require interventions that selectively modify the putative mechanism, demonstrate that the modification precedes behavior change, and account for alternative mediators. The key analytical question is whether craving or cue reactivity adds predictive accuracy beyond prior use, diagnosis, situational access, and treatment context. Such incremental prediction should be evaluated using external validation, calibration, and clinically meaningful decision thresholds. Vafaie & Kober (2022) Evohr et al. (2025)

The distinction among population-level association, calibrated risk prediction, and mechanism identification applies equally to behavioral tests and brain signals. A model may predict group-level differences even when its classification accuracy is too low for individual treatment decisions. Treatment selection additionally requires evidence that a patient with a given profile benefits more from one intervention than from another, a question that standard prognostic studies cannot resolve. Sinha's work on biological correlates of relapse vulnerability emphasizes the complexity of stress and motivational systems; recent imaging syntheses reveal analogous limits in clinical translation. Reliable prospective prediction therefore demands repeated outcome verification, appropriate comparison models, preregistered analysis plans, and evaluation across independent treatment settings. Sinha (2011) Evohr et al. (2025) Vafaie & Kober (2022)

Individual Differences, Substance Class, and Developmental Scope

Individuals exposed to similar cues can show divergent patterns of orienting, craving, approach, avoidance, and subsequent action. Preclinical research on sign tracking and goal tracking provides a mechanistic illustration: animals may learn the same reward contingency while differing in the attraction attributed to a predictive signal. Human variability involves prior dependence severity, history of reinforcement, cue exposure, psychiatric comorbidity, current affect, treatment participation, and personal goals. These variables can alter both the motivational meaning of a cue and the extent to which its effect is visible in a particular measurement. Research should evaluate heterogeneity explicitly instead of interpreting a significant average response as the mechanism for every participant. Flagel et al. (2011) Robinson et al. (2014)

Generalizing across substances is scientifically demanding because their pharmacology, patterns of intoxication and withdrawal, delivery cues, use settings, and treatment contexts differ. Tobacco cues can be frequent and woven into ordinary routines; alcohol cues may carry especially strong social meanings; opioid use may involve distinct withdrawal and treatment states; stimulant cues can be embedded in concentrated episodic seeking. A theory of incentive salience may have cross-substance reach without making every expected temporal trajectory identical. The 2024 review by Bergeria and colleagues reported important differences between animal incubation paradigms and human opioid craving observations. This is evidence for substance-specific model testing and careful separation of craving measured during treatment from experimentally controlled animal drug seeking. Bergeria et al. (2024) Robinson & Berridge (2025)

Age and development can influence how associations are acquired and how behavioral control changes, yet claims about adolescence require substance-specific longitudinal data and careful adjustment for exposure history, family environment, and social opportunities. Similarly, apparent differences between sexes or across cultural populations can depend on sampling, substance accessibility, expected disclosure, and measurement invariance. Current cue-reactivity literature contains substantial heterogeneity in participant selection and tasks, and neural meta-analytic findings show that clinical variables moderate responses. A research program should recruit diverse cohorts, assess whether instruments function similarly across groups, and examine which mechanisms retain predictive validity after transport into a new setting. Fascher et al. (2026) Sayette (2016)

Incubation of Drug Seeking: Models and Historical Evidence

Incubation of drug craving refers to an increase in cue-triggered drug seeking, or an associated motivational response, across an interval of abstinence under specified experimental conditions. Grimm and colleagues reported that rats tested after different periods of withdrawal from cocaine self-administration showed increasing cue-elicited seeking, providing a prominent early demonstration of a time-dependent effect. The phrase 'incubation of craving' applied to these procedures is partly inferential: a lever press or seeking response is observable, whereas the animal's conscious experience of desire is not directly measurable. The experimental result concerns behavior after defined drug-training and withdrawal histories. It generates hypotheses about persistent motivational adaptation but does not automatically establish the same psychological trajectory in human clinical samples. Grimm et al. (2001) Shaham et al. (2003)

Reinstatement paradigms typically train substance self-administration, extinguish seeking under new contingencies, and then test whether cues, drug priming, or stressors can recover the response. Incubation studies add variation in the duration of forced or voluntary abstinence and assess seeking after that interval. These designs provide strong experimental control over drug history, cues, schedules, and response opportunities. Their apparent simplicity should not conceal crucial details: response measures differ with whether reward is delivered during testing, whether extinction has occurred, and whether the subject remains in the original drug-associated environment. A substantial increase in responding may index learned expectation, motivational attraction, frustration, or multiple processes. Shaham and colleagues' historical and methodological review of reinstatement delineated these distinctions and their use in studying recurrence. Shaham et al. (2003)

The strongest mechanistic conclusions arise when a proposed neural pathway or receptor process is manipulated and a selective change in cue-induced seeking follows. Yet interventions applied to animals can affect locomotion, attention, stress, or reinforcement learning as well as the hypothesized incentive component. Drug class, administered dose, duration of exposure, route of use, housing, and withdrawal conditions can shape the result. Translation to a human episode of craving requires independent measures of subjective desire and prospective substance use, neither of which is supplied by a rodent lever-press measure alone. The central scientific accomplishment of incubation research is to make a theoretically important temporal phenomenon testable under controlled conditions. Grimm et al. (2001) Shaham et al. (2003) Bergeria et al. (2024)

The Human Opioid Data and the Translational Mismatch

The growth of cue-induced seeking during abstinence in certain animal models is an influential prediction, but the pattern has not translated uniformly to human clinical measures. Bergeria and colleagues systematically compared opioid-related animal incubation evidence with human research during inpatient and outpatient treatment. Their review found appreciable support for time-related increases in opioid seeking across many preclinical studies, while the human opioid craving studies reviewed did not demonstrate the same increase across abstinence or treatment time. This is an important negative finding for a simplistic generalization: laboratory-defined animal seeking, retrospective human urge, and prospective drug use are separate outcomes measured under substantially different constraints. Bergeria et al. (2024)

Treatment populations introduce environmental changes that matter. An inpatient setting restricts access to substances and alters routines; outpatient treatment may involve pharmacotherapy, social support, expectations, and variable exposure to cues. Self-report is affected by what participants anticipate about use opportunities and by how treatment changes their experience. Therefore, a decline or stable trajectory of reported craving under care does not by itself invalidate every neural sensitization hypothesis. Conversely, postulating unmeasured sensitization cannot substitute for showing a predicted human effect under a valid and sensitive clinical design. The translational task is to specify which human populations, cue types, treatment conditions, abstinence intervals, and behavioral outcomes would count as a meaningful counterpart to preclinical incubation. Bergeria et al. (2024) Robinson & Berridge (2025)

A productive bridge between models would include repeated, standardized cue challenges, within-person craving ratings, markers of actual exposure, and prospective records of use over time. Researchers should distinguish patients in enforced abstinence from individuals with naturalistic access, and should characterize ongoing medications, withdrawal symptoms, psychiatric burden, and social context. If an increase appears only when treatment is absent or when cues signal genuine availability, this would identify an interaction to be explained rather than a failure of measurement to be concealed. Independent replications with prespecified trajectories are required before claiming that incubation is a general clinical law. Bergeria et al. (2024) Vafaie & Kober (2022)

Explicit Craving, Behavior, and the Problem of Dissociation

An observed mismatch between reported craving and drug use should be interpreted using the time scale and response definition of each measure. A person may experience high desire, regard the substance as costly, and refrain from use; another may enter a familiar setting and perform a learned action before articulating an urge. Tiffany's cognitive model anticipated precisely this possibility by separating automatic behavior sequences from conscious processing associated with impediments or conflict. Incentive-sensitization theory similarly distinguishes cue-triggered motivational attraction from the hedonic experience of consumption. Neither account requires subjective desire to be necessary or sufficient for every drug-taking event. The empirical issue is how reliably an urge rating predicts an outcome after accounting for opportunities, learned habits, competing goals, and measurement reliability. Tiffany (1990) Robinson & Berridge (2008)

The fact that urge ratings predict later behavior on average is consistent with occasional individual dissociation. Vafaie and Kober's prospective meta-analysis supports the former proposition, while its moderator findings indicate that associations differ by cue type, measurement, lag to outcome, and substance-related context. A population odds ratio does not specify that every person with a higher craving score will resume use. Prediction is conditional: access, affordability, availability of alternatives, treatment, and social circumstances alter the translation from motivation into action. For causal explanation, it remains necessary to identify whether craving mediates cue exposure effects or merely accompanies a state that also promotes behavior. Vafaie & Kober (2022)

Researchers should avoid both overinterpreting a dissociation and dismissing it as measurement noise. An apparently craving-free episode of consumption may reflect poor temporal sampling, deliberate underreporting, nonconscious action preparation, or genuinely automatic routines. A craving episode without consumption may reflect excellent self-regulation, scarce access, fear of consequences, or immediate environmental constraints. The decisive methods combine intensive sampling, direct observation where ethically feasible, within-person comparison, and sensitivity analyses. Trials that alter one process while measuring other processes independently can reveal whether self-reported urge, cue attention, and actual drug use change together or through different pathways. Sayette (2016) Tiffany (1990) Vafaie & Kober (2022)

Cue Exposure, Extinction Learning, and Contextual Renewal

Extinction procedures repeatedly present a learned drug-related cue without the previously expected drug outcome, often with the goal of weakening its motivational impact. In associative-learning models, the resulting change can involve learning a new cue–no-outcome relation rather than erasure of the original memory. That distinction helps explain recovery of responding after changes in context, passage of time, or renewed drug contact. Conklin and Tiffany's examination of cue-exposure treatments argued that advances in experimental extinction research should inform interventions and highlighted limitations in treatment evidence available at that time. The evidence supports separating demonstrated reductions in session-bound cue response from a claim of durable prevention of relapse. Conklin & Tiffany (2002)

Extinction is inherently dependent on the contingency experienced by the learner. If a cue sometimes predicts access in ordinary life but never does in the clinic, the clinic may become a contextual signal for safety without reducing behavior in high-risk environments. This inference predicts that retrieval cues, varied contexts, and real-world transfer conditions might matter. However, a plausible mechanistic rationale is not an efficacy result. To evaluate cue-exposure therapy, randomized studies need clinically important endpoints, sufficient follow-up, intervention fidelity, and well-specified comparison conditions. The effect of exposure on craving during practice should be reported separately from later substance consumption, retention in treatment, and functioning. Conklin & Tiffany (2002) Vafaie & Kober (2022)

Exposure is also constrained by ethics and patient safety. Some situations involve hazardous withdrawal syndromes, severe psychiatric comorbidity, or access to substances in conditions that would raise immediate risk. Research protocols should define monitoring and clinical support appropriate to the setting. A scientific monograph can examine extinction as a mechanism of learning while recognizing that intervention planning depends on qualified assessment and substance-specific clinical guidance. The generalizable conclusion is methodological: the organism's current learning context and what a cue predicts in that context are essential determinants of the response. Shaham et al. (2003) Conklin & Tiffany (2002)

Cognitive Regulation, Approach Training, and Virtual Reality

Craving can be influenced by cognitive regulation of anticipated outcomes. Kober and colleagues used experimental regulation instructions and functional imaging to investigate neural processes engaged when participants altered their responses to appetitive cues. The study identified interactions involving prefrontal and striatal systems and illustrates that momentary motivational states are dynamically regulated. It does not follow that regulation instructions generalize unaided to every stressful environment or that greater prefrontal activation is always evidence of better long-term recovery. Efficacy depends on whether a strategy can be learned, retrieved under realistic conditions, and sustained as an improvement in clinically meaningful behavior. Kober et al. (2010)

Approach-bias modification is another candidate approach, aimed at modifying automatic tendencies to approach substance-related stimuli. In a randomized study involving alcohol-dependent inpatients, Wiers and colleagues reported changes in trained action tendencies and a favorable one-year outcome pattern; the unadjusted intention-to-treat relapse contrast was 46% versus 59% (p = .057), making the analytic model and missing follow-up data relevant to interpretation. The experimental contrast demonstrates that action tendencies can be manipulated, while later trials and replications are needed to determine generalizability and underlying mediation. A change in a computerized joystick task is a proximal outcome; sustained reduced substance use or improved functioning is a distinct clinical outcome. The claim that approach retraining directly reverses neural sensitization would require additional targeted evidence. Wiers et al. (2011) Robinson & Berridge (2025)

Immersive virtual-reality paradigms provide contextualized visual and social cues for studying and potentially modifying craving. A 2026 systematic review and meta-analysis by Li and colleagues synthesized nine randomized controlled trials with 553 participants across substance-use disorders and reported a moderate pooled reduction in craving following virtual-reality interventions, standardized mean difference −0.52 with 95% confidence interval −0.78 to −0.27. The outcome reflects craving assessments in the included studies. It does not by itself establish durable reduction in relapse or superiority across every substance or treatment environment. Variation in virtual-reality content, exposure intensity, comparator interventions, risk of bias, and follow-up remains central to interpretation. The technology is scientifically promising as a controllable stimulus environment, with its clinical value dependent on robust, replicated outcomes. Li et al. (2026)

Evidence Strength, Replication, and Methodological Confounds

The evidence base has several layers with different strengths. Laboratory cue exposure robustly alters some immediate craving measures; controlled animal studies demonstrate learning and motivational changes under specified histories; prospective human syntheses indicate that cue and craving indicators are associated with later drug use; and neural meta-analyses identify reproducible patterns of activation under particular task contrasts. Each result answers a different scientific question. The causal interpretation of a conditioned response is stronger when the relevant cue contingency is manipulated, whereas the transportability of that result to ordinary life depends on sampling, measurement, and exposure opportunities. High-quality synthesis therefore resists collapsing behavior, subjective experience, neural activity, and clinical outcome into one undifferentiated effect called 'cue reactivity.' Carter & Tiffany (1999) Vafaie & Kober (2022) Evohr et al. (2025)

The most consequential threats to validity include visually unmatched stimuli, variation in abstinence intervals, selective reporting of physiological endpoints, inadequate sample sizes, uncontrolled expectations of drug availability, and incomplete longitudinal outcome data. Fascher and colleagues' modern neural synthesis illustrates how task-design decisions and clinical covariates can change pooled brain maps. Vafaie and Kober's analysis shows a substantial shift in estimated prospective association when accounting for potential publication bias. Together, these findings establish an important hierarchy of claims: a statistically reliable average response requires accurate measurement; a predictive marker must generalize to later behavior; and a mechanism requires targeted causal evidence. The reproducibility agenda includes blinded or preregistered coding, shared stimuli with documented perceptual attributes, reliability estimates, and independent cohort validation. Fascher et al. (2026) Vafaie & Kober (2022)

Clinical-study selection requires particular scrutiny. Participants receiving inpatient treatment, medication, or intensive follow-up differ from people who are actively using substances outside treatment; outcome definitions also range from an isolated lapse to repeated use or severity of symptoms. These differences complicate pooling, even when the studies share the word craving. A rigorous report specifies follow-up windows, missingness, outcome ascertainment, treatment access, and risk of bias. Studies should quantify uncertainty using confidence intervals, analyze heterogeneity, and avoid interpreting a null result with inadequate precision as proof of no mechanism. Conclusions must reflect how and when the underlying observations were made. Bergeria et al. (2024) Evohr et al. (2025) Li et al. (2026)

Clinical Assessment, Intervention Targets, and Safety Boundaries

Craving assessment has practical research value because it records a meaningful experience and can help characterize change across treatment. A clinician or study team may use it in conjunction with substance-use history, frequency and consequences of use, withdrawal symptoms, co-occurring conditions, triggers, functioning, and immediate safety. The prospective meta-analytic evidence supports including craving and cue exposure in assessment, but a high score alone neither establishes a substance-use disorder nor dictates a particular intervention. Clinical decisions depend on substance class, regulatory guidance, severity, patient preference, and the full clinical context. The scientific role of cue-reactivity research is to clarify motivational processes and identify testable treatment targets. Sayette (2016) Vafaie & Kober (2022)

Mechanism-informed interventions can focus on modifying cue-response learning, strengthening cognitive regulation, reorganizing access to high-risk contexts, or altering learned action tendencies. These approaches are not equivalent. Cue exposure targets a conditioned relation; reappraisal targets evaluation of outcomes; approach training targets a response tendency; and virtual-reality methods can supply a controlled or contextualized environment for several approaches. A trial that reduces self-reported craving but leaves meaningful substance-use outcomes unchanged provides evidence for a proximal psychological effect, not necessarily for clinical recovery. The distinction is especially important when broad labels such as 'anti-craving treatment' are applied to heterogeneous protocols. Conklin & Tiffany (2002) Kober et al. (2010) Wiers et al. (2011) Li et al. (2026)

Clinical safety includes recognizing that withdrawal from some substances can require urgent medical management and that craving may occur alongside psychiatric distress or acute substance-related risk. Medication evidence, withdrawal management, and emergency care require dedicated clinical guidelines and individualized professional assessment beyond the explanatory scope of this dossier. Research involving cue exposure should document consent, monitoring, adverse events, and access to appropriate support. External clinical review may be commissioned for texts that issue material treatment recommendations; this work confines itself to research interpretation and the evaluation of published studies. The purpose of the safety boundary is to preserve scientific precision about what an experimental effect demonstrates and which clinical questions need specialized evidence. Sayette (2016) Bergeria et al. (2024)

Open Questions and a Falsifiable Research Agenda

A decisive next generation of studies should distinguish the persistence of sensitization from the persistence of learned expectation. One approach would assess cue-triggered motivation before and after interventions that selectively alter cue–outcome knowledge, while separately measuring hedonic reactions, craving, attention, and instrumental choice. If incentive motivation remains high despite a changed expected outcome under a well-controlled design, that pattern would be more diagnostic of a distinct incentive mechanism. Conversely, if changes in explicit expectancy fully explain behavior under carefully controlled conditions, a simpler associative explanation may suffice for that task. Comparative models need predictions that expose them to possible disconfirmation rather than explanations that can accommodate every result after the fact. Robinson & Berridge (2025) Everitt & Robbins (2016)

Another priority is bridging species and measurement traditions. Preclinical incubation should be linked to human studies with comparable cue histories, abstinence intervals, and repeated tests, while preserving the limitation that self-report cannot be obtained from animals. Human experiments could test controlled variation in drug expectancy and opportunity, paired with prospective EMA and verified substance-use outcomes. Where ethical and feasible, longitudinal measurements of neural or behavioral cue responses can help determine whether stable individual differences and momentary states interact. The challenge raised by human opioid data is not a reason to discard animal experiments; it is a demand for explicit mappings between experimental operations and clinical phenomena. Grimm et al. (2001) Bergeria et al. (2024)

Predictive science must test whether new measures improve decisions. A neural response, an attentional-bias score, and an urge rating should be compared with accessible clinical baselines on discrimination, calibration, and decision benefit in independent samples. Trials can then examine whether an intervention tailored to the marker achieves a better clinical outcome than usual evidence-based care. If a change in cue reactivity precedes and mediates improvement in substance use, the causal account becomes stronger; if outcome improvement occurs without changing the putative process, the hypothesized mediation requires reconsideration. Longitudinal studies across age, drug class, socioeconomic setting, and treatment systems are needed to determine the boundaries of generalization. Evohr et al. (2025) Vafaie & Kober (2022) Li et al. (2026)

Artificial Era: Predictive Cues and the Boundaries of Generalization

The transition From Homo to Artificial creates new research settings for the psychology of cue-mediated behavior, especially when artificial systems organize reminders, recommendations, interfaces, environments, and social interactions. Adaptive technologies can alter when people encounter drug-related imagery, how quickly attention is directed to salient content, and whether competing goals remain accessible. These are questions about human perception, learning, and motivation in technology-mediated environments. Their empirical assessment calls for controlled comparisons of cue exposure, interface design, and consequential human outcomes; extrapolation from a generic account of reinforcement learning does not establish that an artificial system undergoes craving or drug-induced incentive sensitization. Sayette (2016) Vafaie & Kober (2022)

The distinction between mathematical reward optimization and neurobiological incentive sensitization is fundamental to construct validity. A computational agent may modify its policy after repeated feedback or give greater numerical weight to a predictor, but those operations are defined by its architecture and objective function. Human addiction theories concern physiological exposure, learned significance, subjective reports, neurobehavioral adaptation, and situated action. Any proposed comparison should specify its operational mapping and the outcomes that would test it. A meaningful interdisciplinary program would examine how artificial environments modify human cue histories, how digital prompts affect craving measurement, and how systems might assist rigorous longitudinal monitoring under privacy and safety safeguards. Robinson & Berridge (2025) Tiffany (1990)

The research horizon consequently includes digital cues as experimental variables and artificial instruments as measurement systems. Neither development changes the empirical criteria for identifying craving or sensitization in humans. Data quality, access equity, consent, and the potential of monitoring to alter the behavior being measured remain important design issues. A scientific account of the Artificial Era can add new contexts to the study of addiction without replacing the neuropsychological and behavioral evidence on which established claims rest. The operative distinction preserves the independent meaning of the psychological construct while permitting new questions about technological mediation and human–artificial environments. Epstein et al. (2010) Sayette (2016)

Conclusion

Craving, cue reactivity, and incentive sensitization describe interconnected levels of addiction research. Craving is a reportable motivational experience; cue reactivity is a family of measurable responses elicited by learned substance signals; and incentive sensitization is a mechanistic theory of how exposure history can amplify the motivational power of those signals. Their integration explains how a familiar stimulus can acquire unusual importance, how attention and action can be recruited quickly, and why motivational episodes can persist or recur after abstinence. The strongest account combines associative learning, pharmacological and neural adaptation, context-dependent appraisal, habitual action, and available self-regulation. Robinson & Berridge (1993) Tiffany (1990) Robinson & Berridge (2025)

Empirical confidence is greatest for the existence of conditioned cue responses and their prospective association with later use across populations. Important qualifications arise when estimating the size of those associations, attributing them to one neural mechanism, or translating animal incubation procedures into human clinical time courses. Measurement differences and treatment environments help explain why human subjective craving trajectories cannot simply be inferred from animal drug-seeking paradigms. Neuroimaging identifies promising correlates and candidate biomarkers while requiring further prospective validation for individual clinical decisions. Vafaie & Kober (2022) Bergeria et al. (2024) Evohr et al. (2025) Fascher et al. (2026)

The governing scientific conclusion is that cue-dependent motivation is a temporally structured and environmentally situated system. Future progress depends on experimentally discriminating predictions among theories, improving measurement reliability, collecting repeated real-world observations, documenting the conditions of translation across substances and species, and evaluating interventions with meaningful long-term outcomes. The resulting research program has a clear explanatory ambition: to specify when learned signals increase motivation, which processes mediate that increase, and under what conditions motivation becomes substance use. This framework supports rigorous psychological research and provides a foundation for examining new human environments in the Artificial Era. Sayette (2016) Vafaie & Kober (2022) Robinson & Berridge (2025)

References

  • Bergeria, C. L., Gipson, C. D., Smith, K. E., Stoops, W. W., & Strickland, J. C. (2024). Opioid craving does not incubate over time in inpatient or outpatient treatment studies: Is the preclinical incubation of craving model lost in translation? Neuroscience & Biobehavioral Reviews, 160, Article 105618. https://doi.org/10.1016/j.neubiorev.2024.105618
  • Berridge, K. C., & Robinson, T. E. (2016). Liking, wanting, and the incentive-sensitization theory of addiction. American Psychologist, 71(8), 670–679. https://doi.org/10.1037/amp0000059
  • Carter, B. L., & Tiffany, S. T. (1999). Meta-analysis of cue-reactivity in addiction research. Addiction, 94(3), 327–340. https://doi.org/10.1046/j.1360-0443.1999.9433273.x
  • Childress, A. R., Mozley, P. D., McElgin, W., Fitzgerald, J., Reivich, M., & O’Brien, C. P. (1999). Limbic activation during cue-induced cocaine craving. American Journal of Psychiatry, 156(1), 11–18. https://doi.org/10.1176/ajp.156.1.11
  • Conklin, C. A., & Tiffany, S. T. (2002). Applying extinction research and theory to cue-exposure addiction treatments. Addiction, 97(2), 155–167. https://doi.org/10.1046/j.1360-0443.2002.00014.x
  • Drobes, D. J., & Tiffany, S. T. (1997). Induction of smoking urge through imaginal and in vivo procedures: Physiological and self-report manifestations. Journal of Abnormal Psychology, 106(1), 15–25. https://doi.org/10.1037/0021-843X.106.1.15
  • Epstein, D. H., Marrone, G. F., Heishman, S. J., Schmittner, J., & Preston, K. L. (2010). Tobacco, cocaine, and heroin: Craving and use during daily life. Addictive Behaviors, 35(4), 318–324. https://doi.org/10.1016/j.addbeh.2009.11.003
  • Everitt, B. J., & Robbins, T. W. (2016). Drug addiction: Updating actions to habits to compulsions ten years on. Annual Review of Psychology, 67, 23–50. https://doi.org/10.1146/annurev-psych-122414-033457
  • Evohr, B., Wenzel, S., Chen, N., Wade, C., Ekhtiari, H., Ketron, A. C., Janes, A. C., & Gilman, J. M. (2025). Cue-elicited brain activity and treatment outcomes in substance use disorders: A meta-analysis. JAMA Network Open, 8(12), e2548809. https://doi.org/10.1001/jamanetworkopen.2025.48809
  • Fascher, M., Thomsen, H., & Muehlhan, M. (2026). The modulatory effects of design-related and clinical factors in neural drug cue reactivity: An ALE meta-analysis. Neuroscience & Biobehavioral Reviews, 181, Article 106537. https://doi.org/10.1016/j.neubiorev.2025.106537
  • Flagel, S. B., Clark, J. J., Robinson, T. E., Mayo, L., Czuj, A., Willuhn, I., Akers, C. A., Clinton, S. M., Phillips, P. E. M., & Akil, H. (2011). A selective role for dopamine in stimulus–reward learning. Nature, 469, 53–57. https://doi.org/10.1038/nature09588
  • Grimm, J. W., Hope, B. T., Wise, R. A., & Shaham, Y. (2001). Neuroadaptation: Incubation of cocaine craving after withdrawal. Nature, 412, 141–142. https://doi.org/10.1038/35084134
  • Janak, P. H., & Tye, K. M. (2015). From circuits to behaviour in the amygdala. Nature, 517, 284–292. https://doi.org/10.1038/nature14188
  • Kober, H., Mende-Siedlecki, P., Kross, E. F., Weber, J., Mischel, W., Hart, C. L., & Ochsner, K. N. (2010). Prefrontal–striatal pathway underlies cognitive regulation of craving. Proceedings of the National Academy of Sciences, 107(33), 14811–14816. https://doi.org/10.1073/pnas.1007779107
  • Li, J., Ding, F., Chen, J., Ma, X., Wang, X., Hong, Q., & Li, H. (2026). Effects of virtual reality-based interventions on craving in substance use disorders: A systematic review and meta-analysis of randomized controlled trials. Drug and Alcohol Dependence, 285, Article 113226. https://doi.org/10.1016/j.drugalcdep.2026.113226
  • Preston, K. L., & Epstein, D. H. (2011). Stress in the daily lives of cocaine and heroin users: Relationship to mood, craving, relapse triggers, and cocaine use. Psychopharmacology, 218(1), 29–37. https://doi.org/10.1007/s00213-011-2183-x
  • Robinson, T. E., & Berridge, K. C. (1993). The neural basis of drug craving: An incentive-sensitization theory of addiction. Brain Research Reviews, 18(3), 247–291. https://doi.org/10.1016/0165-0173(93)90013-P
  • Robinson, T. E., & Berridge, K. C. (2008). The incentive sensitization theory of addiction: Some current issues. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1507), 3137–3146. https://doi.org/10.1098/rstb.2008.0093
  • Robinson, T. E., & Berridge, K. C. (2025). The incentive-sensitization theory of addiction 30 years on. Annual Review of Psychology, 76, 29–58. https://doi.org/10.1146/annurev-psych-011624-024031
  • Robinson, T. E., Yager, L. M., Cogan, E. S., & Saunders, B. T. (2014). On the motivational properties of reward cues: Individual differences. Neuropharmacology, 76(Pt B), 450–459. https://doi.org/10.1016/j.neuropharm.2013.05.040
  • Sayette, M. A. (2016). The role of craving in substance use disorders: Theoretical and methodological issues. Annual Review of Clinical Psychology, 12, 407–433. https://doi.org/10.1146/annurev-clinpsy-021815-093351
  • Shaham, Y., Shalev, U., Lu, L., de Wit, H., & Stewart, J. (2003). The reinstatement model of drug relapse: History, methodology and major findings. Psychopharmacology, 168(1–2), 3–20. https://doi.org/10.1007/s00213-002-1224-x
  • Sinha, R. (2011). New findings on biological factors predicting addiction relapse vulnerability. Current Psychiatry Reports, 13(5), 398–405. https://doi.org/10.1007/s11920-011-0224-0
  • Tiffany, S. T. (1990). A cognitive model of drug urges and drug-use behavior: Role of automatic and nonautomatic processes. Psychological Review, 97(2), 147–168. https://doi.org/10.1037/0033-295X.97.2.147
  • Vafaie, N., & Kober, H. (2022). Association of drug cues and craving with drug use and relapse: A systematic review and meta-analysis. JAMA Psychiatry, 79(7), 641–650. https://doi.org/10.1001/jamapsychiatry.2022.1240
  • Volkow, N. D., Wang, G.-J., Telang, F., Fowler, J. S., Logan, J., Childress, A.-R., Jayne, M., Ma, Y., & Wong, C. (2006). Cocaine cues and dopamine in dorsal striatum: Mechanism of craving in cocaine addiction. Journal of Neuroscience, 26(24), 6583–6588. https://doi.org/10.1523/JNEUROSCI.1544-06.2006
  • Wiers, R. W., Eberl, C., Rinck, M., Becker, E. S., & Lindenmeyer, J. (2011). Retraining automatic action tendencies changes alcoholic patients’ approach bias for alcohol and improves treatment outcome. Psychological Science, 22(4), 490–497. https://doi.org/10.1177/0956797611400615