RESEARCH DOSSIER
Intertemporal Choice and Delay Discounting in Addiction
Abstract. Intertemporal choice is the allocation of behavior between consequences available at different times, while delay discounting describes the reduction in a delayed consequence's present subjective value. This research monograph critically synthesizes historical choice theories, original studies of tobacco, opioids, alcohol, cannabis, and gambling, contemporary meta-analyses, longitudinal evidence, quantitative models, neural valuation research, and experimental interventions. The evidence establishes a reproducible but heterogeneous association between steeper discounting and addictive behavior. Prospective prediction and clinical treatment-outcome findings are less consistent, while experimentally manipulated nicotine deprivation and episodic future thinking demonstrate context-sensitive valuation. Measurement reliability, reward-domain comparability, competing causal pathways, environmental alternatives, and cross-population generalizability constrain interpretation. The concluding synthesis proposes discriminating tests for measurement, causal identification, and clinically meaningful transfer. This is a critical narrative synthesis grounded in identifiable research literature; it is not a registered systematic review, meta-analysis, experimental report, diagnostic protocol, or individualized treatment recommendation.
Keywords: intertemporal choice; delay discounting; addiction psychology; substance use disorder; behavioral economics; hyperbolic discounting; impulsive choice; reward valuation; reinforcer pathology; episodic future thinking; substance use; longitudinal prediction; psychometrics; cognitive neuroscience; clinical outcome validation
The Scientific Problem of Intertemporal Choice in Addiction
Intertemporal choice concerns decisions whose consequences occur at different times. In addiction research, the central question is how an individual allocates behavior between immediately accessible reinforcement and outcomes that require waiting, repeated effort, or sustained abstinence. Delay discounting describes the systematic decline in the present subjective value assigned to a consequence as its delivery is postponed. It is observable through choices among timed alternatives, and its relationship with addiction is empirically important because many substance-use decisions occur amid acute rewards and later health, social, and economic consequences. The construct gives investigators a quantitative language for studying a limited component of addictive decision making rather than an exhaustive explanation of a disorder Green & Myerson (2004) Bickel et al. (2014).
The research object is a distribution of contextually situated choices, valuation parameters, and trajectories across time. A person who chooses an immediately available smaller amount in one laboratory trial has produced an observable preference under that trial's conditions; whether that choice indicates persistent vulnerability depends on measurement reliability, reward type, delay units, financial circumstances, and predictive associations with subsequent behavior. Addiction diagnoses additionally require disorder-specific patterns of impaired control and consequences. This monograph treats discounting as an explanatory and potentially predictive dimension whose clinical meaning must be established independently rather than inferred from any single choice MacKillop et al. (2011) Exum et al. (2023).
Four questions organize the analysis. How did temporal choice become a formal experimental object? Which models account for the shape and variability of declining subjective value? How consistently does discounting covary with substance use, gambling, severity, and longitudinal outcomes? Which experimental changes in the choice environment modify measured discounting and improve meaningful behavioral outcomes? Evidence is drawn from classical theory, original choice experiments, neural studies, prospective cohorts, meta-analyses, and methodological reviews. The synthesis is a critical narrative research monograph; it neither reports a new empirical experiment nor purports to be a registered systematic review Rung & Madden (2018) Gelino et al. (2024).
Conceptual Boundaries: Impulsivity, Valuation, and Disorder
Impulsivity is a family of partially distinct constructs encompassing urgent action, deficient response inhibition, sensation seeking, and preference for sooner rewards. Delay discounting is one operationalization of the last component. The term impulsive choice is useful when specifying an actual comparison of sooner and later alternatives, but it should not be extended automatically to every feature of clinical impulsivity. A steep fitted curve describes discounting in a given measurement domain; it does not reveal motivation, awareness, diagnostic status, or the quality of other forms of inhibitory control. This separation is essential when different instruments are combined in risk prediction or when neurocognitive findings are interpreted Green & Myerson (2004) Kluwe-Schiavon et al. (2020).
Temporal discounting is also conceptually distinct from risk and probability discounting. Delayed outcomes can be delivered with high certainty or considerable uncertainty, and the expectation that an opportunity will disappear may rationally affect preference. Likewise, an immediate drug may be a directly consumable commodity while a delayed monetary amount is a flexible medium that can purchase many future outcomes. Comparisons between those options confound temporal valuation with the commodity's uses, storage properties, unit size, subjective need, and contextual risk unless experiments control them. Green and Myerson's integrative review specifically cautioned against treating superficially similar discount functions as evidence of one underlying process Green & Myerson (2004) Estle et al. (2007).
The addiction-specific object therefore includes substance-related timing, alternative reinforcers, availability, clinical severity, and repeated decisions during intoxication, withdrawal, recovery, and changing social circumstances. Broader judgment and decision-making science supplies principles, whereas this dossier tests their scope for addiction. Correlational evidence, longitudinal prediction, experimentally manipulated choice, and clinical improvement represent different inferential levels. A measured association between discounting and problem use is compatible with several causal pathways, including shared vulnerability, experience-dependent shifts in valuation, acute pharmacological effects, and exposure to unequal opportunity Amlung et al. (2017) Audrain-McGovern et al. (2009).
Historical Foundations of Temporal Choice and Self-Control
The modern conceptual history has roots in behavior analysis, psychological theories of self-control, and economic work on preference across time. George Ainslie's 1975 account of specious reward brought intertemporal preference reversals and patterns of impulsive choice into a prominent psychological framework. His argument drew attention to the possibility that relative preference changes as a delayed outcome approaches, making commitment, self-control, and sequential choice legitimate explanatory targets. This work is historically influential because it connects the geometry of valuation with a behavioral phenomenon: a person can plan to wait yet choose differently when a nearer reward becomes available Ainslie (1975).
Subsequent experiments developed adjusting-amount and adjusting-delay procedures to identify indifference points and compare proposed discounting functions. The central empirical advance was to replace undifferentiated descriptions of weak will with a measurable schedule of preferences at specified delays. Quantitative models allow a stronger test than a single binary choice because they predict the relationships among amount, duration, and reversals. Green and Myerson synthesized evidence concerning delayed and probabilistic rewards and emphasized that functional form, scaling, and reward magnitude all constrain interpretation Green & Myerson (2004).
The field entered addiction research through controlled comparisons involving alcohol, opioids, tobacco, and gambling. These studies were motivated by hypotheses about preference for immediate reinforcement, but their historical contribution is experimental: they measured choice with common tasks and compared groups with different drug-use histories. A theoretical interpretation can endure even when particular effect estimates change under replication, reanalysis, or better sampling. For this reason, early landmark results are examined as evidence about the tested participants and conditions, not as timeless diagnostic laws or proof of a single neural cause of addiction Kirby et al. (1999) Bickel et al. (1999) Petry & Casarella (1999).
Foundational Human Studies in Opioid Use, Tobacco, and Gambling
Kirby, Petry, and Bickel compared 56 people using heroin with 60 age-matched controls in monetary choice tasks involving smaller immediately available amounts and larger later amounts across intervals from one week to six months. The heroin group showed greater discounting in the tested comparisons. The study became a central example of laboratory intertemporal choice differences associated with opioid dependence. Its group comparison, however, cannot establish whether steeper discounting preceded drug use, resulted from drug exposure or withdrawal, reflected social adversity, or arose through several interacting pathways Kirby et al. (1999).
Bickel, Odum, and Madden subsequently compared 23 current smokers, 22 never-smokers, and 21 former smokers. Current smokers discounted hypothetical money more steeply than the comparison groups, whereas the two nonsmoking groups did not differ significantly. When smokers evaluated cigarettes, the delayed commodity was discounted more rapidly than money. These results are consistent with an addiction-linked change in temporal valuation, but the design was cross-sectional and the reward domains were not identical in economic function. The investigators explicitly identified selection and possible reversibility as competing explanations for the smoking-group differences Bickel et al. (1999).
Petry and Casarella examined delayed-reward discounting among substance-using participants with gambling problems, helping establish the importance of co-occurring patterns of problematic reinforcement. Kirby and Petry later compared people using heroin or cocaine with individuals meeting criteria for alcohol dependence and nondrug-using controls. Across these early investigations, substance class and problem severity were empirically relevant, while groups differed in ways that may matter beyond substance exposure. The enduring lesson is that systematically measured intertemporal preference can characterize clinical samples, while disorder specificity and causal direction require additional designs Petry & Casarella (1999) Kirby & Petry (2004).
Formal Models: Exponential, Hyperbolic, and Flexible Discounting
A basic exponential formulation assigns a delayed reward subjective value V=A exp(−rD), where A is its undiscounted amount, D is delay, and r is the estimated temporal rate. It implies a constant proportional rate of value loss per unit time and, under fixed conditions, dynamically consistent relative rankings. A simple hyperbolic formulation instead uses V=A/(1+kD), where k is a positive fitted rate and decline is relatively steep near the present. The value of these functions lies in the predictions they generate for observed indifference points, not in the presumption that a fitted curve is a literal neural operation Ainslie (1975) Green & Myerson (2004).
Hyperbolic curves permit preference reversals for certain combinations of early and late options when both delays shorten by the same amount. Such reversals help explain why a commitment made at an earlier time may change when a tempting opportunity becomes imminent. Yet many discounting data sets are compatible with more than one function over a restricted experimental range; generalized hyperbolic and quasi-hyperbolic models add flexibility at the cost of additional parameters. Reliable comparison requires specifying whether the aim is in-sample fit, out-of-sample prediction, explanation of preference reversal, or meaningful prediction of later substance-use behavior Green & Myerson (2004) Kable & Glimcher (2007).
The fitted k parameter is sensitive to how delay is represented. A value estimated with delays measured in days cannot be compared numerically with a value fitted to the same choices using months without unit conversion. Magnitude effects, ceiling or floor choices, constrained choice sets, and assumptions about utility curvature likewise alter estimates. Log transforming positive k is common because empirical distributions are skewed; area under the empirical discounting curve offers a partly model-independent summary with different statistical properties. Treating all these indices as interchangeable obscures meaningful methodological variation Johnson & Bickel (2008) Exum et al. (2023).
The Temporal Ecology of Rewards and Alternatives
Addiction-related decisions occur within environments that make some consequences immediate, predictable, and intensely reinforcing while leaving other opportunities delayed, uncertain, costly, or inaccessible. A delay-discounting parameter estimated with monetary options may therefore capture a mixture of preference, expected environmental stability, learned scarcity, and task interpretation. The comparison between immediate substance use and distant health improvement is especially complicated because health benefits arrive gradually, whereas consumption can also produce immediate relief from aversive states. A formal experiment must distinguish the time of delivery from the broader structure of available and expected outcomes Bickel et al. (2011) Green & Myerson (2004).
Research comparing rewards suggests that commodity characteristics matter. Estle and colleagues examined money, beer, candy, and soda and found directly consumable delayed rewards discounted more steeply than money, whereas probability discounting did not show the same commodity distinction. Odum and Rainaud similarly reported stronger delay discounting for hypothetical food and alcohol than for money in adults without reported substance-use problems. Thus the particularly steep discounting of delayed drugs cannot simply be read as a substance-specific pathological signature; fungibility and consumability are alternative explanations that must be modeled Estle et al. (2007) Odum & Rainaud (2003).
Alternative reinforcement is a critical bridge between discounting and behavioral economics. Greater availability of accessible, valuable, non-drug activities may change the competing option set without transforming a stable individual trait. In a study of adolescents, Audrain-McGovern and colleagues identified links among substitute activities, complementary activities, and smoking progression. Experimental and longitudinal designs are needed to determine when changing availability, timing, and expectancy changes substance use itself. A scientifically adequate account treats individual choice and environmental opportunity as jointly specified, rather than attributing a sequence of behaviors to a single scalar preference Audrain-McGovern et al. (2004) Bickel et al. (2011).
Meta-Analytic Evidence Across Addictive Behaviors
MacKillop and colleagues assembled 46 studies with 64 comparisons in a foundational meta-analysis of delayed reward discounting and addictive behavior. The aggregate included 56,013 participants but displayed pronounced heterogeneity; an initial synthesis produced a small standardized difference, d=.15. When very large studies using abbreviated self-report measures were excluded, 57 more methodologically comparable comparisons involving 3,329 participants yielded d=.58. The difference between these two estimates is a substantive result about measurement and design: a headline sample size alone cannot guarantee a comparable construct. Clinical samples showed larger standardized group differences than nonclinical samples within the authors' subsequent analyses MacKillop et al. (2011).
Amlung and colleagues approached the literature through continuous associations rather than categorical group contrasts. Their 2017 meta-analysis found a small association between steep discounting and measures of addictive behavior, approximately r=.14 across the principal synthesis. These estimates answer different questions. Group contrasts can reflect diagnostic selection and the distribution of severity in case-control samples, while correlations measure covariation within the observed range and are affected by restricted variance and reliability. Neither estimate by itself implies an effect of identical magnitude across nicotine, alcohol, cannabis, stimulants, opioids, and gambling Amlung et al. (2017) MacKillop et al. (2011).
A later multilevel meta-analysis by Kluwe-Schiavon and colleagues evaluated delay discounting alongside feedback-based decision measures and found substance-use disorders associated with differences on both domains, while reporting moderation by early onset and psychiatric comorbidity. The overall evidence supports a reproducible association at the population level, with important variation by task, clinical inclusion rules, commodity, and subgroup. The next generation of research must explain why associations vary, rather than assuming that a statistically reliable average describes every person's choices or all mechanistically distinct forms of addiction Kluwe-Schiavon et al. (2020) Strickland et al. (2021).
Tobacco, Nicotine Dependence, and Abstinence States
Tobacco research has an unusually long experimental history because smoking status can be assessed comparatively and abstinence can be manipulated for limited periods. Bickel and colleagues found steeper monetary discounting in current smokers than in either former or never-smokers, with cigarette rewards discounted even more rapidly among smokers. This design identified an association but left unresolved whether discounting was a precursor, consequence, or correlate of smoking. Former-smoker findings are informative because they invite hypotheses about treatment and recovery, but people who successfully quit may differ from continuing smokers before cessation Bickel et al. (1999).
Field and colleagues studied 30 daily smokers under a repeated-measures comparison of at least thirteen hours without cigarettes versus smoking as usual. The abstract reports steeper discounting of both cigarette and monetary rewards after deprivation. This experimentally timed within-person change is evidence that measured discounting can have a state-dependent component; it also shows the importance of assessing craving and withdrawal. The choice outcomes were hypothetical, the sample was small, and manipulation of deprivation does not identify every mediator linking withdrawal to altered valuation Field et al. (2006).
The smoking literature also highlights measurement differences between a laboratory estimate and subsequent behavioral outcomes. Among pregnant women who had stopped smoking, Yoon and colleagues investigated whether delay discounting predicted postpartum relapse. Prospective relationships of this kind improve temporal ordering but depend on the sampled population, the outcome definition, and the accuracy of follow-up. Intertemporal choice remains a useful research predictor when assessed alongside nicotine dependence, social context, cessation support, and repeated exposure opportunities rather than as an autonomous clinical decision rule Yoon et al. (2007) Exum et al. (2023).
Opioids, Stimulants, and the Scope of Drug-Class Generalization
Early comparisons in heroin dependence and later heroin-cocaine-alcohol contrasts gave investigators evidence that addictive substance use could accompany unusually steep monetary discounting. In the original 1999 study by Kirby, Petry, and Bickel, participants selected between explicitly specified immediate and delayed monetary rewards. The careful outcome is a group difference in hypothetical or scheduled monetary choice under the studied procedures. Opioid exposure, dependence severity, socioeconomic history, co-occurring disorders, and acute drug state are potential explanatory contributors that a cross-sectional task alone cannot disentangle Kirby et al. (1999) Kirby & Petry (2004).
The 2004 comparison by Kirby and Petry reported that heroin- and cocaine-using groups discounted delayed rewards more steeply than alcohol-dependent or nondrug-using comparison groups. That result challenges a simplistic assumption of identical temporal preferences across drug classes. At the same time, treatment recruitment, abstinence duration, reward magnitude, and socioeconomic comparability can all influence the observed ranking. Substance-specific differences should be analyzed against common tasks, comparable clinical severity, and standardized modeling assumptions before being interpreted as pharmacologically intrinsic effects Kirby & Petry (2004) Kluwe-Schiavon et al. (2020).
Pathways from temporal preference to opioid or stimulant use plausibly involve expected relief, immediate reinforcement, environmental constraints, and repeated learning, but a plausible pathway is not experimental verification. Temporal discounting is measured at the choice level; brain reward systems, learned drug demand, and executive regulation are measured at additional explanatory levels. Integrating them requires designs in which their predictions differ, preferably longitudinal assessments that incorporate both use behavior and changing environmental conditions. Broader behavioral-economic reviews of reinforcer pathology make this multilevel interpretation explicit and propose candidate mechanisms for direct testing Bickel et al. (2011) Bickel et al. (2014).
Alcohol and Cannabis: Heterogeneous Associations
Alcohol use comprises diverse populations ranging from social consumption to severe use disorder, with substantial variation in acute intoxication, chronic exposure, and clinical recruitment. Group comparisons involving alcohol often identify higher discounting in problematic users, yet the size and practical significance of the effect depend on whether the outcome represents quantity-frequency, harmful consequences, or clinical diagnosis. These measures occupy different levels of a disorder continuum. The 2017 continuous-association synthesis examined these distinctions and cautions against equating an aggregate relationship with precise prediction for any given patient Amlung et al. (2017) MacKillop et al. (2011).
Cannabis offers an instructive empirical limit case. Strickland and colleagues' systematic review and meta-analysis covered 27 studies, 61 effect sizes, and 24,782 participants, reporting an overall correlation of r=.082 between discounting and cannabis-use frequency or severity. The investigators documented stronger associations for incentivized tasks related to severity measures than for hypothetical tasks related to quantity-frequency measures. The effect is small on average and the design-specific moderation matters. Claims that elevated discounting is a universal cannabis phenotype would go substantially beyond these data Strickland et al. (2021).
The cannabis findings also illustrate a general measurement problem. Quantity-frequency indexes how often and how much a substance is used, whereas severity includes loss of control, adverse consequences, and functional impairment. An instrument can correlate with one outcome and weakly with another because the outcomes involve distinct psychological and social processes. Therefore, comparative substance research should specify exposure, disorder criteria, task incentives, and the intended estimand. It should also examine alcohol and cannabis use in the presence of co-use and other concurrent factors rather than treating a primary substance label as a complete behavioral description Strickland et al. (2021) Kluwe-Schiavon et al. (2020).
Gambling, Behavioral Addictions, and Common-Mechanism Claims
Problematic gambling offers a test of intertemporal valuation outside direct pharmacological drug reinforcement. Petry and Casarella examined excessive discounting in people with substance misuse and gambling problems, establishing a basis for studying temporal preference in overlapping behavioral and substance-related patterns. A finding of steeper monetary discounting in gambling does not imply that gambling and opioid use share every causal pathway. Financial risk, variable reward schedules, losses, debt, and exposure to rapidly available bets contribute additional choice mechanisms that laboratory delay tasks must distinguish Petry & Casarella (1999) Green & Myerson (2004).
A 2026 systematic review and meta-analysis by Ahluwalia and colleagues assessed neurocognitive measures in gambling samples. Seventy-six observational studies involving 5,362 participants entered the qualitative review and thirty-five studies involving 2,822 entered quantitative analyses. Delay-discounting results differed by metric: the reported summary for area-under-the-curve measures was g=−1.07, whereas a discount-rate parameter synthesis yielded g=.45. The wide heterogeneity across analyzed cognitive measures and task-specific effects demand care in treating numerical results as equivalent. The review supports a gambling-related association but does not settle its causal role Ahluwalia et al. (2026).
The gambling case clarifies what a cross-domain theory of addiction must achieve: it should specify which features recur across disorders, which depend on a particular reinforcer, and which emerge from contextual opportunity. Gambling choices involve uncertain outcomes in addition to time; collapsing probability and delay into one parameter may hide important distinctions. A computational account that predicts both timed reward choices and independent gambling outcomes would provide a stronger cross-domain test than post hoc similarity in the shape of fitted discounting curves Green & Myerson (2004) Ahluwalia et al. (2026).
Longitudinal Development: Does Discounting Precede Use?
Cross-sectional associations are compatible with discounting preceding drug use, drug use changing discounting, and shared causes affecting both. Audrain-McGovern and colleagues tested temporal ordering in a longitudinal smoking cohort of 947 participants followed from adolescence into young adulthood. Their analyses considered peer and household smoking, depressive symptoms, novelty seeking, attention symptoms, and other substance use. The study investigated whether baseline discounting and subsequent change related to smoking trajectories, demonstrating the importance of measuring both constructs repeatedly instead of assuming a one-way causal relationship from a single baseline score Audrain-McGovern et al. (2009).
Isen, Sparks, and Iacono tested predictive validity in 675 adolescent twins. Hypothetical discounting measured at ages fourteen to fifteen was prospectively associated with later laboratory real-world discounting, but was not consistently related to substance use or trait impulsivity. This result is crucial counterevidence to any claim that discounting must function as a reliable prospective marker of addiction across samples. It shows that temporal-choice behavior can have predictive continuity in its own measurement domain without delivering equivalent prediction of a clinical outcome Isen et al. (2014).
A more recent analysis of the multicenter IMAGEN cohort by Dali and colleagues tracked 1,668 participants across waves at approximately ages fourteen, sixteen, eighteen, and twenty-two. Higher discounting predicted the initiation of occasional smoking and smoking-frequency trajectories, but not the transition from occasional to daily smoking; smoking did not predict subsequent changes in discounting in the reported analysis. The result strengthens one developmental direction in this sample while illustrating outcome specificity. A longitudinal association remains vulnerable to residual confounding and selective attrition, and should not be presented as proof that changing discounting would prevent nicotine dependence Dali et al. (2025).
Natural Variation, Recovery, and Within-Person Dynamics
A useful temporal-choice theory must accommodate differences between people and changes within the same person. Delay-discounting estimates are often treated as relatively persistent, yet experiments involving nicotine deprivation and future-oriented imagery demonstrate that task performance is also susceptible to state and context. This dual requirement has direct implications for studies of relapse and recovery. Investigators should separately estimate stable individual differences, acute within-person fluctuations, and measurement error rather than combine them into a single undifferentiated risk variable Field et al. (2006) Peters & Büchel (2010).
Clinical recovery may change the effective value and availability of alternatives, the expected horizon of abstinence, and the accessibility of non-drug reinforcement. Selection effects complicate interpretation: people who remain in follow-up, maintain abstinence, or complete treatment may have differed from those who do not from the outset. A comparison of former and current smokers therefore warrants a repeated-measures design or credible causal identification strategy before it becomes evidence of a reversible neurocognitive alteration. The classic smoker study explicitly leaves this issue open Bickel et al. (1999) Audrain-McGovern et al. (2009).
Prospective analyses of clinical outcomes provide a stricter validation target than group discrimination alone. Exum and colleagues reviewed seventeen studies measuring delay discounting before substance-use treatment and assessing later outcomes. Approximately forty-seven percent of the studies reported a significant association in some circumstances, with sensitivity to task and parameter choices. The literature therefore contains clinical promise and substantial inconsistency. For future studies, preregistered primary endpoints, explicit control variables, missing-data analyses, and replication across treatment settings are necessary to clarify which changes in discounting carry independent prognostic information Exum et al. (2023) Gelino et al. (2024).
Laboratory Paradigms and the Meaning of Indifference
The usual human discounting task asks participants to choose between a smaller sooner reward and a larger later reward at several delays. In an adjusting-amount procedure, the sooner amount changes according to preceding responses until the participant approaches indifference between the alternatives. Fixed-choice questionnaires instead offer predetermined pairs, often requiring fewer trials. Experiential tasks deliver at least some outcomes within the testing session and may include waiting, response effort, and probabilistic contingencies. Each design measures a different combination of explicit valuation, learning, comprehension, and tolerance for experimental delay. Their estimates should be compared only after procedural differences are made explicit Johnson & Bickel (2008) Exum et al. (2023).
Indifference is a behavioral inference about approximately equal choice probability under specified conditions, not a report of a consciously calculated internal exchange rate. When a participant alternates between options, an investigator may estimate the immediate amount equivalent to a later reward by fitting choice curves or using adaptive rules. The result depends on the range of offered amounts, whether rewards are real or hypothetical, whether participants trust delayed payment, and whether they have encountered equivalent tasks before. Repeated tasks permit within-person analysis but can introduce practice or consistency effects that need modeling Kable & Glimcher (2007) Green & Myerson (2004).
Experimental tasks with nonhuman animals offer stronger control over reinforcement histories and administered delays, while human research can include clinically meaningful histories and extended hypothetical horizons. Translational comparisons require care because a seconds-long food delay for a hungry animal is not equivalent in ecological meaning to a months-long financial delay for a person in treatment. Animal studies are valuable for identifying learning and schedule mechanisms; human clinical prediction must be demonstrated in human populations rather than assumed from formal curve similarity. The central criterion for construct validity remains the connection between measured parameters and the intended independent behavior Rung & Madden (2018) Bickel et al. (2014).
Measurement Quality: Reliability, Validity, and Data Screening
A discount rate derived from a small set of inconsistent choices can be unstable or impossible to interpret. Johnson and Bickel examined non-systematic patterns in archival discounting data and demonstrated a problem with eliminating participants solely on poor R-squared fits: fit quality can correlate with discounting magnitude, creating selective measurement bias. They proposed criteria based on the expected general pattern of decreasing subjective value as delay increases. Their contribution places data-quality judgments before substantive claims about group differences, while preserving the possibility that unusual responding itself may be meaningful if repeatedly observed Johnson & Bickel (2008).
Gelino and colleagues' 2024 systematic review and meta-analysis covered thirty publications, thirty-nine independent samples, and 262 measures of delay or probability discounting. The estimated test-retest reliability was r=.670 with a 95 percent confidence interval of .618 to .716; the pooled within-person stability estimate showed a small nonsignificant mean difference over time. Reliability varied with task domain, reward medium, participant population, and retest interval. This evidence supports meaningful individual-difference research but counsels against treating any single discounting score as a perfectly stable trait, especially when monitoring intervention-related change Gelino et al. (2024).
Construct validation must examine convergence with related choice tasks and discrimination from response inhibition, risk preference, and general distress. Predictive validation asks whether a measure improves forecasts beyond established clinical and contextual variables, ideally in independent samples. Measurement invariance and task comprehension are especially important across age groups, languages, education levels, and economic circumstances. A statistical change in a fitted rate should be interpreted against the test-retest distribution, plausible measurement error, and any manipulation of reward scale. The distinction between reliability and clinical utility is foundational: a reproducible number is not yet a useful treatment target Gelino et al. (2024) Exum et al. (2023).
Parameter Estimation, Model Comparison, and Uncertainty
Fitting a temporal discount function requires choices about likelihood, observation error, response consistency, and parameter constraints. An indifference-point regression treats estimated values as observed data; a trial-level probabilistic model instead predicts the probability of each choice from the subjective-value difference, often including a sensitivity or noise parameter. The latter approach can use more of the available information but introduces additional assumptions. In either case, uncertainty should be reported using confidence or credible intervals rather than implying that a single point estimate uniquely characterizes the participant Green & Myerson (2004) Kable & Glimcher (2007).
Model comparison must distinguish descriptive adequacy from mechanistic explanation. An exponential, simple hyperbolic, generalized hyperbolic, or quasi-hyperbolic function can fit observed choices with different flexibility. Model fit is sensitive to the offered horizons and magnitudes; a more flexible function can improve fit without identifying a unique psychological process. Cross-validation across delays, task formats, or later sessions provides a more stringent criterion for claims of predictive superiority. When competing models perform similarly, a family of plausible parameterizations may be more scientifically informative than a winner selected by one in-sample goodness-of-fit statistic Green & Myerson (2004) Johnson & Bickel (2008).
For clinical prediction, preprocessing and analysis choices should be documented before outcomes are inspected. Researchers should declare exclusion criteria, the treatment of nonsystematic responders, transformation of discount-rate distributions, whether repeated trials are nested within people, and the handling of missing follow-up. Exum and colleagues found that studies using adjusting-choice computerized tasks and log-transformed rates were more likely to report associations with treatment outcomes than studies using other measures. This finding does not prove that those methods are universally superior; it shows that observed prognostic validity is partly methodological and requires direct comparison on shared patient samples Exum et al. (2023).
Neural Valuation and Competing Decision-System Models
Neuroscientific models seek to identify how timing and subjective reward value are represented during intertemporal decisions. McClure and colleagues reported functional MRI evidence that immediately available options elicited preferential activation in limbic-associated regions, whereas lateral prefrontal and posterior parietal regions were engaged in choices regardless of delay. This influential finding motivated accounts of interacting valuation and control systems. The study examined monetary choice rather than addiction diagnosis and does not establish that a single anatomical pathway constitutes an impulsive system or that an observed regional activation causes substance-use disorder McClure et al. (2004).
Kable and Glimcher subsequently identified neural activity correlated with individually estimated subjective value of delayed monetary rewards, notably in ventral striatum, medial prefrontal cortex, and posterior cingulate cortex. Their results support valuation representations that vary across people and choices, offering an alternative emphasis to a rigid separation of immediate versus delayed reward circuitry. The research question changes from which area is active during a choice to how distributed circuits encode a quantity whose behavioral definition is explicit. Neural correlates constrain theory but require intervention or stronger causal designs to identify mechanisms Kable & Glimcher (2007) Peters & Büchel (2011).
Bickel and colleagues developed a competing neurobehavioral decision systems framework and the concept of reinforcer pathology to integrate excessive value assigned to a reinforcer with persistent short-horizon choice. These are theoretically useful because they connect temporal discounting with drug demand and treatment opportunities. Their scientific status is that of integrative explanatory models evaluated against accumulating observations, not anatomical diagnoses of individuals. Specific predictions about network interactions, choice changes, and clinical outcomes should be compared against unitary valuation, context-sensitive learning, and resource-based accounts rather than treating one neural vocabulary as sufficient explanation Bickel et al. (2011) Bickel et al. (2014).
Episodic Future Thinking and Prospective Simulation
Episodic future thinking involves constructing or retrieving personally specific images of possible future experiences. It may influence intertemporal choice by making delayed consequences concrete, affectively salient, and accessible during valuation. Peters and Büchel used subject-specific future event cues during discounting decisions and found reduced discounting accompanied by changes in functional interactions among decision-related and memory-related regions. Their second experiment replicated the behavioral effect and considered alternative explanations. The study identifies a plausible cognitive route through which a distant option gains immediate representational strength at the moment of choice Peters & Büchel (2010).
Future thinking is experimentally appealing because the manipulation can be randomized and the discounting response can be measured promptly. However, laboratory change in monetary choices is an intermediate outcome. It does not establish that a participant has reduced drug use, sustained abstinence, or improved functioning. Valence, vividness, episodic detail, familiarity with the experimenter, expectation of future opportunity, and control-task matching can influence observed effects. A mechanistic trial should measure these features and test whether changes in actual behavior are mediated by changes in temporal valuation Rung & Madden (2018) Ye et al. (2022).
Ye and colleagues synthesized forty-seven studies with sixty-three experimental contrasts and reported an overall effect of episodic future thinking on discounting of Hedges' g=.52, with moderation by future-event valence and task characteristics. Positive future thinking produced a larger pooled effect than neutral or negatively valenced simulations in the reported analyses. These results provide strong evidence of manipulability under studied task conditions, while leaving open the dosage, durability, accessibility, and clinical generalization of interventions in substance-use populations. The difference between shifting an assay and changing a disorder remains a central translational question Ye et al. (2022) Exum et al. (2023).
Timing, Withdrawal, Stress, and State-Dependent Preference
Temporal choice varies with immediate physiological and psychological states that alter the relative attractiveness of available consequences. Acute withdrawal can make relief urgently valuable, while stress and negative affect may narrow effective planning horizons or increase the appeal of fast reinforcement. The within-person nicotine-deprivation experiment by Field and colleagues offers one direct demonstration that experimentally timed abstinence can change measured discounting. Its limited sample and hypothetical tasks require proportionate interpretation, but the result is sufficient to challenge theories that assume a discount rate is entirely fixed Field et al. (2006).
Research during recovery must distinguish a transient craving episode, a persistent reinforcement history, and wider background conditions. For example, financial instability or an unreliable environment can change the expected value of waiting even when an individual understands the nominal amounts in a task. Emotional regulation, social support, access to substitute rewards, and the credibility of delayed outcomes all offer candidate pathways to test. These variables can alter both laboratory choice and real-world behavior through partly independent mechanisms, so a measured shift in discounting should not be asserted to mediate every contextual effect Audrain-McGovern et al. (2004) Bickel et al. (2011).
A strong within-person design samples multiple time points, includes withdrawal and mood measures, and separates between-person means from day-to-day deviations. The investigator can then test whether a rise in steep discounting predicts a later craving, a decision to use, or an independent negative outcome above earlier symptoms and exposure. Ecological assessments and experience sampling are promising because real choices arise amid changing opportunity, but their observational nature requires careful temporal modeling. The methodological objective is to locate discounting within a dynamic system of needs and alternatives rather than to replace the system with one measurement Exum et al. (2023) Gelino et al. (2024).
Behavioral Economics: Demand, Substitution, and Reinforcer Pathology
Delay discounting describes the timing component of valuation; demand procedures describe how consumption changes as price, effort, or another cost increases. These measures need not covary perfectly. A person may highly value a substance but still prefer delayed money in an experimental task, and someone may strongly discount hypothetical money without substantial drug demand. Bickel and colleagues' reinforcer-pathology framework combines high valuation of a commodity with excessive present orientation to explain certain persistent patterns of consumption. Its empirical value rests in whether joint measures improve prediction and identify independently manipulable causal processes Bickel et al. (2011) Bickel et al. (2014).
Substitution is especially important because available alternatives define the choice set. A delayed reward may be objectively large but practically unusable if the person lacks transport, secure housing, or reliable payment mechanisms. Conversely, immediate access to attractive nondrug activities can compete with drug reinforcement. Audrain-McGovern and colleagues' adolescent smoking analysis identified protective substitute reinforcers and complementary activities linked to progression, pointing to the combined relevance of preference and opportunity. Environmental modifications should therefore be studied as independent treatment components rather than assumed to operate exclusively by changing a discount parameter Audrain-McGovern et al. (2004).
Drug-specific and monetary tasks add another distinction. Money is a generalized purchasing medium, whereas cigarettes, food, alcohol, and other consumables can differ in spoilage, satiation, immediacy of use, and associated cues. Estle and colleagues' commodity comparisons show that more rapid discounting of a substance than of cash need not reflect a unique drug-related deficit. Economic measurement that separates demand, reinforcement magnitude, commodity type, and time can distinguish testable hypotheses about addiction and avoid labeling every observed preference difference a manifestation of one pathology Estle et al. (2007) Odum & Rainaud (2003).
Causal Interpretation and Experimental Identification
The central causal question is whether changing temporal valuation can alter addictive behavior, independently of changes in general motivation, financial opportunity, social context, and the treatment relationship. Observational findings do not answer this question by themselves. A discounting score can precede drug use because of common developmental conditions, and drug use can alter the score through intoxication, withdrawal, or lifestyle consequences. The original smoking comparisons, longitudinal cohorts, and acute deprivation experiments represent progressively different causal information; they should be interpreted according to their actual contrasts rather than placed on one undifferentiated evidentiary scale Bickel et al. (1999) Audrain-McGovern et al. (2009) Field et al. (2006).
A randomized study manipulating episodic future thinking can establish an immediate effect of a particular intervention on a choice task when control conditions and measurement are appropriate. A stronger claim that discounting mediates reduced substance use requires additional evidence: the intervention must alter the mediator before the outcome, rival mediators must be considered, and the outcome must be independently measured. Conditioning on posttreatment variables can also introduce bias. Multicomponent therapies create special difficulties because therapeutic engagement, material support, perceived opportunity, and discounting may change together Peters & Büchel (2010) Rung & Madden (2018).
Identification strategies should be matched to the mechanism. Acute, counterbalanced within-person manipulations test state sensitivity; longitudinal cohorts estimate temporal associations subject to confounding; randomized behavioral interventions test the effects of assigned conditions; and mechanistic neuroimaging contributes a different level of explanation. A defensible synthesis can combine these designs without claiming that any one substitutes for the others. Evidence becomes stronger when independent methods produce predictions that converge on a specified causal pathway and when the result generalizes beyond the particular laboratory commodity Isen et al. (2014) Exum et al. (2023).
Clinical Prediction, Treatment Response, and Outcome Validity
Prospective treatment-outcome research has produced a more mixed evidence base than cross-sectional group comparisons. Exum and colleagues' 2023 systematic review included seventeen studies assessing delay discounting before treatment and later abstinence, relapse, use, or adherence outcomes. Eight reported a significant association in at least some analyses, while nine reported no significant associations. Task format and transformation of the fitted parameter appeared related to the frequency of positive findings. Consequently, a claim that steep discounting reliably forecasts treatment failure across addictions is not supported by the complete reviewed pattern Exum et al. (2023).
Clinical utility is a higher standard than demonstrating an association. A proposed prognostic marker should improve out-of-sample prediction beyond severity, treatment history, dependence, co-occurring symptoms, social resources, and current substance use. Calibration, decision thresholds, and performance across populations matter more than a statistically significant coefficient alone. For individual care, measurement uncertainty can make close scores indistinguishable. A temporal-choice task may contribute to research stratification, but the existing evidence does not warrant using it in isolation to deny services, determine risk categories, or forecast a particular patient's recovery Exum et al. (2023) Gelino et al. (2024).
Treatment outcomes should be defined precisely. Initial cessation, sustained abstinence, reduced harmful consumption, retention, craving, employment, quality of life, and overdose prevention are not interchangeable endpoints. Discounting may relate differently to the ability to initiate change and to the structural conditions required to maintain it. Reliable results require prespecified outcome windows, objective or validated reports when feasible, adequate treatment exposure measures, and careful accounting for participants lost to follow-up. The scientific aim is to determine where temporal choice adds information and where it does not Exum et al. (2023) Amlung et al. (2017).
Interventions and the Limits of Surrogate Outcomes
Researchers have experimented with many procedures intended to reduce steep delay discounting: episodic future thinking, framing, working-memory exercises, contingency arrangements, learning-based training, and modifications in reward presentation. Rung and Madden reviewed ninety-two articles investigating experimental reductions in delay discounting and concluded that numerous manipulations could change task performance, with differences in size and persistence. Their synthesis also stressed methodological limitations and the need to determine which manipulations translate into sustained behavior change. The existence of a manipulable laboratory effect is a promising experimental finding, not sufficient demonstration of treatment efficacy for addiction Rung & Madden (2018).
Episodic future thinking has one of the clearest mechanistic literatures. Peters and Büchel demonstrated experimentally induced preference changes alongside neural correlates of episodic simulation, and Ye and colleagues found a pooled positive effect on discounting across diverse experimental comparisons. Future thinking could potentially help make delayed goals concrete within established treatment, but protocols must be evaluated for applicability to people with trauma, depressive symptoms, severe financial insecurity, or reduced access to rewarding future opportunities. The ethical and clinical target should remain patient-relevant functioning rather than a universally flatter discount curve Peters & Büchel (2010) Ye et al. (2022).
Supportive environmental arrangements may improve real choices even when a questionnaire's fitted k does not change substantially. Providing reliable incentives for attendance, access to substitute reinforcers, housing stability, or predictable payments can modify action possibilities and immediate consequences directly. Behavioral-economic theory suggests that these changes interact with temporal valuation, but their benefits and risks require independent outcome research. No intervention should be described as curing addiction solely because it reduces discounting in a one-session experiment Bickel et al. (2011) Audrain-McGovern et al. (2004) Exum et al. (2023).
Developmental, Cultural, and Economic Generalizability
Intertemporal choice is measured across adolescence and adulthood, but the meaning of waiting depends on developmental experience, resources, and culturally structured expectations. Adolescents may confront short horizons, constrained autonomy, and immediate social rewards, while older adults may face different financial and health contingencies. The longitudinal twin study by Isen and colleagues and the IMAGEN cohort analysis by Dali and colleagues show that prospective prediction differs by outcome and sample. Age-related associations therefore require age-appropriate tasks, repeated measures, and careful separation of maturation from changes in opportunity Isen et al. (2014) Dali et al. (2025).
Cross-cultural comparability demands more than translated instructions. Temporal units, payment credibility, local financial volatility, household obligations, typical reward magnitudes, and expectations about future security can all alter the ecological meaning of a laboratory option. A reward of the same nominal currency amount may differ radically in utility across populations. Studies should test whether estimated discount rates remain comparable after adapting magnitude, comprehension checks, incentives, and socioeconomic measurement. Claims about universal impulsivity require replication across substantially different social and economic environments Green & Myerson (2004) Gelino et al. (2024).
Equity also matters in the clinical use of discounting measures. The decision to take an immediate available resource may be adaptive under unstable conditions; assigning an adverse psychological label without considering risk can reproduce social bias. Research should distinguish locally rational choices under uncertainty from a stable preference for immediacy, and report how poverty, trauma, psychiatric comorbidity, and access to treatment affect associations. An interpretable clinical science explains behavior in its actual opportunity structure, using comparative measures to test rather than erase contextual explanations Kluwe-Schiavon et al. (2020) Audrain-McGovern et al. (2009).
Neural Explanation, Ecological Validity, and Competing Accounts
The strongest neural findings demonstrate value representation and interaction among systems during controlled decisions. McClure and colleagues described differential engagement for immediate versus delayed options; Kable and Glimcher linked activity to subject-specific value; Peters and Büchel showed that episodic simulation can modify intertemporal choices and accompanying coupling signals. Together these studies establish a rich neurocognitive research program. They do not furnish a single diagnostic brain signature of addiction, and reverse inference from regional activation to a psychological mechanism remains vulnerable to the many functions shared by those regions McClure et al. (2004) Kable & Glimcher (2007) Peters & Büchel (2010).
Alternative explanations include model-based evaluation of future outcomes, changes in expected reward reliability, valuation of consumable versus exchangeable rewards, habit learning, and state-dependent motivation. They can produce similar observable preference curves under restricted tasks yet diverge when expected delivery risk, reward type, or environmental alternatives are manipulated. Experiments should therefore create conditions where theories predict different choices. Neuroimaging can be most informative when coupled to individual behavioral models and out-of-sample tests rather than broad verbal mappings from a region to an alleged trait Green & Myerson (2004) Estle et al. (2007) Bickel et al. (2014).
Ecological validity is more than using realistic pictures or longer task delays. Addiction-related decisions recur over minutes, days, and years, often involving multiple interacting people and institutions. Trials with genuine consequences, experience sampling, prospective use measurements, and naturally occurring changes in opportunity can complement laboratory paradigms. Their inferential limitations must be stated: naturalistic correlations are vulnerable to confounding, while carefully controlled laboratory changes can lack direct relevance to prolonged treatment. A durable account of intertemporal choice requires both experimental identification and validated transport to lived settings Exum et al. (2023) Rung & Madden (2018).
Computational Models and the Artificial Era
Formal temporal-choice models can be implemented in artificial agents to examine the mathematical consequences of a specified valuation rule under changing reward schedules. Such simulations allow direct observation of preference reversals, commitment behavior, and policy changes when delay, risk, or available alternatives are modified. Their value is explanatory and methodological: the investigator knows the programmed rule and can compare simulated trajectories with human or animal data. Agreement in a narrow fitted curve is insufficient to infer identity of underlying psychological architecture, because distinct algorithms can produce equivalent choices Green & Myerson (2004) Bickel et al. (2014).
In the Artificial Era, human–artificial systems may mediate the representation of future goals through planning applications, reminders, simulations, and decision aids. The relevant psychological research question is whether particular interfaces alter human forecasting, task comprehension, anticipated outcomes, or real behavior. A digital assistant that presents concrete future scenarios could be evaluated against neutral planning support and ordinary clinical care, with attention to durable outcomes and user autonomy. These possibilities remain empirical research questions, and their effectiveness in treating addiction cannot be inferred from laboratory episodic future thinking alone Peters & Büchel (2010) Ye et al. (2022).
A scientifically productive cross-order approach distinguishes human participants, artificial computational models, and hybrid decision environments as different units of observation. Human clinical evidence concerns persons and their health outcomes; artificial simulations test specified decision policies; interface studies test effects of human–artificial interaction. Explicit labels for these units prevent a behavioral analogy from being mistaken for shared subjective experience or shared diagnosis. The broader horizon of Artificial Era research therefore supplies additional experimental instruments while preserving the validity requirements established for psychological and clinical claims Rung & Madden (2018) Gelino et al. (2024).
Research Priorities and a Program of Critical Tests
Priority one is to improve the comparability of measurement. Multi-site studies should standardize the time units, reward domains, incentive conditions, trial-level analysis, and criteria for non-systematic responses while retaining controlled variation that tests generalizability. Task scores should be accompanied by uncertainty estimates and relevant test-retest benchmarks. Direct comparisons of adjusting, fixed-choice, and experiential tasks within the same participants could explain the method-dependent pattern identified in treatment-outcome reviews and distinguish measurement artifacts from genuine differences in predictive capacity Johnson & Bickel (2008) Gelino et al. (2024) Exum et al. (2023).
Priority two is causal and developmental identification. Cohorts beginning before substance initiation, repeated assessments during changing use patterns, within-person withdrawal experiments, and randomized modifications of future-oriented cognition should be integrated through prespecified causal models. Investigators should compare predictions about onset, escalation, remission, and recurrence rather than collapsing them into a single substance-use variable. The divergent findings in adolescent twins and the later IMAGEN analysis make outcome specificity a central test instead of an inconvenient exception Isen et al. (2014) Dali et al. (2025) Field et al. (2006).
Priority three is translational validation. Studies should examine whether interventions that alter discounting also improve independently assessed substance-use outcomes, whether discounting change mediates those outcomes, and whether benefits persist beyond the laboratory. Evaluation must include practical accessibility, adverse effects, differential responses, and social determinants of opportunity. Comparative work across drugs and gambling should identify what is genuinely common and what depends on commodity characteristics, uncertainty, or disorder severity. Progress will be measured by replicated mechanisms and patient-relevant prediction, not by the multiplication of fitted parameters Rung & Madden (2018) Strickland et al. (2021) Ahluwalia et al. (2026).
Conclusion
Intertemporal choice and delay discounting provide an indispensable quantitative entry point into the psychology of addiction. Foundational choice experiments established that current smokers and participants with certain substance-use disorders often assign lower present value to delayed outcomes than comparison groups. Meta-analyses support a reproducible population-level association, though its magnitude differs between categorical and continuous analyses and across methods. Evidence also demonstrates that directly consumable rewards may be discounted more steeply than money even outside addiction, making the reward's properties an essential part of explanation Kirby et al. (1999) Bickel et al. (1999) MacKillop et al. (2011) Amlung et al. (2017) Estle et al. (2007).
The mature evidence base requires a distinction among association, temporal precedence, experimental manipulability, and clinical efficacy. Longitudinal cohorts yield informative but nonuniform predictions; experimental deprivation and episodic future thinking show that temporal choice can change; and systematic reviews of treatment outcomes report inconsistent prognostic utility. Neuroimaging identifies distributed valuation and future-simulation processes but does not support simple anatomical diagnoses. The construct is scientifically robust precisely when its boundary conditions, inferential limitations, and competing explanations are made explicit Isen et al. (2014) Dali et al. (2025) Peters & Büchel (2010) Exum et al. (2023) Gelino et al. (2024).
The resulting synthesis locates discounting within a broader organization of reinforcement, opportunity, expectation, development, affective state, and clinical context. Future experiments should discriminate among causal pathways, validate outcomes outside the laboratory, and compare human decision-making with specified computational models where scientifically useful. A fundamental account of addiction-related intertemporal choice is thus a theory-guided, evidence-constrained investigation of how rewards and goals acquire value across time, and how that value changes under identifiable conditions Bickel et al. (2011) Green & Myerson (2004) Rung & Madden (2018).
References
- Ahluwalia, Y., Negi, P. S., Singh, S., Sarkar, S., & Balhara, Y. P. S. (2026). Neurocognitive deficits associated with gambling behavior: A systematic review and meta-analysis of executive function, decision-making, delay discounting & reward sensitivity. Indian Journal of Psychological Medicine. Advance online publication. https://doi.org/10.1177/02537176261444941
- Ainslie, G. (1975). Specious reward: A behavioral theory of impulsiveness and impulse control. Psychological Bulletin, 82(4), 463–496. https://doi.org/10.1037/h0076860
- Amlung, M., Vedelago, L., Acker, J., Balodis, I., & MacKillop, J. (2017). Steep delay discounting and addictive behavior: A meta-analysis of continuous associations. Addiction, 112(1), 51–62. https://doi.org/10.1111/add.13535
- Audrain-McGovern, J., Rodriguez, D., Epstein, L. H., Cuevas, J., Rodgers, K., & Wileyto, E. P. (2009). Does delay discounting play an etiological role in smoking or is it a consequence of smoking? Drug and Alcohol Dependence, 103(3), 99–106. https://doi.org/10.1016/j.drugalcdep.2008.12.019
- Audrain-McGovern, J., Rodriguez, D., Tercyak, K. P., Epstein, L. H., Goldman, P., & Wileyto, E. P. (2004). Applying a behavioral economic framework to understanding adolescent smoking. Psychology of Addictive Behaviors, 18(1), 64–73. https://doi.org/10.1037/0893-164X.18.1.64
- Bickel, W. K., Jarmolowicz, D. P., Mueller, E. T., & Gatchalian, K. M. (2011). The behavioral economics and neuroeconomics of reinforcer pathologies: Implications for etiology and treatment of addiction. Current Psychiatry Reports, 13(5), 406–415. https://doi.org/10.1007/s11920-011-0215-1
- Bickel, W. K., Koffarnus, M. N., Moody, L., & Wilson, A. G. (2014). The behavioral- and neuro-economic process of temporal discounting: A candidate behavioral marker of addiction. Neuropharmacology, 76(Pt B), 518–527. https://doi.org/10.1016/j.neuropharm.2013.06.013
- Bickel, W. K., Odum, A. L., & Madden, G. J. (1999). Impulsivity and cigarette smoking: Delay discounting in current, never, and ex-smokers. Psychopharmacology, 146(4), 447–454. https://doi.org/10.1007/PL00005490
- Dali, G., Poulton, A., Banaschewski, T., Bokde, A. L. W., Desrivières, S., Flor, H., Garavan, H., Grigis, A., Heinz, A., Martinot, J.-L., Nees, F., Papadopoulos Orfanos, D., Poustka, L., Smolka, M. N., Hohmann, S., Vaidya, N., Walter, H., Whelan, R., Schumann, G., & Hester, R. (2025). Trajectories of delay discounting and smoking from adolescence to young adulthood. Drug and Alcohol Dependence, 277, Article 112955. https://doi.org/10.1016/j.drugalcdep.2025.112955
- Estle, S. J., Green, L., Myerson, J., & Holt, D. D. (2007). Discounting of monetary and directly consumable rewards. Psychological Science, 18(1), 58–63. https://doi.org/10.1111/j.1467-9280.2007.01849.x
- Exum, A. C., Sutton, C. A., Bellitti, J. S., Yi, R., & Fazzino, T. L. (2023). Delay discounting and substance use treatment outcomes: A systematic review focused on treatment outcomes and discounting methodology. Journal of Substance Use and Addiction Treatment, 149, Article 209037. https://doi.org/10.1016/j.josat.2023.209037
- Field, M., Santarcangelo, M., Sumnall, H., Goudie, A., & Cole, J. (2006). Delay discounting and the behavioural economics of cigarette purchases in smokers: The effects of nicotine deprivation. Psychopharmacology, 186(2), 255–263. https://doi.org/10.1007/s00213-006-0385-4
- Gelino, B. W., Schlitzer, R. D., Reed, D. D., & Strickland, J. C. (2024). A systematic review and meta-analysis of test-retest reliability and stability of delay and probability discounting. Journal of the Experimental Analysis of Behavior, 121(3), 358–372. https://doi.org/10.1002/jeab.910
- Green, L., & Myerson, J. (2004). A discounting framework for choice with delayed and probabilistic rewards. Psychological Bulletin, 130(5), 769–792. https://doi.org/10.1037/0033-2909.130.5.769
- Isen, J. D., Sparks, J. C., & Iacono, W. G. (2014). Predictive validity of delay discounting behavior in adolescence: A longitudinal twin study. Experimental and Clinical Psychopharmacology, 22(5), 434–443. https://doi.org/10.1037/a0037340
- Johnson, M. W., & Bickel, W. K. (2008). An algorithm for identifying nonsystematic delay-discounting data. Experimental and Clinical Psychopharmacology, 16(3), 264–274. https://doi.org/10.1037/1064-1297.16.3.264
- Kable, J. W., & Glimcher, P. W. (2007). The neural correlates of subjective value during intertemporal choice. Nature Neuroscience, 10(12), 1625–1633. https://doi.org/10.1038/nn2007
- Kirby, K. N., & Petry, N. M. (2004). Heroin and cocaine abusers have higher discount rates for delayed rewards than alcoholics or non-drug-using controls. Addiction, 99(4), 461–471. https://doi.org/10.1111/j.1360-0443.2003.00669.x
- Kirby, K. N., Petry, N. M., & Bickel, W. K. (1999). Heroin addicts have higher discount rates for delayed rewards than non-drug-using controls. Journal of Experimental Psychology: General, 128(1), 78–87. https://doi.org/10.1037/0096-3445.128.1.78
- Kluwe-Schiavon, B., Viola, T. W., Sanvicente-Vieira, B., Lumertz, F. S., Salum, G. A., Grassi-Oliveira, R., & Quednow, B. B. (2020). Substance related disorders are associated with impaired valuation of delayed gratification and feedback processing: A multilevel meta-analysis and meta-regression. Neuroscience & Biobehavioral Reviews, 108, 295–307. https://doi.org/10.1016/j.neubiorev.2019.11.016
- MacKillop, J., Amlung, M. T., Few, L. R., Ray, L. A., Sweet, L. H., & Munafò, M. R. (2011). Delayed reward discounting and addictive behavior: A meta-analysis. Psychopharmacology, 216(3), 305–321. https://doi.org/10.1007/s00213-011-2229-0
- McClure, S. M., Laibson, D. I., Loewenstein, G., & Cohen, J. D. (2004). Separate neural systems value immediate and delayed monetary rewards. Science, 306(5695), 503–507. https://doi.org/10.1126/science.1100907
- Odum, A. L., & Rainaud, C. P. (2003). Discounting of delayed hypothetical money, alcohol, and food. Behavioural Processes, 64(3), 305–313. https://doi.org/10.1016/S0376-6357(03)00145-1
- Peters, J., & Büchel, C. (2010). Episodic future thinking reduces reward delay discounting through an enhancement of prefrontal-mediotemporal interactions. Neuron, 66(1), 138–148. https://doi.org/10.1016/j.neuron.2010.03.026
- Peters, J., & Büchel, C. (2011). The neural mechanisms of inter-temporal decision-making: Understanding variability. Trends in Cognitive Sciences, 15(5), 227–239. https://doi.org/10.1016/j.tics.2011.03.002
- Petry, N. M., & Casarella, T. (1999). Excessive discounting of delayed rewards in substance abusers with gambling problems. Drug and Alcohol Dependence, 56(1), 25–32. https://doi.org/10.1016/S0376-8716(99)00010-1
- Rung, J. M., & Madden, G. J. (2018). Experimental reductions of delay discounting and impulsive choice: A systematic review and meta-analysis. Journal of Experimental Psychology: General, 147(9), 1349–1381. https://doi.org/10.1037/xge0000462
- Strickland, J. C., Lee, D. C., Vandrey, R., & Johnson, M. W. (2021). A systematic review and meta-analysis of delay discounting and cannabis use. Experimental and Clinical Psychopharmacology, 29(6), 696–710. https://doi.org/10.1037/pha0000378
- Ye, J.-Y., Ding, Q.-Y., Cui, J.-F., Liu, Z., Jia, L.-X., Qin, X.-J., Xu, H., & Wang, Y. (2022). A meta-analysis of the effects of episodic future thinking on delay discounting. Quarterly Journal of Experimental Psychology, 75(10), 1876–1891. https://doi.org/10.1177/17470218211066282
- Yoon, J. H., Higgins, S. T., Heil, S. H., Sugarbaker, R. J., Thomas, C. S., & Badger, G. J. (2007). Delay discounting predicts postpartum relapse to cigarette smoking among pregnant women. Experimental and Clinical Psychopharmacology, 15(2), 176–186. https://doi.org/10.1037/1064-1297.15.2.186