Reward and executive control imbalances in adolescence: measurement and implications for our understanding of alcohol use

1 Introduction

Adolescence is a developmental period characterized by significant neurobiological, psychological, and behavioral change, including increased novelty-seeking and risk-taking in both non-human mammals (1) and humans worldwide (2). Generally recognized as occurring during the second decade of life (3), adolescence can be categorized into early (ages ∼10–14 years), middle (ages ∼15–17 years), and late (ages ∼18–25 years) stages [e.g., (4, 5)] based on physical and psychosocial factors. While many forms of adolescent risk-taking are adaptive (6–8), risk-taking is maladaptive when it involves socially unacceptable, illegal, or dangerous actions with high likelihoods of unintentional injury and negative consequences, both in the short and long terms. Substance use, including alcohol consumption, is a particularly concerning form of risk-taking that often begins during adolescence. Although prevalence rates have decreased in the United States (US), national estimates of adolescent alcohol use (AU) indicate that 14.6% of youth aged 12–20 years currently consume alcohol, with 8.6% binge drinking (i.e., consuming 4+/5 + drinks within a two-hour period for females/males) at least once in the past month and 1.7% binge drinking on five or more days in the past month (160). The acute consequences of AU during adolescence can be severe. More than a quarter of deaths among young adults in the US are attributable to AU (161). Adolescent AU is also linked to poorer academic performance and decreased educational attainment (9), mental health difficulties and social harm (10), and is strongly associated with engaging in other forms of risky behavior, including the use of other substances, unprotected sex, driving under the influence, and physical altercations (11). Prolonged repercussions of adolescent AU extend well into adulthood (12–14) as the early onset of AU, as well as the escalation of drinking in adolescence, are risk factors for the development of alcohol-related problems in adulthood, including alcohol dependence and AU disorder (15–17). Understanding the mechanistic factors that contribute to these trends is critical for developing targeted prevention and intervention strategies, which can mitigate the long-term adverse effects of early AU.

Several heuristic frameworks, including dual systems models of adolescent development, postulate roles for complementary systems in promoting vulnerability to adolescent risk-taking broadly (18–22) and addictive behaviors and alcohol misuse more specifically (23–26). Within dual-process frameworks, reactive processes (e.g., incentive motivation, reward sensitivity, attentional bias, and approach-avoidance behavior) provide bottom-up drive, while reflective processes (e.g., inhibitory control, working memory, and cognitive flexibility) exert top-down behavioral control. Reactive and reflective processes are not unitary and interact dynamically; bottom-up systems can modulate and optimize one another, prime, activate, or modulate top-down processing, and support top-down regulation by feeding forward goal-relevant information (27). The interplay within and between these multifaceted systems ultimately facilitates self-regulation, or the ability to achieve control over motivational impulses for the purposes of adaptation/goal-striving, thereby encompassing both the recruitment of reflective executive control processes and the extent to which, in any given context, internal reactive states, as well as externally-triggered drives, are regulated, ignored, and/or recruited (19, 27, 28). It is worth noting that there are several versions of the “dual systems model” with variations in the conceptualized systems at play and their impacts on adolescent decision making. These variations have been discussed by others [e.g., (29)].

Over the past several decades, investigations have assessed the developmental trajectories of reward and control systems, suggesting that they are distinct. The reactive reward system is primarily driven by subcortical regions such as the striatum and amygdala, is heavily influenced by dopaminergic activity (30, 31), and can be indexed by self-reported and behavioral measures of reward sensitivity. Though not a direct measure of reward sensitivity, sensation seeking is an overlapping construct that is frequently indexed via self-report measures and is often measured in studies that aim to assess the development of the reward system. Together, these reactive processes are hypothesized to reach maximal functional levels in adolescence (32–35). In contrast, the reflective executive control system, which is governed mainly by prefrontal and other higher association regions, enables the execution of functions such as impulse control, deliberative decision-making, and future planning (19, 27, 36). Various self-report measures and behavioral assessments index executive control processes, which gradually mature well into young adulthood (32, 37–42).

While this literature coheres to suggest largely linear increases in executive control from childhood to adulthood and nonlinear patterns of change in reward sensitivity across the same span, few longitudinal investigations have assessed the developmental trajectories of both reward and control-related systems within the same individuals. Those that have done so provide support for the idea of a developmental “imbalance” between systems, including evidence of earlier peaks in sensation seeking relative to impulsivity (34, 43), earlier maturational asymptotes for reward sensitivity relative to executive functions (32), and similarly patterned volumetric changes in neural gray matter within brain regions that regulate affective vs. executive processes (44). This discrepancy in maturational trajectories is posited to create a window of vulnerability in early to mid-adolescence where heightened sensitivity to rewards and underequipped executive control mechanisms coexist, predisposing adolescents in reward-salient contexts toward risky decision-making (6, 18, 19, 21, 22).

2 Independent associations with alcohol use

During this period of vulnerability, either heightened reward sensitivity or lessened executive control may relate to specific aspects of real-world risk-taking, such as alcohol use. Indeed, there is substantial evidence of the contributions of reward- and control-based processes, when considered separately, to adolescent AU initiation and patterns of use. Longitudinal increases in reward sensitivity, reduced reward-related brain morphology, and heightened BOLD activation in reward-processing regions are associated with the initiation of substance use (including AU) as well as binge drinking initiation in adolescence (45–47). Higher reward sensitivity and reward-related neural reactivity are also concurrently associated with higher levels of adolescent AU (48–52). Additionally, higher levels of self-reported reward sensitivity (53), aberrant patterns of reward-related neural activity (54), and rapid escalation of behaviorally-indexed reward sensitivity (55) in early adolescence predict future escalations and increased frequencies and quantities of AU later in adolescence. Similarly, higher levels of adolescent sensation seeking are associated with early AU onset and higher concurrent levels of adolescent AU (56–58). Adolescent levels of, and increases in, sensation seeking are also predictive of future AU (57, 59, 60). These findings cohere with the dual systems model in suggesting that increased levels of reward sensitivity represent vulnerabilities for alcohol misuse.

As expected, weaknesses in executive function also predict AU initiation (61), are concurrently associated with higher levels of adolescent AU (62, 63), and are prospectively associated with higher levels of use, including binge drinking (61, 64–66), in adolescence. Meta-analysis also suggests that when modeled separately, higher reward sensitivity or sensation seeking, and lower executive control are associated with higher levels of alcohol consumption and problematic drinking, both concurrently and prospectively, in adolescent samples (67).

However, these processes have seldom been measured together within the same individuals, and much of the extant research has investigated the independent associations between reward or control processes and AU outcomes without consideration of the joint or even interactive influence of both systems. Validation of the dual systems framework requires these joint influences to be dynamically assessed as development proceeds. Central to the dual-process framework of adolescent risk-taking is the assertion that the functional “imbalance” between reward and executive control systems varies with development, is maximal during adolescence, and is implicated in adolescent risk-taking (18–20, 22). Although developmental differences are paramount to the proposed mechanisms, many investigations are cross-sectional or have quantified imbalances at single instances in time without consideration of their developmental trajectories (68).

The present theoretical review aims to synthesize recent research centered on the developmental imbalance between reward and executive control systems and its implications for AU and related outcomes during adolescence, focusing on investigations spanning the past two decades that have modeled additive, interactive, or within-person imbalances of reward and executive control systems. Potential studies for discussion were identified from APA PsychINFO and indexed as published or in press from January 2004 to July 2024. Studies were initially included if available in English, published in a peer-reviewed journal, and conducted with human participants. The following search string was used: (adolescen) AND (reward OR sensation) AND (executive OR cognitive OR impuls OR inhibit) AND alcohol. Additionally, we included terms capturing theoretical models or neurodevelopmental mechanisms, such as: (dual systems AND alcohol) OR (dual systems AND risk) OR (adolescen* AND imbalance AND alcohol) OR (adolescen* AND mismatch AND alcohol) OR (cognition AND emotion AND interaction AND adolescen* AND alcohol). The initial search yielded 571 articles, which were screened by the first author. Investigations were included if they contained self-reported, behavioral, neuroimaging, or psychophysiological measures of both reward and executive control systems in the same sample. Investigations were excluded if they (1) were unpublished dissertations, reviews, commentaries, books, case studies, or animal studies (2) did not include indexes of both reward and executive control systems, (3) analytic methods did not include both measures in a single model, (4) only included data for participants 18 years of age or older, (5) included a measure of reward sensitivity that was conflated with risk-taking (e.g., a latent measure of reward sensitivity derived from performance on a risk-taking/gambling task), (6) included a single measure indexing both reward and executive control processes (e.g., impulsive sensation seeking, or a single latent measure of sensation seeking and impulsivity), (7) included a measure of substance use that did not include AU, or (8) included a measure of risk-taking that did not include AU. Figure 1 provides a visual depiction of the review process. Table 1 provides a summary of the resulting 40 articles that are reviewed herein.

Table 1. Measurement details of all studies included in review.

AU, alcohol use; AUD, alcohol use disorder; BAS, behavioral approach system (BAS scale); BD, binge drinking; BIS, Barrat impulsivity scale; BL, baseline; BRIEF, behavior rating inventory of executive function; CNLSY, child and young adult supplement of the national longitudinal survey of youth 79 cohort; CSC, climate schools combined; EATQ-R, early adolescent temperament questionnaire revised; EF, executive function; EPQ, eysenck personality questionnaire; GCM, growth curve modeling; GNG, Go/No-Go task; HED, heavy episodic drinking; HSD, health school and drugs; I₇, impulsiveness and venturesomeness questionnaire; IMP, impulsivity; IYDS, international youth development study; IVE, Eysenck's impulsiveness-venturesomeness-empathy questionnaire; J-EIQ, junior eysenck impulsiveness questionnaire; LTS, longitudinal twin study; MID, monetary incentive delay task; NAcc, nucleus accumbens; NCANDA, National Consortium on Alcohol and Neurodevelopment in Adolescence; NIMH, National Institute of Mental Health; NLSY79, National Longitudinal Study of Youth 79; PFC, prefrontal cortex; QXF, quantity-by-frequency; SEM, structural equation modeling; SS, sensation seeking; SSS, sensation seeking scale; SST, stop signal task; SURPS, substance use risk profile scale; SWM, spatial working memory; TPQ, tridimensional personality questionnaire; UPPS, urgency-(lack of)premeditation-(lack of)perseverance-sensation seeking impulsive behavior scale; UPPS-P, urgency-(lack of)premeditation-(lack of)perseverance-sensation seeking-positive urgency impulsive behavior scale; USA, United States of America; W, wave; ZKPQ, Zuckerman-Kuhlman Personality Questionnaire. Baseline age range in years presented or mean age, standard deviation. Study design lists if the analysis presented include concurrent (C) and/or prospective (P) associations between predictors and outcome measures.

Indicates that studies also included additive modeling.

3 Additive associations of reward sensitivity and executive control with alcohol use

The previously described studies have considered each system in isolation and, as such, are limited in their ability to assess the combined influence of an individual's reward and executive control systems, both in terms of their levels and developmental trajectories, on salient outcomes. In contrast, additive modeling that considers both reward and executive control processes simultaneously in statistical approaches allows for a more nuanced understanding of how these distinct, though related, systems contribute to outcomes of interest. In these sorts of additive models, variables indexing reward and executive control processes are modeled concurrently in the same statistical models, thereby quantifying the unique effect of one variable while controlling for the other. All reviewed studies using additive modeling approaches and a summary of their findings can be found in Table 2.

Table 2. Summary of additive modeling findings.

AU, alcohol use; BD/HED, binge/heavy episodic drinking; EC, executive control; F, frequency; ID, mild intellectual disability; Q, quantity; RS, reward sensitivity; S, status; SS, sensation seeking; SU, substance use. Summary of findings for investigations using additive modeling techniques. Results are organized by use outcome and presented by timeframe of adolescence assessed, including early (ages ∼10–14 years), mid (ages ∼15–17 years), and late (ages ∼18 + years) stages. Studies are listed depending on whether they found that measures of both sensation seeking and executive control, reward sensitivity and executive control, sensation seeking alone, reward sensitivity alone, executive control alone, or neither were associated with use patterns. Special characteristics of sample or findings are noted in parentheses or by symbols.

Indicates developmental changes implicated.

Indicates substance naïve participants at baseline.

Indicates EC related to shared externalizing variance and SS specifically related to substance misuse.

Only a couple of studies found neither reward nor control-related processes to be related to adolescent AU tendencies when modeled at the same time, including one report that focused on the retrospective age of AU initiation (69) and another that assessed predictors of concurrent binge drinking frequency (70). However, most other extant investigations found support for reward and control-related systems as predictors of AU initiation, use rates, binge/heavy episodic drinking, and problematic drinking during adolescence.

A cluster of investigations found that after controlling for the effects of reward-related processes, diminished executive control processes alone were significantly associated with elevated adolescent AU. Self-reported inhibitory control predicted alcohol use initiation in alcohol naïve youth (71) and recent alcohol use status (i.e., the consumption of any alcohol in the past 6 months) in young adolescents (72, 73). Executive control processes measured in early to middle adolescence, including self-reported lack of planning and inhibitory control, as well as behaviorally indexed executive functioning, were uniquely associated with concurrent rates of AU, including recent quantity and frequency of use (74) and binge/heavy episodic drinking (72, 73). Similarly, low levels of self-reported inhibitory control predicted higher levels of future rates of use and binge/heavy episodic drinking in early to middle adolescence (72).

There was also some evidence supporting the notion that after controlling for reward-related processes, lower levels of premorbid executive control (i.e., levels measured before the initiation of AU) predicted higher levels of AU later in adolescence (75). Importantly, weakness in executive control, including high self-reported impulsivity and challenges with behaviorally measured executive functions, appeared to be consistently associated with concurrent and prospective problematic AU in adolescence, including alcohol-related harms (72, 73, 76) and symptoms of AU disorder (74).

In contrast, some investigations suggested that only reward-related processes were significantly associated with adolescent AU patterns after controlling for executive control processes. Higher levels of self-reported reward sensitivity alone were related to earlier retrospective ages of AU initiation (77), and reduced brain activation during reward processing in substance-naïve early adolescents predicted more frequent use of alcohol during middle adolescence (78). Additionally, in males only, higher levels of reward drive were associated with substance use, including recent AU quantity and frequency, in early adolescence (79). Relatedly, higher levels of self-reported sensation seeking were related to prospective AU initiation in alcohol-naïve youth (80), recent AU status in early adolescence (69) with evidence of some sex differences in these associations (81), concurrent frequency of alcohol consumption in late adolescence (74), and prospective rates of use in a normative sample (73) and in a sample with elevated externalizing challenges (82). Of note, higher levels of self-reported sensation seeking, even after considering executive control abilities, were associated with adolescents' tendencies to partake in binge/heavy episodic drinking presently (76) and in the future (73, 83), including in high-risk populations (84, 85).

Finally, a large body of literature implicated both reward and executive control processes in adolescent AU patterns, though most of these investigations used measures of sensation seeking to index reward system activity. The few studies that more directly measured reward sensitivity indicated that heightened reward sensitivity and diminished executive control were both concurrently associated with heightened AU (86) as well as substance use profiles (79). Both reward sensitivity and executive control prospectively predict higher levels of AU, with the prospective effect of reward drive being through positive alcohol expectancies (87), from early to late adolescence. Relatedly, higher levels of sensation seeking and lower levels of executive control have both been associated with AU initiation (72, 73, 88), higher concurrent AU rates in early to middle adolescence (89–91), AU status in early adolescent females (81), higher concurrent binge/heavy episodic drinking in early to middle adolescence (91, 92), and higher concurrent problematic AU use in early adolescence through adulthood in individuals with mild to borderline intellectual disability (93) and typically developing late adolescents (74). Similarly, higher levels of sensation seeking and lower levels of executive control both prospectively predicted higher AU rates (94) and problematic use (73) in early to middle adolescence, as well as higher prospective AU rates across middle to late adolescence (74, 95).

Importantly, few prospective investigations of future rates of AU have considered developmental changes in reward and executive control processes in tandem through additive modeling. On the one hand, Khurana and colleagues (75) demonstrated that in early adolescents, change in self-reported impulsivity over four years was related to AU escalation, while change in self-reported sensation seeking was not. In contrast, in a particularly noteworthy investigation, Quinn & Harden (95) analyzed data from a large sample (n = 5,632) of individuals aged 15–26 years who were assessed biennially up to six times. Multivariate latent growth curve modeling demonstrated that latent self-reported sensation seeking and impulsivity were positively associated with baseline AU frequency and that greater baseline levels of sensation seeking predicted steeper increases in AU (95). Moreover, developmental changes in sensation seeking and impulsivity related to the escalation of AU, such that individuals who decreased less steeply in impulsivity and sensation seeking increased more steeply in AU frequency (95).

Most investigations of the joint associations of reward- and control-based processes with composite measures of substance use, including measures of AU quantity and/or frequency, simultaneously implicate both reward- and control-based processes, though not all do (84). In cross-sectional and prospective studies, both elevated reward and diminished executive control processes appear to contribute to more severe substance use profiles in early (79, 96) and middle (97) adolescence, potentially in sex-specific manners (79, 97). Results suggest that higher levels of early adolescent sensation seeking might be a broad risk factor for substance use, including AU, while higher levels of sensation seeking coupled with higher levels of impulsivity in early adolescence might be a risk factor for more severe polysubstance use during early-to-middle adolescence (98). Moreover, executive control processes may broadly relate to externalizing tendencies, while reward-related processes may be uniquely associated with substance misuse (99, 100).

In aggregate, the reviewed literature regarding the additive effects of reward and control systems yields inconsistent findings. While few studies find that neither reward or executive control process are related to adolescent AU, some implicate only reward processes after accounting for executive control, some implicate only executive control processes after accounting for reward processes, and many implicate both processes significantly in initiation, concurrent use, and prospective alcohol and substance use across adolescence and in different populations. Premorbid relative weaknesses in executive control (75) and heightened reward sensitivity (78) have both been shown to confer risk for later patterns of misuse in adolescence. The disparate pattern of findings observed across studies may be interpreted to suggest that risk factors for AU vary throughout adolescence or as use severity progresses. While sensation seeking consistently emerges as a broad risk factor for AU and substance use, its role is suggested—though not formally tested—to be moderated by the strength of executive control processes, as evidenced by differences in how these factors predict the severity and trajectory of substance use, particularly in high-risk populations (96, 98). Weaknesses in executive control, including high impulsivity and challenges with behaviorally measured executive functions, also appear to be consistently associated with problematic AU in adolescence. Importantly, though few investigations assessed associations between developmental changes in reward and control-related systems and use patterns, the limited extant research suggests that individual heterogeneity in these developmental processes may account for differences in AU patterns among adolescents (75, 95).

There are several limitations that must be considered before drawing strong conclusions from additive models. First, there is apparent variability in findings across samples, timeframes assessed, and the periods of adolescence measured. This variability underscores the importance of methodological rigor and the consideration of developmental timing and potential sex-based differences in understanding how reward sensitivity and executive function contribute to the progression of adolescent AU. Second, the reviewed work is heavily reliant on measures of self-reported sensation seeking and impulsivity (which is actually a measure of executive control), as opposed to reward sensitivity and/or behaviorally indexed measures. Third, several of the findings are from the same studies, with cross-sectional and longitudinal analyses conducted on several outcome measures in the same sample [e.g., (72–74)] or are findings from separate investigations conducted in overlapping samples (e.g., the Preventure and IMAGEN studies). Fourth, while many findings from the additive modeling literature suggest that reward and executive control processes independently, though simultaneously, predict various aspects of adolescent alcohol involvement, this modeling approach does not provide insights into how the effect of one system might moderate the effect of the other, as is hypothesized by dual-process models and suggested by some of the findings. Indeed, regression approaches evaluating the unique effects of each system do not take into account an individual's standing on one variable relative to the other variable and, therefore, do not provide information on the “imbalance” between these systems and whether the imbalance, if present, is meaningful (68).

4 Interactions between reward and executive control processes

Several investigations have examined interactions between reward and executive control systems in relation to adolescent AU via moderated regression and related approaches. All reviewed studies using moderated regression approaches and a summary of their findings can be found in Table 3.

Table 3. Summary of moderated regression findings.

AU, alcohol use; EC, executive control; RS, reward sensitivity; SS, sensation seeking; SU, substance use. Summary of findings for investigations using moderated regression techniques. Results are organized by use outcome and presented by timeframe of adolescence assessed, including early (ages ∼10–14 years), mid (ages ∼15–17 years), and late (ages ∼18 + years) stages. Studies are listed depending on whether they found moderation effects between executive control measures and sensation seeking or reward sensitivity or no moderation effects. Special characteristics of sample or findings are noted in parentheses.

First, cross-sectional investigations show mixed evidence of the moderation effects of executive control on reward-related processes. While one study in early adolescents in the USA suggested that executive processes (inhibitory control) moderated the effects of high levels of reward sensitivity on retrospective age of substance use initiation (101), another study of early-to-late adolescents in Berlin failed to find support for the moderating effect of effortful control on heightened reward sensitivity and retrospective alcohol use initiation (77). However, in this same sample of Berlin adolescents, the concurrent associations between higher levels of reward sensitivity and increased AU rates were moderated by effortful control (77). In contrast, investigations in early (101) and late (74) adolescent samples in the USA did not find that executive processes (inhibitory control in the former sample and self-reported impulsivity and a composite index of executive functioning in the latter sample) moderated the relationship between higher reward processes (reward sensitivity in the former sample and sensation seeking and positive/negative subjective effects of alcohol in the latter sample) and greater concurrent substance use severity (101) or alcohol involvement (74). Unfortunately, inconsistent cross-sectional findings make it challenging to draw conclusions regarding the potential moderation effect of executive control on reward-based processes. Contradictory cross-sectional findings could be due to sample-specific features, cultural influences, the period of adolescence sampled (early vs. early-to-late vs. late), and various measurement differences (sensation seeking vs. reward sensitivity, behavioral vs. self-reported, composite measure vs. single index).

Extant prospective investigations do not lend support to the notion that executive processes moderate heightened reward processes' impact on future alcohol use patterns. In early-to-middle adolescence, baseline approach tendencies toward alcohol (measured via the Stimulus Response Compatibility task) positively predicted subsequent increases in AU frequency one year later, though this association was not moderated by working memory capacity (performance on self-ordered pointing task) (102). Similarly, in a Canadian sample of low socioeconomic status males, inhibitory control measured in early adolescence failed to moderate the impact of sensation seeking on future AU, drugs tried, and substance use problems during middle adolescence, though sensation seeking was associated with higher cannabis use at high levels of impulsivity (94).

Overall, these investigations present a fragmented understanding of the interactive effects of reward and executive control processes on adolescent alcohol use due to their reliance on cross-sectional designs and inconsistent findings. Moreover, low statistical power (94), limited variance in alcohol use measures, idiosyncrasies in tasks used, short intervals between assessments, and lack of developmental assessment of predictors (102) are notable limitations. Indeed, even the cited longitudinal investigations (94, 103) did not present data on developmental changes in reward and control systems and potential moderation effects over time. Thus, there is not yet strong evidence to suggest that executive control processes moderate the relationship between reward processes and patterns of adolescent AU.

Investigations that have focused on other forms of risk-taking across adolescence have implicated moderation effects of control processes on approach behaviors. For example, in a longitudinal sample of early adolescents, high levels of behaviorally measured reward sensitivity (via the Point Scoring Reaction Time Task) prospectively predicted high levels of rule-breaking one year later but only at low levels of behavioral inhibitory control (via stop signal task performance) (104). Findings from a cross-sectional sample of early-to-late adolescents revealed that adolescents who reported difficulties with regulatory processing coupled with heightened reactive processes were significantly more likely to experience loss of control over eating (105). Explorations of interaction effects during emerging adulthood further suggest that moderation between the two systems may play a role in the use of other substances and consequences from substance use. For example, a cross-sectional investigation of college-aged individuals found that, after controlling for sex, age, and race, lower self-reported impulsivity enhanced the effect of higher self-reported sensation seeking on more frequent drug use (cannabis and illicit drugs) and alcohol and drug use consequences, but was unrelated to individual differences in antisocial behaviors (aggression and rule-breaking behavior) and past-year AU rate (106).

Despite some apparent consistencies across these studies, moderation effects are not always supported. For example, in a large (n = 5,200) cross-sectional multi-country sample spanning early adolescence to early adulthood (ages 10–30 years), Duell and colleagues found that, after controlling for country, sex, socio-economic status, and IQ, greater reward-seeking (self-reported sensation seeking and performance on a modified Iowa Gambling Task that indexed reward sensitivity) and lower executive control (self-reported planning and behaviorally indexed inhibitory control and response inhibition) demonstrated independent positive effects on risk-taking (self-reported risk preference and performance on the Balloon Analogue Risk Task and Stoplight Task) that did not vary by age. Moreover, the effect of reward-seeking on risk-taking was not moderated by executive control nor was there evidence for a three-way interaction with age (107). The authors noted that the cross-sectional nature of their investigation may have impeded their ability to identify significant interaction effects and pointed to a need for future research to examine changes in the development of reward-seeking and executive control in relation to risk-taking.

Together, while moderation models provide some evidence that imbalances between reward and executive control systems might impact adolescent risk-taking, including alcohol and substance use patterns, findings are markedly variable, and longitudinal assessments that have assessed the impacts of developmental interactions on these outcomes are starkly lacking. Additionally, though several investigators have asserted that assessing interactions between reward and executive control systems is a valid test of the imbalance hypothesis (101, 103, 104, 106–108), others have discussed the methodological limitations of moderated regression approaches (68). First, as noted, several of these regression approaches do not provide information on the development of either system. To do so within a regression framework would require a three-way interaction with age to be tested, as in Duell et al. (107). However, most investigations have not done so, likely because large well-powered samples are needed to detect these sorts of effects. Such an approach would be most informative using multi-wave longitudinal assessments that can capture within-person three-way moderation effects of age. Second, and more concerning, is that interaction terms in moderated regression approaches might not capture the imbalance between systems. As discussed in further detail elsewhere (68, 109), moderated regression approaches inadequately estimate the hypothesized differences between reward and executive processes and are not sufficient substitutes for difference score approaches. Specifically, when continuous measures are indexed, product terms used in moderated regression approaches do not represent the effect of differences between component measures (109) and individuals may have different predicted outcome values (e.g., AU quantity) despite having the same direction and extent of imbalance between reward and executive control systems (e.g., the same value for a difference score between reward sensitivity and impulsivity) (68, 109). In their review of methods used in developmental dual processes investigations of adolescent risk-taking, Meisel and colleagues (68) provided a compelling example of this problem and urged researchers to exercise caution when interpreting interaction results.

Given these limitations, the present body of literature using interaction models is limited in its ability to support or refute dual systems models and their relevance for adolescent risk-taking and substance use patterns. The lack of longitudinal data on how these interactions dynamically evolve over time is a limitation given the patterns of age-related change that are proposed by the dual systems framework. “Within-person” methods, such as standardized or latent difference score approaches or latent class or growth mixture modeling, as opposed to investigations of these systems in isolation or moderated regression approaches, may more appropriately capture developmental imbalances between systems within the same individual and impacts of these imbalances on risk taking and substance use patterns across adolescence.

5 Within-person models of reward and executive control processes

“Within-person” analytic designs, which can assess relations between components within the same person at a given time point (110) and how these relations change over time, particularly those that incorporate modeling of developmental changes in both reward and executive control systems across adolescence, are well suited to elucidate the role that developmental imbalances between reward and executive control systems might play in adolescent AU patterns. Reviewed studies using within-person modeling approaches and a summary of their findings can be found in Table 4.

Table 4. Summary of within-person modeling findings.

AU, alcohol use; EC, executive control; RS, reward sensitivity; SS, sensation seeking; SU, substance use. Summary of findings for investigations using various within-person techniques. Results are organized by outcome and presented by timeframe of adolescence assessed, including early (ages ∼10–14 years), mid (ages ∼15–17 years), and late (ages ∼18 + years) stages. Studies are listed depending on whether they found support for associations between sensation seeking and executive control or reward sensitivity and executive control imbalances and outcomes. Special characteristics of sample or findings are noted in parentheses or by symbols.

Indicates developmental processes implicated.

Indicates substance naïve participants at baseline.

As an example of this approach albeit in the context of a cross-sectional study of a large (n = 15,839) sample of individuals spanning early adolescence to early adulthood (ages 12–28 years), Vazsonyi and Ksinan (111) used a standardized difference score approach to demonstrate that after controlling for age-related effects and other covariates, higher discrepancy scores (more self-reported sensation seeking relative to self-reported impulsivity) were associated with elevations in concurrent norm-violating behaviors. Age moderation was not tested, limiting the conclusions that could be drawn regarding developmental changes. Though not yet used to predict risk-taking behavior or AU, others have used standardized difference scores to investigate developmental changes of within-person imbalances across adolescence. In a community sample (n = 166) of individuals spanning early adolescence to early adulthood (ages 9–30 years) who were assessed biennially up to five times, adolescence was shown to be, on average, a unique period of heightened self-reported reward sensitivity relative to laboratory-measured executive control, whereas early adulthood was a period of heightened executive control relative to reward sensitivity (32). In this study, substantial individual variation in initial levels of imbalance and developmental change was observed.

In work specifically related to adolescent AU patterns, others have used within-person approaches to cross-sectionally investigate the reward-executive control imbalance. Using cross-sectional data from a Dutch sample of early to middle adolescents (n = 341; ages 13–17 years), Van Malderen and colleagues (112) used model-based clustering to identify three behavioral profiles: an “impulsive/under-controlled” profile (high impulsivity and reward sensitivity), an “anxious” profile (moderate impulsivity and reward sensitivity, and high punishment sensitivity), and a “protective” profile (low impulsivity, low reward sensitivity, and low punishment sensitivity). The “impulsive/under-controlled” profile was characterized by the highest prevalence and severity of cannabis use, the most severe AU, and the most addictive behaviors. Additionally, the “impulsive/under-controlled” and “anxious” profiles showed the highest binge eating scores (112). Similar to Vazsonyi and Ksinan (111), this work utilized a within-person approach to suggest that high levels of executive control can be protective against high approach tendencies.

Other work has focused on substance use initiation. Malmberg and colleagues (80) conducted a latent profile analysis of substance-naïve adolescents (ages 11–14) who were studied longitudinally based on their self-reported hopelessness, anxiety sensitivity, impulsivity, and sensation seeking at baseline. There were no significant differences in substance use initiation (alcohol, smoking/tobacco, and/or cannabis status) 20 months later based on group membership. Various aspects of the study's measurement strategy and analytical techniques may account for the negative findings. Specifically, substance use was measured dichotomously in terms of whether participants were users or nonusers of each substance (113). In addition, the use of posterior probabilities (80) may have lowered the likelihood of finding significant parameter estimates. Moreover, the authors noted that their profiled subgroups might be distinctive for other substance-related behaviors, such as escalation of use. Crucially, profile membership was based on reported behavior at one assessment wave, much like Van Malderen and colleagues' (112) and Vazsonyi and Ksinan's (111) work, which fails to account for developmental changes in reward-based and executive control processes in relation to risk-taking behaviors.

In a unique example of a within-person investigation that quantified developmental changes, Mills and colleagues (44) assessed a small group of individuals (n = 33) whose imbalance in structural brain development was prospectively assessed from late childhood into early adulthood and then related to retrospective reports of adolescent risk-taking behaviors provided during adulthood. The authors found evidence for a mismatch in structural brain development such that subcortical volumes (nucleus accumbens and amygdala) matured earlier than prefrontal volumes at both group and individual levels (44). However, the correlations between adolescent structural neurodevelopmental imbalance status and retrospective adolescent risk-taking (including behaviors such as illicit drug use, AU, and unsafe sexual behavior) reported at the close of the study were not significant (44). This investigation's provision of evidence of imbalances between cortical and subcortical structural development was a significant contribution to the literature. However, considerable caution is advised when interpreting the correlations with adolescent risk-taking, given that retrospective assessments of behaviors were collected several years after the behaviors being measured; moreover, the sample size is very small (44).

In a novel approach that used longitudinal modeling that accounts for within-person differences between reward and control systems and developmental processes in relation to prospective substance use, Meisel and colleagues (68) conducted latent difference score growth curve modeling in a longitudinal sample of adolescents. In their investigation, assessments of behavioral response inhibition and reward sensitivity were used to create latent difference scores measuring the imbalance between reward and control systems. The authors did not find significant associations between initial levels of imbalance in early adolescence (intercept at age 12) or change (decrease) in imbalance across early adolescence (slope from age 12 through 14) and the probability or level of AU (quantity by frequency) at age 12 or the growth in the probability or level of AU through age 20 (68). Similar findings resulted when investigating the probability and level of cannabis use from ages 13–20 years (68).

Using the same sample, Meisel and colleagues (68) then conducted growth mixture modeling to simultaneously model growth in reward and executive control measures and to identify groups that shared similar patterns of growth. This resulted in two classes, both of which demonstrated heightened reward sensitivity relative to response inhibition initially (ages 12–13) and then had heightened response inhibition relative to reward sensitivity by age 14, though one class had a more prominent imbalance. Similar to their prior finding, when class membership during early adolescence was examined in relation to substance use through late adolescence, no significant prospective associations were found between class membership and either the probability or levels of alcohol or cannabis use, suggesting that individual differences in patterns of imbalance in early adolescence were not associated with prospective growth in substance use (68).

Though a strong application of developmental within-person measures of imbalance, Meisel and colleagues (68) may have failed to find associations between initial levels and developmental change of imbalance and substance use outcomes because of the limited and young age during which they measured reward- and executive control-related processes. Indeed, other investigations suggest that associations between reward and control measures and substance outcomes vary with age (74, 76). To this end, McCabe and colleagues (2021) applied latent difference score growth curve modeling to a sample of late adolescents (ages 17–20) from the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA) study and found that greater within-person imbalances between self-reported sensation seeking and impulsivity at age 17 (intercept) were associated with greater levels of heavy episodic drinking by age 20 years.

Using a type of growth mixture modeling, Meeus and colleagues (114) applied a latent class growth analysis to a sample from the National Longitudinal Study of Youth and Children and Young Adults Study, which included seven waves of assessments spanning early to late adolescence (ages 12–25 years). Classes were defined based on the joint developmental trajectories of self-reported sensation seeking and impulse control. Class membership was then associated with intercept (age 14–15) and slope (change through age 25) of tobacco, cannabis, and AU. Results indicated that a group characterized by higher levels of sensation seeking relative to impulse control across all ages showed the highest initial AU frequency and the highest level of substance use overall (114). However, contrary to hypotheses, the group that demonstrated the greatest increase in AU frequency over time was characterized by balanced levels of sensation seeking and impulse control in early to mid-adolescence followed by dominance of impulse control over sensation seeking in late adolescence to early adulthood (114). The authors argue that these findings provide some evidence for developmental dual process models of adolescent substance use in that some individuals demonstrated a developmental imbalance where reward processes dominated over control processes in early to middle adolescence, and these individuals engaged in the highest levels of risk behavior (114). They also assert that an imbalance in either direction (sensation seeking dominating inhibitory control or vice versa) confers risk (114). Notably, though, the authors are not clear what the values in their 9-point scale used to report AU frequency represent (i.e., they only provided definitions of the anchors at each extreme: 1-did not drink and 9-drank daily). As such, it is challenging to decipher the predicted levels of AU by emerging adulthood. It is possible that those who demonstrated the greatest increase in AU over the course of the study may have consumed alcohol at rather infrequent rates (e.g., once per month). Such clarifications would be helpful to interpretation, as well as investigations that include measures of other AU outcomes, such as quantity, heavy episodic drinking, or alcohol-related consequences, to better understand the conferred risk of all profiles.

Using similar methods, Wojciechowski (115) used longitudinal data spanning early to late adolescence (ages 14–25 years) from the Pathways to Desistance Study, which includes justice-involved youth (n = 1354) who were followed for seven years (assessed 11 times) after being adjudicated for a serious offense. Wojciechowski (115) used standardized difference scores of self-reported sensation seeking and impulsivity with group-based trajectory modeling to assign individuals to one of three groups. Regressions were then used to predict the presence of different risk-taking behaviors at any point during the study. Results demonstrated that a trajectory characterized by consistently high sensation seeking and consistently low impulse control was associated with increased risk for reporting engagement in all assessed outcomes, including heavy episodic drinking, cannabis use, general offending, and violent offending (115). Interestingly, all three of the groups identified by Wojciechowski's analysis were characterized by stable levels of sensation seeking and impulsivity over time. That is, no group demonstrated developmental changes in these measures with age. Regardless, those with high levels of sensation seeking relative to inhibitory control across development were at maximal risk for engagement in maladaptive behaviors.

In summary, the reviewed within-person methods allow for nuanced examinations of adolescents' reward processes relative to their own executive control tendencies and associations with adolescent risk-taking behavior and AU. Across investigations using within-person methods, there are inconsistencies in findings. Of the seven studies that examined adolescent AU using within-person methods, three reported null associations. However, several methodological limitations may have contributed to these null results. For instance, Malmberg et al. (80) used dichotomous substance use outcomes, Mills et al. (44) relied on retrospective reports of behavior from a small sample of individuals across a large temporal gap, and Meisel et al. (68) assessed reward and control processes at relatively young ages and across a narrow developmental window. In contrast, studies that modeled imbalance trajectories across developmental periods, such as McCabe et al. (116), Meeus et al. (114), and Wojciechowski (115), provide support for the imbalance hypothesis. Yet, while within-person investigations on both theoretical and methodological levels allow for the most robust assessment of imbalances and while there appears to be some coherence of findings through such studies when they are well designed to test the imbalance hypothesis, there are general limitations of within-person investigations, including the cited papers. Within this literature, there is an over-representation of studies using sensation seeking as an index of the reward system and there have been no longitudinal investigations to date that have examined developmental within-person imbalances between reward sensitivity and executive control measures from early through late adolescence in relation to AU patterns across that same developmental period. Additionally, these types of investigations are few in number, and only some have incorporated a consideration of developmental processes into their designs. Without consideration of age-related change, it remains unclear if associations with AU are developmental in nature or due to stable individual differences.

There are also methodological considerations when applying and interpreting algebraic difference score approaches, including the theoretical justification and choice of difference score expression, reliability of the difference score and component measures, and the nature of correlation patterns between difference scores and component measures and/or outcomes measures [e.g., (68, 109, 110, 117, 118)]. In line with these considerations, researchers should explicitly define the theoretical justification for difference scores and evaluate the construct validity and reliability of control and reward measures. To psychometrically justify the use of difference scores, these component measures should be uncorrelated, have equal variances, and be related to outcome measures in roughly equal though opposite manners (109, 110, 118). When these assumptions are met and when these approaches are applied within longitudinal samples that allow for effective testing of age-varying associations between directly quantified within-person measures of imbalance, they can provide a robust means of testing developmental imbalance hypotheses and can be used to differentiate developmental changes from more stable individual differences.

However, standardized difference scores used in regression frameworks are not without limitations. The same imbalance score will predict the same outcome value (e.g., amount of alcohol consumed), regardless of whether the difference score value arises from, for example, slightly elevated reward sensitivity vs. neutral executive function or from highly elevated reward sensitivity vs. somewhat strong executive function. Whether these distinct profiles in reality confer equivalent levels of risk is unclear. Alternative within-person methods—such as modeling the joint developmental trajectories of reward and control systems using growth mixture modeling and investigating how the resultant developmental profiles relate to patterns of alcohol use—may clarify the significance of different patterns that contribute to the same relative reward/control imbalance. However, these alternative modeling approaches require complex analyses and large, longitudinal samples, making them challenging to implement on a broad scale.

6 Discussion

In this review, the question of how best to validate the dual systems model of adolescent development has been considered. The dual systems framework not only suggests important roles for both executive control and reward sensitivity in vulnerability to risk-taking behaviors, but its central tenet is that there are developmental changes in each system's level of function relative to the other. Notably, a functional imbalance is hypothesized to occur during specific phases of adolescence whereby reward sensitivity is elevated relative to executive control abilities, leading to a bias toward unrestrained reward seeking. Here, we have evaluated additive, interactive/moderation, and within-person approaches for assessing the intricate developmental dynamics between reward and executive control systems in relation to adolescent AU. Specifically, this review has considered whether and how reward and executive control processes simultaneously contribute to adolescent AU patterns, how the effect of one might moderate the effect of another, and how developmental imbalances between reward-seeking behaviors and regulatory functions are implicated in adolescent risk-taking and AU.

Table 5 provides a simplified summary of findings across methods and periods of adolescence, with an emphasis on general patterns observed for each method. Notably, we have not included effect sizes in the figure given that our primary critique of this literature extends beyond the magnitude of reported effects, emphasizing methodological strengths and weaknesses as well as theoretical factors.

Table 5. Patterns observed in the reviewed literature.

Note. This table provides a simplified summary of findings across the 40 reviewed studies, organized by modeling approach and alcohol use outcome. It visually represents general patterns of support for relations between reward sensitivity/sensation seeking, executive control, and their interplay in predicting alcohol use. For a more nuanced understanding of these findings, including variability in study design and measurement, we direct readers to Tables 2–4 and the manuscript text. Shading indicates accumulated strength of association across investigations, where grey cells represent significant associations, hashed cells represent contradictory findings, black cells represent null findings, and white cells represent no such studies. Symbol indicates measurement method, where check marks represent single timepoint measurement of predictors and clocks represent measurement of developmental change of predictors. Multiple check marks denote coherence of findings across three or more studies. Early to middle adolescence = ages ∼10-17 years and late adolescence =18+ years. Abbreviations: BD/HED, binge drinking/heavy episodic drinking; EC, executive control; RS, reward sensitivity; SS, sensation seeking

No method appears ideal for modeling the hypothesized functional imbalance between systems. However, as indicated, each analytic approach yields findings, albeit inconsistent at times, that support the contention that variations in both executive control and reward sensitivity are vulnerability factors for the initiation and progression of adolescent alcohol use. Nonetheless, only the within-person approach, particularly examinations of joint developmental trajectories of reward and control systems and associations between resulting profiles and behaviors, is optimized to differentiate individual difference factors from developmental change. We argue for this approach based strongly on theoretical grounds, as it is the only method used in the literature that accurately captures the systematic imbalances that are hypothesized to contribute to risk-taking tendencies in adolescence. In evaluating this literature, conclusions can be offered regarding various stages of AU and whether high reward/low executive control contributes to variations in behavior during each stage. The literature can also be evaluated according to distinct phases of adolescence and whether high reward/low executive control is important for alcohol use during specific age periods.

6.1 High reward sensitivity and low executive control are associated with adolescent AU

Substantial evidence suggests that when considered in tandem, both reward and executive control processes are associated with adolescent substance use patterns broadly (74, 79, 94, 96–101), potentially in a moderated fashion (77, 94, 106) though moderation findings are very inconsistent. Importantly, despite some variability in findings, within-person imbalances between reward and executive control systems, including heightened reward processes relative to executive control processes during middle and late adolescence, provide a strong test of the dual systems model, and some findings suggest that reward biases are associated with elevated adolescent AU (112, 114–116).

6.1.1 Alcohol use initiation

In considering investigations that have assessed both processes simultaneously, it is unclear if reward sensitivity (80) or executive control (71) jointly confer risk for the prospective initiation of AU. While both reward and executive control processes are associated with retrospective age of initiation, sensation seeking, in particular, appears strongly related (77, 80, 82, 84). Causal interpretations of these findings must be avoided since findings could also represent consequences of longer exposure to alcohol. It is unclear if executive processes moderate these effects.

6.1.2 Concurrent and prospective alcohol use

Both reward and executive control processes, particularly during early to middle adolescence, are implicated in concurrent rates of AU (74, 86, 87, 89–91, 95). However, moderation models suggest that strong executive control processes may buffer the effect of heightened reward processes (77). Additionally, though both high reward processes and low executive control may be risk factors for subsequent use (72), lower executive control, in particular, may be implicated in subsequent use rates (72, 75, 95).

6.1.3 Problematic alcohol use

Additionally, executive control processes may play a more prominent role in binge/heavy episodic drinking in early adolescence (72, 73), while reward-based processes may play a more prominent role during middle adolescence and onwards (76, 83, 85, 91, 92). Notably, the imbalance between reward and control processes has been implicated in heavy episodic drinking (115, 116). Importantly, both systems appear to contribute to alcohol misuse/severity (74, 94, 119), though executive control processes may be more strongly implicated in experiencing alcohol-related problems in adolescence (72, 73, 76).

6.2 Benefits of longitudinal within-person approaches

Though findings across samples, periods of adolescence assessed, predictor indices, and specific AU outcomes are variable, collectively, the results suggest that both elevations in reward and weaknesses in executive control processes confer risk for AU patterns in adolescence. Importantly, this evidence underscores the value of considering both systems jointly. While many studies have adopted additive models, which are an improvement over single-system approaches, such methods fall short in capturing the relative imbalance between reward and executive control systems within individuals. Moderation analyses bring us closer to approximating these dynamics, but methodological considerations constrain their interpretability and the absence of three-way interactions with age in these investigations hinders our ability to examine how the impact of the proposed reward-control imbalance varies with maturation. In contrast, within-person approaches, regardless of whether they yield positive findings, offer a direct and developmentally grounded test of dual system frameworks when applied to large samples that span the full range of adolescence. These methods uniquely capture dynamic, individual-level imbalances between reward and executive control systems and their changing impact on behavior across development. Though still relatively rare, and though there are inconsistencies in findings, well-designed within-person studies suggest that developmental imbalances between reward and executive control systems, including heightened reward processes relative to executive control processes, are related to elevated adolescent AU, especially when assessed in middle-to-late adolescence. These studies, and others (32), indicate that there are marked heterogeneities among individuals in the developmental trajectories of the reward-control imbalance. In combination with findings from additive models of developmental changes in reward and executive control systems, it seems that developmental processes, in conjunction with individual differences in the level and pace of changes in reward and control systems, are likely implicated in accelerations in adolescent AU, as well as other forms of risk-taking. As such, further longitudinal investigations using within-person designs are necessary to further elucidate the complex interplay between individual differences and developmental processes when considering reward and executive control systems' impact on adolescents' AU and risk-taking tendencies.

6.3 Methodological and interpretative challenges

Despite progress in this area, several gaps remain in the current understanding of these processes, and some approaches in the extant literature contribute to these shortcomings. We also note that while this review focuses narrowly on control and incentive motivational processes, it must be acknowledged that other systems contribute to adolescent risk-taking, including alcohol use. For instance, the dual systems model does not invoke aspects of behavioral avoidance or stress reactivity (120), which may be important predictors of alcohol use onset [c.f., (121)]. Various neurochemical systems also interact with the reward system to influence responses to alcohol (122, 123). Other methodological and interpretive challenges are evident.

First, extending beyond the reviewed literature to consider the state of the research broadly, the accuracy with which we refer to our measures is of paramount importance, as the terminology used to describe what a measure assesses can significantly impact the interpretation of research findings and the validity of conclusions drawn. Measures such as the Iowa Gambling Task, Balloon Analog Risk Task, and Stoplight Task, all of which have been used in adolescent samples, are complex, multifactorial paradigms designed to be measures of risk-taking. However, researchers in the past have attributed performance on these tasks solely to “reward sensitivity” [e.g., (103, 124)], and some have then used these measures to predict real-world risk-taking behaviors. While these tasks are limited in their ability to function as implicit proxies for real world risk behavior (125), this sort of content contamination, when a measure designed to assess a particular psychological construct is inextricably linked to the behavior or construct it is intended to predict, can lead to inflated or misleading conclusions about the relationship between constructs, in this case, between reward sensitivity and risk-taking. Moreover, performance on the same task is sometimes described and interpreted in terms of different underlying constructs (126). For example, delay discounting performance has been described as representing reward [e.g., (127)] and executive control processes [e.g., (128)], and sensation seeking has been referred to as a measure of impulsive personality [e.g., (100)]. These measurement issues have important implications for interpreting the existing literature. Inconsistent construct definitions and overlapping use of terms like sensation seeking, reward sensitivity, and impulsivity make it difficult to compare findings across studies and to draw clear conclusions about the underlying mechanisms of adolescent AU. The use of measures that blend predictors with outcomes, such as risk-taking tasks labeled as proxies for reward sensitivity, may inflate observed associations and limit causal inferences. Greater precision in measurement and careful attention to the language used to discuss measures are essential for advancing theory, improving cross-study comparability, and drawing valid conclusions about the mechanisms involved in adolescent AU.

Second, some measures are frequently used to index each system, but they may not actually fully capture constructs of interest. Most of the reviewed studies that examined reward and control systems jointly relied on measures of sensation seeking and inhibitory control (impulsivity, lack of premeditation), often conceptualized as impulsive personality traits linked to substance use [e.g., (129, 130)]; as personality risk factors for substance abuse alongside hopelessness and anxiety sensitivity [e.g., (131)]; or from a dual systems/dual pathway approach, sensation seeking as an index of the reward system [e.g., (74, 116)]. However, there is little consensus on the optimal measurement of the reward and control systems, particularly for reward processes. There are clear differences between sensation seeking and reward sensitivity at conceptual, behavioral, and neural levels (124). Sensation seeking, as defined by Zuckerman, is “the need for varied, novel, and complex sensations and experiences, and the willingness to take physical and social risk for the sake of such experiences” [(132), p.10], reflecting a preference for intense, arousing experiences (133). In contrast, reward sensitivity more specifically captures incentive motivation: anticipating, pursuing, and experiencing pleasure from positive stimuli (19, 134). Though both constructs are related to each other and to risk-taking (124), sensation seeking and reward sensitivity are likely mediated by different, though overlapping, neural systems and expressed differently in behavior (135, 136). While sensation seeking has been consistently associated with adolescent AU, when investigating developmental dual systems models grounded in limbic and reward-related brain systems (18), the fundamental process of adolescents' specific responses to positive stimuli (19), reflected by variations in individuals' motivation to seek out and enjoy rewarding experiences (134), may be of more interest. The use of sensation seeking scales in this work is further complicated by shared method variance between sensation seeking and the predicted risk-taking outcomes, as sensation seeking, by definition, encompasses risk-seeking behaviors (132). Similar to the interpretive problems that emerge when measures of risk-taking are treated as indices of the reward system, findings resulting from the use of sensation seeking measures may reflect tautological relationships, given that content contamination and shared item covariance may lead to inflated associations rather than the isolation of unique effects (137). Similarly, measures of inhibitory control dominate most executive control assessments, though other cognitive control functions contribute to risk-taking (138, 139). A full review of the measures used in the field to index theoretical constructs and their limitations is beyond the scope of the present discussion, though these considerations are paramount. To advance theoretical models and accurately characterize developmental pathways to AU, future research must refine and diversify measurement strategies to better distinguish between related but distinct constructs and to capture the behavioral dynamics underlying reward-control imbalances and relationships to AU across development.

Third, most research has relied on self-report measures, with fewer studies incorporating behavioral tasks or investigations of neurodevelopmental imbalances, including imbalances in brain structure and function and their associations with AU patterns. Although questionnaire-based measures of reward sensitivity, like the BAS scales, aim to assess this construct in a focused way, creating laboratory-based performance tasks that measure reward sensitivity without also tapping into other processes, such as risk-taking, decision-making, motor speed, executive function, impulsivity, threat sensitivity, or arousal, is inherently challenging. Such a reality likely contributes to the reliance on self-report measures, at least in relation to reward processes. Notably, different measurement methods across studies likely contribute to inconsistent findings. For example, self-report and behavioral measures of impulsivity often show weak or no correlation (140, 141), likely reflecting differences between trait-level (self-report) and state-level (task-based) characteristics (5, 126). Future research may benefit from utilizing ecologically valid (125), multi-method, and latent variable (126, 142) approaches to better capture the complexity of reward and control processes at trait, behavioral, and neural-based levels. For example, multimodal designs that integrate neuroimaging indices of circuit-level function, performance-based measures of reward sensitivity and executive function, and ecological momentary assessment of affective states and real-world behavior would enable the simultaneous capture of the tonic and phasic dynamics of these systems. Such approaches would allow researchers to parse dissociable components of incentive motivation, regulatory control, and action selection that may otherwise be conflated when relying on a single measurement modality. Critically, integrating data collected at different temporal densities and levels (e.g., neural, cognitive, affective, behavioral) allows for the assessment of coordinated interactions among systems within individuals over time as well as an examination of how alterations in one process of interest (e.g., reward sensitivity) might influence subsequent behaviors (e.g., recruitment of executive control). This type of multimodal, temporally sensitive modeling has the potential to more precisely characterize how these systems interact in the moment and over time to jointly organize decision-making processes that confer vulnerability for adolescent alcohol use.

Fourth, potential differences in the risk conferred due to heightened reward relative to executive control processes due to sex assigned at birth is not fully understood. Though not covered extensively above, there is reason to anticipate potential sex-specific associations between reward and executive control processes with adolescent risk-taking and AU. Firstly, males demonstrate higher rates of certain risk-related behaviors, including fatal accidents (143), gambling (144), and crime (145), though recent evidence suggests a shift in AU patterns with adolescent females now surpassing males in reported levels of use (160). Secondly, while sex differences in executive functioning are not strongly supported (146), findings related to reward processes are more variable and complex, often demonstrating sex-specific developmental patterns, links to psychopathology, and associations with reward-related neural correlates (34, 48, 147–149). Males also tend to show elevated sensation seeking and impulsivity across age groups (e.g., 162–164). Importantly, while some additive investigations reviewed suggest that the effects of reward- and control-related systems may vary by sex (79, 81, 97), such effects have not been sufficiently explored in interactive or within-person designs.

Fifth, throughout much of this literature, determining the directionality of effects is difficult. Concurrent associations between predictors and retrospective reports of AU initiation age or recent AU patterns cannot clarify causality. Even prospective studies face challenges, as alcohol use can alter the very systems that predict its use (150), leading to difficulties in dissociating risk factors from their consequences. For example, adolescent AU is linked to abnormal neural activity during executive function and reward tasks (151), deviations from expected patterns of structural brain development (152), disruptions in normative development of behavioral control as adolescents transition from no use to frequent drinking in adolescence (78), smaller developmental decreases in impulsivity and larger increases in sensation seeking in youth who initiated use (153), increased reward sensitivity after initiation (154), and bidirectional relationships with impulsivity and sensation seeking (74). As such, developmental cascades (155) following AU onset may further disrupt the developmental patterns implicated in use patterns, heightening risk for alcohol-related disorders and other forms psychopathology over time. Indeed, on a theoretical level, cycles of alcohol use disorders involve escalating disruptions to reward, stress, and executive systems, fueling the cyclic nature of addiction (156). Given these complexities, study designs and modeling considerations that account for the bidirectional effects of approach drive, regulatory capabilities, and alcohol consumption are necessary and strongly recommended.

6.4 Developmental vulnerability or individual differences?

Finally, several investigations suggest that risk factors change with age (74–76). Whether these age-related shifts are due to developmental effects, differences in the severity of use that is typical at different ages, or due to other factors that change with age, such as access to substances, social acceptability of use, peer dynamics, or parental permissiveness, is not entirely clear. However, several of the investigations reviewed (68, 95, 114, 116), and others (32), have demonstrated marked heterogeneity in the development of these systems through adolescence. While some research found that the rate of change of imbalance (68, 116) was not predictive of future patterns of use or that adolescents were characterized by stable imbalances that did not appear to shift with development (though these imbalances were still predictive of patterns of use) (115), studies implicate individual differences in developmental trajectories of both systems in adolescent AU patterns [e.g., (95, 114)].

Taken together, developmental processes, combined with individual variation in the level of imbalance, timing, and degree of changes in reward and executive control systems, are likely involved in adolescent AU patterns and other forms of risky decision-making. Crucially, disentangling these sources of influence remains essential, since individual differences can be targeted through prevention efforts at any age, whereas developmentally-specific associations are limited to certain windows in time that may represent intervals when intervention efforts will be most successful. A central goal of this research, and one we hoped to address when we initiated this review, is to identify developmental windows of heightened vulnerability. As we have argued herein, additive and interaction/moderation-based approaches are abundant in the literature but are not well-suited to differentiate the roles of reward and executive control, or to identify their imbalances. Because within-person longitudinal assessments are relatively few in number, the current literature is limited in allowing us to draw firm conclusions regarding when risk for substance misuse is greatest based on an imbalance between reward sensitivity and executive control. If a period can be identified, intervention strategies such as cognitive remediation to boost executive capacities might be most effective during this time. This is speculative, and the clinical significance of such approaches is still debated (157, 158). Ultimately, it remains unclear whether the highest risk lies among adolescents who show an early and persistent imbalance across development, or among those whose imbalance emerges, intensifies or simply fails to decline in late adolescence. Distinguishing among these trajectories is critical, as each implies different timing and targets for prevention and intervention strategies. More longitudinal work that explicitly models divergent developmental courses through the use of techniques such as latent class analysis and growth mixture modeling (159) is needed to clarify these possibilities and determine when, and for whom, preventive efforts may be most effective.

7 Conclusion

Collectively, findings across studies underscore roles for both reward and executive control processes in predicting aspects of adolescent alcohol use. However, despite decades of research, the field has yet to adequately capture how these systems interact dynamically within individuals over time to confer risk. This limitation stems largely from an overreliance on cross-sectional designs and insufficient modeling approaches, which cannot disentangle developmental change from stable individual differences or test the central tenet of dual systems framework models—that the relative dominance of reward over executive control processes changes over time within a person to confer developmentally-specific patterns of risk. To move the field forward, future research must prioritize well-powered, longitudinal, within-person designs, which are uniquely suited to identify whether and when there are sensitive periods when biases of reward sensitivity over executive control confer heightened risk. Advancing this line of inquiry will require careful attention to study design, construct measurement, and analytic approaches that align with the developmental nature of these processes. Elucidating these within-person dynamics is critical, not only for refining theoretical models of addiction vulnerability, but also for informing when and how to intervene in order to mitigate alcohol-related harms during adolescence and beyond.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Alessandra Maria Passarotti, University of Illinois Chicago, Chicago, United States

Reviewed by: Valentina Castelli, University of Palermo, Italy

MoonSoo Lee, Korea University, Republic of Korea

Daniel P. Keating, University of Michigan, Ann Arbor, United States