Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Mar 1.
Published in final edited form as: Drug Alcohol Depend. 2023 Dec 28;256:111068. doi: 10.1016/j.drugalcdep.2023.111068

Initial evidence of delay discounting’s predictive utility for alcohol self-administration in ecologically valid contexts among young adults who drink heavily

Jillian M Rung a,*, Benjamin L Berey b,*, Robert F Leeman a,c
PMCID: PMC10922828  NIHMSID: NIHMS1960213  PMID: 38290204

Abstract

Background:

While delay discounting is robustly associated with alcohol use disorder, whether discounting predicts real-time alcohol use behaviors is unclear. Existing support comes from laboratory studies using intravenous alcohol self-administration methods, thus limiting ecological validity and generalizability. The present study evaluated whether delay discounting predicted real-time alcohol use in naturalistic settings with and without probabilistic negative consequences for consuming larger amounts of alcohol.

Methods:

This secondary analysis utilized data from three laboratory alcohol self-administration studies with young adults who engaged in frequent heavy drinking (N=206, 45% female). Participants completed a delay discounting measure before an alcohol self-administration session in an actual or simulated bar with (n=187) or without (n=19) probabilistic negative consequences (compensation loss) tied to performance on cognitive and psychomotor tasks after alcohol self-administration. Bootstrapped (unstandardized) coefficient estimates and 95% confidence intervals were utilized due to the sample size discrepancy.

Results:

Multiple regressions revealed that delay discounting did not significantly predict estimated blood alcohol concentration (eBAC) or number of drinks consumed when procedures included probabilistic negative consequences. Among participants who completed procedures without probabilistic negative consequences, delay discounting was positively associated with peak eBAC.

Conclusion:

Counter to hypotheses, steeper delay discounting did not predict real-time alcohol use in contexts with probabilistic negative consequences, whereas preliminary evidence suggests that delay discounting predicts real-time alcohol use behaviors in contexts without probabilistic negative consequences. The specific discounting task may have impacted study findings, thus future research should consider how the sign (gain vs. loss), outcome certainty, and delay relate to alcohol consumption.

Keywords: BAC, alcohol use, experimental, impulsivity, external validity

1. Introduction

Extensive research indicates that rapid devaluation of outcomes due to delay in receipt (i.e., steep delay discounting) is a transdiagnostic indicator of myriad psychiatric disorders (Amlung et al., 2019), including alcohol and other substance use disorders (AUD/SUD, respectively) (Amlung et al., 2017; MacKillop et al., 2011). Contemporary models of addiction emphasize distorted decision-making processes characterized by overvaluing proximal reinforcers (e.g., alcohol) and devaluing larger, distal rewards (Bickel et al., 2014). Often characterized as a behavioral choice impulsivity domain (King et al., 2014), delay discounting involves making a series of discrete choices between receiving smaller, immediate rewards versus larger, delayed rewards at various times in the future (e.g., $10 right now or $20 in 10 days). Individuals with AUD discount delayed outcomes more than those without AUD (Petry, 2001), and those with more severe AUD discount more than those with less severe AUD (i.e., those with fewer AUD symptoms; MacKillop et al., 2011). As such, delay discounting is thought to reflect lack of sensitivity to delayed, negative consequences that co-occur with AUD (Bickel et al., 2014).

While many studies have evaluated associations among delay discounting, AUD, and retrospective self-reports of alcohol quantity-frequency (e.g., drinks per drinking day), few have examined whether delay discounting predicts subsequent alcohol use behaviors in real-time. Existing support linking delay discounting to real-time alcohol consumption is limited to laboratory studies employing intravenous self-administration methods following priming or alcohol challenge (i.e., a fixed alcohol dose; Gowin et al., 2017; Grodin et al., 2021; Sloan et al., 2020; Stangl et al., 2017). Collectively, results from these studies provide evidence that steeper delay discounting is associated with greater alcohol self-administration (Grodin et al., 2017; Stangl et al., 2017), higher breath alcohol concentrations (BrACs) (Stangl et al., 2017; Grodin et al., 2017), and higher probability of consuming alcohol to legal intoxication (i.e., BrAC > 0.08%) (Gowin et al., 2017; Sloan et al., 2019), even after controlling for relevant factors such as biological sex and family history of AUD status.

While promising, it is difficult to ascertain the generalizability of these results to more naturalistic contexts when initial consumption is self-initiated. Other features of administration/self-administration experiments diverge from real world drinking, including artificial laboratory contexts and presence of study staff. Moreover, participants consumed alcohol alone in all prior studies examining links between delay discounting and real-time alcohol use behaviors. This is notable limitation given that alcohol is typically consumed with peers in social settings like bars or parties (de Wit & Sayette, 2018). Equally important, contingencies around laboratory self-administration also tend to differ sharply from real world conditions. While delay discounting captures preferences for smaller, immediate versus larger, delayed outcomes, laboratory alcohol administration studies rarely include immediate or delayed consequences tied to alcohol use behaviors. Thus, tendencies to discount steeply may relate more closely to alcohol consumption in naturalistic lab paradigms that include negative consequences as possible outcomes. In contrast, in contexts where alcohol consumption occurs without consequences, outcomes are less salient, perhaps making relationships to delay discounting more difficult to discern.

The present pre-registered (https://osf.io/wdcb3) secondary analysis sought to address a notable literature gap by examining the predictive utility of delay discounting on real-time alcohol use behaviors in ecologically valid drinking contexts. To this end, we combined data from three human laboratory studies where young adults who engaged in frequent heavy drinking had the opportunity to self-administer alcohol in small group settings (Leeman, 2013; 2018; 2022, herein referred to Study 1, 2, and 3, respectively). The vast majority of participants engaged in a self-administration paradigm including probabilistic, delayed negative consequences tied indirectly to greater alcohol consumption. A small subset of participants engaged in self-administration in the same physical context absent these consequences. We hypothesized that delay discounting would predict alcohol self-administration to a greater extent in contexts with probabilistic negative consequences than without.

2. Material and methods

2.1. Participants

Each parent study enrolled young adults between the ages of 21–25. Participants first underwent a pre-screening via the web or phone and in-person screening appointment. In the prior 30 days, eligible participants reported: a) consuming alcohol on >10 days, b) heavy drinking (i.e., 4/5+ standard alcoholic beverages for males/females, respectively) on >4 days, and c) an estimated blood alcohol concentration (eBAC) >0.10% one or more times. Inclusion criteria across studies were identical except that Study 1 required past-month alcohol consumption on >12 days. Exclusion criteria across studies were identical except that participants in Study 3 who used moderate-drinking digital apps in the past-year were ineligible. Exclusion criteria included: a) current (i.e., past-year) treatment-seeking or substance use treatment; b) positive urine toxicology for illegal drugs, excluding cannabis; c) current (i.e., past-year) substance dependence, excluding alcohol, based on DSM-IV (American Psychiatric Association, 2000); d) current withdrawal (i.e., CIWA [Sullivan et al., 1989] score > 8) or prior history of medically-assisted detoxification; e) two positive BrAC readings measured at the outset of study appointments; f) medical issues contraindicating alcohol use; g) body mass index <18.5 or >35; h) pregnancy, lactation or engaging in sexual activity without birth control in those of child bearing potential; i) current prescription for or use of psychotropic medications; j) psychosis/severe psychiatric conditions, and; k) disliking beer. Eligible participants based on the in-person screening appointment were invited to participate in the full study.

All participants who completed the delay discounting measure and an alcohol self-administration session (N=206) were included in the present analyses (Study 1 n=39, Study 2 n=69, Study 3 n=98). Although alcohol self-administration procedures took place in similar ecologically valid physical contexts across all three studies, only participants in Study 1 were randomized to complete the alcohol self-administration paradigm with (n=20) or without (n=19) probabilistic negative consequences. All participants in Studies 2 and 3 completed the self-administration paradigm that included probabilistic, delayed negative consequences tied indirectly to greater alcohol consumption.

2.2. Procedure

Full study procedures are reported in detail elsewhere (see Leeman et al., 2013; 2018; 2022). Across all studies, participants provided written informed consent at an initial in-person appointment followed by baseline assessments including delay discounting. Afterward, participants unlikely to know each other attended an alcohol self-administration session in groups of 2–4 at a local bar (Studies 1 and 2) or simulated bar laboratory (Study 3) on a separate day. Study 1 established the internal validity of the novel alcohol self-administration paradigm modelling impaired control over alcohol (Leeman, 2013) that was implemented in all three studies. In Study 2, participants were randomized to automatic action tendency retraining or control for five days before attending the alcohol self-administration session. There were no statistically significant differences in any reported alcohol self-administration variable based on retraining or control condition assignment (Leeman, 2018). In Study 3, participants completed brief blood alcohol concentration-focused counseling after baseline and before attending the alcohol self-administration session. Participants also had the opportunity to utilize technology with potential efficacy in reducing drinking during the self-administration period, however there were no statistically significant difference in any reported alcohol self-administration variable based on technology assignment (Leeman et al., 2022).

Across all three studies, participants could order non-alcoholic or alcoholic (i.e., three beer options at 4–5% abv and similar caloric content) beverages ad libitum from 17:00–20:00. However, participants were unable to consume additional beers resulting in an eBAC >.10. Customized charts were calculated for each participant that contained eBACs based on possible alcohol consumption in 5-minute increments using a standardized formula based on the Widmark equation: (((number of beers/2) * (constant of 9 for females and 7.5 for males/weight)) – (number of hours × .016)) (Matthews & Miller, 1979). Across the three studies, 54 participants had to be cut off from further drinking at some point during the alcohol self-administration period whereas 152 participants were not. There was a statistically significant difference whereby those who were cut off at some point were almost 10 lbs lighter (M=153.85, SD=28.16) than those who were not cut off (M=163.38, SD=29.95), t(204)=2.04, p=.043. There was no statistically significant sex difference between those who were and were not cut off, X2(N=206)=0.84, p=.358. Participants also provided breath alcohol samples at the end of the alcohol self-administration periods and hourly thereafter until dismissal. We opted not to include breath alcohol readings during the alcohol self-administration period to maintain ecological validity and to avoid interrupting self-administration.

Depending on study and random assignment, participants either completed self-administration with (n=187; 46% female) or without (n=19; 32% female) probabilistic negative consequences for consuming greater amounts of alcohol. The consequence was probabilistic, partial loss of study compensation dependent on performance on four cognitive/psychomotor tasks sensitive to alcohol, which were the Hopkins Verbal Learning Test (Brandt, 1991), time production task (Chait & Perry, 1994), grooved pegboard (Brumback et al., 2007), and digit symbol substitution test (Brumback et al., 2007). All procedures were approved by the local overseeing institutional review boards.

2.3. Measures

2.3.1. Alcohol Consumption

During in-person screening, participants completed the Timeline Followback (TLFB; Sobell and Sobell, 1992) to assess past 30-day alcohol, nicotine, and cannabis use. On days participants reported any alcohol use, they provided the total number of standard alcoholic beverages they consumed and the overall duration (in hours) of alcohol consumption. These data were used to compute average drinks per drinking day and eBAC. During the self-administration sessions, we operationalized alcohol consumption in terms of: a) total number of standard alcoholic drinks consumed, and b) peak eBAC during the session. In all instances, we calculated eBAC based on the Widmark equation specified above in Section 2.2, which accounted for number of drinks consumed, hours spent drinking, biological sex, weight in pounds, and population averaged hourly alcohol metabolism rate.

2.3.2. Delay Discounting

The Monetary Choice Questionnaire (MCQ) includes 27 items where participants chose between a smaller, immediate monetary reward or a larger, delayed monetary reward between 7–186 days in the future (Kirby et al., 1999). The difference between immediate and delayed monetary amounts differs across trials, as does their magnitude and duration of delay to the larger amount. The MCQ includes three subgroupings, each with nine items, based on the larger, delayed reward magnitude. Small, medium, and large magnitude items correspond to larger, delayed rewards from $25–35, $50–60, and $75–85, respectively. We used an automated scoring calculator to derive discounting rates (the overall k) (Kaplan et al., 2016), which were natural log-transformed for all analyses. Larger log(k) values indicate steeper discounting. We chose overall k from the MCQ for analyses, as opposed to k values specific to certain reward magnitudes, because we had no explicit hypotheses regarding magnitude effects and sought to evaluate the hypothesized associations broadly. Finally, to determine the generalizability of findings across metrics, the percentage of larger-later choices (%LLR) was also used to quantify discounting. Lower percentages of larger-later choices correspond with steeper discounting.

Two participants who completed the alcohol self-administration sessions with probabilistic negative consequences had missing data on the MCQ (1 item each), which precluded estimation of k for these participants. As such, these participants were excluded from analyses involving (log) k as a predictor, but retained in analyses using %LLR as the predictor. The denominator for the number of questions answered was adjusted accordingly.

2.4. Data Analysis

Focal study hypotheses were tested using multiple linear regression to predict alcohol use behaviors (i.e., number of alcoholic drinks self-administered, eBAC) from discounting rate (i.e., k and %LLR) and covariates (i.e., biological sex, baseline levels of each outcome [average baseline eBAC and drinks per drinking day from TLFB], and student status). While including a delay discounting by drinking context interaction would have enabled direct tests of the moderating role of probabilistic consequences on alcohol use behaviors, separate models were tested for each context due to the sample size imbalance across the two consequence conditions (approximately 10:1 for negative consequence versus no consequence).

All regression models were subject to standard diagnostic procedures. After fitting each model, observations producing large residuals (>3 SDs) or high influence on regression weights (Cook’s distance, D>4/n) were excluded and the model was refit. Results from models with and without these observations are reported. To help circumvent loss of statistical power due to removal of overly-influential observations in analyses focused on the paradigm without consequences (a much smaller sample), we generated bootstrapped (unstandardized) coefficient estimates and 95% confidence intervals as a robustness check of the models using all observations.

With the limited sample size and comprehensive analytic approach, corrections to p-values would obscure all but large effects. As such, effect sizes (standardized coefficients) are provided for significant model results and confidence intervals for bootstrapped (unstandardized) estimates to illustrate the magnitude and robustness of findings. Several higher-powered analyses using multilevel modeling were attempted in part to address power concerns, but collinearity issues precluded meaningful estimation of model parameters. Analysis code is available in the preregistration (https://osf.io/wdcb3), and study materials are available by contacting the corresponding author.

3. Results

The sample had a mean age of 22.7 (SD=1.3) and was majority White, non-Hispanic (73%), followed by Hispanic (13.8%), Asian (5.9%), Black (4.4%), and Other (2.9%). Demographic characteristics and descriptive statistics for baseline alcohol use variables are provided in Supplemental Table 1. In each condition, the majority of participants were male (54% with and 68% without potential negative consequences). Participants assigned to the negative consequences condition reported about one less drink per drinking day on average (M=5.46 drinks) and were more likely to be students (80%) than those without potential negative consequences (M=6.71 drinks, 47% students). However, average baseline eBAC did not differ across groups.

Table 1 includes model results including all observations and without overly-influential observations for each outcome (number of drinks consumed, peak eBAC) and discounting metric (k, %LLR). Among those completing alcohol self-administration with a probabilistic negative consequence for consuming higher amounts of alcohol, there was no significant relation between delay discounting and alcohol consumption regardless of discounting metric and whether influential observations were included (all ps > .30).

Table 1.

Multiple regression models testing delay discounting measures as predictors of number of alcoholic drinks self-administered and peak estimated blood alcohol concentration in the laboratory with probabilistic negative consequences for greater alcohol consumption.

Outcome Discounting Index Model N F df p R 2A B (SE) t P
Number of Drinks Log(k)
All observations 185 15.97 4,180 <.001 0.25
Intercept 2.00 (0.48) 4.14 <.001
Sex (Male) .91 (0.20) 4.53 <.001
Student Status (Student) .10 (0.24) 0.41 .69
Average Drinks per Drinking Day .23 (0.05) 4.85 <.001
Log(k) −.02 (0.07) −0.24 .81
Without Influential Obs. 165 22.33 4,170 <.001 0.33
Intercept 1.36 (0.45) 3.03 <.001
Sex (Male) .91 (0.18) 5.07 <.001
Student Status (Student) .26 (0.22) 1.20 .23
Average Drinks per Drinking Day .28 (0.05) 5.92 <.001
Log(k) −.07 (0.07) −1.02 .31
Proportion Larger-Later Choices All observations 187 15.18 4,182 <.001 0.23
Intercept 2.14 (0.44) 4.84 <.001
Sex (Male) .88 (0.20) 4.38 <.001
Student Status (Student) .03 (0.24) 0.13 .90
Average Drinks per Drinking Day .23 (0.05) 4.76 <.001
Proportion Larger-Later Choices .08 (0.63) 0.14 .890
Without Influential Obs. 176 20.47 4,171 <.001 0.31
Intercept 1.41 (0.42) 3.37 <.001
Sex (Male) .83 (0.18) 4.54 <.001
Student Status (Student) .26 (0.22) 1.16 .25
Average Drinks per Drinking Day .29 (0.05) 5.95 <.001
Proportion Larger-Later Choices .61 (0.58) 1.05 .30
Peak estimated BAC Log(k)
All observations 185 4.61 4,180 .001 0.07
Intercept .04 (0.01) 4.10 <.001
Sex (Male) .00 (0.00) −0.92 .36
Student Status (Student) .00 (0.00) 0.46 .65
Baseline average estimated BAC .19 (0.05) 3.71 <.001
Log(k) .00 (0.00) −0.34 .73
Without Influential Obs. 180 7.20 4,173 <.001 0.12
Intercept .04 (0.01) 3.96 <.001
Sex (Male) .00 (0.00) −1.18 .240
Student Status (Student) .00 (0.00) 0.83 .410
Baseline average estimated BAC .23 (0.05) 4.53 <.001
Log(k) .00 (0.00) 0.10 .920
Proportion Larger-Later Choices All observations 187 4.35 4,182 .002 0.07
Intercept .04 (0.01) 4.92 <.001
Sex (Male) .00 (0.00) −1.12 .260
Student Status (Student) .00 (0.00) 0.23 .820
Baseline average estimated BAC .18 (0.05) 3.51 <.001
Proportion Larger-Later Choices .00 (0.01) 0.17 .860
Without Influential Obs. 178 7.78 4,173 <.001 0.13
Intercept .04 (0.01) 4.64 <.001
Sex (Male) .00 (0.00) −1.28 .200
Student Status (Student) .00 (0.00) 0.55 .590
Baseline average estimated BAC .23 (0.05) 4.66 <.001
Proportion Larger-Later Choices −.01 (0.01) −0.50 .620
Open in a new tab

In contrast, steeper discounting was consistently associated with greater alcohol consumption when there were no consequences for higher consumption, although significance was primarily confined to eBAC as an outcome. Specifically, k and %LLR significantly predicted peak eBAC in models with all observations (β=.66 and −.69, ps < .028). The exploratory bootstrapping analyses supported model results for peak eBAC with all observations such that the coefficient estimates for k and %LLR were similar to those with all observations (B=.005 and −.05, respectively) and their confidence intervals did not contain zero ([0.002, 0.01] and [−0.09, −0.02], respectively). After removing overly-influential values, k remained a significant predictor but %LLR did not. Likewise, steeper discounting, indicated by k but not %LLR, was associated with a higher number of alcoholic drinks self-administered, but only with overly-influential values removed (β=0.29, p=.04). However, bootstrapping with the full sample showed this result was not robust (B=0.32, 95% CI [−.11, .73]). Scatterplots depicting relationships between laboratory alcohol self-administration and delay discounting variables are provided in Supplemental Figures 1–2.

4. Discussion

The present study sought to extend prior research linking delay discounting and AUD risk by accounting for multiple, important contextual characteristics. Specifically, we examined whether monetary delay discounting predicted real-time alcohol use in ecologically valid contexts with and without probabilistic negative consequences tied to drinking behaviors, on a preliminary basis. Counter to study hypotheses, steeper delay discounting significantly predicted higher peak eBAC when participants completed the session without probabilistic negative consequences. This predictive association was directionally similar when predicting number of drinks consumed, but was not consistently statistically significant across these models. Conversely, delay discounting was not significantly associated with alcohol self-administration behaviors when participants completed the session with probabilistic negative consequences. This result is surprising since delay discounting captures preferences between immediate and delayed outcomes – in this case immediately available alcohol versus future monetary compensation – which were made explicit to participants in the condition including negative consequences.

Notably, results from the group without probabilistic negative consequences align with those from prior experimental laboratory IV alcohol administration studies testing links between delay discounting and real-time alcohol self-administration behaviors, which also lacked probabilistic negative consequences. For example, in the study by Stangl and colleagues (2017), steeper baseline delay discounting – as indexed by log(k) – was positively associated with average and peak BrACs during a 150-minute ad libitum IV self-administration period. Likewise, the study by Grodin and colleagues (2020) grouped participants based on their motivation to continue self-administering alcohol following an initial priming dose and subsequent BrACs during the 120-minute ad libitum self-administration portion of the session. Baseline delay discounting – also indexed by log(k) – significantly predicted group membership, such that participants who continued to exert effort for additional “drinks” and achieved higher BrACs had significantly steeper delay discounting than participants who did not continue to exert progressively more effort for additional “drinks” and who achieved lower BrACs.

Likewise, results from the group without probabilistic negative consequences replicated those from experimental laboratory IV alcohol administration studies despite key methodological differences. Specifically, these differences included: route of administration (i.e., oral vs. IV alcohol), and drinking sessions in ecologically valid contexts with relevant visual and auditory cues (e.g., neon signs, beer tap handles, liquor bottles, music) that took place in small groups. Alternatively, prior studies testing links between delay discounting and real-time alcohol self-administration behaviors took place in sterile laboratory settings and without a social component. While the availability of other particularly salient reinforcers (i.e., participants’ ability to socialize with peers while consuming alcohol) was a notable strength of the present study, results suggest that specific contingencies (i.e, probabilistic negative consequences) may be most relevant for delay discounting and its relation to alcohol use behaviors.

While more research is needed to understand why the presence of probabilistic negative consequences impacted delay discounting’s relation to alcohol use behaviors in the way it did, there were germane differences between choice characteristics in the delay discounting measure and choices surrounding alcohol self-administration. Specifically, the MCQ captures delay discounting by comparing preferences for smaller, immediate versus larger, delayed monetary rewards that are guaranteed. Alternatively, this secondary analysis culled data from three parent experimental studies that included aspects of reward and loss discounting that were probabilistic in nature (e.g., immediate reward of alcohol coupled with probabilistic monetary losses stemming from greater alcohol consumption). Using a measure that accounts for uncertain outcomes to predict real-time alcohol use may have aligned better conceptually with procedures from the three parent experimental studies (e.g., Probability Discounting Task [Richards et al., 1999], Experiential Discounting Task [Reynolds et al., 2004]), but considerably less research has focused on probability (vs. delay) discounting’s relation to health behaviors (MacKillop et al., 2015). Moreover, in prior studies using both delay and probability discounting measures, delay discounting has displayed more consistent links than probability discounting with germane outcomes including alcohol use frequency (Takahashi et al., 2009), sexual risk-taking while drinking (MacKillop et al., 2014), individual AUD symptoms (Acuff et al., 2023).

Most of the prior literature on delay discounting and substance use has focused on monetary rewards as the hypothetical outcome. Yet, evidence suggests that delay discounting is influenced by the commodity being assessed (Chapman & Elstein, 1995; Raineri & Rachlin, 1993), and delay discounting is more strongly related to risk behaviors when assessing outcomes that are most relevant to the specific population (Johnson & Bruner, 2012; Lawyer & Schoepflin, 2013; Rasmussen et al., 2010). For example, in a recent study, college students who reported alcohol-only or alcohol and cannabis use completed three versions of a discounting task with money, alcohol, or cannabis as hypothetical outcomes (Naudé et al., 2021). Participants in the dual use (i.e., alcohol and cannabis) group reported significantly steeper alcohol discounting than the alcohol-only group, whereas the two groups did not significantly differ in monetary discounting. Moreover, participants in the dual use group evinced significantly steeper discounting for alcohol and cannabis relative to money.

Several limitations may have contributed to the null findings in the present study, such as the secondary nature of analyses and uneven sample sizes across drinking paradigms. Future, targeted research on this topic should recruit larger and balanced groups, which will allow direct evaluation of context by discounting associations. Replication and additional studies will also help determine the robustness of the predictive associations found herein. The significant associations between discounting and drinking behaviors found in the present study occurred in conditions more similar to those in prior intravenous self-administration studies (i.e., no probabilistic negative consequences). As such, this provides some assurance that delay discounting’s predictive association with certain alcohol use outcomes is not simply a result of idiosyncrasies of the sample, as does the consistency of findings within groups and across drinking outcome measures (i.e., that discounting was generally significantly predictive of peak eBAC across models and by bootstrapped estimates). Regardless, further replication is needed with larger and more diverse samples.

It is also possible that the payment reduction consequences were highly salient and influenced self-administration decisions among participants across a range of discounting, thus limiting our ability to show relationships between discounting and alcohol self-administration in this condition. As such, researchers should carefully consider and select discounting measures that align with the experimental paradigm being used and overarching study goals. Lastly, the present study used data from a sample of young adults ages 21–25 who reported frequent heavy drinking; thus it is unclear whether these results would generalize to adolescents or adults older than 25 years, or individuals with less frequent alcohol use. For safety reasons, a limitation was placed on alcohol self-administration based on eBAC that affected participants of lower weight to a greater extent than larger participants, however all participants had the opportunity to self-administer multiple alcoholic drinks and we believed this to be a better option than introducing a limitation based on number of alcoholic drinks that likely would have introduced a larger confound, disproportionately affecting heavier participants.

Determining which types of discounting (e.g., delay, probability), commodities (e.g., alcohol vs. money), indices (e.g., log[k], %LLR, area under the curve [AUC]) and outcomes (i.e., rewards, losses) are the strongest predictor of real-time alcohol use behaviors will be an important future direction for this line of research. While discounting of gains and losses is correlated, discounting of gains tends to be steeper than losses and appears to be a separate process (McKerchar et al., 2013). Likewise, it appeared the results slightly differed across the different discounting indices in the no probabilistic consequence group when predicting number of drinks and peak eBAC; however, when considering the robustness checks, the inconsistency was more closely tied to the modelled outcome. Peak eBAC may have better power as an outcome given that it is more granular than number of drinks consumed and incorporates sources of individual differences not otherwise captured by covariates (i.e., participants’ weight). Future research that includes multiple distinct discounting measures or that manipulate the probability, delay, and/or type of future rewards will advance our understanding of how different decision-making processes relate to AUD risk as well as a range of alcohol use behaviors beyond frequency and quantity that capture more nuanced individual differences (e.g., actual or estimated BrAC).

This secondary analysis contributes to our understanding of how discounting relates to alcohol use in situations with greater ecological validity and across different contingencies that could moderate its impact (i.e., presence or absence of negative consequences). These results also support the potential utility of delay discounting to understand how young adults make discrete choices related to real-time alcohol use behaviors. Further, this study highlights germane considerations for future discounting research seeking to understand real-world alcohol-related decision-making. This knowledge is paramount for designing efficacious interventions given that strategies to reduce discounting have the potential to decrease and prevent harmful behaviors (Rung et al., 2019; Rung and Madden, 2018).

Supplementary Material

1

Table 2.

Multiple regression models testing delay discounting measures as predictors of number of alcoholic drinks self-administered and peak estimated blood alcohol concentration (eBAC) in the laboratory without probabilistic negative consequences for greater alcohol consumption.

Outcome Discounting Index Model N F df p R 2A B (SE) t p
Number of Drinks Log(k)
All observations 19 9.15 4,14 <.001 0.64
Intercept 4.49 (1.05) 4.26 <.001
Sex (Male) 0.77 (0.74) 1.04 .32
Student Status (Student) −1.01 (0.46) −2.19 .05
Average Drinks per Drinking Day 0.36 (0.13) 2.78 .01
Log(k) 0.32 (0.16) 2.02 .06
Without Influential Obs. 16 15.6 4,11 <.001 0.80
Intercept 4.33 (0.80) 5.45 <.001
Sex (Male) .37 (0.61) 0.60 .56
Student Status (Student) −1.00 (0.35) −2.85 .02
Average Drinks per Drinking Day 0.38 (0.09) 4.10 <.001
Log(k) 0.27 (0.12) 2.31 .04
Proportion Larger-Later Choices All observations 19 8.81 4,14 <.001 0.63
Intercept 4.10 (0.95) 4.32 <.001
Sex (Male) .76 (0.75) 1.01 .33
Student Status (Student) −1.00 (0.47) −2.14 .05
Average Drinks per Drinking Day .37 (0.13) 2.80 .01
Proportion Larger-Later Choices −2.59 (1.37) −1.90 .08
Without Influential Obs. 17 14.63 4,12 <.001 0.77
Intercept 3.97 (0.72) 5.52 <.001
Sex (Male) −0.00 (0.53) 0.00 >.99
Student Status (Student) −.85 (0.33) −2.57 .02
Average Drinks per Drinking Day .42 (0.09) 4.83 <.001
Proportion Larger-Later Choices −2.14 (1.02) −2.10 .06
Peak estimated BAC Log(k)
All observations 19 5.08 4,14 .01 0.48
Intercept .11 (0.01) 12.16 <.001
Sex (Male) −.02 (0.00) −3.21 .01
Student Status (Student) −.01 (0.00) −1.67 .12
Baseline average estimated BAC .16 (0.07) 2.26 .04
Log(k) .01 (0.00) 3.57 .003
Without Influential Obs. 16 2.28 4,11 .13 0.26
Intercept .11 (0.01) 14.10 <.001
Sex (Male) −.01 (0.00) −2.89 .01
Student Status (Student) −.01 (0.00) −2.00 .07
Baseline average estimated BAC .15 (0.07) 2.06 .06
Log(k) .00 (0.00) 2.44 .03
Proportion Larger-Later Choices All observations 19 5.78 4,14 .006 0.52
Intercept .11 (0.01) 13.58 <.001
Sex (Male) −.02 (0.00) −3.33 .005
Student Status (Student) −.01 (0.00) −1.77 .10
Baseline average estimated BAC .17 (0.07) 2.49 .03
Proportion Larger-Later Choices −.05 (0.01) −3.86 .002
Without Influential Obs. 17 1.08 4,12 .41 0.02
Intercept .10 (0.01) 11.39 <.001
Sex (Male) −.01 (0.01) −1.83 .09
Student Status (Student) .00 (0.01) −0.55 .59
Baseline average estimated BAC .08 (0.10) 0.82 .43
Proportion Larger-Later Choices −.03 (0.02) −1.21 .25
Open in a new tab

Highlights.

  • Delay discounting (DD) reflects a lack of sensitivity to future consequences

  • Alcohol administration studies rarely include consequences tied to alcohol use behaviors

  • DD predicted peak estimated blood alcohol concentration (eBAC) in the absence of probabilistic negative consequences

  • DD did not predict eBAC or number of drinks consumed in the presence of probabilistic negative consequences

Funding:

Jillian Rung’s time was supported by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) [T32AA025877, K99AA029732]. Benjamin Berey’s time was supported by NIAAA [T32AA007459]. Robert Leeman’s time was supported by NIAAA [R21AA026918, R34AA029224, R01 AA029488], the state of Florida and the Lane Professorship. The original studies were supported by NIAAA [K01AA019694, R03AA022232, R21AA023368]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIAAA.

Footnotes

Conflict of Interest

None

Declarations of interest: none

CRediT authorship contribution statement

JMR had a primary role in conceptualization, methodology, formal analysis, writing – original draft, writing – review & editing, and visualization. BLB had a primary role in writing – original draft, and writing – review and editing. RFL had a primary role in conceptualization, methodology, investigation, resources, data curation, writing – original draft, writing – review & editing, visualization, supervision, project administration, and funding acquisition.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Acuff SF, Boness CL, McDowell Y, Murphy JG, & Sher KJ (2023). Contextual decision-making and alcohol use disorder criteria: Delayed reward, delayed loss, and probabilistic reward discounting. Psychology of Addictive Behaviors, 37(1), 121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Amlung M, Marsden E, Holshausen K, Morris V, Patel H, Vedelago L, ... & McCabe RE (2019). Delay discounting as a transdiagnostic process in psychiatric disorders: A meta-analysis. JAMA Psychiatry, 76(11), 1176–1186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Leeman RF, Corbin WR, Nogueira C, Krishnan-Sarin S, Potenza MN, & O’Malley SS (2013). A human alcohol self-administration paradigm to model individual differences in impaired control over alcohol use. Experimental and clinical psychopharmacology, 21(4), 303. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Leeman RF, Nogueira C, Wiers RW, Cousijn J, Serafini K, DeMartini KS, … & O’Malley SS (2018). A test of multisession automatic action tendency retraining to reduce alcohol consumption among young adults in the context of a human laboratory paradigm. Alcoholism: clinical and experimental research, 42(4), 803–814. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Leeman RF, Berey BL, Frohe T, Rowland BH, Martens MP, Fucito LM, … & O’Malley SS (2022). A combined laboratory and field test of a smartphone breath alcohol device and blood alcohol concentration estimator to facilitate moderate drinking among young adults. Psychology of addictive behaviors, 36(6), 710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bickel WK, Koffarnus MN, Moody L, Wilson AG, 2014. The behavioral- and neuro-economic process of temporal discounting: A candidate behavioral marker of addiction. Neuropharmacology, 76, 518–527. 10.1016/j.neuropharm.2013.06.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Brandt J (1991). The Hopkins Verbal Learning Test: Development of a new memory test with six equivalent forms. The clinical neuropsychologist, 5(2), 125–142. [Google Scholar]
  9. Brumback T, Cao D, King A, 2007. Effects of alcohol on psychomotor performance and perceived impairment in heavy binge social drinkers. Drug and Alcohol Dependence, 91, 10–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Chait LD, & Perry JL (1994). Acute and residual effects of alcohol and marijuana, alone and in combination, on mood and performance. Psychopharmacology, 115, 340–349. [DOI] [PubMed] [Google Scholar]
  11. Chapman GB, & Elstein AS (1995). Valuing the future: Temporal discounting of health and money. Medical Decision Making, 15, 373–386. [DOI] [PubMed] [Google Scholar]
  12. de Wit H, & Sayette M (2018). Considering the context: social factors in responses to drugs in humans. Psychopharmacology, 235, 935–945. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Gowin JL, Sloa ME, Stangl BL, Vatsalya V, Ramchandani VA, 2017. Vulnerability for alcohol use disorder and rate of alcohol consumption. American Journal of Psychiatry, 174, 1094–1101. 10.1176/appi.ajp.2017.16101180 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Grodin EN, Montoya AK, Bujarski S, Ray LA, 2021. Modeling motivation for alcohol in humans using traditional and machine learning approaches. Addiction Biology, 26(3), e12949. 10.1111/adb.12949 [DOI] [PubMed] [Google Scholar]
  15. Johnson MW, & Bruner NR (2012). The Sexual Discounting Task: HIV risk behavior and the discounting of delayed sexual rewards in cocaine dependence. Drug and Alcohol Dependence, 123(1–3), 15–21 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Kaplan BA, Amlung M, Reed DD, Jarmolowicz DP, McKerchar TL, Lemley SM, 2016. Automating scoring of delay discounting for the 21-and 27-item monetary choice questionnaires. The Behavior Analyst, 39, 293–304. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. King KM, Patock-Peckham JA, Dager AD, Thimm K, & Gates JR (2014). On the mismeasurement of impulsivity: Trait, behavioral, and neural models in alcohol research among adolescents and young adults. Current Addiction Reports, 1, 19–32. [Google Scholar]
  18. Kirby KN, Petry NM, Bickel WK, 1999. Heroin addicts have higher discount rates for delayed rewards than non-drug-using controls. Journal of Experimental Psychology: General, 128(1), 78–87. 10.1037/0096-3445.128.1.78 [DOI] [PubMed] [Google Scholar]
  19. Lawyer SR, & Schoepflin FJ (2013). Predicting domain-specific outcomes using delay and probability discounting for sexual versus monetary outcomes. Behavioural Processes, 96, 71–78. [DOI] [PubMed] [Google Scholar]
  20. MacKillop J, Amlung MT, Few LR, Ray LA, Sweet LH, Munafò MR, 2011. Delayed reward discounting and addictive behavior: A meta-analysis. Psychopharmacology, 216, 305–321. 10.1007/s00213-011-2229-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. MacKillop J, Celio MA, Mastroleo NR, Kahler CW, Operario D, Colby SM, ... & Monti PM (2015). Behavioral economic decision making and alcohol-related sexual risk behavior. AIDS and Behavior, 19, 450–458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Matthews DB, Miller WR, 1979. Estimating blood alcohol concentration: two computer programs and their applications in therapy and research. Addictive Behaviors, 4(1), 55–60. 10.1016/0306-4603(79)90021-2 [DOI] [PubMed] [Google Scholar]
  23. McKerchar TL, Pickford S, Robertson SE, 2013. Hyperboloid Discounting of Delayed Outcomes: Magnitude Effects and the Gain-Loss Asymmetry. The Psychological Record, 63, 441–451. 10.11133/j.tpr.2013.63.3.003 [DOI] [Google Scholar]
  24. Naudé GP, Reed DD, Thornton TJ, & Amlung M (2021). Dual use of alcohol and cannabis among college students: A reinforcer pathologies approach. Experimental and Clinical Psychopharmacology, 29(4), 407. [DOI] [PubMed] [Google Scholar]
  25. Petry NM, 2001. Delay discounting of money and alcohol in actively using alcoholics, currently abstinent alcoholics, and controls. Psychopharmacology, 154, 243–250. 10.1007/s002130000638 [DOI] [PubMed] [Google Scholar]
  26. Raineri A, & Rachlin H (1993). The effect of temporal constraints on the value of money and other commodities. Journal of Behavioral Decision Making, 6(2), 77–94. [Google Scholar]
  27. Rasmussen EB, Lawyer SR, & Reilly W (2010). Percent body fat is related to delay and probability discounting for food in humans. Behavioural Processes, 83(1), 23–30. [DOI] [PubMed] [Google Scholar]
  28. Reynolds B, & Schiffbauer R (2004). Measuring state changes in human delay discounting: an experiential discounting task. Behavioural Processes, 67(3), 343–356. [DOI] [PubMed] [Google Scholar]
  29. Richards JB, Zhang L, Mitchell SH, & De Wit H (1999). Delay or probability discounting in a model of impulsive behavior: effect of alcohol. Journal of the Experimental Analysis of Behavior, 71(2), 121–143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Rung JM, & Madden GJ (2018). Experimental reductions of delay discounting and impulsive choice: A systematic review and meta-analysis. Journal of Experimental Psychology: General, 147(9), 1349. 10.1037/xge0000462 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Rung JM, Peck S, Hinnenkamp JE, Preston E, & Madden GJ (2019). Changing delay discounting and impulsive choice: Implications for addictions, prevention, and human health. Perspectives on Behavior Science, 42, 397–417. 10.1007/s40614-019-00200-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Sloan ME, Gowin JL, Janakiraman R, Ester CD, Stoddard J, Stangl B, & Ramchandani VA (2020). High-risk social drinkers and heavy drinkers display similar rates of alcohol consumption. Addiction biology, 25(2), e12734. 10.1111/adb.12734 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Sobell LC, Sobell MB, 1992. Timeline Followback: A technique for assessing self-reported alcohol consumption, in: Litten R, Allen J (Eds.), Measuring Alcohol Consumption. Humana Press, Totowa, NJ, pp. 41–72. [Google Scholar]
  34. Stangl BL, Vatsalya V, Zametkin MR, Cooke ME, Plawecki MH, O’Connor S, & Ramchandani VA (2017). Exposure-response relationships during free-access intravenous alcohol self-administration in nondependent drinkers: influence of alcohol expectancies and impulsivity. International Journal of Neuropsychopharmacology, 20(1), 31–39. 10.1093/ijnp/pyw090 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sullivan JT, Sykora K, Schneiderman J, Naranjo CA, & Sellers EM (1989). Assessment of alcohol withdrawal: the revised clinical institute withdrawal assessment for alcohol scale (CIWA-Ar). British Journal of Addiction, 84(11), 1353–1357. [DOI] [PubMed] [Google Scholar]
  36. Takahashi T, Ohmura Y, Oono H, & Radford M (2009). Alcohol use and discounting of delayed and probabilistic gain and loss. Neuroendocrinology Letters, 30(6), 749. [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1960213-supplement-1.docx (1.3MB, docx)

RESOURCES