Evaluating effects of episodic future thinking on valuation of delayed reward in cocaine use disorder: a pilot study

Abstract

Background:

Episodic future thinking (EFT; i.e., envisioning oneself in future contexts) has been demonstrated to reduce discounting of future reward in healthy adults. While this approach has the potential to support future-oriented decision-making in substance use recovery, the impact of EFT on discounting behavior in illicit stimulant users has not yet been evaluated.

Objectives:

This pilot study aimed to (1) assess the feasibility of utilizing EFT methods in individuals with cocaine use disorder (CUD) and (2) conduct preliminary measurement of the EFT effect on discounting behavior in this population.

Methods:

Eighteen treatment-seeking individuals with CUD (17 males) were interviewed about positive and neutral events expected to occur at a range of future latencies. Future event information identified by participants was subsequently included on a subset of trials in an intertemporal choice task to promote EFT; within-subject differences in discounting between standard and EFT conditions were evaluated.

Results:

Participants identified relevant events and demonstrated decreased discounting of future reward when event descriptors were included (relative to discounting without event descriptors; p = .039). It was further noted that most events identified by participants were goals, rather than plans or significant dates.

Conclusion:

While methods previously used to study the effect of EFT on discounting behavior in healthy individuals are also effective in individuals with CUD, methodological factors – including types of events identified – should be carefully considered in future work. These preliminary findings suggest that EFT can reduce impulsive decision-making in cocaine use disorder and may therefore have therapeutic value.

Introduction

The need to forego immediate reinforcement in pursuit of delayed reward is an all too common obstacle on the path to positive behavior change. Intertemporal choice paradigms are frequently used to simulate this fundamental dilemma in laboratory settings – e.g., by requiring participants to decide between smaller, immediate rewards and delayed rewards of higher value. Importantly, individual differences in the rate at which reward loses subjective value with increasing temporal distance – a phenomenon known as delay discounting – have been associated with real-life health and financial behaviors (1), and more extreme discounting behavior has been observed in disorders of impaired impulse control including antisocial personality disorder, ADHD, and substance use disorders (2–5). Recently, interventions that support the valuation of delayed rewards by increasing the psychological proximity of these outcomes have attracted attention (6) but have not yet been vetted in key clinical populations of interest.

The intuitive appeal of intertemporal choice paradigms as a model for problematic substance use is especially strong. At a fundamental level, substance use disorders entail the pursuit of short-term benefits of use (e.g., immediate positive and/or negative reinforcement) at the forfeiture of longer-term positive consequences associated with recovery (e.g., improved health, relationships, finances, etc.). The putative neurobiological basis of these conflicting drives has been articulated in various dual systems models of addiction (4,7–10), with studies of intertemporal choice frequently cited as neural and behavioral evidence of intersystem imbalance (11–16). This work has consistently demonstrated aberrant engagement of neural networks underlying cognitive control and prospection during intertemporal choice in individuals with substance use disorders (14,15,17–20). Importantly, evidence suggests that a tendency toward greater discounting of future reward is a vulnerability factor for the initial development of problematic substance use (21–23). Neurocognitive consequences of habitual substance use may further promote pursuit of short-term rewards, giving rise to increasingly impulsive discounting behavior over time (24).

While the extent to which substance use directly alters human intertemporal choice is not known (likely varying by substance and other factors; 25), findings are generally consistent with a pattern of steeper discounting in association with greater substance use severity (5). Similarly, more extreme delay discounting has also been shown to predict worse substance use outcomes – including poorer abstinence and/or retention during treatment (26,27), lower abstinence self-efficacy (28), and more rapid resumption of use following treatment (29). Conversely, individuals with substance use disorders who exhibit more gradual devaluation of delayed reward may be more likely to achieve and maintain sobriety. Provided that intertemporal choice effectively models decisions to use versus abstain in early recovery, strategies to alter this choice behavior have become a focus of great interest in recent years (6,30–32).

Evidence that intertemporal choice behavior is malleable has been available for some time. For example, Loewenstein (33) demonstrated that subtle variations in how a choice context is “framed” – such as whether one is asked to postpone immediate reward versus speed-up delayed reward – can significantly impact choice behavior. Subsequent work on framing in intertemporal choice has revealed multiple factors that may temper impulsive decision-making. Explicit zero manipulations, for example, emphasize that no reward will be received at the unchosen latency (e.g., 10 USD today and 0 USD in one month versus 0 USD today and 30 USD in one month) and have been shown to significantly reduce delay discounting in healthy adults (34–37). Similar reductions in discounting have been shown for manipulations that frame the difference between larger, later and smaller, sooner rewards as a bonus for choosing delayed reward (38) or penalty for choosing immediate gratification (39). Intriguingly, seemingly inconsequential details – such as specifying the timing of delayed reward as a date rather than delay latency can meaningfully influence choice behavior. Not only has this “date-delay framing” effect been shown to reduce discounting in healthy individuals (40, 41; although also see, 42–44), the effect of this manipulation in substance users was large enough that resulting choice behavior was comparable to a healthy comparison group (45).

Similar to the “date-delay framing” effect, discounting behavior in intertemporal choice tasks also appears to be less impulsive when the timing of delayed reward is explicitly linked to a personally relevant future event (e.g., an upcoming vacation). Unlike framing manipulations which superficially alter context, this “episodic future thinking” (EFT) effect is thought to dilate the temporal window in which the decision-maker evaluates potential reinforcers, supporting more robust representation and valuation of delayed reward (46). This phenomenon has been associated with activation of brain regions underlying self-referential mental simulation of future events and is positively correlated with the subjective vividness of associated mental imagery (47). The EFT effect on discounting behavior has now replicated over 20 times and recent meta-analyses highlight EFT-related manipulations as the most promising class of interventions to enhance the subjective valuation of delayed reward in intertemporal choice contexts (6,31).

EFT manipulations have already been successfully utilized to reduce delay discounting in smokers (48), cannabis users (49), and individuals with alcohol use disorder (50). Moreover, corresponding evidence that EFT manipulations reduced actual (48) and hypothetical (50) substance use in these samples: (a) support the ecological validity of delay discounting as a proxy for use-related decision-making and (b) highlight the potential of EFT-based intervention strategies in substance use treatment. However, important questions remain. Given the evidence that future thinking is more greatly impaired in hard drug versus alcohol users (51), it is unclear if benefits of EFT will translate to other substance using populations. Indeed, while Snider and colleagues (50) demonstrated an EFT effect on discounting in alcohol users, this effect was primarily evident in individuals with less severe pathology. Because EFT engages neural substrates of memory, self-referential processing, and cognitive control (52) – which may be impacted differently based on substance use type and severity – further investigation of these factors will clarify the therapeutic potential of EFT.

Cocaine use impacts neurocognitive capacities underlying EFT (53) and evidence from intertemporal choice paradigms suggests this population suffers a significantly constricted temporal horizon (54). Naturally, diminished future orientation can both invite and reflect circumstances that perpetuate future uncertainty (e.g., housing instability, increased risk of adverse legal and medical consequences). Chronic cocaine users commonly present with such circumstances and may therefore have greater difficulty identifying future event information needed to evoke EFT. The primary objective of the current pilot project was therefore to determine if treatment-seeking individuals with cocaine use disorder were able to identify a series of personally relevant future events, expected to occur at latencies ranging from 1-week to 1-year. Features of participant-generated events that may impact effectiveness in evoking EFT (e.g., event type, valence) were also considered to inform future work. Our secondary objective was to conduct a preliminary investigation of the EFT effect in cocaine use disorder using methods configured to capture individual differences. Specifically, unlike previous studies of EFT in substance using populations, we emulated the approach used by Peters and Büchel (47) which omitted methods used to guide or prime EFT. While such methods may be particularly important to support EFT in substance users, they may also obscure individual variation in EFT proclivity. Methods optimized to measure this variation could inform precision delivery of interventions targeting EFT but have not yet been evaluated in individuals with substance use disorders.

Methods

Participants

Participants were recruited through the VA Pittsburgh Healthcare System (VAPHS) Center for Treatment of Addictive Disorders upon referral to an outpatient Contingency Management program for cocaine use disorder. Participants were U.S. military Veterans aged 18–70 with past-month cocaine use and normal or corrected-to-normal vision. Exclusion criteria included the history of traumatic brain injury, neurological disease, bipolar disorder, or psychotic illness. Participants were required to abstain from illicit substances for at least 72 hours (verified by urine- and/or saliva-based drug screens) and provide a negative alcohol breath test prior to assessment. Participants provided written, informed consent and all procedures were approved by the VAPHS Institutional Review Board. Monetary compensation for time and travel was provided. Participants additionally received a monetary bonus of 20-50, USD based on a randomly selected trial from the intertemporal choice task, to encourage naturalistic decision-making.

Procedures and measures

Participants completed an in-person baseline testing session prior to initiating a 24 session course of Contingency Management at VAPHS involving twice-weekly urinalysis and face-to-face visits with a provider. Baseline testing included laptop-based administration of a personalized intertemporal choice task and N-Back working memory task, as well as administration of the Addiction Severity Index and Timeline Follow-Back procedure. Self-report questionnaires and an electroencephalography procedure were additionally completed at this visit (data to be described elsewhere). Our personalized intertemporal choice procedure was based on work by Peters and Büchel (47) and included discounting trials presented under both standard and episodic future thinking (EFT) conditions to investigate differences in discounting behavior using a within-subjects design. Each participant completed a brief in-person interview about upcoming positive and/or neutral events at latencies ranging from approximately one week to one year, which were used to create “event tags” for the EFT condition. Participants viewed a calendar during the interview and were verbally prompted to identify meaningful dates and holidays, as well as other specific events they could foresee occurring during the upcoming year. Importantly, participants were not primed or guided to engage in EFT through episodic specificity induction (for example, see 55) or other elaborative or imaginal strategies used in similar previous work (48–50). Participants subsequently rated each event on personal relevance, valence, and arousal/excitement using a six-point scale.

Six events were selected with latencies approximating delays of 1 week, 2 weeks, 1 month, 3 months, 6 months, and 1 year. When multiple events were generated, events with ratings similar to those of other events included for that participant were favored – in keeping with methods described by Peters and Büchel (47). An “event tag” describing each event (e.g., “New Year’s Day”) was then collaboratively determined for use in a personalized intertemporal choice task created in E-Prime (Psychology Software Tools, Pittsburgh, PA). For each participant, this task included decision trials with (EFT condition) and without (Standard condition) event tag information for each target latency. In the standard condition, a string of hashes (i.e., “############”) was presented as a placeholder for event information, as described by Peters and Buchel (47). Examples of the task interface for EFT and Standard decision trials are depicted in Figure 1. On each trial, participants were able to select their preferred choice (i.e., 12 USD available “today” or a larger amount available at some delay) using the laptop touchpad at their own pace (i.e., there was no time limit for responding).

Figure 1.

Example stimuli representing decision trials from standard (left) and episodic future thinking (right) conditions of the personalized intertemporal choice task.

Each of the 12 conditions (2 Event Conditions (EFT, Standard) x 6 Latencies) were randomly presented in a blockwise-fashion and involved a sequence of trials, beginning with a choice between a value of 12 USD today versus 14 USD at the delay latency. The value of the delayed amount was then incrementally increased until the delayed amount was preferred on two or more consecutive trials. The “switch point” (i.e., value at which preference shifted from immediate to delayed reward) was then evaluated in the opposite direction – starting with the largest possible value of delayed reward (i.e., 50 USD) and incrementally decreasing until preference shifted back to the immediately available option. In both EFT and standard conditions, delays were specified in days. Standard delays were 7, 14, 30, 90, 180, and 365 days. Six distinct EFT delays corresponded with latencies of events identified by participants during the interview.

Data analysis

In order to characterize types of future events identified by participants, each event selected for inclusion in the task was categorized as a plan, goal, or other significant date according to pre-specified definitions for each event type (see Table 1). Participant-identified future event descriptors were subsequently reviewed, using a data-driven approach to extract key themes; a single, best-fitting descriptive theme was later assigned to each event. To validate the reliability of ratings based on pre-specified event type definitions and descriptive themes extracted by the initial reviewer (SF), an independent rater (AO) also reviewed and coded future event information identified by participants. Interrater reliability (IRR) was computed using the approach described by Miles and Huberman (56), yielding an IRR of 0.81 for event type and 0.85 for event theme – suggesting an acceptable level of rating consistency. Likert-scale ratings of personal relevance, valence, and arousal/excitement for events described by participants were additionally summarized by event type and theme using descriptive statistics.

Table 1.

A priori definitions for event tag categories and data-driven event tag themes.

Categories
	Plan	Social or recreational plan made ahead of time.
	Goal	A goal related to personal growth, self-improvement, quality of life, or other personal milestone
	Significant Date	Significant dates including birthdays, holidays, or other scheduled event that is not related to a goal or plan.
Themes
	Recreational	Recreational events including leisure activities, hobbies, and vacations
	Substance Use	Event, activity, or milestone directly related to substance use recovery
	Self-Improvement	Event, activity, or milestone related to self-improvement in mental, physical, or spiritual health domains
	Vocational/Educational	Event, activity, or milestone related to vocational, educational, or career development efforts
	Quality of Life	Event, activity, or milestone related to basic needs (e.g., food, shelter, transportation) and/or Quality of Life
	Social	Event, activity, or milestone related to strengthening/maintaining relationships with family, friends, or community
	Holidays	Holidays or other Seasonal Events for which a specific plan or goal has not been described
	Personal Dates	Significant dates including birthdays and anniversaries of significant life events
	Other	Other significant dates or events (e.g., pay days, due dates, dates of legal proceedings or decisions)

With respect to discounting behavior, “switch points” were determined for each condition by averaging the maximum delayed value resulting in immediate amount selection and the minimum delayed value resulting in delayed amount selection. A hyperbolic discounting model was then applied using R (https://www.R-project.org/), wherein the subjective value of delayed reward (V) is a function of the reward amount (A), delay (D), and discounting rate (k) of the form: V = A/(1+ kD) (see 57). For each participant and condition (EFT, Standard), the parameter k that minimized least square errors across all switch points was determined (see Figure 2 for example discounting functions). Root-mean-square error (RMSE) was subsequently computed for each model, with lower values reflecting better model fit. Paired t-tests were conducted to compare discounting rates and RMSE derived from models applied to EFT versus standard discounting conditions. Due to the skewed distribution of discounting rates, the natural log of k was used for all statistical comparisons.

Figure 2.

Example discounting functions with actual and predicted data representing four individual participants. Participant data depicted in the top left illustrate the Episodic Future Thinking (EFT) effect, as evidenced by more rapid devaluation of delayed reward (i.e., steeper slope of the discounting function) in Standard relative to EFT conditions. In contrast, participants depicted in the bottom left did not exhibit an EFT effect – rather, demonstrating similar discounting behavior across both conditions. Please note that individual participants included in the current figure are identified using markers of the same color in Figure 3 (e.g., cyan data markers represent Subject 07 in both figures) to better inform interpretation of log(k) values depicted therein. Discounting functions representing median actual and predicted values for the sample under EFT and Standard conditions are also presented in black at the right side of the figure.

Within-subject differences in discounting behavior between EFT and standard conditions were evaluated using a paired t-test. The magnitude of the EFT effect was also quantified at the individual-level by subtracting the natural log of k computed for the EFT condition from the natural log of k computed for the standard condition, consistent with methods described by Peters and Büchel (47). The resulting value (i.e., log(k)_{Standard –} log(k)_EFT) will hereafter be referenced as Δlog(k)_Standard-EFT. The relationship between Δlog(k)_Standard-EFT and treatment outcome (i.e., % cocaine-negative urines during Contingency Management), as well as between Δlog(k)_Standard-EFT and working memory function (i.e., N-Back task performance), were also considered in exploratory Pearson correlation analyses. With respect to the latter, the sensitivity index (i.e., the difference between the z-transformed hit rate and the z-transformed false alarm rate, also known as d’) from the 2-Back condition of the task was used as a measure of working memory function. Additional exploratory correlational analyses considered whether subjective ratings of events (i.e., personal relevance, valence, arousal/excitement, averaged across events for each participant) or the proportion of goal-type events contributed to individual variability in the EFT effect.

Results

Twenty-one Veterans enrolled in the current pilot and 18 were fully eligible to proceed with baseline testing. This sample was predominantly male (1 female), with a mean age of 51 (SD = 14). Participants reported an average of 5.3 previous episodes of substance use treatment (SD = 2.8), 14.2 lifetime years of cocaine use (SD = 13.6), and 1,133 USD (range: 20 USD-7,000 USD; SD = 1,543 USD) in cocaine use during the month preceding the current episode-of-care. All but one participant was concurrently in treatment for one or more other substance use disorders (opioid use disorder: 10/18; cannabis use disorder: 7/18; alcohol use disorder: 6/18; sedative, hypnotic, or anxiolytic use disorder, 2/18; amphetamine use disorder: 1/18) and 9/18 were in medication-assisted treatment for opioid use disorder at the time of research enrollment. All participants completed the future events interview during the baseline assessment visit and were subsequently followed throughout a 24 session course of Contingency Management. Fourteen participants from the current sample also successfully completed the personalized intertemporal choice task at baseline, in which information from the future events interview was incorporated in the form of event tags in the EFT condition. Of the four participants who did not successfully complete the task, three were impacted by technical malfunctions that resulted in data loss and one declined to participate due to fatigue. Participants who completed the personalized intertemporal choice task provided an average of 12/24 (specifically, 51.8% (SD = 33)) cocaine-negative urines during the Contingency Management treatment interval (range: 4% to 100%).

Participants successfully identified 107 out of 108 total events requested; a single participant was unable to identify an event for the one-week target latency and the suggestion to use an upcoming holiday was provided. Mean latencies (reported here in days) roughly approximated target latencies, with greater variability noted for more distal targets: 1 week (M = 5.17, SD = 2.66), 2 weeks (M = 14.94, SD = 2.51), 1 month (M = 35.78, SD = 5.98), 3 months (M = 90.44, SD = 27.89), 6 months (M = 184.67, SD = 20.55), 1 year (M = 339.61, SD = 47.24). Evaluation of events identified by participants revealed that goals were the most frequently identified event type, comprising 51.8% of total events – with plans and significant dates representing another 27.8% and 20.4%, respectively. Event ratings with respect to personal relevance, valence, and arousal/excitement are further summarized by event type and theme in Table 2, along with corresponding examples of participant-generated events. Across the sample, event ratings ranged from 1 to 6 across all three domains but were high, on average, consistent with successful identification of future events with strong personal relevance (M = 5.42, SD = 0.94), positive valence (M = 5.39, SD = 0.95), and high levels of excitement/arousal (M = 5.16, SD = 1.20).

Table 2.

Summary of future event categories, themes, and subjective ratings.

			Average Rating
Category	Theme	Proportion	Personal Relevance	Valence	Arousal/Excitement	Example
Plan		30/108	5.6 (0.7)	5.6 (0.7)	5.6 (0.9)
	Recreational	14/30	5.5 (0.9)	5.2 (0.9)	5.5 (1.2)	Attending first high school football game of the season.
	Social	16/30	5.7 (0.6)	5.8 (0.4)	5.7 (0.6)	Celebrating Valentine’s Day with my girlfriend.
Goal		56/108	5.5 (0.9)	5.4 (1.0)_	5.1 (1.2)
	Substance Use	13/56	5.3 (0.9)	5.4 (0.8)	4.5 (1.5)	Celebrating one year sober.
	Self-Improvement	8/56	5.9 (0.4)	5.4 (1.1)	4.8 (1.8)	Getting a new gym membership.
	Vocational/Educational	15/56	5.5 (1.1)	5.6 (0.9)	5.3 (1.1)	Starting full-time employment.
	Quality of Life	13/56	5.5 (1.0)	5.2 (1.2)	5.3 (0.8)	Buying a new car.
	Social	4/56	6.0 (0.0)	6.0 (0.0)	6.0 (0.0)	Volunteering to work the polls at the next election.
Significant Dates		22/108	5.0 (1.2)	5.1 (1.1)	4.8 (1.4)
	Holidays	7/22	4.7 (1.8)	4.6 (1.7)	4.6 (1.8)	Thanksgiving Day.
	Personal Dates	11/22	5.0 (0.9)	5.5 (0.7)	5.3 (0.8)	My granddaughter’s birthday.
	Other	4/22	5.3 (0.5)	5.0 (0.0)	3.8 (1.5)	Expecting a favorable outcome in court.

RMSE values representing hyperbolic discounting model fit were acceptable for models fit to datapoints derived from both EFT (M = 6.9, SD = 3.0) and Standard (M = 6.5, SD = 3.2) conditions, approximately corresponding to 10% of the range of possible target values (i.e., subjective value of 50 USD, ranging from 0-50 USD). Importantly, there was also no difference in model fit, as represented by RMSE, between the two experimental conditions (t(13) = 0.595, p = .562, two-tailed).

To evaluate the presence of an EFT effect in the sample, the log-transformed discounting parameter, k, was compared between EFT and standard conditions. When considering log-transformed values of k, more positive values reflect a steeper discounting slope and, therefore, more rapid devaluation of delayed reward. Consequently, a positive difference in log-transformed k-values between standard and EFT conditions (i.e., Δlog(k)_Standard-EFT) reflects reduced discounting under the EFT condition. As depicted in Figure 3, while the overall distribution of log(k) values was comparable for EFT and Standard conditions, log(k)_EFT was more negative than log(k)_Standard for the majority of participants, resulting in predominantly positive values of Δlog(k)_Standard-EFT. Statistical comparison further revealed a significant reduction in log(k) values derived from the EFT condition, relative to the Standard condition – reflecting reduced discounting of delayed reward in the presence of personally relevant future event tags (t(13) = 2.29, p = .039, two-tailed; d = 0.612).

Figure 3.

Plot of individual discounting rates (log(k)) by condition (Panel A) and difference in discounting rate between standard and EFT conditions (Δlog(k)_Standard-EFT; Panel B). Less negative log(k) values represent steeper discounting and positive values of Δlog(k)_Standard-EFT reflect a larger Episodic Future Thinking (EFT) effect on discounting behavior. Each unique marker color used in Panels A and B represents the same individual participant.

Correlation coefficients were near zero for the correlation between Δlog(k)_Standard-EFT and both % cocaine-negative urines during treatment and the sensitivity index (d’) from the 2-Back working memory task. Correlations between magnitude of the EFT effect and average subjective ratings of events with respect to personal relevance, valence, and arousal/excitement, as well as proportion of future events categorized as goals, were also nonsignificant (both with and without correction for multiple comparisons). An analogous exploratory correlation analysis targeting standard discounting behavior (i.e., log(k)_Standard) similarly identified no statistically significant linear associations.

Discussion

Results of the current pilot study demonstrate preliminary feasibility of EFT paradigms in treatment-seeking individuals with cocaine use disorder, including successful generation of personally relevant future events at latencies ranging from one-week to one-year. Moreover, future events identified by our participants effectively reduced discounting of delayed reward in the current sample. Most remarkably, a significant effect was observed despite omission of episodic specificity induction or related elaborative/imaginal strategies – which have been used in all previous work examining EFT in substance using samples. It was further identified that most future events generated by treatment-seeking individuals with cocaine use disorder were goals (rather than plans or other significant dates). Goals have previously been demonstrated to more robustly invoke EFT-related reductions in delay discounting than other types of future events (58). The special significance of goal-oriented EFT has been highlighted in previous research targeting substance using samples (48,50) but has not been directly investigated in previous work.

While goals were frequently identified as future events in the current sample, individuals who identified more goals did not clearly manifest a stronger EFT effect. Though this pilot study has a modest sample size and is therefore underpowered to detect anything but relatively large effects, this observation may still inform future work. Specifically, it may be that treatment-seeking substance users are more likely to engage in EFT referencing an ideal future self with respect to goals and, perhaps, other future events. Goal-oriented EFT has already been shown to more robustly engage brain regions underlying prospection and cognitive control (59) – potentially supporting larger reductions in impulsive decision-making noted elsewhere (58,60). Furthermore, because the ideal future self is a goal-state of its own accord (61) any future event envisioned from this perspective (e.g., performing a mundane task) may afford additional self-regulatory and motivational benefits. Adding subjective assessment of the degree to which future event cues evoke the ideal future self may help to clarify this factor in future work targeting treatment-seeking substance users.

The strength of the EFT effect on discounting behavior was also unrelated to substance use treatment outcomes and working memory function in the current sample. These results require replication; however, some potentially interesting implications are nonetheless worth noting. First, while we avoided using strategies like episodic specificity induction to probe individual proclivity toward EFT, such techniques enable measurement of individual capacity for EFT – which could map more strongly to functional outcomes and presumed cognitive correlates. For example, measuring the strength of the EFT effect in the context of an exercise that enhances this type of cognition (i.e., episodic specificity induction) may better predict response to substance use interventions that similarly aim to support EFT. Alternatively, omitting episodic specificity induction and related elaborative/imaginal strategies may provide for more naturalistic assessment of EFT tendencies with promise to identify individuals for whom EFT-oriented interventions could be most beneficial. Next, while working memory appears to moderate the EFT effect on delay discounting in healthy individuals, this relationship was primarily evident for neutral, as opposed to positively valenced, future events (62). Notably, events identified by our sample were primarily positive (average rating of 5.39 out of 6). It is also possible that other factors that are uniquely represented in clinical samples contribute to individual differences in the EFT effect in a manner independent of working memory function (e.g., increased mortality salience, see 63, trauma exposure, see 64). Such factors should be considered in future work investigating the relationship between working memory function and EFT in clinical populations.

The current pilot study provides preliminary evidence of the EFT effect in treatment-seeking individuals with cocaine use disorder and several limitations will be addressed in a larger-scale implementation. This work will consider other potential mediators and moderators of the EFT effect in this population – including the possible role of demand characteristics in shaping EFT-related discounting behavior, as suggested by Rung and Madden (6). In keeping with the methods of Peters and Büchel (47), the current pilot did not investigate EFT relative to an episodic recent thinking control condition (see 65). However, we intend to specifically consider the EFT effect relative to a more robust control condition incorporating date-delay framing (see Introduction), as well as episodic recent thinking in the future.

Conclusion

The current work provides evidence that existing EFT manipulations targeting discounting behavior can be successfully implemented in individuals with cocaine use disorder. Furthermore, we provide preliminary evidence of reduced delay discounting in the context of EFT in this population, as well as the first evidence of an EFT effect in substance users, in the absence of elaborative and imaginal strategies used to promote EFT. Based on observations from the current pilot study, future events related to personal goals (or representing the ideal future self, more broadly) should be carefully considered in future work. Treatment-seeking substance users may be especially likely to think about the future from the perspective of an ideal future self and this may have important implications for the empirical study of EFT, as well as development of substance use interventions integrating EFT techniques.