Written or drawn episodic future thinking cues improves delay discounting in adults

Abstract

Delay discounting (DD) describes choices between small, immediate rewards and larger, delayed rewards. Individuals who are high in DD favor small, immediate rewards, and this preference is related to health behaviors including higher energy intake, smoking and less physical activity. Episodic future thinking (EFT) is an intervention in which one thinks about personal positive future events and this decreases DD in adults and children. In previous studies episodic events have been presented as written or auditory cues. Episodic future images are also imagined visually, but the impact of personal visual cues has not been tested. Research examining sensory modality and semantic memory has shown drawn items are associated with better recall than writing or viewing provided images. This study compared drawn versus written episodic future or recent cues on DD. Sixty-nine adults were randomized to one of three groups; EFT-written, EFT-drawn or Episodic recent thinking (ERT)-written cues, and completed a computerized adjusting amount DD task cued with episodic events. Results showed both written and drawn EFT cues had a larger effect on DD than ERT-written cues and individual differences in immediate time perspective moderated this effect. This suggests that drawn and written cues can have similar effects on DD, providing future clinical work flexibility in how to present cues in the field. In addition, presenting drawn cues may improve DD for individuals who have an immediate time perspective.

1. Introduction

Delay discounting (DD) is a measure of one’s preference for receiving smaller immediate rewards versus larger rewards after a delay (Rung & Madden, 2018). Individuals who prefer smaller immediate rewards are described as high discounters, as they devalue future rewards. DD has been related to health behaviors, including energy intake (Appelhans et al., 2012) and obesity (Amlung, Petker, Jackson, Balodis, & MacKillop, 2016), smoking (Baker, Johnson, & Bickel, 2003; Sheffer et al., 2012), and exercise frequency (Daugherty & Brase, 2010; Epstein et al., 2020). DD has been described as a trans-disease process (Bickel et al., 2019), due to its relationship with health behaviors, making it an important potential target for changing behavior.

There are several experimental manipulations that can modify one’s DD in a state-dependent manner (Rung & Madden, 2018; Scholten et al., 2019). Episodic future thinking (EFT) is an intervention that reliably modifies DD (Rung & Madden, 2018; Schacter, Benoit, & Szpunar, 2017), as well as modifying energy intake (O’Neill, Daniel, & Epstein, 2016), online grocery shopping (Hollis-Hansen, Seidman, O’Donnell, & Epstein, 2018), smoking demand (Stein et al., 2016), alcohol use (Snider, LaConte, & Bickel, 2016) and environmental conservation intentions (Lee, Sung, Wu, Ho, & Chiou, 2018). EFT involves asking adults to imagine and describe positive, personal events within different future temporal windows (Atance & O’Neill, 2001; Peters & Buchel, 2010). Engaging in future episodic thought can decrease discounting of the future and shift preferences from smaller, immediate rewards towards larger, delayed rewards (Daniel et al., 2013a, 2013b).

Episodic future thinking is based in part on retrieving previous experiences from autobiographical memory (Schacter, Addis, & Buckner, 2007; Wang, Yue, & Huang, 2016) and integrating this information into future events using prospective memory (Schacter et al., 2017). Memory research suggests that cues with high visual imagery can stimulate greater memory specificity when engaging in autobiographical memory retrieval (Williams, Healy, & Ellis, 1999). Along with this, research suggests that drawing, as compared to writing and viewing picture cues, improves semantic memory (Fernandes, Wammes, & Meade, 2018). In a series of eight studies, Fernandes et al. (2018) have shown that drawing, versus writing or mentally visualizing, improved memory for word lists and scientific concepts. Current EFT approaches use written or audio cues presented during DD tasks, during which individuals are asked to visualize their events (Daniel et al., 2013a, 2013b; O’Donnell, Oluyomi Daniel, & Epstein, 2017), but drawn cues have not been examined.

Time perspective is a general form of prospective thinking, describing the extent to which a person considers the future versus the present when making everyday decisions (Strathman, Gleicher, Boninger, & Edwards, 1994). Individual differences in time perspective are related to differences in the production of episodic future events with greater richness/experiential presence in those with greater future orientation (Gollner, Ballhausen, Kliegel, & Forstmeier, 2017). Future time perspective is also inversely related to DD, as greater future orientation is related to less future discounting (Acuff et al., 2017; Gollner et al., 2017). Considering individual differences in time perspective may be important to EFT’s effects on DD.

Given that the effects of drawing episodic or prospective memories has not been examined, visual imagery may stimulate greater memory recall than verbal cues (Dando, 2013; Williams et al., 1999) and EFT is based in part on retrospective memory (Schacter et al., 2007; Wang et al., 2016), we wanted to compare the effects of drawn episodic future cues and written episodic future cues to written episodic recent cues as the control group. Episodic recent thinking (ERT) is a common control group in EFT studies as it controls for episodic thinking, and creation of cues, but does not focus on the future, or rely on retrospective memory to create cues. We hypothesized that both the drawn and written future cues would result in lower discounting compared to the recent thinking control group. We also explored whether individual differences in time perspective moderated the effect of EFT on DD, as a lower immediate time perspective is related to both episodic future cue vividness and DD.

2. Methods

2.1. Participants

Participants were 72 adults (30 males, 42 females) ages 18–55 years old. Adults were recruited for high DD, assessed by an adjusting delay task (Koffarnus & Bickel, 2014) to prevent ceiling effects for the future thinking groups (Hollis-Hansen et al., 2018). Sample characteristics are presented in Table 1. Exclusionary criteria included 1) psychopathology (e.g. depression), 2) adjusting delay DD k-value < 0.00194 (equivalent to choosing the immediate reward when the delay is 1-year), 3) not currently pregnant 4) conditions that interfere with drawing/handwriting or using a computer for about 1 h, and 5) recent participation in a study using future thinking methods (previous 6 months). A DD threshold was used as inclusionary criteria to prevent a ceiling effect, whereas participants who do not discount the future have little room for improvement with EFT cues, a criterion used in previous studies of EFT (Hollis-Hansen et al., 2018; Mansouri, Crandall, & Temple, 2020; O’Neill et al., 2016). Pregnant women were excluded as there is concern that their episodic and verbal memory may not reflect the general population (Henry & Rendell, 2007; Wilson et al., 2011). One hundred and seventy-eight adults were screened for eligibility, with 65 being ineligible for participation (32 with low DD, 28 with psychological disorders, 3 pregnant, 1 recent experience with EFT, 1 reporting difficulties with physical writing), 8 declined to participate, and 19 were eligible but unable to contact and 12 were on hold due to filled stratification cells. Three participants scored less than 0.00194 (representing no discounting on rewards at a 1 year delay) on the adjusting-delay discounting task during the laboratory session after randomization and were excluded from data analysis for a final n = 69.

Table 1.

Participant Characteristics

	Overall	EFT Writing	ERT Writing	EFT Draw	p
Sex (M/F)	30/39	10/13	11/12	9/14	0.84
Age (years)	32.2 ± 10.0	32.0 ± 9.2	33.2 ± 10.6	31.4 ± 10.6	0.83
Age Range	19.5 – 55.4	20.9 – 53.6	19.5 – 52.6	20.0 – 55.4	--
BMI	27.9 ± 7.6	28.7 ± 9.5	27.7 ± 6.2	27.3 ± 6.9	0.82
Income ($US)	81235 ± 75043	71424 ± 67195	79102 ± 56092*	93087 ± 96841	0.62
Education (years)	14.9 ± 2.1	15.2 ± 2.4	14.9 ± 1.9	14.6 ± 2.1	0.59
PHQ-9	2.3 ± 2.0	2.0 ± 1.8	2.5 ± 2.0	2.3 ± 2.1	0.63
Adjusting Delay Discounting k-value	0.055 ± 0.121	0.052 ± 0.114	0.052 ± 0.062	0.061 ± 0.171*	0.96
Student Status
Full time	38	13	11	14	0.66
Part time	28	9	10	9
Not a student	3	1	2	0
Race/Ethnicity
White and non-Hispanic	47	16	16	16	0.94
Non-White and/or Hispanic	22	7	7	8
Consideration of future Consequences	53.7 ± 7.4	53.9 ± 8.5	53.4 ± 8.7	53.8 ± 4.8	0.97
Immediate subscale	16.6 ± 5.0	16.5 ± 5.4	17.0 ± 6.2	16.3 ± 3.0	0.90
Future subscale	28.3 ± 3.4	28.4 ± 4.1	28.4 ± 3.4	28.2 ± 2.5	0.97
Cue Ratings during DD Task
Average cue frequency	4.4 ± 0.6	4.5 ± 0.6	4.4 ± 0.6	4.2 ± 0.7	0.33
Average vividness of cues	4.3 ± 0.7	4.3 ± 0.7	4.2 ± 0.6	4.2 ± 0.8	0.76

Mean ± Standard Deviation,

^*One missing data point

2.2. Power and sample size

Sample size was determined based on effect sizes of EFT versus ERT differences on area under the curve (AUC), ranging from 0.65 to 1.25 (Hollis-Hansen, O’Donnell, Seidman, Brande, & Epstein, 2019). Using the median effect size of Hedges g = 0.83, α = 0.05 and 80% power, differences between EFT and ERT can be seen with 24 per group, or 72 participants. Power to detect the effects of drawing versus writing was also examined and based on the effects of drawing versus visualizing on semantic memory (Cohen’s d = 0.87), which can be seen with 22 per cell (α = 0.05, power = 80 %) for a total of 66 participants.

2.3. Procedure

Participants were recruited from a database of individuals involved in previous studies at the University at Buffalo Behavioral Medicine laboratory, online advertisements (e.g. Facebook, Craigslist) and ads on the University at Buffalo Clinical Trials website from August 2019 – February 2020 in Buffalo NY. Participants first completed an online or phone screening questionnaire to determine eligibility. Eligible participants were scheduled for an appointment and randomized to one of three groups (EFT-written, EFT-drawn, ERT-written) in pre-randomized blocks of six (rand in Microsoft Excel (Microsoft Inc, Redmond WA)), stratified by gender and race/ethnicity (White/non-Hispanic, non-White or Hispanic) with a 1:1:1 group allocation ratio. Eligible participants were assigned consecutive identification numbers upon scheduling an appointment. After signing consent forms, participants completed the Patient Health Questionnaire – 9 (PHQ-9) and the 5-trial adjusting delay discounting task (Koffarnus & Bickel, 2014) as eligibility checks, in addition to the consideration of future consequences scale (Joireman, Shaffer, Balliet, & Strathman, 2012).

The PHQ-9 was used to screen for depression (Kroenke & Spitzer, 2002), as depression is associated with deficits in one’s ability to experience anticipatory pleasure (Hallford, Barry et al., 2020) and is associated with difficulties in forming prospective memories (Roepke & Seligman, 2016). No participants were excluded for PHQ-9 scores greater than 15 as used in previous EFT studies (Stein et al., 2017). The minute 5-trial adjusting delay discounting task (Koffarnus & Bickel, 2014) was used to screen individuals during the appointment prior to data collection. Individuals were asked to choose between hypothetical monetary rewards of $50 now or $100 at various delays. The delay to the larger reward was adjusted across 5 trials to score a k-value. Individuals with k-values smaller than or equal to 0.00194, equivalent to choosing the delayed reward when given a choice between immediate and 1-year rewards, were considered ineligible and were not scheduled for appointments. If a participant re-screened during the appointment scored less than 0.00194, they were excluded from data analysis (n = 3).

Participants then created a set of five episodic cues guided by the experimenter as described below. During the DD task, cues were displayed on a stand next to the computer screen and participants were instructed to think about their cue for approximately 30 s before each delay time point. Participants completed vividness and frequency questions after each time point before the next cue, indicating how vividly and frequently they thought about their cues during the DD task. At the end of the session, participants completed a demographics questionnaire and had their height and weight measured. Participants were compensated $25 for their time. All procedures were approved by the University at Buffalo’s Internal Review Board.

2.4. Episodic cue groups

Both the EFT and ERT groups followed procedures outlined by O’Donnell et al. (2017). In brief, participants were guided by the experimenter and asked to identify and describe five positive, personal events. Each event was rated on a 1–5 Likert-like scale (1 – not at all, 5 – very much) for liking/enjoyment, importance, excitement, and two vividness questions. If participants rated cue vividness less than 3, they were asked to choose a different event. After rating their initial event cue, participants were asked to provide details including who they were with, what they were doing, where they were and how they were feeling. Participants and experimenters filled out a checklist for each detailed cue, indicating use of the correct tense, inclusion of specific details and a focus on positives. Participants were not prompted to discuss any specific type of event, besides being positive and matching the specified time period.

The EFT cue generation task asked participants to imagine and describe a vivid, positive future event they were looking forward to for five time periods of 1 month, 6 months, 1 year, 2 years and 5 years. The ERT cue generation task asked participants to remember and describe a vivid, positive recent event they enjoyed for five time periods of 1 day, 2 days 3 days, 4 days and 5 days ago. ERT controls for personalization and episodic thinking, but does not orient them to the future or activate retrospective memory. The time periods for ERT were chosen to be close in time to the present, so that retrospective thinking does not activate prospection (Hollis-Hansen, O’Donnell, Seidman, Brande, & Epstein, 2019; Schacter & Addis, 2007).

2.4.1. Cue creation modalities

Immediately after creating each cue, participants were instructed to either draw their event or write their event by hand based on their randomized group. All groups were provided a blank sheet of white paper inscribed with a large rectangle, pens, pencils and a set of colored pencils. Participants were read instructions by the experimenter and given three minutes to create a written or drawn cue. During the three minutes, the written cue was displayed on the computer screen along with a 3-minute timer. Participants were asked to move on whether or not they completed their drawn/written event when the time was up. After generating the handwritten cue, a manipulation check asked if participants could read/identify their cue.

2.4.2. Drawn cues

Participants in the drawn cue condition were asked to draw out their event after creating the cue with the experimenter. They were provided with the instructions below:

• Instructions: Now you will draw your event on the sheet of paper in front of you. This drawing will be used in the next task, so please include the details in your event description. Try to focus on the positive parts of your event. You’ll have 3 min to draw this event. You may use any of the pens/pencils that you want, but you cannot start over. Your drawing should be something that you can recognize, it will not be judged on quality.

2.4.3. Written cues

Participants in the written cue condition were asked to copy their event verbatim. They were asked to include all the details from their event, but were allowed to add additional details.

2.5. Measures

2.5.1. Adjusting amount DD

Adjusting amount DD (Frye, Galizio, Friedel, DeHart, & Odum, 2016) was used with five time points, 1 month, 6 months, 1 year, 2 years and 5 years, with a delayed reward of $100. Each time point included five adjusting questions for the immediate reward depending on their choice for the previous iteration, to calculate an indifference point for each time point. During the task, the personalized cues were displayed on a stand next to the computer screen before each time period and during each choice. Participants were first asked to think about the cue for at least 30 s before each delay period, then make their choices about the different delayed rewards. Participants were also asked two questions as an attention check, $0 in 1 day or $50 now, and $100 in 1 day or $0 now. After each time point, participants were asked to rate how frequently and vividly they thought about their cues during the previous time point (1 – not at all, 5 – very much). Area under the curve (AUC) was calculated (Myerson, Green, & Warusawitharana, 2001) as the dependent measure.

2.5.2. Consideration of future consequences scale

The consideration of future consequences scale (CFCS) is a 14-item scale measuring individual differences in how influenced one’s behavior is by both possible present and future outcomes of action (Joireman et al., 2012; Strathman et al., 1994). CFCS includes two subscales, immediate consequences (Cronbach’s alpha = 0.78, e.g. I only act to satisfy immediate concerns, figuring that I will take care of future problems that may occur at a later date) and future orientation (Cronbach’s alpha = 0.68, e.g. I am willing to sacrifice my immediate happiness or well-being in order to achieve future outcomes) with comparable Cronbach’s alphas to previously reported estimates (0.80 for immediate subscale, 0.82 for future subscale) (Joireman et al., 2012). The future and immediate subscales have been shown to have differential correlations with the future subscale being positively correlated with promotion orientation, while the immediate subscale is negatively correlated with promotion orientation and the future subscale is correlated with health eating and exercise intentions, while the immediate subscale is not (Joireman et al., 2012). The future and immediate subscales have been shown to have acceptable test-retest reliability (immediate, r = 0.77, future, r = 0.74) (Nigro, Cosenza, Ciccarelli, & Joireman, 2016).

2.5.3. Demographics

Race/ethnicity, employment status, household income and education level, in addition to subjective community/national social status, were assessed using a standard questionnaire (Stewart, 2004).

2.5.4. Anthropometrics

Weight was assessed by a digital scale (TANITA Corporation of America Inc, Arlington Heights, IL) and height using a digital stadiometer (Measurement Concepts & Quick Medical, North Bend, WA). Body mass index (BMI) was calculated according to the formula: BMI = kg/m².

2.6. Analytic plan

Participant demographics and anthropometrics were examined by group using analysis of variance for continuous variables and chi-square for categorical variables. Group differences in participant characteristics were considered as potential covariates in subsequent analyses. Group differences in frequency of thinking about cues and vividness of cues during the DD task was conducted using a one-way ANOVA. Next, an ANOVA was used to examine main effects of group on DD quantified as area under the curve (AUC), with post-hoc linear contrasts to compare individual groups and Cohen’s d effect sizes were calculated for group contrasts. To examine null effects between groups, Bayesian ANOVA was conducted with non-informative priors (r scale fixed effects = 0.5), in JASP (JASP Team, 2020). Post-hoc tests examined Bayes factor B10, comparing the model versus the null hypothesis, with the posterior odds corrected for multiple testing and individual comparisons completed with a Cauchy (0, r = 1/SQRT(2)) prior. To examine moderators of the effect of group on AUC, regression models with group dummy coded (ERT-written as the reference group), were used. A Bayesian linear regression was first used to specify the model, using uniform model priors and JZS (r scale = 0.354) coefficient prior distribution in JASP (JASP Team, 2020). In the first step, the moderator was added to a regression model with two dummy-coded variables. In the second step, the interactions between the moderator and dummy coded variables were included. Incremental f-test was used to test significant increases in variance between step 1 and step 2 and eta squared (η2) effect sizes were calculated for the interactions. Immediate and future time perspective (consideration of future consequences scale, CFCS), age, sex and BMI were examined as potential moderators of the effect of group on AUC. Moderators were centered before analysis. Simple slopes were used to examine significant moderators of group by examining the main effects of group at ±1 standard deviation of the centered moderator and the group contrasts were re-examined with EFT-drawn as the reference group as outlined by West, Aiken, & Krull (1996). Analyses were conducted in JASP 0.14.1 (JASP Team, 2020).

3. Results

Participant demographics and anthropometrics by group are presented in Table 1. There were no significant differences between the groups on any participant characteristics. The ANOVA showed significant between group differences in AUC (F(2,66) = 3.45, p = 0.038). Post-hoc linear contrasts showed that EFT-written (0.493 ± 0.293) and EFT-drawn (0.491 ± 0.272) were significantly different from ERT-written (0.307 ± 0.261) (t(,66) = 2.63, p = 0.011, Cohen’s d = 0.68), and there was not a difference between EFT-drawn and EFT-written (t(66) = 0.02, p = 0.98, Cohen’s d = 0.01). The comparison between EFT-written and EFT-drawn was also examined using a Bayesian ANOVA, and the comparison between EFT drawn and EFT written had a Bayes Factor B10, (uncorrected) of 0.292 (Lee & Wagenmakers, 2013; Wagenmakers et al., 2018), indicating that there is moderate evidence for the null hypothesis, or that the groups are not different. Fig. 1A shows the indifference points for each group.

Figure 1.

A. Delay discounting curves by group. Average group indifference points for each time delay (±SEM) are displayed in US dollars ($), with a delayed reward of $100. The drawn and written groups have been slightly offset for ease of viewing. B. Interaction between immediate time perspective and group on DD. Area under the curve was estimated using the linear regression model with group dummy coded and ERT written as the reference group. Contrasts were calculated using simple slopes by re-centering immediate time perspective at ± 1 Standard deviation *p<0.05

When examining the potential moderation models, a Bayesian regression analysis was used with all two-way interactions for potential moderation of group by immediate time perspective, future time perspective, BMI, sex and age. The model including the interaction between group and immediate time perspective, covarying sex and BMI, was 1.26 times better than the next best fit including the interactions between group and immediate time perspective and group and sex. It was also 47.2 times better than the model including only the main effects of group and immediate time perspective, covarying BMI and sex. Moderation analysis showed CFCS immediate subscale significantly moderated the effect of group on AUC and increased the model variance accounted for (FINC(2, 61) = 4.42, p = 0.003, Table 2), with significant differences between EFT drawn versus EFT written (b = −0.03, CI = 0.006, 0.055, t = 2.47, p = 0.016), and between EFT drawn and ERT written (b = −0.06, CI = −0.095, −0.022, t =3.17, p = 0.002), but not EFT written versus ERT written (b = 0.03, CI = −0.010, 0.066, t = 1.47, p = 0.146).

Table 2.

Interaction between group and CFCS immediate time perspective

	B (SE)	CI	t	p
Step 1
Constant	0.052 (0.126)	−0.200, 0.305	0.42	0.68
BMI	0.006 (0.004)	−0.002, 0.014	1.45	0.15
Sex	0.145 (0.062)	0.021, 0.269	2.34	0.022
CFCS-Immediate	−0.018 (0.006)	−0.031, −0.006	2.96	0.004
EFT written vs. ERT written	0.187 (0.075)	0.038, 0.337	2.50	0.015
EFT drawn vs. ERT written	0.207 (0.075)	0.057, 0.356	2.76	0.008
Model F-test F(5, 68) = 4.77, p < 0.001, R2 = 0.275
Step 2
CFCS-Immediate × (EFT written vs. ERT written)	0.028 (0.019)	−0.010, 0.066	1.47	0.15
CFCS-Immediate × (EFT drawn vs. ERT written)	0.058 (0.018)	0.022, 0.095	3.17	0.002
Model F-test F(7, 68) = 5.81, p < 0.001, R2 = 0.400, ΔR2 = 0.125, FINC(2, 61) = 6.38, p = 0.003

^*CFCS-immediate, Consideration of future consequences immediate time perspective; EFT written group versus ERT written dummy coded variable; EFT drawn group versus ERT written dummy coded variable

As seen in Fig. 1B, when immediate time perspective was high there was a significant main effect of group between EFT drawn and ERT written (b = 0.50, CI = 0.27, 0.74, t = 4.28, p < 0.001), EFT written and ERT written (b = 0.34 CI = 0.10, 0.58, t = 2.79, p = 0.007) but not between EFT written and EFT drawn (b = −0.17, CI = −0.35, 0.02, t = 1.81, p = 0.08). When immediate time perspective was low, there were no main effects of group, EFT written versus ERT written (b = 0.06, CI = −0.18, 0.29, t = 0.48, p = 0.63), EFT drawn and ERT written (b = −0.08, CI = −0.31, 0.14, t = 0.74, p = 0.46), and EFT written and EFT drawn (b = 0.14, CI = −0.05, 0.33, t = 1.48, p = 0.14). Following this, time perspective was significant for EFT written (b = −0.03, CI = −0.048, −0.012, t = 3.25, p = 0.002) and ERT written (b = −0.058, CI = −0.091, −0.025, t = 3.50, p < 0.001), but not EFT drawn group (b = 0.001, CI = −0.016, 0.017, t = 0.05, p = 0.96). Future time perspective, age, BMI and sex did not moderate the effect of group on AUC.

4. Discussion

This study replicates several studies showing that episodic future thinking (EFT) improves discounting of delayed monetary rewards (Daniel et al., 2013a, 2013b; Mansouri et al., 2020; O’Donnell et al., 2017; Stein et al., 2016), and extended the use of EFT to include both written and drawn cues. Drawing personally relevant future events has several advantages over other types of visual cuing, including allowing individuals to include personally relevant and positive details, and encoding and elaboration of the event during the drawing process. There is a parallel between the traditional approach of allowing participants to write about their future events and include relevant details in their own words, with allowing participants to choose details in their drawings in a way that is personally relevant. DD was lower for both drawn and written EFT cues compared to a written ERT cue control group, and the EFT groups were not different from each other. While previous research has shown that visual imagery can stimulate greater details when engaged in autobiographical memory retrieval (Goddard, Pring, & Felmingham, 2005; Williams et al., 1999), we did not observe visual cues were superior to written cues to modify DD. This is significant, as it suggests participants have the flexibility in choosing how to present their cues in larger studies of clinical utility based on their preferences. Clinical applications can be considered for health behaviors that have been related to delay discounting, including physical activity (Sofis, Carrillo, & Jarmolowicz, 2017), obesity-related behaviors (Amlung et al., 2016; Tang, Chrzanowski-Smith, Hutchinson, Kee, & Hunter, 2019) and smoking (Audrain-McGovern et al., 2009).

We also found that immediate time perspective interacted with the form of EFT cues, as the EFT drawing group had similar discounting rates for both high and low immediate time perspectives, while the EFT and ERT written groups showed similar increases in DD for greater immediate time perspective (i.e. lower AUC). The moderation of the effect of EFT on DD by immediate time perspective suggests that using visual cues to stimulate episodic future thinking can reduce discounting of the future for people across a broad range of time perspectives. We replicated recent research showing that the consideration of future consequences immediate time perspective was correlated with DD (Macaskill, Hunt, & Milfont, 2019), as adults with an immediate time perspective discounted the future more steeply.

The creation of a novel episodic future cue involves retrospection to recall previous events that may inform the novel episode, semantic memory to identify general rules surrounding a novel event, and prospection to project oneself into the future (Schacter & Addis, 2007; Wang et al., 2016). Thus, there are many pathways for EFT or drawing to improve episodic prospection and reduce the value of immediate rewards. For example, EFT has been shown to improve prospective memory in children (Nigro, Brandimonte, Cicogna, & Cosenza, 2014; Terrett et al., 2019) and young adults (Terrett et al., 2016), and adolescents asked to imagine future actions performed better on a prospective memory task (Altgassen, Kretschmer, & Schnitzspahn, 2017). Drawing can improve semantic memory recall in a variety of populations, including ones in which there is a serious memory deficit (Fernandes et al., 2018), and visual cues have aided in increasing details in episodic memory and episodic future thinking in children with autistic spectrum disorder (Anger et al., 2019).

Retrospection and prospection are also related to time perspective, providing a possible explanation for novel interaction between immediate time perspective and EFT cue presentation. Adults with high immediate time perspective who were presented written EFT cues, showed a preference for smaller immediate rewards than those with low immediate time perspective. However, for adults presented with drawn EFT cues, similar discounting was seen for a wide range of immediate time perspective. In other words, drawn EFT cues lowered discounting rates similarly for both low and high immediate time perspectives. Immediate time perspective is related to lower self-report measures of retrospective and prospective memory in habitual gamblers, including both problematic and nonproblematic users (Nigro, D’Olimpio, Ciccarelli, & Cosenza, 2019) and in a sample of undergraduate students (Cinan & Dogan, 2013). A possible mechanism for the moderation of the EFT effect by immediate time perspective is that drawn cues may improve prospective memory, which may be diminished in individuals with higher immediate time perspective, and should be tested in future research.

There are several groups for which drawn cues may provide utility as well as practicality when using EFT as a way to modify discounting or health behaviors. One population is children. While EFT cues can modify both DD and energy intake in children (Daniel, Said, Stanton, & Epstein, 2015), children produce future events that are less detailed than adolescents (Gott & Lah, 2014). This is relevant, as richness of details is related to EFT effectiveness (Calluso, Tosoni, Cannito, & Committeri, 2019). A personally drawn EFT cue may increase the effect of EFT on DD and other behaviors for children by increasing the richness of details for episodic events. Adolescents have also been shown to have improved DD when presented with EFT cues (Bromberg, Lobatcheva, & Peters, 2017). It is theorized that adolescent’s high discounting rates may be related to an underdeveloped future orientation, rather than an increased sensitivity to reward (van den Bos, Rodriguez, Schweitzer, & McClure, 2015), suggesting that if drawing increases prospective memory, this population would benefit from an EFT procedure that specifically increases ability to prospect. This suggests interesting future research questions to explore the mechanisms of the effect of drawing on prospection and DD, in addition to examining if drawing enhances EFT in specific populations.

There are populations of adults that may also benefit from drawing rather than writing EFT cues. Adults with reading difficulties, including low-literacy (Perin, 1998) and dyslexia (Nergard-Nilssen & Hulme, 2014), may benefit from an intervention that does not rely on expressive writing. There are several clinical groups that have had associated difficulties in recalling episodic memory with a recent meta-analysis showing deficits in depression, bipolar disorder and schizophrenia (Hallford, Austin, Takano, & Raes, 2018). Clinical groups with deficits in either episodic of prospective memory may benefit from drawing EFT cues, if drawing improves prospective memory. Depression is also related to overgeneralization of memories and less positive memory (Holmes, Blackwell, Burnett Heyes, Renner, & Raes, 2016), making it a good candidate for an intervention that improves memory specificity (Hallford, Sharma, & Austin, 2020). Dyslexia, in addition to reading difficulties, is also related to problems with prospective memory (Smith-Spark, 2018). Episodic future thinking is a possible treatment component for several of these clinical populations, and drawn cues may increase efficacy.

When considering EFT’s clinical utility, it is important to note that its effects on DD have been shown to improve decision making when presented simultaneously with the decision (Daniel et al., 2013b; Rung & Epstein, 2020). One study has shown that repeated use of EFT cues over 1–3 months continues to improve DD in adults (Mellis, Snider, Deshpande, LaConte, & Bickel, 2019). Further research should be done examining how to increase the length of the effects of creating and imagining EFT cues, the effects of repeated practice using cues, and examine if different cue presentations influences these effects.

Despite testing a novel intervention format for EFT, this study has limitations. Perhaps the major limitation is absence of a drawing recent memory group, which would have controlled for the effects of drawing itself on DD. Drawing was also used as a way of representing the cue, rather than in cue creation, as all groups created their event cues aloud guided by an experimenter to equate that experience. Future research should examine how drawing effects cue generation, which could be implemented by having everyone draw their cues, and then write them, but differentially use the drawn or written cues in the DD task. We also chose to use drawn cues, rather than presenting images or photos, so parsing out the importance of drawing the cues is a question to explore in future studies.

While this comparison was not included as a separate group, we chose to study drawing as drawing versus visualizing and seeing images have been studied in semantic memory (Fernandes et al., 2018) and motoric components enhanced semantic memory retention (Wammes, Jonker, & Fernandes, 2019). Future research may want to examine the importance of the self-created images versus provided visual cues for personal future events. Participants were screened for high discounting and were largely students, influencing generalizability of the research to low discounters and a non-student population. It might be expected that individuals who are already low discounters would have little space to improve, limiting the comparison between two different types of EFT manipulations. There was no increase in discounting across all the time points, which may be due to EFT modifying the shape of the traditional delay discounting curve, rather than non-systematic responding. Researchers have suggested that all data be included when EFT is used to modify delay discounting (Bickel et al., 2020).

This study suggests that both written and drawn EFT cues can improve monetary decision making. DD has been correlated with multiple health behaviors, and EFT has been used to influence demand for snack foods (Sze, Stein, Bickel, Paluch, & Epstein, 2017) and cigarettes (Stein et al., 2016; Stein, Tegge, Turner, & Bickel, 2018). This research should be extended by examining how EFT cue modalities influence discounting of non-monetary commodities, including food or cigarettes, or how it influences demand for these commodities. Previous research suggests one’s monetary discounting rate is related to discounting of other commodities (Odum et al., 2020). We would hypothesize that EFT cues would have similar effects on other commodities, and future research should extend the generalizability of these findings.

Episodic future thinking has been shown to be a reliable way to reduce discounting (Rung & Madden, 2018). Drawing provides new options for implementing EFT interventions in a variety of populations, such as children, individuals with reading or writing deficits and potentially populations that have reduced prospective memory. However, rather than focus on populations with deficits in prospection, it may be interesting to study people who regularly express themselves visually, like artists, architects, industrial designers, or interior designers. Focusing on strengths in using visual cues, rather than weaknesses, may provide novel insights into how visual cues can facilitate and strengthen episodic forethought. Future research may examine differences in verbal and visual learning and memory as a potential moderator of the effects of cue modality on DD and examine its efficacy in populations for which written EFT cues may be a barrier. Drawing to create visual cues is a novel use of the EFT paradigm to improve preferences for larger delayed rewards.

Data statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.