Working Memory Training Decreases Delay Discounting Among Stimulant Addicts Warren K

Remember the Future: Working Memory Training Decreases Delay Discounting Among Stimulant Addicts Warren K. Bickel, Richard Yi, Reid D. Landes, Paul F. Hill, and Carole Baxter Background: Excessive discounting of future rewards has been observed in a variety of disorders and has been linked both to valuation of the past and to memory of past events.

Methods: To explore the functionality of discounting and memory, we examined whether training of working memory would result in less discounting of future rewards. In this study, 27 adults in treatment for stimulant use were randomly assigned to receive either working memory training or control training according to a yoked experimental design. Measures of delay discounting and several other cognitive behaviors were assessed pre- and posttraining. Results: Rates of discounting of delayed rewards were signiﬁcantly reduced among those who received memory training but were unchanged among those who received control training; other cognitive assessments were not affected by memory training. Discount rates were positively correlated with memory training performance measures. Conclusions: To our knowledge, this is the ﬁrst study demonstrating that neurocognitive training on working memory decreases delay discounting. These results offer further evidence of a functional relationship between delay discounting and working memory.

Key Words: Addiction, delay discounting, neurobehavioral deci- tion of the past and future are linked (17,18), then it might be sions systems theory of addiction, neurocognitive rehabilitation, possible to decrease an individual’s discounting of future events by stimulants, working memory increasing his or her ability to remember past events. In this study, we employed neurocognitive rehabilitation approaches that are proven to be effective with individuals with ne neurobehavioral process evident in a variety of disor- schizophrenia (19,20) to examine the effects of working memory ders and suboptimal behaviors is a high level of delay dis- training on measures of delay discounting, working memory, and counting (1). Delay discounting refers to the decrease in O related assessments in individuals in treatment for psychomotor value of a reward as a function of the delay to its receipt (2). An stimulants (e.g., cocaine, methamphetamine). Frequent or heavy individual’s rate of discounting can be measured by assessing pref- users of stimulants have been shown to exhibit neurocognitive erences between a sooner, smaller reward or a later, larger one. deficits, including deficiencies in working memory (21,22) and high Higher rates of discounting are associated with drug dependence rates of delay discounting (23,24). (2), problem gambling (3), obesity (4,5), and human immunodefi- In this study, participants receiving treatment for their stimulant ciency virus risk behaviors (6,7). Given that delay discounting un- use received either experimental or control memory training. Ex- derlies a wide variety of disadvantageous behaviors, it may function perimental (Active) Training consisted of working memory tasks as a transdisease process (8). with monetary reinforcement for performance. Control Training The correlation between rates of delay discounting and subop- consisted of presenting the same working memory tasks and cue- timal behaviors has led, in part, to efforts to gain a better under- ing the correct response. Reinforcement for each participant in the standing of discounting in terms of its relationship to other deci- control group was yoked to performance of a participant in the sion-making and neurocognitive processes. Consideration and active group; yoking ensured that the amount of reinforcement valuation of the future has been shown to overlap with processes obtained by each experimental group participant during each ses- and brain regions associated with memory or valuation of the past sion is also obtained by each participant in the control group. Pre- (9,10). For example, discounting of past and future rewards have and posttraining assessments on a variety of decision-making and been found to be qualitatively and quantitatively comparable (11– cognitive functions quantified the effects of training and deter- 13) by conforming to the same signature hyperbolic function and mined whether the effects were selective or more general. magnitude effect, and bilateral damage to the hippocampus impairs the ability to remember the past and to imagine future personal experiences (14). More directly, recent demonstrations indi- Methods and Materials cate significant correlations between measures of working memory Participants and delay discounting (15,16). Thus, if consideration of and valua- Twenty-seven participants (20 male, seven female, mean age ϭ 38.6 years) being treated for stimulant use at a substance-abuse From the Department of Psychiatry (WKB, RY, PFH), Center for Addiction treatment facility enrolled and completed all assessments. Stimu- Research, and Department of Biostatistics (RDL), University of Arkansas lant abuse/dependence diagnosis was determined by clinical staff for Medical Sciences, and Recovery Centers of Arkansas (CB), Little Rock, at the treatment facility using criteria established by the DSM-IV-TR Arkansas. (25) and documented to be consistent with the findings from as- Address correspondence to Warren K. Bickel, Ph.D., Department of Psychia- pects of Addiction Severity Index—5th Edition administered to all try, Center for Addiction Research, University of Arkansas for Medical individuals in the facility as a component of intake. All participants Sciences, Little Rock, AK 72205-7199; E-mail: [email protected]. met the criteria for stimulant dependence, except two participants Received Feb 19, 2010; revised Aug 9, 2010; accepted Aug 10, 2010. in the active group and three in the control group, who met criteria

Table 1. Participant Characteristics

Control Active Test Statistic and (n ϭ 13) (n ϭ 14) p Value

Years of Education, Mean (SD) 12.0 (1.7) 12.4 (1.7) t25 ϭ .64; p ϭ .528

Quick Test IQ, Mean (SD) 37.1 (5.1) 37.4 (3.0) t25 ϭ .17; p ϭ .862 a 2 Years Using Stimulants, Median (IQR) 14.5 (11.5, 20.5) 12 (3, 13) ␹ (1) ϭ 3.24; p ϭ .085 Primary Stimulant of Abuse, %b Cocaine Only 77 71 2 Methamphetamine Only 15 22 ␹ (2) ϭ .16; p ϭ .922 Both Cocaine and Methamphetamine 8 7 2 History of Nonstimulant Use, % 77 79 ␹ (1) ϭ .01; p ϭ .918 IQR, interquartile range. aData were not collected for two participants, one from each group. bA chi-square test of independence with degrees of freedom (df) ϭ 2. at minimum for stimulant abuse, but dependence could not be nize as many words as possible. The word lists differed be- determined. Participants were free from unmanaged symptoms of tween the pre- and posttraining assessments. DSM-IV Axis I and II disorders and had no history of major brain 7. Delay Discounting (2). Delay discounting was assessed using a insult. Additional participant characteristics are provided in Table 1. computerized, binary choice procedure. In each trial, participants chose between an immediate, smaller and a later, larger Materials amount of money. The dollar amount of the immediate out- Pre- and Posttraining Assessments. Seven assessments were come was adjusted from trial to trial according to a decreas- collected in sessions that occurred before and after completion of ing-adjustment algorithm while the delayed outcome re- the training sessions. Participants were compensated $20 and $10 mained constant, until the participant’s response indicated for the pre- and posttraining sessions, respectively. indifference between the sooner, smaller reward and the later, 1. Frontal Systems Behavior Scale (FrSBe) (26). On this 46-item larger reward. This procedure was employed to determine measure, participants rate themselves on a 5-point scale and indifference points for each of three temporal discounting assesses behavioral sequela related to frontal lobe damage; it conditions: real $100, hypothetical $100, and $1000. The includes three behavioral syndromes: apathy, disinhibition, larger later reward was available for the real $100 task at four and executive dysfunction. delays (1 day, 1 week, 1 month, and 6 months) and for hypo- 2. Letter–Number Sequencing (LNS) (27). In this working mem- thetical monetary amounts at seven delays (1 day, 1 week, 1 ory assessment, a mixed combination of letters and numbers month, 6 months, 1 year, 5 years, and 25 years). One of the is presented to participants who must put them in sequential choice trials from the real $100 condition was selected ran- order, consisting of numbers first (ascending) followed by domly at the conclusion of the session, and the participant letters (alphabetical). The letter–number string presentation received the outcome he or she picked for that trial. increases in difficulty (number of items) until the participant is Training Program. Four commercially available memory- no longer able to sequence correctly three strings of equiva- training programs (PSSCogReHab), modified to permit detailed lent difficulty. measurement for the yoking design, were used during the training 3. Balloon Analog Risk Task (BART) (28). For each trial of the sessions. These were as follows: computerized BART, a uninflated balloon appears on the monitor; 5¢ (hypothetical) are earned and collected in a temporary reserve by inflating the balloon (via one unit per mouse 1. Sequence recall of digits—auditory (SRD-A): participants were click). This sum can be emptied into a permanent bank that aurally presented a series of numbers and required to memo- cannot be lost. However, if the balloon continues to be in- rize them in the order in which they were presented. The flated (adding money to the temporary reserve) and the bal- program began with a sequence of three digits, increasing by loon bursts, the amount in the temporary reserve is lost. one digit upon a correct response, up to a maximum of 10 4. Go/No-Go Task (29). In this computerized procedure, four of digits. An incorrect response resulted in a different sequence eight 2-digit numbers are arbitrarily designated positive stim- of the same number of digits. Five missed sequences resulted uli (Sϩ) and four other 2-digit numbers are designated nega- in the end of the program. tive stimuli (S–). The participants’ task is to learn by trial and 2. Sequenced Recall Reversed Digits—Auditory (SRRD-A): this error to respond (within 2 sec) to Sϩ and not to respond to S–. program is identical to the SRD-A, except that participants 5. Phone Message Task. Located within the Memory II program were required to recall the digits in the reverse order of of the PSSCogReHab (Psychological Software Services, India- presentation. napolis, Indiana), this program assesses an individual’s epi- 3. Sequenced Recall of Words—Visual (SRW-V): participants sodic memory by visually and aurally presenting a detailed were shown a list of four-letter words on a computer screen. phone message. After a brief study period, participants an- After a brief study period, participants were instructed to find swer a series of multiple-choice questions regarding details in the presented words in the correct sequence from a list of 16 the message. words. The program began with three words, increasing by 6. The Hopkins Verbal Learning Test—Revised (HVLT-R) (30). one word upon a correct response, up to a maximum of 11 This standardized assessment of verbal learning and memory words. An incorrect response resulted in a new list of the same requires memorization of a list of words aurally presented. number of words. Five missed sequences terminated the Following a delay, participants are then asked to recall/recog- program.

www.sobp.org/journal 262 BIOL PSYCHIATRY 2011;69:260–265 W.K. Bickel et al.

Table 2. Spearman Correlation Coefficients Between Discount Rates and v ϭ eϪk͙d Memory Training Scores in Active Training Group d where v is the value of the sooner amount (reported as a propor- First Training Last Training d tion of the delayed amount), d is the delay, and k is the index of Session (p Value) Session (p Value) discounting. This model of discounting (31) provided a better fit Pretraining Discounting (i.e., had a lower error sum of squares) of the delay discounting data Hyp $100 –.65 (.01) –.58 (.03) in this study compared with Mazur’s hyperbolic model in 73% (118 Hyp $1000 –.45 (.11) –.42 (.14) of 162) of all data sets (32). Because of positively skewed distribu- Real $100 –.38 (.20) –.29 (.32) tions of discounting coefficients, natural logarithm-transformed k Posttraining Discounting values were estimated and employed in the analyses of discount- Hyp $100 –.44 (.11) –.52 (.06) ing. Hyp $1000 –.54 (.05) –.61 (.02) Each pre- and posttraining assessment was analyzed in a linear Real $100 –.35 (.22) –.37 (.19) mixed model, with Treatment, Time, Subscale (for FrSBe, HVLT-R, Hyp, hypothetical. and delay discounting), and all interactions as fixed effects and participant pair as a random effect. Because the Treatment–Time (pre to post) interaction is the effect of interest, we report these 4. Verbal memory—categorizing (VM-C): Participants were pre- 2 results and corresponding ␩G effect sizes. Composite performance sented with 20 words falling into four categories (e.g., colors, measures of the training session for the Active Training participants vegetables). Participants were instructed to place each word were calculated by adding scores from the SRD-A, SRRD-A, SRW-V, into its correct category and then choose the original words and one third of the VM-C (to equate its weight to the other three from a list of 180. scales).

Procedures Results All participants completed 1 pretraining session, 4 to 15 training sessions, and 1 posttraining session. The range of time lapsed be- Active and control training groups did not statistically differ on tween pre- and posttraining sessions was 9 to 44 days, with a mean any demographic variables. Because of a marginal nonsignificant of 25 days. Participants were assigned into either the Active Train- difference between the mean ages of the two groups (35.7 for ing (14 participants) or Control Training (13 participants) conditions active vs. 41.6 for control, t25 ϭ 1.89, p ϭ .070), age was included as at the conclusion of the pretraining session according to our yoked a covariate in subsequent analyses. When it was found not to affect design. Specifically, the first participant was assigned to the Active inferences, age as a covariate was dropped from the final models. Training condition. Each subsequent participant was evaluated to ascertain whether he or she met the matching criteria (gender and Effects of Training memory [LNS] score) of any participant in the Active Training con- No differential effect of training condition was observed for any of the following pre- to posttraining assessments: FrSBe [F(1,275) ϭ dition. If a given participant met the matching criteria, he or she was 2 2 .41, p ϭ .524, ␩G ϭ .001], LNS [F(1,25) ϭ .13, p ϭ .721, ␩G ϭ .005], assigned to the Control Training condition. If the participant did not 2 BART [F(1,25) ϭ .43, p ϭ .520, ␩G ϭ .017], Go/No-Go Task [F(1,25) ϭ match any of the existing Active Training participants, he or she was 2 .01, p ϭ .921, ␩G ϭ .000], Phone Message Task [F(1,25) ϭ .50, p ϭ assigned to the Active Training condition until 14 participants had 2 2 been assigned to the Active Training condition. .486, ␩G ϭ .020], and HVLT-R [F(1,125) ϭ 1.84, p ϭ .178, ␩G ϭ .014]. Active Training. In each training session, participants com- In contrast, an effect of training was observed with measures of delay discounting (see Table 2). The treatment-by-time interaction pleted each of the four memory training programs twice. The num- 2 ber of training sessions was determined by each participant’s was significant [F(1,122) ϭ 9.01, p ϭ .003, ␩G ϭ .069; Figure 1], with progress in the training programs; three consecutive sessions with- out an increase in performance on any two programs resulted in the conclusion of training, with a minimum of four and a maximum of 15 sessions. Participants were compensated $10 for attending each session and an additional $10 per session and for study completion. Participants were also compensated according to increasing schedules for performance. These schedules varied across training modules because of variations in performance thresholds and scoring, with a minimum of $.00 and a maximum of $30.85 per session. Control Training. The Control Training program was identical to the Active Training program in all essential features (e.g., stimu- lus, response, number of trials, progression, and feedback) with two exceptions. First, the Control Training program identified the correct answers to the participants, so that they did not need to en- gage working memory to obtain correct responses. Second, progression through modules and associated compensation for each Control Training participant was yoked to an individual in the Active Training Group.

Statistical Methods Delay discounting rates were calculated using the exponential- Figure 1. Differences in discounting (and 95% confidence intervals) from power discounting function, pre- to posttraining for active and control groups. www.sobp.org/journal W.K. Bickel et al. BIOL PSYCHIATRY 2011;69:260–265 263 no evidence that this effect differed across the three discounting conditions tested [F(2,126) ϭ .08, p ϭ .925]. Those undergoing Active Training significantly decreased their discounting rate k by 50% on average. Although not significant, the control group participants increased their rate by 50%. A nonparametric approach was also used to evaluate whether memory training affected a decrease in discounting. Specifically, the number of discounting tasks on which each participant showed a decrease at the posttreatment assessment was counted (three being the maximum possible). The two treatments were compared on their distributions of these counts with a (one-sided) Wilcoxon– Mann–Whitney test. The active group exhibited decreases in more of the three discounting tasks examined [␹2(1) ϭ 5.54, p ϭ .013]. The distributions of the counts of participants’ decreases in delay discounting are given in Table 3. Figure 2 shows the change in discounting rate from pre- to posttraining in the $100 hypothetical monetary discounting procedure for each individual. In this measure, nine participants in the active group showed a decrease (indi- Figure 2. Change in discounting ln(k) for individual participants in the active cated by positive change values), and two participants in the con- and control groups, calculated as pretraining minus posttraining. Positive trol group showed a decrease. values indicate a decrease in discounting. framing effects (e.g., the hidden-zero effect) (33) or situational Relationship Between Memory and Discounting changes (e.g., reinforcer deprivation level) (34) to produce change. Pre- and posttraining assessments were correlated with com- To our knowledge, this study is the first to demonstrate that neuro- posite working memory by obtaining Spearman’s correlation coef- cognitive training of working memory can decrease delay discount- ficients between pretraining discount assessments and the com- ing. This result is consistent with data from studies demonstrating a posite working memory score from the first training session and correlation between working memory performance and delay dis- between the posttraining discount assessments and the composite counting (21,22). The findings also corroborate studies supporting working memory score from the last training session. These analy- a strong relationship between valuation of the past and valuation of ses revealed significant correlations between delay discounting the future (11–13) and other studies using functional magnetic and working memory performance (Table 2). Overall, these data imaging indicating an overlap of brain regions (such as the dorso- illustrate that posttraining rates of delay discounting were signifi- lateral prefrontal cortex) that are activated during working memory cantly associated with the composite working memory perfor- and delay discounting tasks (35–37). mance scores from the last session in the Active Training group. Second, these findings support the competing neurobehavioral Moreover, the number of training sessions was significantly nega- decision systems hypothesis of addiction (11). According to this tively correlated with posttraining discounting rate only in the Ac- neuroeconomic approach, decisions are made on the basis of two tive Training group (r ϭ –.54, p ϭ .048). decision systems. One, referred to as the impulsive decision system, Discussion is embodied in the limbic and paralimbic brain regions and is associated with the acquisition of more immediate reinforcers. The This study suggests that working memory training among stim- other, referred to as the executive system, is embodied in the pre- ulant-dependent individuals results in a decrease in discounting of frontal cortex and is associated with planning and deferred out- delayed rewards consistent with previous reports of a relationship come. According to this hypothesis, addiction results from a hyper- between working memory and delay discounting. This effect of active impulsive system and a hypoactive executive decision training working memory was selective and did not affect any other system. Interestingly, McClure et al. (36) showed that delay dis- pre- and posttraining measures. Moreover, the absence of any sig- counting may provide a summary measure of the relative control of nificant effects in the control group indicates that these effects are these two systems. Specifically, he and his colleagues measured not related to attention, exposure to the stimuli associated with brain activation during choices between immediate and delayed working memory training, or the reinforcers delivered. Instead, the rewards and found greater relative activation of the impulsive sys- results support that the change in discounting resulted from rein- tem when the immediate option was chosen and greater relative forced working memory training. We offer four comments regard- activation of the executive system when the latter option was cho- ing these findings. sen. By providing a summary of the relative preference for immedi- First, these results, if replicable, support a new strategy or inter- ate or delayed options, delay discounting rates may serve as a vention by which to decrease the discounting of delayed rewards. summary measure of the relative control by these two brain sys- Several studies have demonstrated that discounting measures can tems. On the basis of that view, greater discounting among addicts be manipulated. Many of these studies have been focused on either is consistent with the notion of a hyperactive impulsive system and a hypoactive executive system specified by the competing neu- Table 3. Participants in Control and Active Groups Showing Decrease in robehavioral decision systems hypothesis. Moreover, our observed Delay Discounting decrease in the rate of discounting following working memory No. Discounting Procedures with Decreased k training is consistent with an increase in relative activation of the executive system. Count (Row %) 0123Third, the results of this study suggest several lines of future inquiry. For example, an important question is the durability of the Control 7 (54) 4 (31) 1 (7.5) 1 (7.5) effect examined here. Does this change in discounting persist or Active 3 (21.5) 2 (14) 3 (21.5) 6 (43) dissipate? If the effect decays over time, can booster working mem-

www.sobp.org/journal 264 BIOL PSYCHIATRY 2011;69:260–265 W.K. Bickel et al. ory training sessions continue the effect? Pursuant to this, is there a 3. Reynolds B (2006): A review of delay-discounting research with humans: ceiling on the effects of training? Another set of questions concern Relations to drug use and gambling. Behav Pharmacol 17:651–667. whether the training effect would be observed in other diagnostic 4. Epstein LH, Dearing KK, Temple JL, Cavanaugh MD (2008): Food rein- groups with greater rates of discounting than controls. Such a per- forcement and impulsivity in overweight children and their parents. Eat Behav 9:319–327. vasive effect would provide supporting evidence that excessive 5. Weller RE, Cook EW III, Avsar KB, Cox JE (2008): Obese women show delay discounting is a transdisease process. However, given that greater delay discounting than healthy-weight women. Appetite 51: previous studies among smokers have shown that individuals who 563–569. discount less exhibit better treatment outcomes or are less likely to 6. Odum AL, Madden GJ, Badger GJ, Bickel WK (2000): Needle sharing in relapse (38–40), perhaps the most important question is whether a opioid-dependent outpatients: Psychological processes underlying decrease in the rate of discounting also results in improvements in risk. Drug Alcohol Depend 60:259–266. 7. Chesson HW, Leichliter JS, Zimet GD, Rosenthal SL, Bernstein DI, Fife KH treatment outcomes. (2006): Discount rates and risky sexual behavior among teenagers and Finally, this study supports the viability of neurocognitive reha- young adults. J Risk Uncertainty 32:217–230. bilitation to improve at least one aspect of executive function 8. Bickel WK, Mueller ET (2009): Toward the study of trans-disease pro- among addicts. Such approaches have been shown to be effective cesses: A novel approach with special reference to the study of co- for brain injury (depending on the magnitude of the injury) and for morbidity. J Dual Diagn 5:131–138. schizophrenia (19,20). A prior effort to improve executive function 9. Bickel WK, Miller ML, Yi R, Kowal BP, Lindquist DM, Pitcock JA (2007): in addiction resulted in limited improvement (41). Our study differs Behavioral and neuroeconomics of drug addiction: Competing neural systems and temporal discounting processes. Drug Alcohol Depend 90S: from previous approaches in that only a single executive function S85–S91. (i.e., working memory) was trained, but assessments examining the 10. Buckner RL, Carroll DC (2007): Self-projection and the brain. Trends Cogn effects of that training included a variety of other decision and Sci 11:49–57. executive function tasks. Perhaps isolating executive functions for 11. Bickel WK, Yi R, Kowal BP, Gatchalian KM (2008): Cigarette smokers training would reveal other interesting effects among other execu- discount past and future rewards symmetrically and more than con- tive functions. Importantly, these changes in executive function are trols: Is discounting a measure of impulsivity? Drug Alcohol Depend 96:256–262. consistent with the notion of neuroplasticity and suggest that at 12. Yi R, Gatchalian KM, Bickel WK (2006): Discounting of past outcomes. Exp least some of the neurocognitive deficits related to addiction might Clin Psychopharmacol 14:311–317. be reversible. 13. Dixon MR, Holton B (2009): Altering the magnitude of delay discounting We note that although memory performance during training by pathological gamblers. J Appl Behav Anal 42:269–275. was significantly correlated with discounting, the pre- and post- 14. Hassabis D, Kumaran D, Vann SD, Maguire EA (2007): Patients with training measures of memory were not different. This is likely due to hippocampal amnesia cannot imagine new experiences. Proc Natl Acad a lack of sensitivity of the LNS, the scoring algorithm of which may Sci USA 104:1726–1731. 15. Bobova L, Finn PR, Rickert ME, Lucas J (2009): Disinhibitory psychopa- not be able to detect anything less than large changes in perfor- thology and delay discounting in alcohol dependence: Personality and mance. Alternative explanations are that working memory training cognitive correlates. Exp Clin Psychopharmacol 17:51–61. does not improve working memory performance, that LNS and 16. Shamosh NA, DeYoung CG, Green AE, Reis DL, Johnson MR, Conway training programs are sufficiently different so that training on one ARA, et al. (2008): Individual differences in delay discounting: Relation to will not influence performance on the other even if working mem- intelligence, working memory, and anterior prefrontal cortex. Psychol ory is improved, or that there was insufficient duration to document Sci 19:904–911. 17. Rachlin H (2000): The Science of Self-Control. Cambridge, MA: Harvard working memory improvement. In particular, the potential for in- University Press. sufficient working memory training may be related to the variability 18. Trope Y, Liberman N (2003): Temporal construal. Psychol Review 110: in training duration across participants. If the latter is the case, then 403–421. providing sufficient training may also change the degree of selec- 19. Krabbendam L, Aleman A (2003): Cognitive rehabilitation in schizophre- tivity observed here. However, clarification of these various out- nia: A quantitative analysis of controlled studies. Psychopharmacology comes will await further examination with alternative measures of 169:376–382. working memory such as the O-Span Working Memory measure, 20. Twamley EW, Jeste DV, Bellack AS (2003): A review of cognitive training in schizophrenia. Schizophr Bull 29:359–382. which has been shown to be more sensitive to interventions (42). 21. Ardila A, Rosselli M, Strumwasser S (1991): Neuropsychological deficits However, we note again that working memory measures of perfor- in chronic cocaine abusers. Int J Neurosci 57:73–79. mance in the training modules were appropriately correlated with 22. Simon SL, Domier CP, Sim T, Richardson K, Rawson RA, Ling W (2002): delay discounting scores. Cognitive performance of current methamphetamine and cocaine abusers. J Add Dis 21:61–74. ThisstudywassupportedbytheNationalInstitutesofHealth(Grant 23. Coffey SF, Gudleski GD, Saladin ME, Brady KT (2003): Impulsivity and rapid discounting of delayed hypothetical rewards in cocaine-depen- Nos. R01DA024080, R01DA022386, R01DA11692, and UL1RR029884), dent individuals. Exp Clin Psychopharmacol 11:18–25. the Arkansas Biosciences Institute, the major research component of 24. Monterosso JR, Ainslie G, Xu J, Cordova X, Domier CP, London ED (2007): the Arkansas Tobacco Settlement Proceeds Act of 2000, and the Wilbur Frontoparietal cortical activity of methamphetamine-dependent and D. Mills Endowment. Dr. Bickel is a principal of HealthSim LLC. We comparison subjects performing a delay discounting task. Hum Brain thank the Recovery Centers of Arkansas and Annie Carter for their Mapp 28:383–393. assistance. 25. American Psychiatric Association (2000): Diagnostic and Statistical Man- All authors report no biomedical financial interests or potential ual of Mental Disorders, 4th ed., text revision (DSM-IV-TR). Washington, DC: American Psychiatric Publishing. conflicts of interest. 26. Grace J, Stout JC, Malloy PF (1999): Assessing frontal lobe behavioral syndromes with the frontal lobe personality scale. Assessment 6:269– 1. Madden GJ, Bickel WK (2009): Impulsivity: The Behavioral and Neurologi- 284. cal Science of Discounting. Washington, DC: American Psychological As- 27. Wechsler D (1997a): WAIS-III WMS-III Technical Manual. San Antonio, TX: sociation. Psychological Corporation. 2. Bickel WK, Marsch LA (2001): Toward a behavioral economics under- 28. Lejuez CW, Read JP, Kahler CW, Richards JB, Ramsey SE, Stuart GL, et al. standing of drug dependence: Delay discounting processes. Addiction (2002): Evaluation of a behavioral measure of risk taking: the balloon 96:73–86. analogue risk task (Bart). J Exp Psychol Appl 8:75–84. www.sobp.org/journal W.K. Bickel et al. BIOL PSYCHIATRY 2011;69:260–265 265

29. Newman JP, Widom CS, Nathan S (1985): Passive avoidance in syn- 37. McClure SM, Ericson KM, Laibson DI, Loewenstein G, Cohen JD dromes of disinhibition: Psychopathy and extraversion. J Personal Soc (2007): Time discounting for primary rewards. JNeurosci27:5796– Psychol 48:1316–1327. 5804. 30. Brandt J, Benedict RHB (2001): Hopkins Verbal Learning Test—Revised. 38. Dallery J, Raiff B (2007): Delay discounting predicts cigarette smoking in Professional Manual. Odessa, FL: Psychological Assessment Resources. a laboratory model of abstinence reinforcement. Psychopharmacology 31. Yi R, Bickel WK, Landes RD (2009): Novel models of intertemporal valu- 190:485–496. ation: Past and future outcomes. J Neurosci Psychol Econ 2:102. 39. Mueller ET, Landes RD, Kowal BP, Yi R, Stitzer ML, Burnett CA, et al. (2009): 32. Angeletos GM, Laibson D, Repetto A, Tobacman J, Weinberg S (2001): Delay of smoking gratiﬁcation as a laboratory model of relapse: Effects The hyperbolic consumption model: Calibration, simulation, and empir- of incentives for not smoking, and relationship with measures of exec- ical evaluation. J Econ Perspect 15:47–68. utive function. Behav Pharmacol 20:461–473. 33. Magen E, Dweck CS, Gross JJ (2008): The hidden-zero effect: Represent- 40. Yoon JH, Higgins ST, Heil SH, Sugarbaker RJ, Thomas CS, Badger GJ ing a single choice as an extended sequence reduces impulsive choice. Psychol Sci 19:648–649. (2007): Delay discounting predicts postpartum relapse to cigarette 34. Field M, Santarcangelo M, Sumnall H, Goudie A, Cole J (2006): Delay dis- smoking among pregnant women. Exp Clin Psychopharmacol 15:176– counting and the behavioural economics of cigarette purchases in smok- 186. ers: The effects of nicotine deprivation. Psychopharmacology 186:255–263. 41. Fals-Stewart W, Lucente S (1994): The effect of cognitive rehabilita- 35. Owen AM, McMillan KM, Laird AR, Bullmore E (2005): N-back working tion on the neuropsychological status of patients in drug abuse memory paradigm: A meta-analysis of normative functional neuroim- treatment who display neurocognitive impairment. Rehabil Psychol aging studies. Hum Brain Mapp 25:46–59. 39:75–94. 36. McClure SM, Laibson DI, Loewenstein G, Cohen JD (2004): Separate 42. Schoofs D, Wolf OT, Smeets T (2009): Cold pressor stress impairs perfor- neural systems value immediate and delayed monetary rewards. Sci- mance on working memory tasks requiring executive functions in ence 306:503–507. healthy young men. Behav Neurosci 123:1066–1075.