Contextual Control of Delay Discounting by Pathological Gamblers

CONTEXTUAL CONTROL OF DELAYDISCOUNTING BY PATHOLOGICAL GAMBLERS

MARK R. DIXON, ERIC A. JACOBS, AND SCOTT SANDERS

SOUTHERN ILLINOIS UNIVERSITY

The present study demonstrated the relative impact of gambling and nongambling contexts onthe degree of delay discounting by pathological gamblers. We used a delay-discounting task with20 pathological gamblers in and out of the natural context in which they regularly gambled. For16 of the 20 participants, it appeared that the difference of context altered the subjective value ofdelayed rewards, thereby producing relative changes in delay-discounting rates that weregenerally consistent with a hyperbolic model of intertemporal choice. The current data suggestthat empirically derived k values from delay-discounting tasks are context sensitive and are notconstant across various settings for the individual. Implications for future transitional research onaddictive disorders generally, and gambling specifically, are discussed.

DESCRIPTORS: choice, self-control, impulsivity, delay discounting, establishing opera-tion, gambling

_______________________________________________________________________________

Recent estimates suggest that approximately3% to 5% of the population gambles more thanis financially responsible (National GamblingImpact Study Commission Report, 1999),which is up from estimates of 1% reported 20years ago (Ladouceur, Boisvert, Pepin, Loran-ger, & Sylvain, 1994). Excessive or pathologicalgambling is considered a type of impulsecontrol disorder by the Diagnostic and StatisticalManual of Mental Disorders (DSM-IV-TR;American Psychiatric Association, 2000),whereby the gambler displays high levels ofimpulsivity. Behavioral conceptualizations ofthe construct of impulsivity have been noted asthe individual’s selecting an immediate smallerreinforcer over a larger delayed reinforcer(Critchfield & Kollins, 2001). Choosing thelatter alternative is termed self-control (Rachlin& Green, 1972). Over the past decade, there

has been a growing literature conceptualizingsome maladaptive behavior as problems ofintertemporal choice (Bickel & Vuchinich,2002; Critchfield & Kollins; Herrnstein &Prelec, 1992). Behaviors such as compulsivegambling and habitual substance abuse, forexample, have been conceptualized as repeatedchoices between the relatively punctuated andproximal consequences of engaging in thebehavior (i.e., intoxication or winning a jackpot)and the relatively diffuse and distal conse-quences of abstaining from the problembehavior (i.e., a healthier lifestyle or fiscalsecurity). Consistent with this interpretation,a number of studies have demonstrated thatthese behavioral problems are often correlatedwith diminished sensitivity to the larger yetdelayed outcomes (Dixon, Marley, & Jacobs,2003; Madden, Petry, Badger, & Bickel, 1997;Petry & Casarella, 1999).

In these studies, researchers often comparedthe delay discounting of people diagnosed withimpulse control disorders (i.e., substance de-pendence or compulsive gambling) to thediscounting of matched control participantsusing a hypothetical choice task that wasdeveloped by Rachlin, Raineri, and Cross(1991). Using this task, participants are re-quired to make repeated choices between

This study was completed in partial fulfillment of theMS degree in behavior analysis and therapy by the thirdauthor under the direction of the first and second authorsand Brandon F. Greene. We thank the UniversityTeletrack of Carbondale, Illinois, for allowing this researchto take place.

Address all correspondence to Mark R. Dixon, BehaviorAnalysis and Therapy Program, Rehabilitation Institute,Southern Illinois University, Carbondale, Illinois 62901(e-mail: [email protected]).

doi: 10.1901/jaba.2006.173-05

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2006, 39, 413–422 NUMBER 4 (WINTER 2006)

413

immediate and delayed hypothetical conse-quences (e.g., hypothetical amounts of moneyor drugs). The delay to the deferred outcome isheld constant within conditions but is variedacross conditions. Within each delay condition,the amount of the immediate consequence isvaried across choices to determine the point ofsubjective equivalence (i.e., an indifferencepoint) between the immediate and delayedconsequences.

The resulting indifference curves are gener-ally consistent with the following hyperbolicmodel introduced by Mazur (1987):

V ~ A=(1 z kD): ð1ÞIn Equation 1, V is the subjective value of thedelayed consequence (i.e., the indifferencepoint), A is the nominal amount of delayedconsequence, D is the delay to deferred con-sequence, and k is a free parameter thatdescribes sensitivity to change in delay. Thedegree of discounting parameter, k, providesa convenient index of sensitivity to delayedconsequences. Higher k values indicate lowersensitivity to delayed consequences. That is, forany given amount and delay, lower indifferencepoints will yield higher derived k values.

Consistent with the interpretation that somemaladaptive behavior may be functionally relatedto diminished sensitivity to delayed or diffuseconsequences, derived k values for individualswith impulse control disorders tend to be higherthan those for participants who do not displaythe maladaptive behavior. Opioid-dependentindividuals, for example, discount delayed con-sequences more severely than do matched controlparticipants (Madden et al., 1997). Likewise,self-identified heavy drinkers discount delayedconsequences more severely than self-identifiedlight drinkers (Vuchinich & Simpson, 1999),cigarette smokers discount more severely thannonsmokers (Odum, Madden, & Bickel, 2002),and pathological gamblers discount more severe-ly than matched control participants (Dixon etal., 2003). Moreover, it appears that differencesin delay discounting may also be correlated with

the severity of the maladaptive behavior. Opioid-dependent individuals with comorbid gamblingproblems, for example, discount delayed con-sequences more severely than opioid-dependentindividuals without the dual diagnosis (Petry &Casarella, 1999). Likewise, opioid-dependentindividuals who express willingness to usea hypodermic needle after someone else has usedit discount delayed monetary and heroin out-comes more severely than those who refuse to usethe dirty needle (Odum, Madden, Badger, &Bickel, 2000). The converging pattern ofevidence does much to support the contentionthat maladaptive impulse control disorders maybe functionally related to a general deficit insensitivity to delayed or diffuse consequences.

Although the aforementioned data greatlyextend the external validity of the hypotheticalchoice task (Rachlin et al., 1991) and Mazur’s(1987) hyperbolic model of temporal discount-ing, many questions remain regarding theorigins of these individual differences intemporal discounting and the dynamics oftemporal discounting. A number of reinforce-ment variables, for example, have been shownto influence delay discounting, including re-inforcement magnitude, commodity type,schedule dynamics, and deprivation. In thehypothetical choice task, adults typically dis-count smaller money amounts more rapidlythan they discount larger money amounts(Green, Myerson, & McFadden, 1997). More-over, qualitatively different reinforcers aresometimes discounted to different degrees.Substance-dependent individuals, for example,tend to discount delayed drug reinforcers toa greater degree than a yoked amount of delayedmoney (Madden, Bickel, & Jacobs, 1999;Madden et al., 1997). Schedule manipulationshave also been shown to affect delay discount-ing. Ostaszewski, Green, and Meyerson (1998),for example, demonstrated that currencies sub-ject to rapid inflation are discounted moreseverely than yoked amounts of a more eco-nomically stable currency. Deprivation has also

414 MARK R. DIXON et al.

been shown to affect delay discounting. Gior-dano et al. (2002) studied the effects of drugdeprivation on delay discounting in opioid-dependent outpatients receiving buprenorphinepharmacotherapy. Degree of deprivation wasmanipulated within participants by asking themto complete hypothetical choice assessmentseither before (i.e., while experiencing mildopioid withdrawal) or 2 hr following (i.e., whileexperiencing peak drug effects) buprenorphineadministration. Participants discounted delayedmonetary and drug reinforcers to a greaterdegree under the deprived conditions. Each ofthese reinforcement variables suggest an area ofconcern for those interested in developinginterventions to treat impulse control disordersthat may be related to delay discounting.

From a therapeutic perspective, understandingthe effects of antecedent control of delaydiscounting may be of critical importance indeveloping effective interventions for impulsecontrol disorders such as drug dependence orcompulsive gambling. For someone strugglingwith these behavioral problems, merely encoun-tering contexts that have been correlated withreinforcement of the problem behavior may besufficient to induce relapse. For example, whilein a weekly therapy session, a pathologicalgambler may have every intention of quittinggambling, and believes that his life will nowchange. However, after removal from the therapysetting, now on his way home passing a casino,quitting gambling seems much less possible. Thepropensity to relapse towards engaging in themaladaptive behavior may be indicated byvarying degrees of discounting. What seemsperplexing to researchers and treatment providersin the area of gambling addiction is that whileclaiming to want and need to stop gambling, theproblem gambler often fails at doing so. Failedattempts by the pathological gambler may beconsidered the result of a disease, an impulsivepersonality, or lack of willpower. Yet, behavior-ally, it may be the case that the pathologicalgambler discounts the delayed outcomes of

terminating gambling, and the degree to whichthey do varies with the environmental arrange-ment in which he finds himself. Thus, while inthe therapist’s office or at home with a familymember, the future consequences of quittinggambling are more psychologically present thanwhile only footsteps from a casino.

In the present study we examined the effectsof context on delay discounting of problemgamblers. Specifically, we performed within-participant comparisons of delay discounting ingambling (an off-track betting facility) andnongambling (e.g., a coffee shop) contexts todetermine if delayed hypothetical moneyamounts were discounted to a higher degreein the gambling context.

METHOD

Participants and Setting

Twenty participants were recruited to partic-ipate (19 men and 1 woman, mean age 5 37years). Each participant completed the SouthOaks Gambling Scale (SOGS), a 20-itemquestionnaire that assesses pathological gam-bling based on DSM-IV-TR (2000) criteria andhas been shown to have acceptable psychomet-ric properties (Lesieur & Blume, 1987). Scoreson the SOGS range from 0 to 14, with a scoreof 4 or more indicative of potential pathologicalgambling. The mean SOGS score for the 20participants in the present study was 6.6 (range,5 to 12).

The gambling context in which sessions tookplace was an off-track betting facility. Thefacility housed two bars, a number of tables andchairs, and approximately 30 televisions broad-casting horse-racing events from all over theworld. The nongambling contexts includedcoffee shops, restaurants, or a preferred businessor public location where the participant wouldbe able to perform the delay-discounting task.

Materials

A personal computer running a delay-dis-counting software program was used for data

DELAY DISCOUNTING 415

collection. This program was written in VisualBasic.NET by the first author (see Dixon &MacLin, 2003). All reward choices were madeby the participant pointing directly at a hypo-thetical amount of money displayed on thecomputer screen. The time delays were 1 week,2 weeks, 1 month, 6 months, 1 year, 3 years,and 10 years. The monetary reward amountswere $1,000, $990, $960, $920, $850, $800,$750, $700, $650, $600, $550, $500, $450,$400, $350, $300, $250, $200, $150, $100,$80, $60, $40, $20, and $10; these amountshave been used frequently in other studies (e.g.,Bickel, Odum, & Madden, 1999; Critchfield &Kollins, 2001; Madden et al., 1999; Petry &Casarella, 1999; Rachlin et al., 1991).

Procedure

Each participant completed a preexperimentalassessment using the SOGS in the gamblingcontext and two experimental sessions of thedelay-discounting task (one in the gamblingcontext and one in the nongambling context),each lasting approximately15 to 25 min. At theonset of the preexperimental session, whichtook place in the gambling context whereparticipants were recruited, the participant reada consent form and agreed to complete thestudy. The participants then were verballyadministered the SOGS. The SOGS was thenscored, and participants who did not score 4 ormore (indicative of probable pathologicalgambling) were dismissed and excluded fromthe remainder of the study. Further questioningof the participants revealed that they wereregular patrons of the betting facility, had neverattended a Gamblers Anonymous meeting, andhad never sought help for a gambling problem.Participants were randomly assigned to contextorder. Participants who were assigned to thegambling context first completed the delay-discounting task. Participants who were as-signed to the nongambling context first werescheduled to meet the experimenter in sucha context within the next few days to completethe delay-discounting task. After the completion

of the first administration of the delay-dis-counting task, all participants rescheduled thenext session with the experimenter to becompleted in the other context within approx-imately 1 week.

At the beginning of the first experimentalsession, the researcher read to participantsa script describing the delay-discounting com-puter program. In the program, participantswere presented with multiple consecutive mon-etary choices. The participants was instructed topoint to the hypothetical amount of money heor she would rather have: either a larger delayedreward or a smaller immediate reward. After theparticipant pointed to the desired selection onthe screen, the response was recorded. Thissequence continued for all hypothetical mone-tary rewards on the smaller immediate side. Forexample, the initial choice presented to theparticipant was between $1,000 available nowor $1,000 available after 1 week. After a partic-ipant selected an amount, the next choice (i.e.,$990 now or $1,000 after 1 week) waspresented. The amount of money availablecontinued to decrease to $10 along the valuesdescribed earlier. The process was then repeatedin ascending order (i.e., amounts increasingfrom $10 to $1,000 now vs. $1,000 1 weeklater). The ascending and descending sequenceswere then repeated at the next larger delay value(e.g., $1,000 available now vs. $1,000 availablein 2 weeks).

RESULTS

The indifference points obtained at each ofthe seven delays are shown in Table 1 undergambling and nongambling conditions for eachparticipant. Within each assessment, the in-difference points usually decreased or remainedthe same across successive delay values. Eachobtained indifference curve was evaluatedaccording to the following criteria to determineconsistency with delay discounting, broadlyconstrued. The mean of the indifference pointsfrom the three shortest delay conditions had to


exceed the mean of the indifference points fromthe three longest delay conditions, and in-difference points could not increase acrosssuccessive delays more than once. The obtainedindifference curves for all participants under allconditions met these criteria.

Table 1 also contains the empirically deriveddiscounting parameters (k values) and theindividual proportions of variance (R2) ac-counted for by the hyperbolic model for eachparticipant under each condition. In general,

the hyperbolic model provided an adequatedescription of the majority of the individualindifference curves. Across all participants andconditions, the median R2 value was 0.86 andthe interquartile range was 0.64 to 0.94. In 12instances involving 9 participants, however, theproportion of variance accounted for byEquation 1 fell below 0.7. Such poor fits oftenoccurred despite orderly indifference curves(e.g., see Participant 103 under the gamblingcondition). Thus, although Equation 1 ade-

Table 1

Indifference Points, Derived k Values, Proportions of Variance Accounted for (R 2) by Equation 1, and Area Under the

Curve (AUC) Measures for Each Participant Under Gambling (G) and Nongambling (NG) Conditions

Participant Condition

Delay (weeks)

k R 2 AUC1 2 4 25 52 156 520

101 G 750 700 450 250 100 60 60 .2440 0.96 .087NG 1,000 850 850 550 350 150 150 .0349 0.97 .214

102 G 1,000 1,000 10 10 10 10 10 .2942 0.71 .016NG 1,000 1,000 1,000 500 400 175 10 .0307 0.98 .184

103 G 850 800 700 550 450 350 250 .0258 0.55 .347NG 920 850 800 700 550 400 300 .0127 0.75 .409

104 G 700 700 700 500 100 100 60 .1002 0.8 .121NG 800 800 775 650 200 100 100 .0454 0.87 .157

105 G 1,000 1,000 1,000 650 500 500 500 .0085 0.55 .521NG 1,000 1,000 1,000 700 550 500 500 .0070 0.61 .529

106 G 1,000 1,000 920 800 650 600 500 .0040 0.68 .590NG 800 800 700 700 500 400 400 .0137 0.16 .436

107 G 990 990 990 990 990 990 500 .0012 0.82 .819NG 920 1,000 920 850 800 650 550 .0025 0.71 .651

108 G 850 850 800 700 500 400 10 .0176 0.87 .302NG 800 750 700 650 600 500 500 .0150 22.72 .526

109 G 920 850 800 750 550 10 10 .0236 0.89 .135NG 700 700 650 550 500 350 200 .0268 20.18 .335

110 G 920 920 800 750 550 450 10 .0137 0.9 .333NG 920 920 850 800 700 400 100 .0101 0.95 .364

201 G 875 850 700 550 250 10 10 .0576 0.94 .085NG 920 920 920 750 550 10 10 .0220 0.93 .138

202 G 550 550 500 500 10 10 10 .3242 0.55 .047NG 960 920 1,000 500 500 500 10 .0186 0.86 .342

203 G 700 700 800 10 10 10 10 .1877 0.88 .032NG 1,000 1,000 500 600 10 10 10 .0799 0.83 .054

204 G 850 850 800 700 700 650 550 .0032 21.47 .628NG 850 850 800 800 700 700 550 .0026 21.09 .655

205 G 920 850 920 600 300 200 100 .0326 0.97 .216NG 850 850 990 770 700 300 100 .0128 0.92 .321

206 G 600 500 400 250 100 60 10 .4509 0.86 .067NG 920 750 500 300 100 40 10 .1665 0.97 .064

207 G 1,000 1,000 1,000 850 650 300 250 .0101 0.97 .371NG 1,000 1,000 920 800 600 400 250 .0099 0.97 .406

208 G 750 700 600 600 450 10 10 .0484 0.65 .110NG 920 850 750 700 650 10 10 .0223 0.85 .144

209 G 850 850 700 500 250 200 200 .0525 0.87 .235NG 500 400 400 200 200 200 200 .6232 20.52 .206

210 G 1,000 1,000 1,000 500 300 10 10 .0398 0.96 .097NG 1,000 1,000 920 700 500 100 10 .0224 0.97 .170


quately described most of the individual in-difference curves, a subset of the indifferencecurves may not have been formally consistentwith the hyperbolic model.

Nonetheless, the k parameter provides a use-ful index of degree of discounting, as higher kvalues indicate greater sensitivity to delay toreinforcement. If the gambling context increasesthe degree of delay discounting, then the kparameters derived from data collected in thegambling context should exceed those derivedfrom data collected in the nongambling context.Data from the 11 participants for whomEquation 1 provided an adequate description(i.e., R2 .0.7) of the indifference curves fromboth contexts are shown in Figure 1. For 10 ofthese 11 participants, the k value derived fromdata obtained in the gambling context exceededthat derived from data obtained in thenongambling context.

Area under the indifference curve (AUC),another measure of delay discounting, is alsopresented in Table 1 for each participant underboth conditions. AUC is a theoretically neutralmeasure of delay discounting that requires noa priori assumptions regarding the form of theindifference curve (Myerson, Green, & War-usawitharana, 2001). As such, the AUCmeasure is applicable to a wider range of

indifference curves than other quantitativemodels, such as those based on hyperbolic orexponential delay discounting, that make ex-plicit assumptions regarding the form of theindifference curve. The AUC ranges from0 (steep discounting) to 1 (little or nodiscounting).

If the gambling context exacerbates delaydiscounting, then the AUC obtained in thegambling context should be lower than theAUC obtained in the nongambling context.Indeed, for 16 of the 20 participants, the AUCwas lower in the gambling context (seeFigure 2). An analysis of variance was con-ducted to assess the effects of context andsequence of conditions on AUC. The effect ofsequence was not statistically significant, F(1,18) 5 2.4480, p 5 .14, nor was the interactionbetween sequence and context, F(1, 18) 5

0.0220, p 5 .88. However, the main effect ofcontext on AUC was statistically significant,F(1, 18) 5 4.9788, p 5 .04.

Figure 3 shows aggregate discounting curvesfor all participants in gambling and nongam-bling contexts. The data points represent themedians of the individual indifference points ateach delay (see Table 1). Equation 1 provideda good description of the aggregate indifferencecurves. The proportions of variance accounted

Figure 1. k values derived from data obtained in thegambling and nongambling contexts for select participants(see text for details). The dashed reference line depicts

equality between the k values.

Figure 2. Area under the indifference curve obtainedin the gambling and nongambling contexts for eachparticipant. The dashed reference line depicts equality

between the AUC measures.


for by the hyperbolic model were 0.96 and 0.93in the gambling and nongambling contexts,respectively. The derived k value for theaggregate data was higher for participants inthe gambling context (k 5 .0358) than in thenongambling context (k 5 .0171). The areaunder the aggregate curve was also smaller inthe gambling context (AUC 5 .1762) than inthe nongambling context (AUC 5 .3118).

DISCUSSION

The current findings illustrate that mostpathological gamblers discounted delayed re-wards to a greater degree in a gambling context.For the majority of participants, k valuesderived from fitting Equation 1 to individualindifference curves were larger in the gamblingcontext than in the nongambling context,indicating more severe discounting when gam-bling. It should be noted, however, thatEquation 1 did not provide a good fit to theindividual indifference curves for a number ofparticipants, despite relatively orderly discount-ing data. Individual AUC measures were alsosmaller in the gambling context than in thenongambling context for the majority of

participants, again demonstrating steeper tem-poral discounting when gambling. Aggregatedata patterns were in accord with these in-dividual patterns. However, the aggregate datawere well described by Equation 1. Althoughprevious research has demonstrated the utilityof the hyperbolic model (Equation 1) to fitobtained discounting data (e.g., Kirby &Marakovic, 1995; Kirby, Petry, & Bickel,1999; Madden et al., 1997; Mazur, 1987; Petry& Bickel, 1998; Rachlin et al., 1991; Raineri &Rachlin, 1993), our present findings revealedmany instances in which this model provideda relatively poor fit to the obtained individualdata. The difficulties we had with Equation 1were similar to those reported by Dixon et al.(2003) in a comparison of delay discounting inpathological gamblers and matched controlsand by Dixon et al. (2005) in a comparison ofdelay discounting in persons with acquiredbrain injuries and matched controls. As in theseprevious studies, AUC measures were deemeduseful for data analysis. The strength of theAUC measure is that it is not tied to anyparticular theory about the form of the in-difference curve (i.e., hyperbolic or exponentialdiscounting). As such, it can be used to evaluatedata that do not conform to existing quantita-tive models. One weakness of the AUCmeasure, however, is that similar areas may beobtained from very different discounting curves.Nonetheless, AUC analyses are gaining popu-larity for analyzing discounting data that maynot fit hyperbolic decay functions (Dixon et al.,2003, 2005; Myerson et al., 2001), and futureresearchers may wish to compare the twometrics with various other clinical populationsto further establish their concordance.

The present study also underscores the valueof conducting within-participant analyses ofvariables that may affect intertemporal choice.Sixteen of our 20 participants showed greaterdegrees of delay discounting when assessedwithin a gambling environment compared toa nongambling environment. Although Fig-

Figure 3. Aggregate indifference curves for partici-pants in the gambling and nongambling contexts. Datapoints represent medians of the individual indifference

points. Error bars represent the interquartile range of theindividual indifference points at each delay. The solid lineshows the best fit in the gambling context, and the dashed

line shows the best fit in the nongambling context.


ure 3 displays aggregate differences in discount-ing across contexts, the group analysis mayobscure the magnitude and consistency of theeffect. The individual data displayed in Table 1and Figure 2 provide a more refined within-participant analysis of the subtle changes indiscounting that occurred for each participant.For example, the interquartile ranges displayedin Figure 2 illustrate considerable overlap acrosscontexts and could be considered an indicationthat the effect of context was not robust.However, it is the relative change that occurredwithin each individual participant’s discountingpattern that should be considered our mostimportant finding. Consider Participant 101,who displayed an AUC value of .08 in thegambling context and .21 in the nongamblingcontext, and Participant 103, who displayedAUCs of .34 and .40 in those respectivecontexts. Both of Participant 101’s AUCs weremuch higher than 103’s (indicating generallygreater discounting by 103), but the relativechange within each participant was similar. Thisillustrates the true value of within-participantanalysis. Although both participants had differ-ent degrees of delay discounting, each wasrelatively equally sensitive to the contextualmanipulations of the present experiment.

The present findings add to a growing line ofresearch on variables that affect delay discount-ing. Previous research has shown that severity ofdiscounting can be affected by conditions ofdeprivation (Giordano et al., 2002), reinforcermagnitude (Green et al., 1997), reinforcer type(Bickel et al., 1999; Madden et al., 1997,1999), and history (Bickel et al.). Thesefindings are important because they suggestthat such variables need to be controlled whenconducting discounting assessments and theyprovide a rationale for conducting multipleassessments within individual participants.

The present data also further our understand-ing of the discounting behavior of pathologicalgamblers, who have been shown to discount togreater degrees than matched controls (Dixon et

al., 2003; Petry & Casarella, 1999). Our datashow that altering the context in which the delay-discounting task is completed can result indifferent degrees of discounting. To date, theimpact of context on altering the behavior ofpathological gamblers has only been speculative(Dixon et al., 2003); however, the current datademonstrate empirical support for this assertion.

The present study was conducted in a naturalsetting and, as such, suffers from some potentialthreats to internal validity that should berecognized. First, our sample may not havebeen representative of pathological gamblers ingeneral. Participants were self-selected volun-teers, and the method of identifying patholog-ical gamblers relied on self-report rather thandirect observation of relevant behavior. Al-though the SOGS is the primary screeninginstrument in the gambling literature, oursample may have been restricted to thosepathological gamblers who responded accurate-ly to the questionnaire. In addition, the samplemay have included gamblers who underesti-mated the prevalence of their gambling. Thissampling problem seems more likely in thatmany pathological gamblers may not wish todisclose the severity of their disorder to anunknown researcher. Furthermore, we usedonly pathological gamblers who were currentlygambling and not seeking treatment for thisdisorder. Thus, our participants may actuallyrepresent the most severe type of pathologicalgambler, one who is currently gambling and notseeking treatment. Another potential confound-ing effect is that our current participants werefree to consume alcohol within the gamblingestablishment, and drug use may have occurredprior to entering this establishment. Therefore,we are unaware of how substance use may havedifferentially affected our participants’ respond-ing across the two contexts. Richards, Zhang,Mitchell, and de Wit (1999) reported, however,that alcohol consumption did not affectresponding on the hypothetical money choicetask in a laboratory context. We were also not


aware of, nor did we control for, the amount oftime the participant was in the gamblingestablishment or the number of winning orlosing wagers that had been made beforecompleting our experiment. This latter pointbegs the question regarding the mechanismresponsible for behavior change across contexts.Relevant variables in the gambling contextmight include the presence of others who aregambling, alcohol consumption, and theamount of money available to participants. Itmay have also been the case that the nongam-bling context was less natural or social in naturethan the off-track betting facility. The specificvariables that contributed to the between-context differences observed in the presentstudy are unknown, but it seems safe toconclude that a gambling context containscritical variables that alter delay discounting,at least for pathological gamblers.

Future researchers may wish to furtherexplore various additional parameters such asimmediate financial instability, which may alterpathological gamblers’ discounting. Pietras andHackenberg (2001), for example, have demon-strated that income and budgetary constraintsaffect risky choice in humans in the laboratory.In translating the implications of their work togambling, an experimenter might assess howmoment-by-moment wins or losses alter thedegree to which that person discounts delayedrewards prior to their initiation of the nextwager. Brief discounting tasks could be com-pleted between successive gambles, and com-parisons could be made based on prior winningor losing wagers. Additional research might alsoattempt to evaluate the correspondence betweendiscounting and the actual size of the next wagermade by a pathological gambler in a gamblingcontext. Finally, researchers might investigatethe potential relations that may exist betweengambling severity (as reported using metrics likethe SOGS) and degrees of discounting. It maybe possible that regression models that includefactors such as SOGS score, context, age of

onset of disorder, and current indebtedness maybe predictive of delay discounting.

In summary, the present experiment illus-trates a clear rationale for conducting trans-lational research on delay discounting. Thecontext in which the experimental procedureswere conducted altered the choice patterns formany of the participants. There is a growingliterature on temporal discounting in peoplewith impulse control disorders. To the extentthat insensitivity to delayed or diffuse conse-quences contributes to the provenance andmaintenance of these behavioral problems,understanding the variables that control tem-poral discounting is of critical importance to thedevelopment of effective treatments (see, e.g.,Petry, 2005). Identifying the conditions underwhich an individual is most sensitive to thelong-term consequences of his or her behaviormay enhance the likelihood that he or she willcommit to a course of action that will bringbehavior under control of those extendedoutcomes. Behavior analysts are uniquely poisedto advance this research by examining environ-ment-based controlling variables instead ofaccepting prescientific, organism-based explana-tions for these pervasive and socially significantbehavioral problems.

REFERENCES

American Psychiatric Association. (2000). Diagnostic andstatistical manual of mental disorders (4th ed., textrev.). Washington, DC: Author.

Bickel, W. K., Odum, A. L., & Madden, G. L. (1999).Impulsivity and cigarette smoking: Delay discountingin current, never, and ex-smokers. Psychopharmacolo-gy, 146, 447–454.

Bickel, W. K., & Vuchinich, R. E. (Eds.). (2002).Reframing health behavior change with behavioraleconomics. Mahwah, NJ: Erlbaum.

Critchfield, T. S., & Kollins, S. H. (2001). Temporaldiscounting: Basic research and the analysis of sociallyimportant behavior. Journal of Applied BehaviorAnalysis, 34, 101–122.

Dixon, M. R., Jacobs, E. A., Sanders, S., Guercio, J.,Soldner, J., & Parker-Singler, S., et al. (2005).Impulsivity, self-control, and delay discounting inpersons with acquired brain injury. Behavioral Inter-ventions, 20, 101–120.


Dixon, M. R., & MacLin, O. H. (2003). Visual basic forbehavioral psychologists. Reno, NV: Context Press.

Dixon, M. R., Marley, J., & Jacobs, E. A. (2003). Delaydiscounting by pathological gamblers. Journal ofApplied Behavior Analysis, 36, 449–458.

Giordano, L. A., Bickel, W. K., Loewenstein, G., Jacobs,E. A., Marsch, L. A., & Badger, G. J. (2002). Mildopioid deprivation increases the degree that opioid-dependent outpatients discount delayed heroin andmoney. Psychopharmacology, 163, 174–182.

Green, L., Myerson, J., & McFadden, E. (1997). Rate oftemporal discounting decreases with amount ofreward. Memory & Cognition, 25, 715–723.

Herrnstein, R. J., & Prelec, D. (1992). A theory ofaddiction. In G. Loewenstein & J. Elster (Eds.),Choice over time (pp. 331–360). New York: Sage.

Kirby, K. N., & Marakovic, N. N. (1995). Modelingmyopic decisions: Evidence for hyperbolic delay-discounting within subjects and amounts. Organiza-tional Behavior and Human Decision Processes, 64,22–30.

Kirby, K. N., Petry, N. M., & Bickel, W. K. (1999).Heroin addicts have higher discount rates for delayedrewards than non-drug-using controls. Journal ofExperimental Psychology: General, 128, 78–87.

Ladouceur, R., Boisvert, J. M., Pepin, M., Loranger, M.,& Sylvain, C. (1994). Social cost of pathologicalgambling. Journal of Gambling Studies, 10, 399–409.

Lesieur, H. R., & Blume, S. B. (1987). The South OaksGambling Screen (SOGS): A new instrument for theidentification of pathological gamblers. AmericanJournal of Psychiatry, 144, 1184–1188.

Madden, G. J., Bickel, W. K., & Jacobs, E. A. (1999).Discounting of delayed rewards in opioid-dependentoutpatients: Exponential or hyperbolic discountingfunctions. Experimental & Clinical Psychopharmacol-ogy, 7, 284–293.

Madden, G. J., Petry, N. M., Badger, G. J., & Bickel, W.K. (1997). Impulsive and self-control choices inopioid-dependent patients and non-drug-using con-trol participants: Drug and monetary rewards.Experimental and Clinical Psychopharmacology, 5,256–262.

Mazur, J. E. (1987). An adjusting procedure for studyingdelayed reinforcement. In M. L. Commons, J. E.Mazur, J. A. Nevin, & H. Rachlin (Eds.), Quantita-tive analyses of behavior: Vol. 5. The effect of delay andof intervening events on reinforcement value (pp. 55–73). Hillsdale, NJ: Erlbaum.

Myerson, J., Green, L., & Warusawitharana, M. (2001).Area under the curve as a measure of discounting.Journal of the Experimental Analysis of Behavior, 76,235–243.

National Gambling Impact Study Commission. (1999).Final report. Retrieved June 6, 2006, from http://govinfo.library.unt.edu/ngisc/reports/fullrpt.html

Odum, A. L., Madden, G. J., Badger, G. J., & Bickel, W.K. (2000). Needle sharing in opioid-dependentoutpatients: Psychological processes underlying risk.Drug and Alcohol Dependence, 60, 259–266.

Odum, A. L., Madden, G. J., & Bickel, W. K. (2002).Discounting of delayed health gains and losses bycurrent, never- and ex-smokers of cigarettes. Nicotine& Tobacco Research, 4, 295–303.

Ostaszewski, P., Green, L., & Myerson, J. (1998). Effectsof inflation on the subjective value of delayed andprobabilistic rewards. Psychonomic Bulletin & Review,5, 324–333.

Petry, N. M. (2005). Pathological gambling. Washington,DC: APA Press.

Petry, N. M., & Bickel, W. K. (1998). Polydrug abuse inheroin addicts: A behavioral economic analysis.Addiction, 93, 729–738.

Petry, N. M., & Casarella, T. (1999). Excessive discount-ing of delayed rewards in substance abusers withgambling problems. Drug and Alcohol Dependence, 56,25–32.

Pietras, C. J., & Hackenberg, T. D. (2001). Risk-sensitivity in humans as a function of earningsbudget. Journal of the Experimental Analysis ofBehavior, 76, 1–19.

Rachlin, H., & Green, L. (1972). Commitment, choiceand self-control. Journal of the Experimental Analysis ofBehavior, 17, 15–22.

Rachlin, H., Raineri, A., & Cross, D. (1991). Subjectiveprobability and delay. Journal of the ExperimentalAnalysis of Behavior, 55, 233–244.

Raineri, A., & Rachlin, H. (1993). The effect of temporalconstraints on the value of money and othercommodities. Journal of Behavioral Decision Making,6, 77–94.

Richards, J. B., Zhang, L., Mitchell, S. H., & de Wit, H.(1999). Delay or probability discounting in a modelof impulsive behavior: Effect of alcohol. Journal of theExperimental Analysis of Behavior, 71, 121–143.

Vuchinich, R. E., & Simpson, C. A. (1999). Delayed-reward discounting in alcohol abuse. In F. J.Chaloupka, M. Grossman, W. K. Bickel, & H. Saffer(Eds.), The economic analysis of substance use and abuse(pp. 103–122). Chicago: The University of ChicagoPress.

Received November 29, 2005Final acceptance June 7, 2006Action Editor, James E. Carr


Documents

Contextual Control of Delay Discounting by Pathological Gamblers