neuroeconomics and adolescent substance abuse: individual ... · functional magnetic resonance...

NEW RESEARCH

JOURNAL

VOLUM

Neuroeconomics and Adolescent SubstanceAbuse: Individual Differences in Neural

Networks and Delay DiscountingCatherine Stanger, Ph.D., Amanda Elton, Ph.D., Stacy R. Ryan, Ph.D.,G. Andrew James, Ph.D., Alan J. Budney, Ph.D., Clinton D. Kilts, Ph.D.

Objective: Many adolescents with substance use problems show poor response to evidence-based treatments. Treatment outcome has been associated with individual differences inimpulsive decision making as reflected by delay discounting (DD) rates (preference for imme-diate rewards). Adolescents with higher rates of DD were expected to show greater neuralactivation in brain regions mediating impulsive/habitual behavioral choices and less activationin regions mediating reflective/executive behavioral choices. Method: Thirty adolescentsbeing treated for substance abuse completed a DD task optimized to balance choices of imme-diate versus delayed rewards, and a control condition accounted for activation during magni-tude valuation. A group independent component analysis on functional magnetic resonanceimaging time courses identified neural networks engaged during DD. Network activity wascorrelated with individual differences in discounting rate. Results: Higher discounting rateswere associated with diminished engagement of an executive attention control networkinvolving the dorsolateral prefrontal cortex, dorsomedial prefrontal cortex, inferior parietalcortex, cingulate cortex, and precuneus. Higher discounting rates also were associated with lessdeactivation in a “bottom-up” reward valuation network involving the amygdala, hippo-campus, insula, and ventromedial prefrontal cortex. These 2 networks were significantly nega-tively correlated. Conclusions: Results support relations between competing executive andreward valuation neural networks and temporal decision making, an important, potentiallymodifiable risk factor relevant for the prevention and treatment of adolescent substance abuse.Clinical trial registration information—The Neuroeconomics of Behavioral Therapies forAdolescent Substance Abuse, http://clinicaltrials.gov/, NCT01093898. J. Am. Acad. Child Adolesc.Psychiatry, 2013;52(7):747–755. Key Words: adolescent substance abuse, delay discounting,functional magnetic resonance imaging, neuroeconomics

primary model of decision making usedto explain substance use behavior is
A intertemporal decision making or choices
between 2 alternatives that occur at differentpoints in time.1 There is a general tendency forrewards to lose value the further away they are inthe future, a phenomenon referred to as delaydiscounting (DD). DD rates generally followa hyperbolic function, in which reward valuationdecreases very rapidly across short delays andthen more slowly across longer delays.2 DD ishypothesized to be particularly relevant to

Supplemental material cited in this article is available online.

OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATR

E 52 NUMBER 7 JULY 2013

substance use because substance use can becharacterized as a choice between the tangibleand immediate rewards of consumption and thedelayed rewards of abstinence. There is a largeliterature supporting the association between DDrate and adult and adolescent substance abuseonset and severity.1,3,4 Further, studies have re-ported worse adult and adolescent substanceabuse treatment outcomes for high discounters.5-7

For example, the authors previously reported thattreatment-enrolled teens with higher DD rateswere less likely to achieve abstinence.8

NEURAL MECHANISMS OF DDA meta-analysis identified 25 regions of signifi-cant neural activation during DD tasks.9 Three

Y

www.jaacap.org 747

http://clinicaltrials.gov/

http://www.jaacap.org

STANGER et al.

primary regions of robust activation includevalue-related regions (ventral striatum), valueconsideration regions (medial prefrontal cortex[PFC]), and future forecasting regions (posteriorcingulate cortex). These regions are consistentwith the valuation network proposed by Petersand Buchel,10 who also proposed 2 additionalnetworks important in DD: a cognitive controlnetwork, involving activation of the anteriorcingulate cortex and decreased top-down regu-lation of the medial PFC by the dorsolateral PFC,and a prospection/episodic imagery network,involving activity in the medial temporal lobe(hippocampus and amygdala). However, thereare developmental differences between adoles-cent and adults in these regions that may affectDD. Adolescents show maturation similar toadults in limbic and paralimbic “bottom-up”brain regions that function with respect toprimary reinforcers11,12 and slower maturationof the “top-down” frontal cortex and PFC, whichregulate executive function and decision mak-ing.11,13 This asymmetric development is theo-rized to be related to riskier decision makingin adolescents than in adults.14,15 This combina-tion of heightened neural response to rewardand motivational cues and delayed behavioraland cortical control may contribute to adolescentpreferences for immediate rewards.16

There are relatively few studies of neuralmechanisms of DD in adolescence. Several stud-ies have examined age-related functional andstructural brain changes related to DD, and 2have identified relations between neural functionand structural connectivity and DD rates thatwere independent of age-related changes.17,18 Forexample, strengthening of functional couplingamong the ventromedial (vm) PFC, ventralstriatum, anterior cingulate cortex, and temporallobe was associated with decreased discounting,suggesting that developing connectivity betweenthe vmPFC and ventral striatal systems mayaccount for individual differences in DD rates.17

Increased ventral PFC white matter organizationalso is associated with decreased DD rates.18

These results suggest that there may be indi-vidual brain and behavioral differences evident inadolescence that confer risk independent ofdevelopmental changes.

Several studies also have documented neuralstructural and activation differences betweenadolescent substance users and controls.19,20

There is longitudinal evidence that alcohol usein adolescence may negatively affect memory

JOURN

748 www.jaacap.org

and attention21 and evidence of neural activationdifferences between substance users even at theearliest stages of tobacco use and demographi-cally matched same-age peers.22 However, mostinformative for treatment development is identi-fying the utility of individual neural differencesin youth who display problem use and/or whomeet diagnostic criteria in predicting individualdifferences in treatment-relevant constructs suchas decision making to ultimately improve treat-ment outcomes.

The present study was designed to identifyindividual differences in neural network activa-tion related to decision making (DD) in adoles-cents with substance use problems. Adolescentsubstance users were assessed at treatment entryusing laboratory and functional magnetic reso-nance imaging (fMRI) methods while makingintertemporal choice decisions. Analyses exploredrelations between the neural processing patternsthat occur when making choices between imme-diate and delayed rewards and DD rate. Theauthors hypothesized that the DD task wouldactivate neural networks consistent with rewardvaluation and cognitive control, and that thepatterns of activation in these networks would becorrelated with individual differences in DD.

METHODParticipantsParticipants were recruited from 2 ongoing studiesinvestigating behavioral treatments for adolescentsubstance abuse (marijuana trial and alcohol trial).Fifty-two subjects enrolled in the treatment studiesduring recruitment for the present study. Two teensrefused screening and 9 screened eligible but declinedto participate in this study. Six subjects were not eligi-ble for MRI owing to metal in their body (most oftenbraces), and 2 reported claustrophobia and were notscanned. In addition, data for 3 scanned subjects wereremoved from the dataset because of head movement(n ¼ 1), incomplete discounting data (n ¼ 1), andremoval from the scanner owing to claustrophobia(n ¼ 1). Therefore, 30 scanned subjects were in-cluded in the analyses. These participants were 12 to18 years old (mean age 15.7 years, SD 1.7 years; 80%male; 63.3% Caucasian and 36.7% African American).

Teens in the marijuana trial (n ¼ 14) reportedmarijuana use in the past 30 days or provided a tetra-hydrocannabinol-positive urine test and met DSMcriteria for marijuana abuse or dependence. Teens inthe alcohol trial (n ¼ 16) reported alcohol use in thepast 30 days and met DSM criteria for alcohol abuseor dependence or had had 1 binge episode (�5drinks in 1 day) in the past 90 days. A bachelor’s level

AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY

VOLUME 52 NUMBER 7 JULY 2013


NEUROECONOMICS AND SUBSTANCE ABUSE

research assistant administered the Vermont Struc-tured Diagnostic Interview to assess DSM-IV substanceuse and mental health disorders.23 Interviewers weretrained to administer the instruments by manualreview, observation, and supervised practice inter-views. The interview has shown good psychometricproperties.23 Binge drinking was assessed using theTime-Line Follow-Back method for 90 days beforeintake.24 Alcohol-dependent youth were excludedfrom the marijuana trial and were assigned to thealcohol trial. Youth eligible for the 2 trials wereassigned to the alcohol trial. Eighteen adolescents(60%) met the criteria for marijuana abuse or depen-dence only, 3 (10%) met the criteria for alcohol abuseor dependence only, 6 (20%) met the criteria for mari-juana abuse or dependence and alcohol abuse or de-pendence, and 3 (10%) reported binge drinking only.On average, adolescents reported smoking marijuanaon 9.40 days (SD 9.66 days) and drinking alcoholon 1.87 days (SD 2.97 days) in the 30 days before theintake appointment. In addition, adolescents reporteddrinking an average of 3.29 drinks (SD 4.74 drinks,range ¼ 0–16 drinks) per drinking day in the 30 daysbefore the intake appointment. Based on caregiverand/or teen report, subjects also met criteria for 1 ormore of these disorders: attention-deficit/hyperactivitydisorder (caregiver: 36.7%, teen: 13.3%; caregiver and/or teen: 36.7%), conduct disorder (caregiver: 16.7%,teen: 6.7%; caregiver and/or teen: 16.7%), oppositionaldefiant disorder (caregiver: 36.7%; teen: 16.7%; care-giver and/or teen: 43.3%), major depression (caregiver:6.7%; teen: 13.3%; caregiver and/or teen: 16.7%), andgeneralized anxiety disorder (caregiver: 10.0%; teen:13.3%; caregiver and/or teen: 16.7%). Tables S1 and S2,available online, present individual and group-leveldemographic, DD, diagnostic, and substance use data.

Measurement of DDA DD task was administered to each subject immedi-ately before MRI acquisition using a computerizedprogram. Adolescents were asked to choose betweenreceiving (hypothetically) $1,000 after a delay andreceiving a smaller amount of money immediately. Foreach delay, the subject was presentedwith 6 consecutivedecision-making trials. The delay intervals were 1 day, 1week, 1 month, 6 months, 1 year, 5 years, and 25 years.For each delay, the amount of money offered “now”

started at $500, and the amount increased or decreasedbased on the subject’s choice for trials 2 through 6.

DD rate was estimated using Mazur’s equation2:Vd ¼ V/(1 þ kD), where Vd represents the discountedvalue at D delay, V is the undiscounted amount, and kis the estimated discounting parameter. High valuesof k indicate greater discounting or preference forimmediate rewards. Vd was derived by calculatingindividuals’ indifference point, which is the value ofthe immediate reward that is considered as attractiveas the $1,000 delayed reward. Indifference points

JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATR


were calculated for each delay and fit to the hyper-bolic model of DD rate (k) and then log trans-formed (lnk).

fMRI ProceduresDD Task. The DD task completed in the scanner wasoptimized for functional neuroimaging using an event-related trial design. Functional T2*-weighted echopla-nar images were acquired using a Philips Achieva3.0 Tesla X-series MRI and an 8-channel head coil(Philips Healthcare, Andover, MA) with the followingparameters: 3 � 3 � 3-mm3 isotropic voxels, 2,000-msrepetition time, 30-ms echo time, 240 � 240-mm fieldof view, 90� flip angle, 80 � 80 matrix, and 37 slices. T1-weighted structural images were acquired for align-ment and tissue segmentation purposes using anMPRAGE sequence (matrix 192 � 192, 160 slices,repetition time 2,600 ms, echo time 3.02 ms, flip angle8�, final resolution 1 � 1 � 1-mm3). In the scanner, thesubject was presented with discounting trials for eachof 4 delay intervals (1 month, 6 months, 1 year, and 5years) plus control trials (choice between 2 differentamounts of money, both received “today”) presentedin a randomized sequence. The fMRI DD task usedeach individual’s indifference points from the pre-MRIDD task as the starting value for the smaller, sooner(SS) amount at each delay. This starting point wasdesigned to produce equal numbers of SS and larger,later (LL) choices at each delay, making the task simi-larly challenging for all subjects. The smaller value wasalways offered “today” and presented on the left sideof the screen and $1,000 was offered at 1 of the delayintervals and displayed on the right. The subjectmade a decision by pressing 1 of 2 buttons on a buttonbox corresponding with the subject’s choice. Afterthe decision, the selected option was surroundedwith a bold rectangle for 1 second to confirm that aresponse was recorded. Choices involved the samedelay intervals and LL amount ($1,000) for all subjects,with variable SS amounts based on the subject’sstarting indifference point and subsequent choices ateach delay.

The task was divided into 2 runs, each consistingof 2 sets of 25 trials and 3 sets of rest periods(25 seconds), for a total of 100 decision-making trials.The task was self-paced and lasted approximately20 minutes. Each set of 25 trials consisted of 5 trialsof each of the 4 delays and 5 control trials with afixed interstimulus interval of 5 seconds. The subject’sresponse on each trial of each delay determinedwhether or not the smaller amount offered “today”on the next trial of that delay increased (prior selec-tion of $1,000) or decreased (prior selection of smalleramount; algorithm available on request from thefirst author).

Data Processing and Analyses. Image preprocessingand statistical analyses were performed using Analysisof Functional Neuroimages (AFNI; http://afni.nimh.

Y

www.jaacap.org 749

http://afni.nimh.nih.gov


STANGER et al.

nih.gov) software. Functional images underwent thefollowing preprocessing steps: slice time correction,de-obliquing, motion correction, de-spiking, align-ment to the subject’s structural image, warping toMontreal Neurological Institute standardized space,removal of signal fluctuations in white matter andcerebral spinal fluid from voxel time courses, spatialsmoothing with a 6-mm full-width at half-maximumgaussian kernel, scaling to percentage of signalchange, and, for voxel-wise contrasts, masking of non-gray matter voxels with masks created using FSLsoftware (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) fromthe subject’s structural image.

Voxel-wise general linear model analyses wereconducted in AFNI. Decision trials were modeled asepochs beginning at the presentation of choices andending with a response. Participants’ mean responsetime ranged from 1.69 to 6.96 seconds (mean 3.31seconds, SD 1.16 seconds). Neural activation associ-ated with making 2 types of decisions, SS or LL, wascompared with that of control “no delay” trials(CON) in voxel-wise contrasts. In addition, DD rate(lnk value) was correlated with neural activity whilemaking DD decisions in the scanner (i.e., SS versusCON, LL versus CON). Results from whole-brainvoxel-wise analyses were subjected to multiplecomparison correction based on 10,000 Monte Carlosimulations conducted using the 3dClustSim com-mand in AFNI. To obtain a corrected p value of .05,only clusters composed of 17 or more voxelssurviving a p-value threshold of .005 were consideredsignificant.

In addition, a group independent component anal-ysis (ICA) on fMRI time courses25 was conducted withgroup ICA of fMRI toolbox (GIFT) in MATLAB,26

solving for 20 components. ICA was conducted withthe infomax algorithm and the following selectedoptions: data entry (2 runs per subject), no dummyscans, using spatial temporal regression for back-reconstruction, removal of image mean at each timepoint, standard principal component analysis withstacked datasets, 2-step data reduction, no batch esti-mation, and scaling values to z scores. ICA was re-peated 5 times using the ICASSO algorithm to identifythe most reliable and stable components across all5 ICA iterations.

For each subject, the experimental design wasconvolved with the SPM27 canonical hemodynamicresponse function to estimate the blood oxygenationlevel-dependent signal response for 3 decision types:SS, LL, and CON. Estimated blood oxygenation level-dependent responses for each trial type were modeledin regression analyses in GIFT as predictors of eachcomponent time course, with 6 directions of headmotion included as covariates. This method is analo-gous to the general linear model approach to voxel-wise analysis of fMRI task data, except thatdata were reduced from thousands of voxels to 20independent components. The association of each

JOURN

750 www.jaacap.org

component with each trial type (SS, LL, and CON) wasrepresented by b estimates and t values. Thirteencomponents representing head motion, noise, andsensory processing were excluded from further anal-ysis. For the remaining 7 components, contrast valueswere calculated for each subject for SS-CON trials andLL-CON trials using b estimates from the regressionanalysis. A positive contrast value indicates that thenetwork was more active during SS or LL than duringCON trials, and a negative contrast value indicatesthat the network was more active during CON than SSor LL trials. Individual SS-CON and LL-CON con-trast values for each of the 7 retained componentswere correlated with lnk. Additional analyses con-trolled for age, sex, and past 30-day substance use(number of alcoholic drinks, number of days of mari-juana and/or K2 [synthetic cannabis] use, and numberof days of tobacco use). The authors hypothesized thatadolescents with substance abuse problems who havehigher discounting rates, reflecting a greater prefer-ence for immediate rewards, would show greateractivation in neural regions mediating impulsive/ha-bitual behavioral choices and less activation in neuralregions mediating more reflective/executive behav-ioral choices compared with individuals exhibitinglower rates of discounting.

RESULTSLocalization of Whole-Brain Activations Related toIntertemporal Choice BehaviorPairwise Contrasts. Voxel-wise, brain-wide plan-ned contrasts of SS-CON and LL-CON choicesresulted in distributed neural activations typi-cally attributed to impulsive and deliberativechoice behavior. Task-related activations are re-ported in Tables S3 and S4, available online, andrepresent those that survived a cluster-levelcorrection for multiple comparisons.

SS Versus CON. Compared with the judgmentof relative monetary amount in the CON trials,the choice of SS rewards was associated withactivation of the dorsomedial, dorsolateral, andpolar PFCs, precuneus, posterior cingulate cortex,fusiform and lingual gyri, superior temporalgyrus, cerebellum, and paracentral lobule(Table S3, available online). Compared with CONchoices, SS choices were associated with lessactivation of the inferior parietal cortex, rightprecentral gyrus extending into the inferiorfrontal cortex, bilateral parahippocampal gyri,and temporal and parietal cortices.

LL Versus CON. Compared with CON choices,the choice of LL rewards was associated withactivation of the presupplementary motor area,polar and dorsolateral PFCs, cuneus extendinginto the inferior occipital gyrus, cerebellum,



http://afni.nimh.nih.gov

http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/



posterior cingulate cortex, bilateral dorsal cau-date nucleus, and parietal cortex (Table S4,available online). Choice of LL rewards comparedwith CON choices were associated with lessactivation of the bilateral middle temporal gyrus,right inferior frontal cortex extending into theamygdala, right dorsolateral PFC, inferior pari-etal cortex, cuneus, middle cingulate cortex, leftinferior frontal cortex, and left amygdala.

SS Versus LL. Compared with the LL decisiontrials, SS choices were associated with greateractivation of the right lingual gyrus (7.5, �70.5,�6.5 mm) and occipital cortex (34.5, �85.5,17.5 mm). No other relative activations for LLversus SS trials survived multiple comparisoncorrection.

Regression AnalysesBrain-wide analyses assessed the relation bet-ween individual differences in discounting rate(lnk) and the magnitude of task-related regionalbrain activation. Significantly correlated brainregions (p < .05, corrected) are reported inTable S5, available online. For SS-CON, lnkwas significantly correlated with activation ofthe dorsomedial PFC, bilateral middle/posteriorinsula, right posterior superior temporal sul-cus, precuneus, and posterior parietal cortex. ForLL-CON, lnk was significantly correlated with

FIGURE 1 Valuation network: larger, later (LL) choice trials vregions were coactivated by task, as identified from group indresonance imaging time courses. (B) Correlation between log-activation. Individual contrast values for LL versus control trials (pnegative value indicates less activity for LL versus control) are o



activation of the bilateral middle insula, superiorand middle temporal gyri, dorsomedial PFC, andprecuneus. For SS-LL, no regions of activationwere significantly correlated with lnk after whole-brain correction.

Group ICA ResultsOf the 7 components of activation assessed incorrelation analyses for their relation to indi-vidual differences in discounting rates, 2 com-ponents were significantly correlated with lnk(p < .05), although neither survived a Bonferronicorrection for the number of components tested(p < .05/[2 contrast types {SS versus CON and LLversus CON} � 7 components ¼ 14] ¼ 0.00357).Those 2 components are described below, andactivation patterns for the other 5 networks areshown in Figure S1, available online.

Valuation Network. As shown in Figure 1B, lnkpositively correlated (r ¼ 0.45, p ¼ .013) withrelative activation for LL versus CON choices fora component comprising the amygdala andhippocampus, paralimbic cortex involving thevmPFC, insula and posterior cingulate cortices,and ventral striatum (Figure 1A). These coac-tivated regions are involved in motivation, valu-ation, prospection, and salience processing,10,28,29

suggesting that this component represents avaluation network. This association remained

ersus control trials. Note: (A) Network map. Depictingependent component analysis of functional magnetictransformed delay discounting rate (lnk) and networkositive value indicates greater activity for LL versus control;n the y axis; individual values of lnk are on the x axis.

Y

www.jaacap.org 751


FIGURE 2 Cognitive control/executive function network: smaller, sooner (SS) choice trials versus control trials. Note:(A) Network map. A second network of coactivated brain regions as determined from group independent componentanalysis of functional neuroimaging time courses. (B) Correlation between log-transformed delay discounting rate (lnk)and network activation. Individual contrast values for SS versus control trials (positive value indicates greater activity forSS versus control; negative value indicates less activity for SS versus control) are on the y axis; individual values of lnk areon the x axis.

STANGER et al.

after controlling for age, sex, and substance usefrequency (r ¼ 0.56, p ¼ .004).

Cognitive Control/Executive Function Network.As shown in Figure 2B, lnk negatively correlated(r ¼ �0.41, p ¼ .023) with activation for SS-CONchoices in a bilateral frontal-parietal network(right > left). Coactivated regions included theventrolateral and dorsolateral PFCs, superiorparietal cortex, and precuneus (Figure 2A). Theseregions are involved in executive functions suchas goal representation, cognitive control, andresponse selection.10,30,31 This associationremained after controlling for age, sex, and sub-stance use frequency (r ¼ �0.44, p ¼ .029).

Network CorrelationsTo assess the role of functional network interac-tions in DD, the correlation was computedbetween activity in these 2 networks duringdecision-making trials (mean activation duringSS and LL choices). As shown in Figure 3, activityin these networks was highly negatively corre-lated (r ¼ �0.67, p < .0001), suggesting that the2 networks function in a reciprocal manner incontributing to individual differences in DD.

DISCUSSIONWhole-brain pairwise contrasts indicated exten-sive activations broadly consistent with many

JOURN

752 www.jaacap.org

other DD fMRI studies.9 Group ICA yielded 7components reflecting theoretically consistentregions of neural coactivation during all trials.Activation in 2 components or networks showedsignificant relations with individual DD rates.However, neither effect survived the Bonferronicorrection, supporting the need to replicate thesefindings. One network, a putative valuationnetwork, showed ventral limbic activationsinvolving the amygdala and hippocampus,paralimbic cortex involving the vmPFC, insula,and posterior cingulate cortex, and the ventralstriatum. These regions are involved in multiplecognitive functions integral to intertemporaldecision making, including valuation, pro-spection/future forecasting, and episodicimagery.9,10 Others have reported similar rela-tions between ventral striatal activity andadolescent risk taking.32 These findings uniquelydemonstrate by ICA that these separate neuralprocesses related to DD are organized intoa higher-order network. Further, activation ofthis higher-order network predicted individualdifferences in DD rates. There was less activityamong lower discounters in this “bottom-up”reward valuation network when choosingLL (compared with simple immediate rewardmagnitude valuation) and lesser suppression ofactivity among higher discounters.




FIGURE 3 Correlation between valuation and cognitivecontrol networks. Note: Individuals’ mean brain activityfor the cognitive control network (x axis) plotted againstmean brain activity for the valuation network (y axis) forall decision making trials (irrespective of choice) versuscontrol trials.


Activation in a network that likely reflectscognitive control and executive function alsowas related to individual DD rates. Similar tothe cognitive control, the regulatory networkproposed by Peters and Buchel,10 this networkshowed bilateral frontoparietal activations en-compassing the ventrolateral and dorsolateralPFCs, superior parietal cortex, and precuneus.Higher DD teens showed less engagement in thisexecutive network than lower DD teens duringimpulsive decision making, with SS choicesreflecting less involvement of this cognitive con-trol network for higher discounters. In otherwords, SS choices involve more executive pro-cessing for lower discounters.

The group ICA results are consistent with 2independent neural processing networks medi-ating the valuation and choice processes relatedto DD with decreasing activation of a fronto-parietal network and increasing activation ofa limbic-paralimbic network, both predictinggreater discounting. These results were consis-tent with a study showing that higher risktaking on a gambling task was associated withdecreased activity in control-related regions inthe dorsal medial PFC and greater activity inreward (valuation) regions in the vmPFC innonreferred adolescents.33 In the present study,similar relations were observed between DDrate and activation in these regions in the dorsalmedial PFC (integrated into a larger cognitivecontrol network) and in these regions in thevmPFC (integrated into a larger valuationnetwork).



The strong negative correlation observed be-tween the 2 networks suggests that the balancein activity between these networks influencestemporal decision making. Several studies havesuggested that there are interacting (competing)networks involved in reward valuation,34,35 withan evolutionarily older impulsive system (limbicand paralimbic regions) primarily involved inthe valuation of immediate rewards and themore recently developed executive system(prefrontal regions) involved in the considerationof the future and the selection of delayedrewards. The balance (or imbalance) in activationand connectivity between these competingvaluation systems is hypothesized to underlieindividual DD rates.1 Alternatively, others havesuggested that reward valuation is betterconceptualized as reflecting the activity ofa single neural network that tracks subjectivevalue at all delays.36 Because analyses collapsedacross all delays, the role of delay in the activityof these networks is unknown. However, thepresent results do support the involvement of 2distinct neural networks that function in oppo-sition during temporal decision making, withlower discounters showing greater activity inexecutive control regions and greater suppres-sion of activity in valuation regions.

The period of mid-adolescence (14–16 yearsold) appears to be the time of greatest develop-mental change in DD,15 suggesting that adoles-cence might be a unique and ideal time to attemptto decrease DD. Interventions such as contin-gency management that attempt to shift prefer-ences to delayed rewards might be most effectiveduring this developmental period. The use ofrewards also may influence decision makingand its neural correlates in adolescent substanceusers. For example, adolescents with substanceuse problems have shown greater activation thancontrol adolescents in prefrontal cognitive controlregions during an inhibition task when rewardswere available.37 Similarly, the use of rewardsfacilitates cognitive control in adolescents toa greater extent than in adults.38 Thus, treatmentapproaches that offer consistent and tangiblerewards might be particularly effective inadolescence, and the mechanism for such en-hanced effects might be enhanced engagement ofcognitive control or executive brain regionsrelated to DD.

Working memory training also may influenceneural function related to DD, leading to de-creases in substance use. For example, individual

Y

www.jaacap.org 753


STANGER et al.

differences in DD among healthy adults arecorrelated with activity in the left anterior PFCwhile performing a working memory task.39

Further, Bickel et al.40 showed that workingmemory training resulted in lower DD rates inadult stimulant abusers. Similarly, Houben et al.41

showed working memory training led to signifi-cant decreases in alcohol intake in problemdrinkers. These results suggest that interven-tions involving working memory training mightenhance treatment response by influencing theactivity or functional balance of neural networksthat underlie DD.

Overall, this study was intended as an initialexploratory study, and these results should beconsidered preliminary until they are replicatedwith a larger, independent sample. Beyond thetested variables of age, sex, and recent substanceuse frequency, the sample was heterogeneousfor other substance use and/or mental healthhistory and problems, which may have influ-enced the findings. Overall, the sample sizeprecluded a cross-sectional assessment ofdevelopmental changes in these networks or inDD. It also was unable to control exposure tospecific substances in specific quantities. Thisstudy tested individual differences amongsubstance users entering treatment. It will beimportant in future studies to assess the gener-alizability of these networks to teens who donot use substances, teens with substance useproblems but not seeking treatment, and simi-lar samples of adults. There are also differentmethods and fMRI task parameters that havebeen used to study DD that may contribute todifferences in activation across studies. Forexample, ensuring comparability in decisiondifficulty across subjects and balancing fre-quency of SS and LL choices may maximizerelations between behavioral measurements andneural activity,42 but may minimize differencesin neural activity for SS versus LL.

DD is related to many forms of substanceabuse, and it may be an informative marker ofindividual differences that could predict treat-ment response and/or improve as a result oftreatment. Two neural processing networks were

JOURN

754 www.jaacap.org

found to relate to individual differences in DDrates. These bottom-up (e.g., limbic and para-limbic) and top-down (e.g., parietal andprefrontal) networks functioned in oppositionwhile subjects made temporal decisions aboutrewards. Developmental differences in thematuration of these networks may make teensmore vulnerable to impulsive decision makingand substance use and more responsive to inter-ventions targeting these systems. Neuroeconomicapproaches can contribute to the understand-ing of neural mechanisms that underlie DDbehavior and thereby may offer additional cluesto better direct prevention or treatment ap-proaches. These results showing discounting-related differences in neural activation areconsistent with competing neurobehavioral deci-sion systems theory. Interventions to modify DDand its underlying neural mechanisms includingthose targeting working memory might leadto enhanced treatment outcome in adolescentsubstance users. &

AL

Accepted April 23, 2013.

Drs. Stanger and Budney are with the Geisel School of Medicine atDartmouth. Dr. Elton is with the University of North Carolina at ChapelHill. Dr. Ryan is with the University of Texas Health Science Center atSan Antonio. Drs. James and Kilts are with the University of Arkansas forMedical Sciences.

This research was funded by National Institute on Drug Abuse grantsDA029442, DA015186, and DA022981, by National Instituteon Alcohol Abuse and Alcoholism grant AA016917, and byUL1TR000039 and KL2TR000063 through the NIH National Centerfor Research Resources and the National Center for AdvancingTranslational Sciences.

Disclosure: Dr. Stanger has received support from the National Insti-tutes of Health (NIH) and the Sturgis Foundation. Dr. Budney hasreceived support from the NIH. Dr. Kilts has received support from theNIH and has attended a scientific advisory board meeting for Allerganand a national advisory board meeting for a mental health facility(Skyland Trail). He is a co-holder of U.S. patent 6,373,990 (“Methodand Device for the Transdermal Delivery of Lithium”). Drs. Elton, James,and Ryan report no biomedical financial interests or potential conflictsof interest.

Correspondence to Catherine Stanger, Ph.D., Department of Psychi-atry, Geisel School of Medicine at Dartmouth, 85 Mechanic Street,Suite B3-1, Lebanon, NH 03756; e-mail: [email protected]

0890-8567/$36.00/ª2013 American Academy of Child andAdolescent Psychiatry

http://dx.doi.org/10.1016/j.jaac.2013.04.013

REFERENCES
1. Bickel WK, Miller ML, Yi R, Kowal BP, Lindquist DM, Pitcock JA.
Behavioral and neuroeconomics of drug addiction: competingneural systems and temporal discounting processes. Drug AlcoholDepend. 2007;90:S85-S91.

2. Mazur JE. An adjusting procedure for studying delayed reinforce-ment. In: Commons ML, Mazar JE, Nevin JA, Rachlin H, eds.

Quantitative Analysis of Behavior: The Effect of Delay and of Inter-vening Events on Reinforcement Value. Vol 5. Hillside, NJ:Psychology Press; 1987:55–73.

3. Field M, Christiansen P, Cole J, Goudie A. Delay discounting andthe alcohol Stroop in heavy drinking adolescents. Addiction. 2007;102:579-586.

OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY


mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.jaac.2013.04.013

http://refhub.elsevier.com/S0890-8567(13)00263-3/sref1









4. Audrain-McGovern J, Rodriguez D, Tercyak KP, Epstein LH,Goldman P, Wileyto EP. Applying a behavioral economic frame-work to understanding adolescent smoking. Psychol AddictBehav. 2004;18:64-73.

5. Yoon JH, Higgins ST, Heil SH, Sugarbaker RJ, Thomas CS,Badger GJ. Delay discounting predicts postpartum relapse tocigarette smoking among pregnant women. Exp Clin Psycho-pharmacol. 2007;15:176-186.

6. Washio Y, Higgins ST, Heil SH, et al. Delay discountingis associated with treatment response among cocaine-dependentoutpatients. Exp Clin Psychopharmacol. 2011;19:243-248.

7. Krishnan-Sarin S, Reynolds B, Duhig AM et al. Behavioralimpulsivity predicts treatment outcome in a smoking cessationprogram for adolescent smokers. Drug Alcohol Depend. 17 2007;88:79–82.

8. Stanger C, Ryan S, Fu H, et al. Delay discounting predicts treat-ment outcome among adolescent substance abusers. Exp ClinPsychopharmacol. 2012;20:205-212.

9. Carter R, Meyer J, Huettel S. Functional neuroimaging of inter-temporal choice models: a review. J Neurosci Psychol Econ. 2010;3:27-45.

10. Peters J, Buchel C. The neural mechanisms of inter-temporaldecision-making: understanding variability. Trends Cogn Sci.2011;15:227-239.

11. Blakemore SJ, Choudhury S. Development of the adolescent brain:implications for executive function and social cognition. J ChildPsychol Psychiatry. 2006;47:296-312.

12. Giedd JN, Blumenthal J, Jeffries NO, et al. Brain developmentduring childhood and adolescence: a longitudinal MRI study. NatNeurosci. 1999;2:861-863.

13. Gogtay N, Giedd JN, Lusk L, et al. Dynamic mapping of humancortical development during childhood through early adulthood.Proc Natl Acad Sci U S A. 2004;101:8174-8179.

14. Green L, Fry AF, Myerson J. Discounting of delayed rewards:a life-span comparison. Psychol Sci.. 1994;5:33-36.

15. Steinberg L, Graham S, O’Brien L, Woolard J, Cauffman E,Banich M. Age differences in future orientation and delay dis-counting. Child Dev. 2009;80:28-44.

16. Casey BJ, Jones RM, Somerville LH. Braking and accelerating ofthe adolescent brain. J Res Adolesc. 2011;21:21-33.

17. Christakou A, Brammer M, Rubia K. Maturation of limbic corti-costriatal activation and connectivity associated with develop-mental changes in temporal discounting. Neuroimage. 2011;54:1344-1354.

18. Olson EA, Collins PF, Hooper CJ, Muetzel R, Lim KO, Luciana M.White matter integrity predicts delay discounting behavior in 9- to23-year-olds: a diffusion tensor imaging study. J Cogn Neurosci.2009;21:1406-1421.

19. Abdullaev Y, Posner MI, Nunnally R, Dishion TJ. Functional MRIevidence for inefficient attentional control in adolescent chroniccannabis abuse. Behav Brain Res. 2010;215:45-57.

20. Lopez-Larson MPL-LMP, Bogorodzki P, Rogowska J, et al. Alteredprefrontal and insular cortical thickness in adolescent marijuanausers. Behav Brain Res. 2011;220:164-172.

21. Tapert SF, Granholm E, Leedy NG, Brown SA. Substance use andwithdrawal: neuropsychological functioning over 8 years inyouth. J Int Neuropsychol Soc. 2002;8:873-883.

22. Rubinstein ML, Luks TL, Moscicki AB, Dryden W, Rait MA,Simpson GV. Smoking-related cue-induced brain activation inadolescent light smokers. J Adolesc Health. 2011;48:7-12.

23. Hudziak JJ, Copeland W, Stanger C, Wadsworth M. Screening forDSM-IV externalizing disorders with the Child Behavior



Checklist: a receiver-operating characteristic analysis. J ChildPsychol Psychiatry. 2004;45:1299-1307.

24. Sobell LC, Sobell MB. Timeline follow-back: a technique forassessing self-reported alcohol consumption. In: Litten RZ,Allen JP, eds. Measuring Alcohol Consumption: Psychosocial andBiological Methods. Totowa, NJ: Humana Press; 1992:41-72.

25. Calhoun VD, Adali T, Pearlson GD, Pekar JJ. A method for makinggroup inferences from functional MRI data using independentcomponent analysis. Hum Brain Mapp. 2001;14:140-151.

26. MATLAB [computer program]. Version 6.5.1. Natick, MA: TheMathworks; 2003.

27. Frackowiak RSJ, Friston KJ, Frith CD, Dolan RJ, Mazziotta JC, eds.Human Brain Function. Academic Press; San Diego, CA; 1997.

28. Zink CF, Pagnoni G, Martin-Skurski ME, Chappelow JC, BernsGS. Human striatal responses to monetary reward depend onsaliency. Neuron. 13 2004;42:509–517.

29. Benoit RG, Gilbert SJ, Burgess PW. A neural mechanism mediatingthe impact of episodic prospection on farsighted decisions.J Neurosci. 2011;31:6771-6779.

30. Ridderinkhof K, van den Wildenberg W, Segalowitz S, Carter C.Neurocognitive mechanisms of cognitive control: the role ofprefrontal cortex in action selection, response inhibition, perfor-mance monitoring, and reward based learning. Brain Cogn. 20042004;56:129–140.

31. Hare TA, Camerer CF, Rangel A. Self-control in decision-makinginvolves modulation of the vmPFC valuation system. Science.2009;324:646-648.

32. Galvan A, Hare T, Voss H, Glover G, Casey BJ. Risk-taking andthe adolescent brain: who is at risk? Dev Sci. 2007;10:F8-F14.

33. Van Leijenhorst L, Moor BG, Op de Macks ZA, Rombouts SARB,Westenberg PM, Crone EA. Adolescent risky decision-making:neurocognitive development of reward and control regions.Neuroimage. 2010;51:345-355.

34. Bechara A. Decision making, impulse control and loss of will-power to resist drugs: a neurocognitive perspective. Nat Neurosci.2005;8:1458-1463.

35. McClure SM, Laibson DI, Loewenstein G, Cohen JD. Separateneural systems value immediate and delayed monetary rewards.Science. 2004;306:503-507.

36. Monterosso JR, Luo S. An argument against dual valuation systemcompetition: cognitive capacities supporting future orientationmediate rather than compete with visceral motivations. J NeurosciPsychol Econ. 2010;3:1-14.

37. Chung T, Geier C, Luna B, et al. Enhancing response inhibition byincentive: comparison of adolescents with and without substanceuse disorder. Drug Alcohol Depend. 2011;115:43-50.

38. Jazbec S, Hardin M, Schroth E, McClure E, Pine D, Ernst M. Age-related influence of contingencies on a saccade task. Exp Brain Res.2006;174:754-762.

39. Shamosh NA, DeYoung CG, Green AE, et al. Individual differencesin delay discounting relation to intelligence, working memory, andanterior prefrontal cortex. Psychol Sci. 2008;19:904-911.

40. Bickel WK, Yi R, Landes RD, Hill PF, Baxter C. Remember thefuture: working memory training decreases delay discountingamong stimulant addicts. Biol Psychiatry. 2011;69:260-265.

41. Houben K, Wiers RW, Jansen A. Getting a grip on drinkingbehavior: training working memory to reduce alcohol abuse.Psychol Sci. 2011;22:968-975.

42. Monterosso JR, Ainslie G, Xu JS, Cordova X, Domier CP,London ED. Frontoparietal cortical activity of methamphetamine-dependent and comparison subjects performing a delay dis-counting task. Hum Brain Mapp. 2007;28:383-393.

Y

www.jaacap.org 755


















































































































TABLE S1 Demographic Information, Substance Use, and Mental Health Diagnoses

ID lnk Age Gender

Days UsedTobacco 30 dBefore fMRI (n)

Drinks 30 dBefore fMRI (n)

Days UsedCannabis 30 dBefore fMRI (n) Diagnosis

1 �5.610 17 M 30 48 0 alcohol abuse, ODD2 �3.260 18 F 0 10 0 none3 �10.200 17 F 30 0 13 alcohol abuse, cannabis dependence,

ADHD, ODD, MD, GAD4 �7.630 16 M 0 0 0 alcohol abuse, ODD5 �7.800 16 F 4 6 0 alcohol dependence, MD, GAD6 �1.860 13 M 0 0 2 cannabis abuse, ADHD, ODD7 �6.030 14 M 9 0 3 cannabis abuse, ADHD, ODD, CD8 �4.350 17 M 0 6 7 cannabis abuse, ADHD, ODD, CD9 �2.680 13 M 0 0 8 cannabis abuse

10 �3.340 17 M 30 0 14 cannabis abuse11 �3.950 17 M 0 0 0 cannabis abuse12 �2.280 15 M 6 0 6 cannabis abuse13 �9.680 17 M 0 7 2 cannabis abuse14 �5.760 14 M 0 0 0 cannabis abuse, ADHD, ODD, CD,

MD, GAD15 �3.610 18 M 25 0 25 cannabis abuse16 �3.530 12 F 0 0 0 cannabis abuse, ADHD, ODD, MD17 �9.200 16 M 0 8 6 cannabis abuse, CD18 �4.530 15 M 30 0 1 cannabis abuse19 �2.870 15 M 29 26 17 cannabis dependence, ADHD, ODD,

CD, MD, GAD20 �4.290 15 F 0 0 0 cannabis abuse21 �3.330 14 M 0 0 5 cannabis abuse22 �5.980 14 M 12 0 4 cannabis abuse, GAD23 �7.980 18 F 0 0 0 none24 �8.920 18 M 30 119 4 alcohol dependence, cannabis abuse,

ADHD, ODD25 �3.600 16 M 0 8 6 alcohol abuse, cannabis dependence26 �3.260 13 M 0 1 0 ADHD27 �6.570 17 M 22 51 15 cannabis abuse28 �7.360 17 M 30 17 30 alcohol abuse, cannabis abuse, opiate

abuse, ADHD, ODD29 �4.660 14 M 7 5 3 alcohol abuse, cannabis abuse, ODD30 �4.980 17 M 15 2 1 alcohol abuse, cannabis abuse,

ADHD, ODD

Note: Tobacco use includes cigarettes and smokeless tobacco. One participant (ID 1) used smokeless tobacco on 30 days (d) and cigarettes on 8 days.No other participants reported smokeless tobacco use. One participant (ID 3) used synthetic cannabis on 3 days and marijuana on 10 days in the 30days before the scan. No other subjects reported synthetic cannabis use. ID 28 did not use opiates in the 30 days before the scan. ADHD ¼ attention-deficit/hyperactivity disorder; CD ¼ conduct disorder; F ¼ female; fMRI ¼ functional magnetic resonance imaging; GAD ¼ generalized anxietydisorder; M ¼ male; MD ¼ major depressive disorder; ODD ¼ oppositional defiant disorder.

STANGER et al.

JOURN

755.e1 www.jaacap.org


TABLE S2 Rate of Delay Discounting (lnk) and Substance Use in the 30 Days (d) Before the Functional MagneticResonance Imaging Scan by Age and Sex

lnk Tobacco (d) Alcohol (Drinks) Marijuana/Synthetic Cannabis (d)Mean � SD Mean, Median, Range Mean, Median, Range Mean, Median, Range

All (N ¼ 30) �5.30 � 2.36 10.3, 2.0, 0e30 10.5, 0.0, 0e119 5.7, 3.0, 0e30Ages 12e13 (n ¼ 4) �2.83 � 0.74 0.0, 0.0, 0e0 0.3, 0.0, 0e1 2.5, 1.0, 0e8Ages 14e15 (n ¼ 9) �4.41 � 1.37 10.3, 7.0, 0e30 3.4, 0.0, 0e26 4.3, 3.0, 0e17Ages 16e17 (n ¼ 13) �6.48 � 2.37 12.4, 4.0, 0e30 11.8, 6.0, 0e51 7.2, 6.0, 0e30Age 18 (n ¼ 4) �5.94 � 2.93 13.8, 12.5, 0e30 32.3, 5.0, 0e119 7.3, 2.0, 0e25Males (n ¼ 24) �5.08 � 2.23 11.5, 6.5, 0e30 12.4, 0.5, 0e119 6.6, 4.0, 0e30Females (n ¼ 6) �6.18 � 2.87 5.7, 0.0, 0e30 2.7, 0.0, 0e10 2.2, 0.0, 0e13

JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY

VOLUME 52 NUMBER 7 JULY 2013 www.jaacap.org 755.e2



TABLE S3 Significant Areas of Task Activation When Choosing a Smaller Amount “Now” Versus Control Choices

Voxels (n) t Statistic x y z BA Significant Area of Activation

194 9.45 1.5 13.5 50.5 6 dorsal anterior cingulate cortex217 8.12 22.5 �85.5 �12.5 18 right fusiform gyrus363 7.93 �13.5 �67.5 23.5 31 left precuneus

7.82 4.5 �70.5 41.5 7 right precuneus6.68 7.5 �52.5 8.5 30 posterior cingulate cortex

24 6.90 �4.5 �19.5 29.5 23 left middle cingulate cortex49 6.45 �34.5 49.5 20.5 10 left middle frontal gyrus59 6.30 25.5 55.5 14.5 10 right superior frontal gyrus

225 6.07 �10.5 �91.5 �6.5 17 left lingual gyrus18 5.96 �64.5 �16.5 8.5 42 left superior temporal gyrus17 5.56 40.5 �58.5 �33.5 right cerebellum18 4.90 �7.5 �31.5 68.5 6 left paracentral lobule42 4.52 1.5 �43.5 44.5 7 precuneus/cingulate cortex24 �8.68 31.5 �49.5 44.5 7 right superior parietal lobule

250 �8.00 �43.5 �70.5 5.5 38, 19 left middle temporal gyrus244 �7.78 43.5 �1.5 41.5 6 right middle frontal gyrus

�6.28 55.5 22.5 20.5 45 right inferior frontal gyrus�5.87 43.5 22.5 �12.5 47 right inferior frontal gyrus

86 �7.07 �22.5 �46.5 �6.5 37 left parahippocampal gyrus450 �6.55 58.5 �49.5 29.5 40 right supramarginal gyrus

�5.52 61.5 �40.5 �9.5 21 right middle temporal gyrus48 �6.36 25.5 �46.5 �6.5 37 right parahippocampal gyrus

196 �6.28 55.5 22.5 20.5 45 right inferior frontal gyrus�5.92 49.5 22.5 �0.5 47 right inferior frontal gyrus

45 �5.93 �61.5 �34.5 �12.5 21 left middle temporal gyrus35 �5.49 �52.5 7.5 23.5 44 left inferior frontal gyrus23 �5.47 52.5 4.5 �21.5 21 right middle temporal gyrus38 �4.94 4.5 55.5 41.5 8 right dorsomedial prefrontal cortex18 �4.86 �34.5 �13.5 �30.5 20 left fusiform gyrus19 �4.30 �61.5 �43.5 35.5 40 left supramarginal gyrus

Note: Areas are identified according to description for the areas of activation, Brodmann’s classification, and Talairach-Tourneaux coordinates. Activationmagnitude is indicated by t statistics. BA ¼ Brodmann area.


755.e3 www.jaacap.org VOLUME 52 NUMBER 7 JULY 2013

STANGER et al.


TABLE S4 Significant Areas of Task Activation When Choosing a Larger, Later Amount Versus Control Choices

Voxels (n) t Statistic x y z BA Significant Area of Activation

206 9.27 1.5 13.5 50.5 6 supplementary motor area32 dorsal anterior cingulate cortex

76 9.16 25.5 55.5 11.5 10 right superior frontal gyrus451 8.11 22.5 �94.5 2.5 17 right ventral cuneus

7.47 �37.5 �88.5 �3.5 18 left inferior occipital gyrus51 8.08 40.5 �55.5 �33.5 right cerebellum61 7.84 �31.5 49.5 23.5 10 left superior frontal gyrus

291 7.66 10.5 �58.5 17.5 30 right posterior cingulate cortex6.87 �7.5 �73.5 50.5 7 precuneus

18 5.98 31.5 25.5 44.5 8 right middle frontal gyrus26 5.66 13.5 22.5 11.5 right caudate24 4.86 �13.5 19.5 8.5 left caudate22 4.61 �43.5 �55.5 �33.5 left cerebellum20 4.54 �43.5 �79.5 29.5 39 left angular gyrus22 4.12 1.5 �40.5 38.5 31 precuneus/cingulate cortex81 �8.14 �61.5 �25.5 �12.5 21 left middle temporal gyrus

339 �7.48 49.5 31.5 5.5 45 right inferior frontal gyrus�6.77 40.5 19.5 �12.5 47 right inferior frontal gyrus�6.14 19.5 �1.5 �12.5 right parahippocampal gyrus/amygdala

53 �7.23 7.5 46.5 35.5 8 right medial frontal gyrus19 �7.20 31.5 �49.5 44.5 7 right superior parietal lobule64 �7.14 58.5 1.5 �21.5 21 right middle temporal gyrus

147 �6.85 7.5 �88.5 20.5 18 right dorsal cuneus28 �6.47 43.5 �1.5 41.5 6 right middle frontal gyrus55 �6.43 �25.5 �46.5 �6.5 19 left parahippocampal gyrus87 �6.36 �43.5 �70.5 5.5 19, 37 left middle temporal gyrus/inferior temporal gyrus

387 �6.30 55.5 �19.5 �6.5 21 right middle temporal gyrus�6.21 52.5 �55.5 23.5 39 right superior temporal gyrus�5.42 61.5 �52.5 �9.5 37 right inferior temporal gyrus

37 �5.88 �55.5 �49.5 2.5 left middle temporal gyrus18 �5.71 �13.5 �19.5 38.5 31 left middle cingulate cortex19 �5.61 �1.5 �1.5 38.5 24 left middle cingulate cortex64 �5.50 19.5 �52.5 �9.5 19 right parahippocampal gyrus59 �5.23 1.5 40.5 17.5 32 left medial frontal gyrus26 �4.75 �52.5 7.5 23.5 44 left inferior frontal gyrus20 �4.75 �22.5 �4.5 �12.5 left parahippocampal gyrus/amygdala41 �4.70 �46.5 34.5 �0.5 47 left inferior frontal gyrus26 �4.27 �37.5 22.5 �15.5 47 left inferior frontal gyrus (pars orbitalis)






TABLE S5 Areas of Task Activation When Delay Discounting Rate (lnk) Regresses to Choices of Smaller, Sooner andLarger, Later

Voxels (n) t Statistic x y z BA Significant Areas of Activation

Smaller/sooner vs. control trials45 4.16 1.5 �19.5 50.5 6 middle cingulate cortex

3.56 1.5 �13.5 38.5 32 dorsal anterior cingulate cortex36 4.15 46.5 �13.5 2.5 22, 13 right insula30 5.12 �46.5 �4.5 �6.5 22 left insula30 4.17 40.5 �55.5 11.5 22 right superior temporal gyrus26 4.99 4.5 �49.5 65.5 7 right precuneus26 4.24 �61.5 �28.5 20.5 40, 42 left postcentral gyrus/superior temporal gyrus18 4.89 28.5 �49.5 56.5 7 right superior parietal lobule

Larger/later vs. control trials46 5.60 �46.5 �4.5 �6.5 22 left insula25 5.01 �58.5 �31.5 20.5 42 left superior temporal gyrus25 4.36 �4.5 �16.5 50.5 6 middle cingulate cortex23 4.27 61.5 �46.5 2.5 22 right middle temporal gyrus21 4.63 �13.5 �49.5 50.5 7 left precuneus20 4.34 34.5 4.5 17.5 13 right insula



755.e5 www.jaacap.org VOLUME 52 NUMBER 7 JULY 2013

STANGER et al.


FIGURE S1 Additional network components. Note: Additional networks identified by group independent componentanalysis of functional magnetic resonance imaging time courses that were not significantly related to delaydiscounting rate.





neuroeconomics and adolescent substance abuse: individual ... · functional magnetic resonance...

Documents