Prefrontal–striatal pathway underlies cognitive
regulation of craving
Hedy Kobera,b,1, Peter Mende-Siedleckic, Ethan F. Krossd, Jochen Weberb, Walter Mischelb,1, Carl L. Hartb,e,f,
and Kevin N. Ochsnerb,1
aDivision of Substance Abuse, Department of Psychiatry, Yale University School of Medicine, New Haven, CT 06519;bDepartment of Psychology, Columbia
University, New York, NY 10027;cDepartment of Psychology, Princeton University, Princeton, NJ 08544;dDepartment of Psychology, University of Michigan,
Ann Arbor, MI 48109;eDivision on Substance Abuse, New York State Psychiatric Institute, New York, NY 10032; andfDepartment of Psychiatry, College of
Physicians and Surgeons, Columbia University, New York, NY 10032
Contributed by Walter Mischel, June 21, 2010 (sent for review March 1, 2010)
The ability to control craving for substances that offer immediate
at the root of substance use disorders and is critical for mental and
physical health. Despite its importance, the neural systems support-
ing this ability remain unclear. Here, we investigated this issue using
functional imaging to examine neural activity in cigarette smokers,
the most prevalent substance-dependent population in the United
States, as they used cognitive strategies to regulate craving for
craving was associated with (i) activity in regions previously associ-
ated with regulating emotion in particular and cognitive control in
general, including dorsomedial, dorsolateral, and ventrolateral pre-
frontal cortices, and (ii) decreased activity in regions previously as-
sociated with craving, including the ventral striatum, subgenual
cingulate, amygdala, and ventral tegmental area. Decreases in crav-
ing correlated with decreases in ventral striatum activity and
increases in dorsolateral prefrontal cortex activity, with ventral
striatal activity fully mediating the relationship between lateral pre-
frontal cortex and reported craving. These results provide insight
into the mechanisms that enable cognitive strategies to effectively
regulate craving, suggesting that it involves neural dynamics paral-
lel to those involved in regulating other emotions. In so doing, this
studying this ability across substance using populations and devel-
oping more effective treatments for substance use disorders.
drug craving|cigarette smokers|emotion regulation|functional MRI|
impulse to consume a desirable—but, in the long run, un-
healthy—substance is critical to both mental and physical health
(1). This ability is perhaps no more important than for individuals
with substance use disorders (SUDs), which are chronic relapsing
conditions (2, 3) with staggering social costs (4). The strong desire
touse drugs (known as drug craving) and failure tocurb this desire
are central to such disorders and are thought to be primary con-
tributors to drug use in general (5–8). Three types of evidence
directly link both craving and failure to regulate it to drug-taking
behavior. First, across substance-using populations, prospective
studies show that craving predicts relapse to drug taking following
abstinence (9–15). Second, cognitive–behavioral therapies (CBT)
include training in the cognitive regulation of craving, and are ef-
fective for treating SUDs (16). Third, across different forms of
is associated with reduced relapse over time (13, 17–20).
mechanisms underlying substance use by exposing drug users to
conditioned drug-associated cues such as photos, videos, or stories
increases craving and activity in regions previously associated with
emotion and the experience of drug effects, including the ventral
he ability to resist immediate gratification by controlling the
22), amygdala (21, 23), insula (23, 24), medial prefrontal cortex
(mPFC), orbitofrontal cortex (OFC; e.g., refs. 21–23), and anterior
cingulate cortex (ACC; e.g., refs. 21, 22, 24) including its subgenual
portion (sgACC; e.g., refs. 22, 25).
Although this research provides insight into the mechanisms by
which drug cravings are generated, the mechanisms by which crav-
to SUDs and their treatment. The finding that craving depends on
cognitive regulation of craving could be similar to the cognitive
regulation of emotion, a topic of increasing empirical interest. To
date, the majority of studies on the cognitive regulation of emotion
have examined negative emotions elicited by the presentation or
anticipation of aversive stimuli (reviewed in ref. 26). These studies
found that cognitivestrategies can be used to decrease, maintain,or
andactivity in brainsystems involved ingeneratingemotion,such as
implicated in cognitive control (28, 29), including the dorsolateral
prefrontal cortex (dlPFC), dorsomedial prefrontal cortex (dmPFC),
focused on the regulation of responses to positive images, food, or
monetary reward and have reported similar findings (30–33).
Taken together, these two bodies of literature suggest that the
effective regulation of craving might involve the use of systems im-
plicated in emotion regulation, such as lateral PFC, to modulate
VS. Recently, initial steps have been taken toward exploring this
hypothesis by showing videos depicting substance use to cigarette
smokers (34) or cocaine users (35) while asking them to “resist” or
“inhibit” craving, respectively. Although both studies observed ac-
these activations is difficult to interpret for three reasons. First,
neither study observed a significant reduction in craving. Second, in
both studies participants were allowed to self-select one or more
SUDs in clinical contexts. Third, neither study included a control
condition to determine whether regulation of craving for the abused
substance was more or less effective, or involved neural systems
different from regulation of craving for another appetitive stimulus
for which there were no substance use issues. Such a condition is
important if one wants to draw inferences about the specificity or
generality of regulation-related changes in craving.
Author contributions: H.K., E.F.K., W.M., C.L.H., and K.N.O. designed research; H.K. and
P.M.-S. performed research; H.K. and J.W. analyzed data; and H.K. and K.N.O. wrote the
The authors declare no conflict of interest.
psych.columbia.edu, or firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
| August 17, 2010
| vol. 107
| no. 33
the mechanisms by which clinically relevant cognitive strategies
effectively reduce craving remain unclear. First, would the effec-
tive down-regulation of craving using treatment-relevant strategies
be accompanied by increased activity in prefrontal control systems,
related regions? Second, is the regulation of craving for an abused
substance different from an appetitive control stimulus? Third, are
individual differences in the ability to regulate craving related to
increased activity in control-related systems and decreased activity
in craving-related systems? If the answers to the first and third
questions are yes, then a fourth question naturally arises: Is the
hypothesized inverse relationship between PFC control-related re-
gions and self-reported craving mediated by changes in activity in
craving-related regions (30, 36)? Or, put another way, does the PFC
act on the striatum to modulate craving? Although such a relation-
ship would be predicted by theories of both emotion regulation (36)
and substance use (8, 37), this relationship has not been demon-
To address these questions, we collected functional MRI (fMRI)
data while tobacco cigarette smoking participants completed a task
(38). We selected cigarette smokers because cigarette smoking is
the most lethal form of substance use: in the United States, more
deaths are caused each year by tobacco use than by all forms of
illegal drug use, alcohol use, motor vehicle injuries, suicides, and
murders combined (39, 40). In addition, the relationship between
craving and drug taking in this group is particularly robust (re-
viewed in ref. 38).
During the ROC task, participants viewed cigarette- and food-
associated cues on two types of trials. On baseline trials, designed
to isolate the neural correlates of cravings elicited by smoking and
food cues, the instruction “NOW” directed participants to think
about the immediate feelings associated with smoking or eating.
Wehaveshownpreviouslythat smokersreport robustcue-induced
cravings in this condition (38). On regulation trials modeled after
treatments for SUDs (1, 38, 41), the instruction “LATER” di-
rected participants to think about the long-term consequences
associatedwithsmoking and eating high-fat foods. We haveshown
previously that smokers are able to decrease their craving for cig-
arettes and food using this strategy (38). After viewing each ciga-
rette or food cue, and thinking about it in the instructed manner,
participants rated how much they craved the type of stimulus seen
on that trial (trial schematic in Fig. S1).
Behavioral Ratings. Upon arrival, participants reported their crav-
ing for food and cigarettes. These were not significantly correlated
with any of the subsequent in-scanner ratings (detailed in SI Text).
Analyses of in-scanner ratings revealed significant main effects of
both instruction and cue type (Fig. 1): overall, significantly lower
cravingwasreportedon LATERcomparedwithNOW instruction
trials [F(1,20) = 44.16, P < 0.001] and on food compared with
cigarette cue trials [F(1,20) = 11.10, P < 0.01]. These data dem-
onstrate that, in general, regulation with cognitive strategies ef-
fectively reduced craving and that cigarette cues elicited stronger
cravings than food cues, as would be expected for smokers.
Next, we determined how regulation affected craving for each
interaction [F(1,20) = 7.02, P < 0.05] reflecting a greater modu-
lation of craving on NOW vs. LATER trials for cigarettes com-
pared with food cues. Inspection of Fig. 1 shows, however, that
cigarette and food cues differed in baseline craving reported on
NOW trials [t(20) = 4.52, P < 0.001]. To correct for this baseline
difference and to ensure that craving decreases did not appear
larger for cigarette cues simply because reported craving was
overall greater, we calculated regulation-related decreases in
craving as the percentage (rather than the absolute) decrease in
craving on LATER relative to NOW trials. Comparison of these
values showed that the percentage drop in craving due to regula-
tion was only marginally greater for cigarettes than food [33.52%
for cigarettes, 29.76% for food; t(20) = 1.96, P < 0.07]. This result
is important because it suggests that whereas smokers experience
greater cravings for cigarettes than other appetitive cues, when
instructed to regulate their cravings using a cognitive strategy,
for cigarettes as they are for other appetitive stimuli. Finally, none
of these effects differed between functional runs (see SI Text for
Imaging Results. Main effects of regulation. Totestourfirstprediction
—that effective down-regulation of craving increases activity in
prefrontal control systems and decreases activity in craving-related
active when participants regulated craving on LATER trials com-
pared to NOW trials. As shown in Fig. 2 and Table S1, our first
prediction was confirmed: we observed increased activation in re-
and with regulating negative emotion in particular (26), including
dmPFC, dlPFC, and vlPFC (Fig. 2A). This was accompanied by
decreased activation in regions previously associated with emotion
in general (42) and with craving in particular (6), including VS,
not discuss the main effect of cue type (food vs. cigarettes) in this
paper because it is not relevant to the theoretical issues discussed.
The results of this analysis do not change the interpretation of the
data presented here in any way, and will be reported elsewhere.
Whole-brain interaction. To test the second prediction, and to de-
termine whether the effects of craving regulation varied by cue
types,weperformeda whole-braininteraction analysis (CN vs.CL)
interest, and we limited our analysis to identifying regions showing
one of them. First, we were interested in regions showing larger
increases during regulation (i.e., a larger LATER > NOW effect)
indicative of greater demands for control. Only one region showed
such aneffect: Activation in dmPFC was stronger forLATER than
interested in regions showing larger regulation-related drops in
activity(i.e.,alargerNOW>LATEReffect)for one type of stimu-
food) and instruction type (NOW/LATER). Overall, reported craving was sig-
nificantly decreased on LATER vs. NOW instruction trials, across both stimulus
types, and was significantly greater for cigarette compared with food cues,
across both instructions. Error bars represent ±SEs. When expressed as percent
drop relative to craving reported on NOW trials, regulation-related drops in
craving on LATER trials were only marginally different between cigarettes
and food (P < 0.07).
| www.pnas.org/cgi/doi/10.1073/pnas.1007779107Kober et al.
an effect: For food compared with cigarette stimuli, activation in
thepostcentral gyrus was greater onNOWcomparedwithLATER
trials. As we had no a priori expectations about dmPFC being
more active when regulating responses to food or the precentral
gyrus showing greater regulation-related modulation for food, we
simply report these findings and offer no post hoc interpretation
ofthem. Whattheseresultsimportantly show, however, isthatthe
main effects of strategy for control and craving-related regions
(i.e., activations in control-related regions and modulations in
craving-related regions) are not qualified by an interaction with
stimulus type (Table S2).
Correlations with individual differences in regulatory ability. To test our
third prediction—that individual differences in the ability to de-
crease craving would relate to increased activity in control-related
systems and decreased activity in craving-related systems—we
computed a robust whole-brain correlation of activity in the NOW
each participant on LATER trials as compared with NOW trials.
(43), and because the proportional decreases in craving were
similar for cigarette and food cues, we maximized power for this
analysis by first computing this correlation collapsing across cue
types. We found that decreases in self-reported craving correlated
with decreases in VS activity (R2= 0.47) and with increases in left
dlPFC activity (R2= 0.56; Table S3). To then confirm that these
results did not differ across cue types, we used the Fisher z test to
determine whether the observed correlations significantly differed
as a function of cue type. In both the VS and the dlPFC, no sig-
nificant differences were observed (P > 0.1).
Test of prefrontal–striatal mediation of craving. Finally, we addressed
reported above are consistent with either of two possible kinds of
functional relationships between regulation-related drops in craving
and prefrontal or VS activity. The first possibility is that cognitive
strategies exert parallel effects on each brain region, and that pre-
frontal cortex and the VS are independently related to changes in
craving. The second possibility is that the inverse relationship be-
tween dlPFC activity and craving is mediated by changes in VS ac-
tivity, as predict based on prior findings of frontal–subcortical
pathways mediating regulation of aversive emotion (36, 44). Given
that we wished to test a strong a priori hypothesis that prefrontal
cortex would act on the striatum to influence craving, a mediation
analysis, which specifically tests for such a relationship, was deemed
to be most appropriate (relative to connectivity analyses that only
examine covariances between two regions, for example). Therefore,
to test this prediction, we performed a formal mediation analysis,
collapsing across cues types, given that the analysis reported above
found that correlations of brain activity and craving did not vary as
a function of cue type. This analysis explicitly tested whether the re-
VS (denoted M in Fig. 3). We found that the VS fully mediated the
relationship between dlPFC and reported craving (Fig. 3).
Taken together, these findings provide direct support for the hy-
pothesis that clinically relevant cognitive strategies can effec-
tively reduce craving via the effects of prefrontal control systems
on the VS and other systems implicated in reward learning and
emotion. As such, these findings have implications for our basic
understanding of affect-regulatory mechanisms and of substance
use disorders, as well as for studies that translate this work to
With respect to basic mechanisms, the present study builds on
prior work on delay of gratification in children and temporal (or
adults prefer immediate smaller rewards compared with larger
delayed rewards (45, 46), but that cognitive strategies can influ-
ence and even reverse this preference (47). Whereas such studies
for immediate as opposed to delayed rewards, the present work
focuses on how we can actively take control over the affective
impulses that motivate these choices.
have shown that the same prefrontal systems support the effective
regulation of cravings elicited by two different types of appetitive
suggests that the vlPFC and dmPFC regions (that were engaged
during regulation of craving for both cue types), as well as the
dlPFC–striatal pathway (which mediated successful regulation),
play general roles in the regulation of different kinds of appetitive
desires. This fits with prior studies on the regulation of negative
emotion (26) and positive emotion (30, 32) showing activation of
these prefrontal systems, and suggests that common prefrontal–
subcortical dynamics support the use of cognition to regulate
responses to various kinds of affective cues, including drug cues.
Importantly, this mechanism is consistent with a prevailing hy-
pothesis in the substance use literature— that a prefrontal–striatal
pathway is involved in the control of craving in general, and com-
pulsive drug use in particular (8, 37, 48). However, the behavioral
results indicate that although smokers showed greater craving for
cigarettes, they were at least as effective at downregulating these
cravings as they were at downregulating their cravings for another
smoking does not result from a general inability to control appe-
titive impulses but, rather, from the failure to effectively deploy
regulatory strategies that could be effective if one were motivated
showing that the same prefrontal–striatal pathway mediated the
effective regulation of cravings elicited by cigarette and food cues.
With respect to clinical contexts, the present study has at least
three kinds of translational implications. First, it builds on initial
attempts to study the regulation of responses to drug-related
craving.(A) Medial(Left) and lateral (Right) views ofbrain regionsthatshowed
greater activation in the LATER vs. NOW trials, when participants used a cog-
nitive strategy to reduce their craving. Highlighted activations are shown in
regions previously implicated in regulation of aversive emotion. (B) Medial
reported in studies of cue-induced craving or emotion. corr, Corrected for
multiple comparisons; uncorr, uncorrected for multiple comparisons.
Regions active during or modulated by the cognitive regulation of
Kober et al. PNAS
| August 17, 2010
| vol. 107
| no. 33
stimuli (34, 35) that did not observe significant reductions in
craving. Here, we showed that craving could be curbed effectively
via a single, clinically relevant strategy that involved cognitively
that highlighted its potential longer-term deleterious effects on
one’s health. In general, such cognitive strategies are known to be
in children and adults (1, 26), and for treating psychiatric (50) as
well as substance use disorders (18, 51).
Second, the results of this study identify neural mechanisms by
which cognitive strategies reduce craving, and in turn, a potential
mechanism by which cognitive therapies can successfully decrease
drug craving and use over time. In this regard, the present findings
have great practical importance for understanding nicotine de-
pendence in particular, which has been identified as the leading
preventable cause of disease and death in the United States,
contributing to ~430,000 deaths every year (52). However, future
work is needed to replicate and generalize these results across
larger populations of smokers and individuals with other SUDs.
findings as a model, and the method used here as a tool, for testing
the hypothesis that impaired prefrontal control over subcortical
been suggested previously (8, 37, 53). For example, for a given
population, this method could be adapted for comparing and test-
ing the efficacy of specific cognitive strategies for regulating crav-
ings induced either by cues or by negative emotions that also may
beusedtomeasurebaseline differences in regulatoryabilityand to
track the effects of treatment on their function. In either case, the
ultimate goal is to develop more effective treatments for SUDs.
women and 12 men, aged 18–45 y, mean 26.8. SD 8.94). Participants were
recruited via posters and electronic bulletin board ads from the general New
York City population. Participants were considered cigarette smokers if they
reported smoking more than 10 cigarettes/d, 7 d/wk, for at least 1 y (mean
cigarettes/wk 110, SD 35.93, range 70–175). Participants reported starting
to smoke regularly between 14 and 28 y of age (mean age of onset 16.95,
SD 3.27), and smoked between 1 and 30 y (mean years 9.30, SD 8.63). All par-
index (kg/m2) of more than 28, or any of the following conditions: dependence
onsubstances (other thannicotine),neurologicalorpsychiatric disorders, use of
prescription (psychiatric and/or nonpsychiatric) medication that could affect
function, cardiovascular disease or diabetes, head trauma with loss of con-
sciousness for more than 30 min, pregnancy, claustrophobia, or any implants
contraindicated in MRI. Participants also were excluded if they reported any
food allergies or aversion to any of the pictured food stimuli presented in this
experiment. Before participation, all participants gave informed consent in ac-
cordance with the Columbia University Institutional Review Board. After com-
pletion of the study, they were paid $70 for their participation.
ROC Task. In this single-session study, participants were exposed to photo-
graphic images of cigarettes and food that were previously shown to induce
craving for these substances (38). On each trial, participants were directed to
think about these stimuli in one of two ways. NOW cues instructed partic-
ipants to consider the immediate consequence of consuming the pictured
substance. LATER cues instructed participants to consider the long-term
consequences of repeatedly consuming the substance; this strategy is often
used in cognitive–behavioral treatment for SUDs. Our prior work indicates
that the latter strategy decreases reported craving for both food and ciga-
rettes (38). There were 100 trials (50 food, 50 cigarette) with intertrial
intervals jittered around 4 s and a pseudorandomized presentation order.
craving, whereby VS is a complete mediator of the dlPFC–craving relationship. Path coefficients are shown next to arrows indicating each link in the analysis,
with SEs in parentheses. Path a refers to the path from dlPFC to VS; path b refers to the direct link between VS and craving; and path c’ refers to the total
association between dlPFC and craving, without the mediator VS. *P < 0.05, **P < 0.01, ***P < 0.001.
Mediation model for the association between dorsolateral prefrontal cortex (dlPFC), ventral striatum (VS), and regulation-related decreases in
| www.pnas.org/cgi/doi/10.1073/pnas.1007779107Kober et al.
Strategy training. Before scanning, participants were instructed in and trained
touseeach strategy.Participants then experienced eight sample trials, during
which they were asked to practice using the strategies while looking at
photographs of cigarettes and food that were not used during the scanning
session. During training, participants were also trained to use a Likert-type
five-point scale to indicate the extent of their craving after each stimulus
presentation, as has been used previously (24, 38, 54, 55). The scanning
session began once participants indicated to the experimenter that they
could use these strategies while looking at the stimuli, and that the direc-
tions were understood.
Scanning session. To minimize effects of circadian rhythms on cue reactivity,
scans were all performed in the midafternoon (3:00–5:00 PM). In addition,
participants were instructed to abstain from smoking or eating for 2 h before
the scanning session, and confirmed their abstinence status upon arrival via
a carbon monoxide reading (Vitalograph Breath CO monitor 29700; Vitalo-
graph, Inc). During the scanning session, participants completed five func-
tional runs consisting of 20 trials each. As Fig. S1 illustrates, each trial began
with a fixation cross that was displayed for an intertrial interval jittered
around 4 s. This interval was followed by a 2-s instructional cue (NOW or
LATER) followed by a 6-s presentation of the stimulus (a picture of food or
cigarettes). Following another delay period jittered around 3 s, participants
next indicated how much they wanted to consume the substance at that
moment using a rating scale of 1 (not at all) to 5 (very much) that appeared
onscreen for up to 3 s or until the participants indicated a response. Exposure
to study stimuli and the order of the instructional cues and photos was
counterbalanced across participants. A total of 100 trials lasting ≈18 s each
were completed. Finally, participants completed individual difference meas-
ures, rerated all of the stimuli, were paid for their time, and were debriefed.
Data Acquisition and Analysis. Stimulus presentation was controlled using E-
Prime software (PST Inc.). An LCD projector displayed stimuli on a back-
projection screen mounted in the scanner suite. Responses were made with
the right hand on a five-finger button-response unit (Avotec Inc. and Res-
onance Technologies). Response data were subjected to a 2 (Cue: Food vs.
Cigarette cues) × 2 (Strategy: NOW vs. LATER) repeated-measures ANOVA,
with an alpha level of P < 0.05. The relative drop in self-reported craving
between the NOW and LATER conditions was calculated for each subject and
used to correlate with brain activation in subsequent analyses.
fMRI Data Acquisition and Analysis. Participants were scanned in a GE Signa
Twin Speed Excite HD scanner (GE Medical systems). Functional images were
acquired with a T2*-weighed EPI BOLD sequence with a TR of 2,000 ms, TE of
34 ms, flip angle of 90°, 64 × 64 in-plane matrix, field of view of 22.4 cm, and
28 4-mm slices in an ascending-interleaved order. High resolution SPGR
structural imageswere alsoacquiredwithaTRof9,700ms,TEof2,300ms,flip
angle of 20°, 256 × 256 in-plane matrix, field of view 24 cm, and ∼182 1-mm
slices covering the entire brain of the subject (range: 176–192).
Following prior protocols established in our laboratory (36), functional
images were subjected to preprocessing using SPM5 (Wellcome Department
of Cognitive Neurology), warped to the Montreal Neurological Institute
template and smoothed using a 6-mm kernel FWHM. After preprocessing,
we extracted the global time course from white matter in each volume, and
removed “spikes” (e.g., volumes in which the global time course was above
or below 3 SD from the mean, after accounting for low-frequency drifts).
The data were then subjected to first-level statistical analysis with a standard
GLM model. We used a boxcar regressor for the instruction cue and picture
presentation period. The instruction cues were not modeled separately, as
they were highly correlated to the picture period because of their shortness
and regularity (2 s, always right before the stimulus). To address a possible
question about the soundness of this methodology, we modeled the
instructions separately, and found that the data were not substantially dif-
ferent. The rating period was modeled with an event-related regressor, but
is not of interest here. We also used motion parameters and high-pass filter
parameters as additional regressors of no interest. We then performed
second-level random effects analysis using robust regression to localize
regions of activation across subjects. The robust regression approach itera-
tively reweights the regression matrix using a bisquare weighting function
(robustfit.m in the statistics toolbox of MATLAB software, Mathworks). This
approach allows a limited number of strong outliers to be down-weighted
automatically, thus showing a more accurate outcome in cases in which the
sampled measure (DV) cannot be supposed to have a normally distributed
error. Such robust procedures are therefore particularly resistant to outliers
(56) and have been used previously with fMRI data (36, 57). Similarly, whole-
brain robust correlations were computed to assess the relationship between
self-reported craving and brain activation.
To correct for multiple comparisons we used AlphaSim, a Monte Carlo
takes into account the voxel-wise and cluster-volume thresholds to establish
height and cluster level were considered to be significantly activated or
deactivated in the whole-brain analysis. Several regions that were previously
reported in studies of cue-induced drug craving, the VS and amygdala, were
selected as a priori regions of interest and were considered significant at an
uncorrected whole-brain threshold of P < 0.005. In addition, as the IFG and
dlPFC are often reported in studies of emotion regulation, they were also
selected a priori and considered significant at this uncorrected threshold.
Regions in which activity in the NOW vs. LATER contrast correlated with
drop in self-reported craving (e.g., “regulation success”) were subjected to
mediation analyses using the robust regression weights to exclude outliers.
This mediation analysis used the standard three-variable path model (58).
Mediation analyses test whether the relationship between any two variables
(activation in dlPFC and regulation success) can be explained by the values
from a third variable (activation in VS). If VS is a true mediator of the dlPFC-
regulation success relationship, then this relationship will become in-
significant when VS is controlled for in the model. According to standard
convention, “a” refers to the dlPFC-VS effect, “b” refers to the VS-regulation
success effect, and “c” refers to the direct dlPFC-regulation success effect,
controlling for the mediator VS. The product “a*b” tests the significance of
the direct mediator. As is customary, we used a bootstrapping test for the
statistical significance of the product “a*b” (36, 42, 59, 60).
ACKNOWLEDGMENTS. We thank the participants in this study, Michaela
Bamdadfor help in recruiting, Stephen Dashnaw for MRI operations, Tor Wager
for use of mediation software, and members of the SCAN Unit for helpful
commentary. This work was supportedbya National Science Foundation Gradu-
ateResearchFellowship (toH.K.),byNationalInstituteon Drug AbuseGrant R01
DA022541 (to K.N.O.), and by National Institute of Mental Health Grant R01
MH076137 (to K.N.O.).
1. Mischel W, et al. “Willpower” over the life span: Decomposing self-regulation. Soc
Cogn Affect Neurosci, in press.
2. McLellan AT, Lewis DC, O’Brien CP, Kleber HD (2000) Drug dependence, a chronic
medical illness: Implications for treatment, insurance, and outcomes evaluation.
3. Sinha R (2009) Modeling stress and drug craving in the laboratory: Implications for
addiction treatment development. Addict Biol 14:84–98.
4. National Institutes ofHealth(2008) Understanding DrugAbuseandAddiction(National
Institutes of Health, National Institute on Drug Abuse, Bethesda, MD): Info Facts, 1–4.
5. O’Brien CP, Childress AR, Ehrman R, Robbins SJ (1998) Conditioning factors in drug
abuse: Can they explain compulsion? J Psychopharmacol 12:15–22.
6. Sinha R, Li CSR (2007) Imaging stress- and cue-induced drug and alcohol craving:
Association with relapse and clinical implications. Drug Alcohol Rev 26:25–31.
7. Volkow ND, et al. (2006) Cocaine cues and dopamine in dorsal striatum: Mechanism of
craving in cocaine addiction. J Neurosci 26:6583–6588.
8. Volkow ND, Fowler JS, Wang GJ (2003) The addicted human brain: Insights from
imaging studies. J Clin Invest 111:1444–1451.
9. Crits-Christoph P, et al. (2007) Predictors of sustained abstinence during psychosocial
treatments for cocaine dependence. Psychother Res 17:240–252.
10. Doherty K, Kinnunen T, Militello FS, Garvey AJ (1995) Urges to smoke during the first
11. Galloway GP, et al. (2008) How long does craving predict use of methamphetamine?
Assessment of use one to seven weeks after the assessment of craving. Substace Abuse
Res Treat 1:63–79.
and in the field: Predictors of craving in treated alcoholics. Addiction 95:889–900.
13. O’Connell K, Schwartz J, Gerkovich M, Bott M, Shiffman S (2004) Playful and
rebellious states vs. negative affect in explaining the occurrence of temptations and
lapses during smoking cessation. Nicotine Tob Res 6:661–674.
14. Paliwal P, Hyman SM, Sinha R (2008) Craving predicts time to cocaine relapse: Further
validation of the Now and Brief versions of the cocaine craving questionnaire. Drug
Alcohol Depend 93:252–259.
15. Swan GE, Ward MM, Jack LM (1996) Abstinence effects as predictors of 28-day relapse
in smokers. Addict Behav 21:481–490.
16. Dutra L, et al. (2008) A meta-analytic review of psychosocial interventions for
substance use disorders. Am J Psychiatry 165:179–187.
on relapse crisis outcomes after smoking cessation. J Consult Clin Psychol 57:443–449.
18. O’Connell KA, Hosein VL, Schwartz JE, Leibowitz RQ (2007) How does coping help
people resist lapses during smoking cessation? Health Psychol 26:77–84.
19. Shiffman S, Paty JA, Gnys M, Kassel JA, Hickcox M (1996) First lapses to smoking:
Within-subjects analysis of real-time reports. J Consult Clin Psychol 64:366–379.
Kober et al.PNAS
| August 17, 2010
| vol. 107
| no. 33
20. Carroll KM, Nich C, Frankforter TL, Bisighini RM (1999) Do patients change in the way
we intend? Treatment-specific skill acquisition in cocaine-dependent patients using
the Cocaine Risk Response Test. Psychol Assess 11:77–85.
21. Due DL, Huettel SA, Hall WG, Rubin DC (2002) Activation in mesolimbic and
visuospatial neural circuits elicited by smoking cues: Evidence from functional
magnetic resonance imaging. Am J Psychiatry 159:954–960.
22. David SP, et al. (2005) Ventral striatum/nucleus accumbens activation to smoking-
related pictorial cues in smokers and nonsmokers: A functional magnetic resonance
imaging study. Biol Psychiatry 58:488–494.
23. Franklin TR, et al. (2007) Limbic activation to cigarette smoking cues independent
of nicotine withdrawal: A perfusion fMRI study. Neuropsychopharmacology 32:
24. Wilson SJ, Sayette MA, Delgado MR, Fiez JA (2005) Instructed smoking expectancy
modulates cue-elicited neural activity: A preliminary study. Nicotine Tob Res 7:
25. Brody AL, et al. (2002) Brain metabolic changes during cigarette craving. Arch Gen
26. Ochsner KN, Gross JJ (2008) Cognitive emotion regulation: Insights from social
cognitive and affective neuroscience. Curr Dir Psychol Sci 17:153–158.
27. Phillips ML, Ladouceur CD, Drevets WC (2008) A neural model of voluntary and
automatic emotion regulation: Implications for understanding the pathophysiology
and neurodevelopment of bipolar disorder. Mol Psychiatry 13:829, 833–857.
28. Nee DE, Wager TD, Jonides J (2007) Interference resolution: Insights from a meta-
analysis of neuroimaging tasks. Cogn Affect Behav Neurosci 7:1–17.
29. Wager TD, Jonides J, Reading S (2004) Neuroimaging studies of shifting attention: A
meta-analysis. Neuroimage 22:1679–1693.
30. Delgado MR, Gillis MM, Phelps EA (2008) Regulating the expectation of reward via
cognitive strategies. Nat Neurosci 11:880–881.
31. Hare TA, Camerer CF, Rangel A (2009) Self-control in decision-making involves
modulation of the vmPFC valuation system. Science 324:646–648.
32. Kim SH, Hamann S (2007) Neural correlates of positive and negative emotion
regulation. J Cogn Neurosci 19:776–798.
33. Wang GJ, et al. (2009) Evidence of gender differences in the ability to inhibit brain
activation elicited by food stimulation. Proc Natl Acad Sci USA 106:1249–1254.
34. Brody AL, et al. (2007) Neural substrates of resisting craving during cigarette cue
exposure. Biol Psychiatry 62:642–651.
35. Volkow ND, et al. (2010) Cognitive control of drug craving inhibits brain reward
regions in cocaine abusers. Neuroimage 49:2536–2543.
36. Wager TD, Davidson ML, Hughes BL, Lindquist MA, Ochsner KN (2008) Prefrontal-
subcortical pathways mediating successful emotion regulation. Neuron 59:1037–1050.
37. Everitt BJ, Robbins TW (2005) Neural systems of reinforcement for drug addiction:
From actions to habits to compulsion. Nat Neurosci 8:1481–1489.
38. Kober H, Kross EF, Mischel W, Hart CL, Ochsner KN (2010) Regulation of craving by
cognitive strategies in cigarette smokers. Drug Alcohol Depend 106:52–55.
39. Centers for Disease Control and Prevention (CDC) (2005) Annual smoking-attributable
mortality, years of potential life lost, and productivity losses—United States, 1997–
2001. MMWR Morb Mortal Wkly Rep 54:625–628.
40. McGinnis JM, Foege WH (1993) Actual causes of death in the United States. JAMA
41. Carroll K (1998) Therapy Manuals for Drug Addiction. A Cognitive-Behavioral
Approach: Treating Cocaine Addiction (National Institute of Drug Abuse, Rockville,
42. Kober H, et al. (2008) Functional grouping and cortical-subcortical interactions in
emotion: A meta-analysis of neuroimaging studies. Neuroimage 42:998–1031.
43. Lieberman M, Berkman E, Wager T (2009) Correlations in social neuroscience aren’t
voodoo. Perspect Psychol Sci 4:299–307.
44. Delgado MR, Nearing KI, Ledoux JE, Phelps EA (2008) Neural circuitry underlying the
regulation of conditioned fear and its relation to extinction. Neuron 59:829–838.
45. Bickel WK, Marsch LA (2001) Toward a behavioral economic understanding of drug
dependence: Delay discounting processes. Addiction 96:73–86.
46. McClure SM, Laibson DI, Loewenstein G, Cohen JD (2004) Separate neural systems
value immediate and delayed monetary rewards. Science 306:503–507.
47. Mischel W, Baker N (1975) Cognitive appraisals and transformations in delay
behavior. J Pers Soc Psychol 31:254–261.
48. Koob GF, Le Moal M (2001) Drug addiction, dysregulation of reward, and allostasis.
49. Gross JJ, John OP (2003) Individual differences in two emotion regulation processes:
Implications for affect, relationships, and well-being. J Pers Soc Psychol 85:348–362.
50. Butler AC, Chapman JE, Forman EM, Beck AT (2006) The empirical status of cognitive-
behavioral therapy: A review of meta-analyses. Clin Psychol Rev 26:17–31.
51. Carroll KM(2008) Cognitive-Behavioral
Publishing Textbook of Substance Abuse Treatment, eds Galanter M, Kleber HD
(American Psychiatric Publishing, Arlington, VA), 4th Ed, pp 349–360.
52. Centers for Disease Control and Prevention (CDC) (2007) Cigarette smoking among
adults—United States, 2006. MMWR Morb Mortal Wkly Rep 56:1157–1161.
53. Garavan H, Stout JC (2005) Neurocognitive insights into substance abuse. Trends Cogn
54. Lee JH, Lim Y, Wiederhold BK, Graham SJ (2005) A functional magnetic resonance
imaging (FMRI) study of cue-induced smoking craving in virtual environments. Appl
Psychophysiol Biofeedback 30:195–204.
55. McClernon FJ, Kozink RV, Lutz AM, Rose JE (2009) 24-h smoking abstinence
potentiates fMRI-BOLD activation to smoking cues in cerebral cortex and dorsal
striatum. Psychopharmacology (Berl) 204:25–35.
56. Wager TD, Keller MC, Lacey SC, Jonides J (2005) Increased sensitivity in neuroimaging
analyses using robust regression. Neuroimage 26:99–113.
57. Tom SM, Fox CR, Trepel C, Poldrack RA (2007) The neural basis of loss aversion in
decision-making under risk. Science 315:515–518.
58. Baron RM, Kenny DA (1986) The moderator-mediator variable distinction in social
psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc
59. Efron B, Tibshirani RJ (1993) An Introduction to the Bootstrap (Chapman & Hall, New
60. Shrout PE, Bolger N (2002) Mediation in experimental and nonexperimental studies:
New procedures and recommendations. Psychol Methods 7:422–445.
| www.pnas.org/cgi/doi/10.1073/pnas.1007779107Kober et al.