Behavioral and neural correlates of delay of
gratification 40 years later
B. J. Caseya,1, Leah H. Somervillea, Ian H. Gotlibb, Ozlem Aydukc, Nicholas T. Franklina, Mary K. Askrend, John Jonidesd,
Marc G. Bermand, Nicole L. Wilsone, Theresa Teslovicha, Gary Gloverf, Vivian Zayasg,
Walter Mischelh,1, and Yuichi Shodae,1
aSackler Institute for Developmental Psychobiology, Weill Cornell Medical College, New York, NY 10065;bDepartment of Psychology, Stanford University,
Stanford, CA 94305;cDepartment of Psychology, University of California, Berkeley, CA 94720;dDepartment of Psychology, University of Michigan, Ann Arbor,
MI 48109;eDepartment of Psychology, University of Washington, Seattle, WA 98195;fLucas Imaging Center, Department of Radiology, Stanford University
School of Medicine, Stanford, CA 94305;gDepartment of Psychology, Cornell University, Ithaca, NY 14853; andhDepartment of Psychology, Columbia
University, New York, NY 10027
Edited* by Michael Posner, University of Oregon, Eugene, OR, and approved July 26, 2011 (received for review May 27, 2011)
We examined the neural basis of self-regulation in individuals
from a cohort of preschoolers who performed the delay-of-
gratification task 4 decades ago. Nearly 60 individuals, now in
their mid-forties, were tested on “hot” and “cool” versions of a go/
nogo task to assess whether delay of gratification in childhood
predicts impulse control abilities and sensitivity to alluring cues
(happy faces). Individuals who were less able to delay gratification
in preschool and consistently showed low self-control abilities in
their twenties and thirties performed more poorly than did high
delayers when having to suppress a response to a happy face but
not to a neutral or fearful face. This finding suggests that sensi-
tivity to environmental hot cues plays a significant role in individ-
uals’ ability to suppress actions toward such stimuli. A subset of
these participants (n = 26) underwent functional imaging for the
first time to test for biased recruitment of frontostriatal circuitry
when required to suppress responses to alluring cues. Whereas the
prefrontal cortex differentiated between nogo and go trials to
a greater extent in high delayers, the ventral striatum showed
exaggerated recruitment in low delayers. Thus, resistance to temp-
tation as measured originally by the delay-of-gratification task is
a relatively stable individual difference that predicts reliable biases
in frontostriatal circuitries that integrate motivational and control
reward|behavioral suppression|functional MRI|inferior frontal gyrus|
success. Developmentally, this ability has been assessed by
measuring how long a young child can resist an immediate re-
ward (e.g., a cookie) in favor of a larger, later reward (e.g., two
cookies) (1). Even as adults we vary in our ability to resist
temptations. Alluring situations can diminish our control (2–4);
what serves as an alluring situation that requires a capacity to
control our impulses, however, changes as a function of age (e.g.,
from cookies to social acceptance). In the present study we ex-
amined the extent to which individual differences in delay of
gratification assessed when participants were in preschool and in
their 20s and 30s predict control over impulses and sensitivity to
social cues at the behavioral and neural level when the partic-
ipants were in their 40s.
Delay of gratification depends importantly on cognitive con-
trol (5). Cognitive control refers to the ability to suppress com-
peting inappropriate thoughts or actions in favor of appropriate
ones (6–11). Previously, we have shown that performance on the
delay-of-gratification task in childhood predicts the efficiency
with which the same individuals perform a cognitive control task
(the go/nogo task) as adolescents and young adults (5). Indi-
viduals who as preschoolers directed their attention toward re-
warding aspects of the classic delay-of-gratification situation,
such as focusing on the cookies (high-temptation-focus group),
he ability to resist temptation in favor of long-term goals is an
essential component of individual, societal, and economical
had more difficulty suppressing inappropriate actions than did
their low-temptation-focus counterparts, especially for the most
difficult trials. Difficulty was manipulated by increasing the num-
ber of “go” trials preceding a “nogo” trial, thus making the “go”
response more salient and automated. Differences between the
high- and low-temptation-focus groups increased as the number of
preceding “go” trials increased, with the high-temptation-focus
group having more difficulty, reflected in slower response times,
suppressing responses. These findings suggest that performance
in preschool delay of gratification may predict the capacity, in
adulthood, to control thoughts and actions, as reflected in per-
formance on cognitive control tasks, and that the ability to control
one’s thoughts and actions can vary by the potency of interfering
information (12). Likewise, alluring or social contexts can diminish
self-control (4, 13, 14).
Early experiments on delay of gratification demonstrated that
part of the contextual effect was due to the different cognitive
strategies that individuals used. For example, “cooling” the hot,
appealing, or appetitive features of tempting stimuli by reap-
praisal or reframing strategies to focus on their cool, cognitive
features (e.g., to envision the marshmallow as a cloud or a little
cotton ball, rather than as a sweet, delectable treat) has been
shown to be highly effective in enhancing delay of gratification
(e.g., 1, 15–17). The same preschool child who yielded immedi-
ately to the temptation by representing the hot, appetitive fea-
tures of the reward (e.g., its yummy, sweet, chewy taste) could
wait for long periods for the same tempting stimulus by focusing
on its cool qualities (e.g., its shape). At the same time, there
seem to be important, naturally existing individual differences in
the spontaneous use of such strategies (e.g., 5, 18).
Indeed, Metcalfe and Mischel (2) proposed “cool” and “hot”
systems to explain the dynamics of resisting temptation during
the delay-of-gratification task. These two interacting neuro-
cognitive systems are implicated in self-control. Whereas the
first, a “cool” system, involves cognitive control-related neural
circuitry, the second, a “hot” system (19), involves desires and
emotions that are under stimulus control and are associated with
emotional brain regions. Recent brain imaging studies have
provided evidence for dissociable brain systems related to im-
mediate over long-term choice behavior consistent with the no-
Author contributions: B.J.C., O.A., J.J., M.G.B., N.L.W., G.G., V.Z., W.M., and Y.S. designed
research; I.H.G., O.A., M.K.A., J.J., M.G.B., N.L.W., W.M., and Y.S. performed research; B.J.C.,
N.L.W., T.T., G.G., V.Z., W.M., and Y.S. contributed new reagents/experimental tools; B.J.C.,
L.H.S., N.T.F., N.L.W., T.T., W.M., and Y.S. analyzed data; and B.J.C., L.H.S., I.H.G., O.A., J.J.,
M.G.B., N.L.W., V.Z., W.M., and Y.S. wrote the paper.
The authors declare no conflict of interest.
*This Direct Submission article had a prearranged editor.
Freely available online through the PNAS open access option.
1To whom correspondence may be addressed. E-mail: firstname.lastname@example.org, wm@
psych.columbia.edu, or email@example.com.
| September 6, 2011
| vol. 108
| no. 36www.pnas.org/cgi/doi/10.1073/pnas.1108561108
tion of interacting “hot” and “cool” systems. Specifically, whereas
top-down prefrontal regions have been shown to be involved in
cognitive control during delay of rewards, limbic or emotional
brain regions have been shown to be associated with more im-
mediate choices (20–24). Complementary imaging studies have
shown that a region of the prefrontal cortex, the inferior frontal
gyrus, is critically involved in resolving interference among
competing actions [e.g., to go or not to go (10, 25)] and among
competing representations in memory (e.g., 11). In each case,
prepotent information interferes with other goal-specific in-
formation, thus requiring cognitive control processes to resolve
In the present longitudinal study, we manipulated the alluring
qualities of targets in an impulse control task to examine be-
havioral and neural correlates of delay of gratification using
functional magnetic resonance imaging (fMRI). Participants
were individuals whose ability to delay gratification was tested as
4-y-olds on the original delay-of-gratification task and whose self-
control abilities remained consistent in follow-up assessments.
Two experiments were conducted to investigate the ability of
these individuals, now in their 40s, to refrain from responding to
alluring cues. We developed two tasks to examine impulse con-
trol—one in the presence of neutral (“cool”) stimuli and one
containing compelling (“hot”) stimuli. Because marshmallows
and cookies are unlikely to be as rewarding to individuals now as
adults as they were when they were young children, we used the
social cues of faces with emotional expressions (happy faces
relative to neutral and fearful faces), shown to bias behavior
similarly to secondary reinforcers (4, 13, 26). Experiment 1
tested whether individuals who were less able to delay gratifi-
cation as children and young adults (low delayers) would, as
adults in their 40s, show less impulse control in suppression of
a response to “hot” relative to “cool” cues.
In experiment 2, we used fMRI to examine neural correlates
of delay of gratification. We hypothesized that participants with
consistently low levels of self-control from young childhood to
early adulthood (low delayers), compared with their consistently
high-control counterparts, would be characterized by diminished
activity in the right prefrontal cortex, implicated in response
inhibition (13, 27–29), and by amplified activity in the ventral
striatum, implicated in processing of positive or rewarding cues
(13, 26, 30).
Result and Discussion
Experiment 1 Results. In experiment 1, 59 participants classified as
low or high delayers (Table 1) completed a behavioral version of
the “hot” and “cool” impulse control tasks. Reaction times for
trials that required a response (“go” trials) and accuracy for “go”
and “nogo” trials were compared by delay group and task.
Reaction times. There were no effects of delay group on reaction
time measures to correct “go” trials [main effect of group, F
(1,57) = 2.23, P > 0.1; group × task interaction, F(1,57) = 0.002,
P > 0.9].
Accuracy. Participants performed with a high level of accuracy for
correctly responding to “go” trials during both the “cool” (99.8%
correct) and “hot” tasks (99.5% correct). Low and high delayers
performed with comparable accuracy on “go” trials [neither the
main effect of group, F(1,57) = 1.08, P > 0.3, nor the interaction
of group and task, F(1,57) = 0.05, P > 0.8, was significant].
Accuracy for “nogo” trials was more variable [mean false
alarm rate for “cool” task, 9.96%; for “hot” task, 12.2%; main
effect of task, F(1,57) = 7.89, P = 0.007, η2partial= 0.122] and
yielded a significant interaction of group and task [F(1,57) =
4.312, P = 0.042, η2partial= 0.070; Fig. 1]. Whereas low and high
delayers performed comparably on the “cool” task [t(57) =
−0.24, P > 0.8], the low delayers trended toward performing
more poorly on the “hot” task than did the high delayers [t(57) =
1.64, P = 0.11, d = 0.43]. Further, only the low delay group
showed a significant decrement in performance for the “hot”
trials relative to the “cool” trials [t(26) = 3.09, P = 0.005, d =
0.89; high delay group P > 0.5]. Additional planned analyses
separated “hot” task “nogo” trials into fear and happy sub-
categories. The decrement in performance for low delayers was
driven largely by commission errors on the happy “nogo” trials
[low delayers, 15.7% errors; high delayers, 11.2% errors; t(57) =
2.06, P = 0.044, d = 0.55], whereas low and high delaying groups
performed at equivalent levels of accuracy for the fear “nogo”
trials [12.0% and 10.4%, respectively, t(57) = 0.804, P > 0.4].
Experiment 1 Discussion. In this experiment the go/nogo task
produced differences between the two delay groups only in the
presence of emotional (“hot”) cues. Specifically, individuals who,
as a group, had more difficulty delaying gratification at 4 y of age
showed more difficulty as adults in suppressing responses to
happy faces. The findings are consistent with previous work
suggesting that the capacity to resist temptation varies by con-
text; the more tempting the choice for the individual, the more
predictive are the individual differences in people’s ability to
regulate their behavior (e.g., 12). Thus, behavioral correlates of
delay ability are a function not only of cognitive control but also
of the compelling nature of the stimuli that must be suppressed.
Because behavioral differences between the low and high
delayers only emerged on the “hot” task, we scanned participants
during this task in experiment 2.
Experiment 2 Results. Reaction times. As with the behavioral find-
ings outside of the scanner, the two delay groups did not differ
significantly in reaction times to correct “go” trials [t(24) = 0.81,
P > 0.4].
Accuracy. Overall, accuracy rates for the “hot” go/nogo task in the
scanner were uniformly high for “go” trials (mean 98.2% correct
hits), with more variable performance to “nogo” trials (12.4%
false alarm rate). Differences between the two delay groups in
“nogo” accuracy were consistent with the observed differences in
the ”hot” task performance in experiment 1, with low-delay
participants committing more false alarms than high-delay par-
ticipants (low delayers, 14.5%; high delayers, 10.9%; Fig. 1,
Right). This performance difference, however, did not reach
statistical significance [t(24) = 1.08, P = 0.29, d = 0.44], likely
Table 1. Subject demographics
GroupnFemale (n)Age (y)Age 4 y delay scoreAdult self-reportRaven’s score*
Experiment 1 sample
Experiment 2 sample
44.6 ± 2.1
44.3 ± 1.6
332.3 ± 147.0
−284.2 ± 145.2
7.61 ± 0.54
5.91 ± 1.04
10 44.8 ± 1.8
44.2 ± 1.8
304.2 ± 145.2
−222.8 ± 145.3
7.39 ± 0.45
6.18 ± 0.58
25.5 ± 5.0
22.9 ± 4.64
*Raven’s data were collected on the imaging subjects only.
Casey et al. PNAS
| September 6, 2011
| vol. 108
| no. 36
owing to fewer data points given the smaller sample and fewer
trials in the imaging version of the task.
Imaging results. The 2 (delay group: high, low) × 2 (trial type:
nogo, go) × 2 (emotion: happy, fear) voxelwise ANOVA was
conducted to identify brain regions differentially active by task
conditions and delay group. The comparison of “nogo” vs. “go”
trials identified candidate regions differentially engaged as a
function of cognitive control demands, including the right in-
ferior frontal gyrus (x = 35, y = 11, z = −4), showing significantly
greater responses to “nogo” relative to “go” trials (P < 0.05,
whole-brain corrected; Fig. 2A and Table 2). As expected, both
left primary motor cortex (x = −44, y = −23, z = 54) and ipsi-
lateral cerebellum (x = 23, y = −47, z = −19) were more active
for “go” trials, which required a response, vs. “nogo” trials, which
did not (Ps < 0.05, whole-brain corrected; Fig. 2 B and C). No
regions reached significance for the main effect of or two-way
interactions with delay group. Regions identified for the main
effect of emotion included bilateral portions of the temporal
lobes (Table 2).
Although the interaction of delay group and response was not
significant in whole-brain analyses, we hypothesized that the in-
ferior frontal gyrus would show subtle differences in recruitment
to nogo trials as a function of delay group (5). To test this pre-
diction, parameter estimates from the inferior frontal gyrus re-
gion functionally defined by the main effect of “nogo” relative to
“go” trials were extracted for each individual and evaluated for
potential interactions by delay group in post hoc statistical
analyses. This analysis yielded a significant interaction of trial
type and delay group in inferior frontal gyrus recruitment [F
(1,24) = 7.05, P = 0.014, η2partial= 0.23; Fig. 3]. Follow-up t tests
indicated that high-delay participants showed greater polariza-
tion of inferior frontal gyrus response to “nogo” relative to “go”
trials [t(14) = 4.33, P = 0.001, d = 2.6)], whereas differential
inferior frontal gyrus recruitment was less pronounced in low
delayers [t(10) = 1.93, P = 0.082, d = 0.82]. No other region for
the main effect of “nogo” vs. “go” trials differentiated the delay
groups (i.e., parietal cortex; Table 2).
In the primary analysis of interest, we examined group dif-
ferences in the ability to suppress a response to a positive cue, as
tested by the three-way interaction of delay group (high, low),
trial type (nogo, go), and emotion (happy, fear). This analysis
revealed a cluster in the ventral striatum (x = 11, y = 2, z = −4;
P < 0.05 small volume corrected; Fig. 4). Planned follow-up
analyses of this interaction showed greater ventral striatal activity
in low delayers relative to high delayers overall during task runs
in which happy faces served as the “nogo” stimulus [t(24) = 2.20,
P = 0.038, d = 0.90], with low delayers engaging the ventral
striatum more than high delayers specifically for happy “nogo”
trials [t(24) = 1.96, P = 0.06, d = 0.80; Fig. 4].
Experiment 2 Discussion. Resisting temptation, as measured by the
“hot” go/nogo task, was supported by frontostriatal circuitry.
Specifically, the right inferior frontal gyrus was involved in ac-
curately withholding a response. Interrogation of this region
showed that compared with high delayers, low delayers had di-
minished recruitment of this region for correct “nogo” relative to
“go” trials. In addition, the ventral striatum demonstrated a sig-
nificant difference in recruitment between high and low delay
groups. This reward-related region showed a three-way in-
teraction of group × trial type × emotion, with elevated activity
to happy “nogo” trials for low delayers relative to high delayers.
This finding highlights the importance of the qualities of the
stimulus people have to resist, such as its salience or allure, in
modulating cognitive control ability.
General Discussion. The results from both experiments together
suggest three key findings. First, resistance to temptation as
measured originally by the delay-of-gratification task, and in the
z = -19
z = 46
z = -2
was associated with correct inhibition of a response (nogo) relative to
making a correct response (go). Left: Activation map depicting right inferior
frontal gyrus activation, thresholded at P < 0.05, whole-brain corrected,
displayed on a representative high-resolution T1-weighted axial image.
Right: Percentage change in MR signal for “go” and “nogo” trials in the
inferior frontal gyrus. (B) Left primary motor cortex was associated with
making a correct response (go) relative to correctly withholding a response
(nogo). Left: Activation map depicting activation in left precental gyrus,
thresholded at P < 0.05, whole-brain corrected, displayed on a representa-
tive high-resolution T1-weighted axial image. Right: Percentage change in
MR signal for “go” and “nogo” trials in the left precentral gyrus. (C) Left
cerebellum was associated with making a correct response (go) relative
to correctly withholding a response. Left: Activation map depicting left
cerebellum activation, thresholded at P < 0.05, whole-brain corrected, dis-
played on a representative high-resolution T1-weighted axial image. Right:
Percentage change in MR signal for “go” and “nogo” trials in the right
Main effects of the go/nogo task. (A) The right inferior frontal cortex
Cool Task Hot Task Hot Task
False Alarm Rate
not differ in performance on a go/nogo task when cues are “cool” stimuli
(neutral facial expressions), but low delayers make more errors when the
cues are “hot” (emotional faces). Right: Experiment 2 (inside the scanner). A
similar pattern is observed. Error bars denote SEM.
Left: Experiment 1 (outside the scanner). High and low delayers do
| www.pnas.org/cgi/doi/10.1073/pnas.1108561108Casey et al.
present study by a “hot” version of an impulse control task, is
a relatively stable individual difference characteristic. Second,
and consistent with delay experiments on the value of “cooling”
the hot features of temptations (1), behavioral correlates of delay
ability involve not cognitive control in general, but in particular
in response to positive compelling cues. Third, resisting temp-
tation is supported by ventral frontostriatal circuitry, with the
inferior frontal gyrus showing lesser recruitment in low delayers
and the ventral striatum showing exaggerated recruitment in low
delayers when resisting alluring cues. These findings suggest that
sensitivity to positive social cues influences an individual’s ability
to suppress thoughts and actions and thus can undermine self-
The relative stability in the capacity to resist temptation was
shown over a 40-y span. Specifically, individuals who, as a group,
had more difficulty delaying gratification at 4 y of age and con-
tinued to show reduced self-control abilities subsequently
exhibited more difficulty as adults in suppressing responses to
positive social cues. Previous research has documented that
higher delay ability promotes the development of better social–
cognitive and emotional coping in adolescence and buffers
against the development of a variety of dispositional physical and
mental health vulnerabilities in middle age, such as high BMI,
cocaine/crack use, features of borderline personality disorder,
anxious overreactions to rejection, and marital divorce/separation
(17, 18, 31–33), even when controlling for childhood social en-
vironment and child health (34). Thus, our findings confirm the
significance and predictive validity of delay ability in preschoolers
for behaviors in later life. These findings complement prior work
characterizing developmental features of self-regulation (35–37)
by providing further evidence for their predictive validity in
subsequent stages of life (38).
An important construct related to resisting temptation is that
of impulse control (5). The go/nogo task is considered a reliable
measure of impulsivity (25, 39). In the present study, the “cool”
version of the go/nogo task did not differentiate the low- and
high-delay groups. Only in the presence of positive social (“hot”)
cues did the go/nogo task reveal differences between the two
delay groups. These findings are consistent with our previous
study showing associations between performance on the delay of
gratification task and the go/nogo task, which were particularly
apparent in certain contexts (5). In that study, the differences
between groups were most pronounced on difficult trials, “nogo”
trials preceded by several “go” trials, such that the “go” response
became more compelling (28). Thus, the behavioral correlates of
delay ability are a function not only of impulse control but also of
the salience of the stimulus one has to resist.
Two neurocognitive systems that rely on distinct neural sys-
tems have been proposed to enable self-control (2). Whereas the
“cool” system involves cortical control regions, including the
prefrontal cortex, the “hot” system involves deep brain structures
such as the ventral striatum that are implicated in the processing
of desires and rewards. Resisting temptation, as measured by the
“hot” go/nogo task in the present study, supports this view, with
the prefrontal cortex and the ventral striatum differentiating low-
and high-delay participants. The difference in inferior frontal
gyrus recruitment for “nogo” relative to “go” trials was reduced
in low delaying participants, potentially reflecting reduced ability
in these individuals to invoke cognitive control in the context of
emotional or “hot” cues. The ventral striatum has been impli-
cated in reward and in immediate, as opposed to delayed, choice
behavior (20, 21, 26, 40). Thus, sensitivity to environmental cues
influences an individual’s ability to suppress thoughts and
actions, such that control systems may be “hijacked” by a primi-
tive limbic system, rendering control systems unable to appro-
priately modulate behavior. Similar analogies of imbalances
between these neural systems in the literature suggest that ad-
diction (41) and adolescence (13, 42, 43) may be contexts when
cognitive control may be particularly vulnerable to alluring
Resisting temptation in favor of long-term goals is important for
individual, societal, and economic functioning. Our findings pro-
vide a neurobiological basis for differences in this ability. Together
the behavioral and imaging findings provide evidence that the
ability to delay gratification assessed early in life predicts how
well individuals can regulate behavior years later, particularly
Go Trials Nogo Trials
trials is more pronounced in high delayers than in low delayers. The right
inferior gyrus (region shown in Fig. 2A) showed a significant interaction
between group and trial type, with greater polarization of inferior frontal
gyrus response to “nogo” relative to “go” trials for high delayers. Error bars
Differential inferior frontal gyrus recruitment between nogo and go
ventral striatum when inhibiting responses to positive social cues 40 y later.
Left: Activation map for the three-way interaction of task, emotion, and
delay group depicting ventral striatum activity thresholded at P < 0.05, small
volume corrected, displayed on a representative high-resolution T1-
weighted axial image. Right: Ventral striatal response to happy “nogo” trials
in high and low delayers.
Low delay ability in early childhood predicts greater recruitment of
Brain activations observed for main effects of trial type
Trial type (nogo > go)
Inferior parietal cortex
Inferior frontal gyrus
Trial type (go > nogo)
Emotion (fear > happy)
Middle temporal gyrus
Middle temporal gyrus
Threshold: P < 0.05, whole-brain corrected.
Casey et al. PNAS
| September 6, 2011
| vol. 108
| no. 36
when they are required to suppress thoughts and actions toward
alluring social cues.
Materials and Methods
We contacted 117 individuals from more than 500 original participants who
completed the delay-of-gratification task at age 4 y at Stanford’s Bing
Nursery School during the late 1960s and early 1970s and were above or
below average in their original delay-of-gratification performance as well as
in self-report measures of self-control administered in their 20s and 30s (44).
More than half of these individuals agreed to participate in a longitudinal
behavioral study (experiment 1), and of these 59 subjects, nearly half agreed
to be part of a functional neuroimaging study (experiment 2). Both the
behavioral and imaging experiments received institutional review board
approval, and all participants provided consent before testing and scanning.
Experiment 1. Participants. Participants were selected on the basis of their
scores on two types of self-control measures: (i) the length of time they
waited for the more desirable but delayed reward at age 4 y, and (ii) their
score(s) on a self-control scale in adulthood. The time spent delaying a re-
ward at age 4 y (see ref. 1 for review) was recorded in seconds and nor-
malized by calculating a difference score of the participant’s delay time
relative to the average delay time of children in the same experimental
condition (see ref. 33 for details). The self-control scale used in the adult-
hood assessments consisted of eight items [e.g., “Unable to delay grati-
fications; cannot wait for satisfactions” (reverse coded); “Is attentive and
able to concentrate”] from the modified shortened version of the California
Child Q-set (45). This self-report measure was administered in 1993 (when
participants were in their 20s) and in 2003 (when participants were in their
30s). Internal consistency of this scale was relatively high (α = 0.74 in 1993
and 0.82 in 2003).
Of the 562 participants for whom we had preschool delay scores at age 4 y,
155 completed the 1993 follow-up and 135 completed the 2003 follow-up.
For the present study, we targeted individuals who obtained consistently
high or consistently low scores on measures of self-control. The high-delay
group was composed of individuals who scored above the mean in their
preschool delay time as well as in their subsequent self-control assessments in
adulthood (whichever was available, or the average of the two when both
were available), whereas the low-delay group was composed of individuals
met the criterion for the high-delay group, and 57 participants met the
criterion for the low-delay group; all of these individual were sent invitation
letters. The final sample (n = 59) consists of those who accepted our in-
vitation to participate in the present follow-up. As expected, high- and low-
delay groups differed significantly in delay score [t(57) = 16.1, P < 0.001] and
reported self-control as adults [t(57) = 8.06, P < 0.001]; they did not differ in
age [t(50) = 0.69, P > 0.49] (age information was not obtained from four
high delayers and three low delayers, but all participants are known to have
been born between 1964 and 1970) or sex distribution [χ2(1) = 0.07, P > 0.8;
Behavioral tasks. Participants completed two versions of a go/nogo task (4, 46,
47). The “cool” version of the go/nogo task consisted of male and female
stimuli with neutral expressions. Within a single run, both male and female
stimuli were presented, one sex as a “go” (i.e., target) stimulus to which
participants were instructed to press a button, and the other sex as a “nogo”
(i.e., nontarget) stimulus to which participants were instructed to withhold
a button press. Before the onset of each run, a screen appeared indicating
which stimulus category served as the target. Participants were instructed
to respond as quickly and accurately as possible. During the task, a face
appeared for 500 ms, followed by a 1-s interstimulus interval. A total of 160
trials were presented per run in pseudorandomized order (120 go, 40 nogo).
Thus, the task was a 2 (trial type: go, nogo) × 2 (stimulus sex: male, female)
factorial design. Accuracy and response latency data were acquired in four
runs representing each combination of stimulus sex (male, female) and trial
type (go, nogo). The “hot” version of the go/nogo task was identical to
the “cool” version, except that for this task fearful and happy facial ex-
pressions served as stimuli. Both expressions were used as targets (go) and
nontargets (nogo) in a 2 (trial type: go, nogo) × 2 (emotion: fear, happy)
Stimuli and apparatus. Stimuli consisted of happy, fearful, and neutral faces of
unique identities from the NimStim set of facial expressions (48). The tasks
were presented using Inquisit software on laptop computers that were sent
to participants’ homes.
Data analysis. Behavioral data were analyzed for two accuracy measures on
the go/nogo tasks: hit rate for “go” trials (correctly pressing) and correct
suppression rate for “nogo” trials (correctly withholding press). Because
delay score was nonnormally (bimodally) distributed (Table 1), delay score
was treated as a categorical rather than a continuous variable. Analyses of
response accuracy (correct “go,” correct “nogo”) were performed with 2
(task: “hot,” “cool”) × group (low delayer, high delayer) ANOVAs. Reaction
times for correct “go” trials were trimmed for outlying responses within
each participant (±3 SDs from the mean, resulting in a mean 1.23% data
excluded). Mean reaction time data were analyzed at the group level with
a 2 (task: “hot,” “cool”) × group (low delayer, high delayer) ANOVA. Be-
havioral effects relevant to “nogo” trials are presented graphically as false
alarm rates, the mathematical inverse of correct suppression. For significant
effects, effect sizes were calculated using the η2partialmetric for ANOVA and
Cohen’s d for t test results (correcting for shared variance in the case of
paired t tests).
Experiment 2. Participants. Of the 59 individuals who completed experiment 1,
27agreedtocomplete theimaging study.One 41-y-old man, alowdelayeron
the original delay-of-gratification task, was excluded from all analyses be-
cause of poor performance (59% hit rate). The high- and low-delay groups
differed significantly in delay score [t(24) = 9.1, P < 0.001] but did not differ
in age [t(21) = 0.87, P > 0.390] (age information was not obtained from one
high delayer and two low delayers, but all participants are known to have
been born between 1964 and 1970), sex distribution [χ2(1) = 2.3, P > 0.1], or
their scores on Raven’s progressive matrices [t(22) = 1.26, P > 0.2].
Behavioral paradigm. The “hot” version of the go/nogo task was similar to the
task described in experiment 1. Differences were in the timing, number of
trials, and apparatus. The timing of the task included the presentation of the
face stimulus for 500 ms, followed by a jittered intertrial interval ranging
from 2 to 14.5 s in duration (mean, 5.2 s), during which participants rested
while viewing a fixation crosshair. A total of 48 trials were presented per run
in pseudorandomized order (35 “go,” 13 “nogo”). In total, imaging data
were acquired for 26 nogo trials and 70 go trials for each expression. The
stimuli and instructions were identical to those used in experiment 1.
E-Prime software was used to display the task (viewable by a rear projection
screen) and to record button responses and reaction times on a Neuroscan
five-button response pad.
Image acquisition. Participants were scanned with a General Electric Signa 3.0-T
fMRI scanner (General Electric Medical Systems) with an eight-channel head
coil. A high-resolution, T1-weighted anatomical spoiled gradient sequence
(256 × 256 in-plane resolution, 240-mm field of view, 136 × 1.2-mm axial
slices) was used to acquire an anatomical scan for each subject for trans-
formation and localization of data to Talairach grid space. A spiral in-and-
out T2*-weighted sequence (49) was used to acquire four runs of functional
data (repetition time 2,500 ms, echo time 30 ms, flip angle 90°, skip 0, 64 ×
64 matrix) with 34 4-mm slices per volume.
Behavioral data analysis. Behavioral differences between the low- and high-
delay groups were analyzed on two accuracy measures on the go/nogo tasks:
hit rate for “go” trials (correctly pressing) and correct suppression rate for
“nogo” trials (correctly withholding press). Group behavioral analysis was
conducted as described in experiment 1. Mean reaction times for correct
“go” trials were calculated and analyzed as described in experiment 1 (mean
1.24% responses trimmed).
Imaging data processing. The fMRI data analysis was performed using Analysis
of Functional Neuroimages software (50). Functional datasets were corrected
for slice acquisition time and realigned within and across runs to correct
for head motion. All subjects moved less than 2.5 mm in all planes, with the
exception of one participant who had 3.8% of their functional data excluded
for motion during scan runs. Functional datasets were coregistered to each
participant’s high-resolution anatomical scan, converted to percentage sig-
nal change, transformed into the standard coordinate space of Talairach
and Tournoux (51), and smoothed with a 6-mm FWHM Gaussian kernel.
Imagingdataanalysis. Avoxelwise general linear modelanalysiswas conducted
for each participant by convolving task regressors of interest (happy-go,
happy-nogo, fear-go, fear-nogo, and errors) with a γ-variate hemodynamic
response function. Motion parameters, linear and quadratic trends were
modeled as regressors of noninterest. To identify regions of interest of
further analysis, parameter estimate (β) maps for the conditions happy-go,
happy-nogo, fear-go, and fear-nogo were carried to a random-effects
As described earlier, one participant was excluded for excessively poor
behavioral performance on the fMRI version of the task, leaving 26 partic-
ipants for group analyses. A 2 × 2 × 2 group linear mixed-effects model was
conducted with factors of trial type (within subjects: go, no-go), emotion
(within subjects: happy, fearful), and group (between subjects: high-delayer,
low-delayer). Candidate regions of interest were identified by the main ef-
fect of response map and by the trial type × emotion × group interaction
| www.pnas.org/cgi/doi/10.1073/pnas.1108561108Casey et al.
map. Resulting clusters considered statistically significant exceeded whole- Download full-text
brain correction for multiple comparisons using a P value and cluster combi-
nation identified by Monte Carlo simulations to preserve a corrected α < 0.05
(53 voxels at P < 0.005 thresholding). On the basis of our prior work (13), the
striatum was targeted as an a priori structure of interest for voxelwise analysis.
Task effects in the striatum from the present experiment were queried within
an anatomically masked area (1,060 voxels) including the dorsal and ventral
striatum and achieved P < 0.05 corrected thresholding on the basis of this
search volume (13 voxels at P < 0.005, uncorrected). Regions of interest were
created as spheres around peak voxels with a radius of 4 mm, each containing
10 3 × 3 × 3 voxels. Parameter estimates were extracted for the four conditions
for each participant and plotted for descriptive purposes.
ACKNOWLEDGMENTS. This work was supported in part by National Insti-
tutes of Health Grants R01 DA018879 (to B.J.C.), R01 HD069178 (to B.J.C. and
W.M.), and by National Science Foundation Grant 06-509 (to Y.S.).
1. Mischel W, Shoda Y, Rodriguez MI (1989) Delay of gratification in children. Science
2. Metcalfe J, Mischel W (1999) A hot/cool-system analysis of delay of gratification:
dynamics of willpower. Psychol Rev 106:3e19.
3. Somerville LH, Casey BJ (2010) Developmental neurobiology of cognitive control and
motivational systems. Curr Opin Neurobiol 20:236e241.
4. Hare TA, Tottenham N, Davidson MC, Glover GH, Casey BJ (2005) Contributions of
amygdala and striatal activity in emotion regulation. Biol Psychiatry 57:624e632.
5. Eigsti IM, et al. (2006) Predicting cognitive control from preschool to late adolescence
and young adulthood. Psychol Sci 17:478e484.
6. Kahneman D, Treisman A, Burkell J (1983) The cost of visual filtering. J Exp Psychol
Hum Percept Perform 9:510e522.
7. Allport A (1987) Selection for action: Some behavioral and neurophysiological con-
siderations of attention and action. Perspectives on Perception and Action, ed
Sanders HHAF (Laurence Erlbum Associates, Hillsdale, NJ), pp 395e419.
8. Cohen JD, Servan-Schreiber D (1992) Context, cortex, and dopamine: A connectionist
approach to behavior and biology in schizophrenia. Psychol Rev 99:45e77.
9. Casey BJ, et al. (2000) Dissociation of response conflict, attentional control, and ex-
pectancy with functional magnetic resonance imaging. Proc Natl Acad Sci USA 97:
10. Casey BJ, Tottenham N, Fossella J (2002) Clinical, imaging, lesion, and genetic ap-
proaches toward a model of cognitive control. Dev Psychobiol 40:237e254.
11. Jonides J, Nee DE (2006) Brain mechanisms of proactive interference in working
memory. Neuroscience 139:181e193.
12. Shoda Y, Mischel W, Peake PK (1990) Predicting adolescent cognitive and self-regu-
latory competencies from preschool delay of gratification: Identifying diagnostic
conditions. Dev Psychol 26:978e986.
13. Somerville LH, Hare T, Casey BJ (2011) Frontostriatal maturation predicts cognitive
control failure to appetitive cues in adolescents. J Cogn Neurosci 23:2123e2134.
14. Gardner M, Steinberg L (2005) Peer influence on risk taking, risk preference, and risky
decision making in adolescence and adulthood: An experimental study. Dev Psychol
15. Mischel W, Ebbesen EB, Zeiss AR (1972) Cognitive and attentional mechanisms in
delay of gratification. J Pers Soc Psychol 21:204e218.
16. Mischel W, Underwood B (1974) Instrumental ideation in delay of gratification. Child
17. Mischel W, Ayduk O (2004) Willpower in a cognitive-affective processing system: The
dynamics of delay of gratification. Handbook of Self-Regulation: Research, Theory,
and Applications, eds Baumeister RF, Vohs KD (Guilford, New York), pp 99e129.
18. Rodriguez ML, Mischel W, Shoda Y (1989) Cognitive person variables in the delay of
gratification of older children at risk. J Pers Soc Psychol 57:358e367.
19. Abelson RP (1963) Computer simulation of “hot cognition”. Computer Simulation of
Personality, eds Tomkins SS, Mesick S (Wiley, New York), pp 277e302.
20. McClure SM, Ericson KM, Laibson DI, Loewenstein G, Cohen JD (2007) Time dis-
counting for primary rewards. J Neurosci 27:5796e5804.
21. Rangel A, Camerer C, Montague PR (2008) A framework for studying the neurobi-
ology of value-based decision making. Nat Rev Neurosci 9:545e556.
22. Sanfey AG (2007) Social decision-making: Insights from game theory and neurosci-
ence. Science 318:598e602.
23. Ochsner KN, Gross JJ (2005) The cognitive control of emotion. Trends Cogn Sci 9:
24. Heatherton TF, Wagner DD (2011) Cognitive neuroscience of self-regulation failure.
Trends Cogn Sci 15:132e139.
25. Aron AR, Robbins TW, Poldrack RA (2004) Inhibition and the right inferior frontal
cortex. Trends Cogn Sci 8:170e177.
26. Galvan A, et al. (2005) The role of ventral frontostriatal circuitry in reward-based
learning in humans. J Neurosci 25:8650e8656.
27. Casey BJ, et al. (1997) A developmental functional MRI study of prefrontal activation
during performance of a go-no-go task. J Cogn Neurosci 9:835e847.
28. Durston S, Thomas KM, Worden MS, Yang Y, Casey BJ (2002) The effect of preceding
context on inhibition: An event-related fMRI study. Neuroimage 16:449e453.
29. Ochsner KN, et al. (2004) For better or for worse: Neural systems supporting the
cognitive down- and up-regulation of negative emotion. Neuroimage 23:483e499.
30. Delgado MR, Nystrom LE, Fissell C, Noll DC, Fiez JA (2000) Tracking the hemodynamic
responses to reward and punishment in the striatum. J Neurophysiol 84:3072e3077.
31. Ayduk O, et al. (2000) Regulating the interpersonal self: Strategic self-regulation for
coping with rejection sensitivity. J Pers Soc Psychol 79:776e792.
32. Ayduk O, et al. (2008) Rejection sensitivity and executive control: Joint predictors of
borderline personality features. J Res Pers 42:151e168.
33. Mischel W, Shoda Y, Peake PK (1988) The nature of adolescent competencies pre-
dicted by preschool delay of gratification. J Pers Soc Psychol 54:687e696.
34. Kubzansky LD, Martin LT, Buka SL (2009) Early manifestations of personality and
adult health: A life course perspective. Health Psychol 28:125e130.
35. Posne MI, Rothbart MK (2000) Developing mechanisms of self-regulation. Dev
36. Carlson SM, Moses LJ (2001) Individual differences in inhibitory control and children’s
theory of mind. Child Dev 72:1032e1053.
37. Rothbart MK, Ahadi SA (1994) Temperament and the development of personality.
J Abnorm Psychol 103:55e66.
38. Moffitt TE, et al. (2011) A gradient of childhood self-control predicts health, wealth,
and public safety. Proc Natl Acad Sci USA 108:2693e2698.
39. Dalley JW, Everitt BJ, Robbins TW (2011) Impulsivity, compulsivity, and top-down
cognitive control. Neuron 69:680e694.
40. O’Doherty JP, Buchanan TW, Seymour B, Dolan RJ (2006) Predictive neural coding of
reward preference involves dissociable responses in human ventral midbrain and
ventral striatum. Neuron 49:157e166.
41. Bechara A (2005) Decision making, impulse control and loss of willpower to resist
drugs: A neurocognitive perspective. Nat Neurosci 8:1458e1463.
42. Galvan A, et al. (2006) Earlier development of the accumbens relative to orbitofrontal
cortex might underlie risk-taking behavior in adolescents. J Neurosci 26:6885e6892.
43. Chein J, Albert D, O’Brien L, Uckert K, Steinberg L (2011) Peers increase adolescent risk
taking by enhancing activity in the brain’s reward circuitry. Dev Sci 14:F1eF10.
44. Mischel W, et al. (2011) ‘Willpower’ over the life span: Decomposing self-regulation.
Soc Cogn Affect Neurosci 6:252e256.
45. Block J, Block JH (1980) The California Child Q-set (Consulting Psychologists Press, Palo
46. Hare TA, et al. (2008) Biological substrates of emotional reactivity and regulation in
adolescence during an emotional go-nogo task. Biol Psychiatry 63:927e934.
47. Tottenham N, Hare TA, Casey BJ (2011) Behavioral assessment of emotion discrimi-
nation, emotion regulation and cognitive control in childhood, adolescence, and
adulthood. Front Psychol 2:39.
48. Tottenham N, et al. (2009) The NimStim set of facial expressions: Judgments from
untrained research participants. Psychiatry Res 168:242e249.
49. Glover GH, Thomason ME (2004) Improved combination of spiral-in/out images for
BOLD fMRI. Magn Reson Med 51:863e868.
50. Cox RW (1996) AFNI: Software for analysis and visualization of functional magnetic
resonance neuroimages. Comput Biomed Res 29:162e173.
51. Talairach J, Tournoux P (1988) Co-planar Stereotaxic Atlas of the Human Brain
(Thieme Medical Publishers, New York).
Casey et al. PNAS
| September 6, 2011
| vol. 108
| no. 36