Self-Control Relies on Glucose as a Limited Energy Source: Willpower Is More Than a Metaphor
The present work suggests that self-control relies on glucose as a limited energy source. Laboratory tests of self-control (i.e., the Stroop task, thought suppression, emotion regulation, attention control) and of social behaviors (i.e., helping behavior, coping with thoughts of death, stifling prejudice during an interracial interaction) showed that (a) acts of self-control reduced blood glucose levels, (b) low levels of blood glucose after an initial self-control task predicted poor performance on a subsequent self-control task, and (c) initial acts of self-control impaired performance on subsequent self-control tasks, but consuming a glucose drink eliminated these impairments. Self-control requires a certain amount of glucose to operate unimpaired. A single act of self-control causes glucose to drop below optimal levels, thereby impairing subsequent attempts at self-control.
Self-Control Relies on Glucose as a Limited Energy Source: Willpower Is
More Than a Metaphor
Matthew T. Gailliot, Roy F. Baumeister,
C. Nathan DeWall, Jon K. Maner, E. Ashby Plant,
Dianne M. Tice, and Lauren E. Brewer
Florida State University
Brandon J. Schmeichel
Texas A&M University
The present work suggests that self-control relies on glucose as a limited energy source. Laboratory tests
of self-control (i.e., the Stroop task, thought suppression, emotion regulation, attention control) and of
social behaviors (i.e., helping behavior, coping with thoughts of death, stifling prejudice during an
interracial interaction) showed that (a) acts of self-control reduced blood glucose levels, (b) low levels
of blood glucose after an initial self-control task predicted poor performance on a subsequent self-control
task, and (c) initial acts of self-control impaired performance on subsequent self-control tasks, but
consuming a glucose drink eliminated these impairments. Self-control requires a certain amount of
glucose to operate unimpaired. A single act of self-control causes glucose to drop below optimal levels,
thereby impairing subsequent attempts at self-control.
Keywords: self-regulation, glucose, attention, emotion regulation, prejudice
Self-control (or self-regulation) is the ability to control or over-
ride one’s thoughts, emotions, urges, and behavior. Self-control
allows for the flexibility necessary for successful goal attainment,
and it greatly facilitates adherence to morals, laws, social norms,
and other rules and regulations. As such, it is one of the most
important and beneficial processes in the human personality struc-
ture. A burgeoning body of evidence has linked good self-control
to a broad range of desirable outcomes, including healthier inter-
personal relationships, greater popularity, better mental health,
more effective coping skills, reduced aggression, and superior
academic performance, as well as less susceptibility to drug and
alcohol abuse, criminality, and eating disorders (DeWall,
Baumeister, Stillman, & Gailliot, in press; Duckworth & Selig-
man, 2005; Finkel & Campbell, 2001; Gailliot, Schmeichel, &
Baumeister, 2006; Gottfredson & Hirschi, 1990; Kahan, Polivy, &
Herman, 2003; Pratt & Cullen, 2000; Shoda, Mischel, & Peake,
1990; Tangney, Baumeister, & Boone, 2004; Vohs & Heatherton,
Self-control seems to rely on a limited energy or strength, such
that engaging in a single act of self-control impairs subsequent
attempts at self-control, as if some sort of energy had been used up
during the initial act (for reviews, see Baumeister, Gailliot, De-
Wall, & Oaten, in press; Muraven & Baumeister, 2000). Although
viewing self-control as an energy resource has served as a highly
convenient metaphor that explains a broad range of empirical
findings, the precise nature of the energy source of self-control has
remained unspecified. In the present, we examined whether self-
control does indeed rely on an actual energy source, namely, blood
Since Freud (1923/1961a, 1933/1961b), psychological theoriz-
ing about personality or the self has used energy models relatively
sparingly. Yet, the human body is undeniably an energy system,
and its very life depends on ingesting energy and then using it to
fuel its activities, including complex psychological processes. The
human brain consumes 20% of the body’s calories even though it
constitutes only 2% of the body’s mass (Dunbar, 1998). In order
for evolution to have selected in favor of such expensive psycho-
logical processes, those processes must have paid great adaptive
dividends to offset such a high cost in calories. The capacity for
self-control provides numerous benefits (e.g., Baumeister, 2005),
and so it is plausible that self-control may have been one psycho-
logical process that was immensely valuable despite being so
expensive in terms of caloric energy (glucose).
An accumulating amount of evidence is consistent with the
notion that self-control relies on some kind of energy. For instance,
after resisting the temptation to eat freshly baked cookies, partic-
ipants in one study quit sooner on a subsequent task requiring
effortful persistence, compared with participants who did not have
to resist eating the cookies (Baumeister, Bratslavsky, Muraven, &
Tice, 1998). Resisting the temptation to eat the cookies presum-
ably depleted an energy resource that could otherwise have been
used to persist on the subsequent task. A variety of other behaviors
have been found to rely on and deplete this energy source as well,
including managing one’s impression (Vohs, Baumeister, & Cia-
rocco, 2005), suppressing stereotypes and prejudice (Gordijn, Hin-
driks, Koomen, Dijksterhuis, & Van Knippenberg, 2004; Richeson
& Shelton, 2003; Richeson & Trawalter, 2005; Richeson,
Trawalter, & Shelton, 2005), coping with thoughts and fears of
dying (Gailliot et al., 2006), controlling one’s monetary spending
(Vohs & Faber, 2004), restraining aggression (DeWall et al., in
Matthew T. Gailliot, Roy F. Baumeister, C. Nathan DeWall, Jon K.
Maner, E. Ashby Plant, Dianne M. Tice, Lauren E. Brewer, Department of
Psychology, Florida State University; Brandon J. Schmeichel, Department
of Psychology, Texas A&M University.
Correspondence concerning this article should be addressed to Matthew
Gailliot or Roy Baumeister, Department of Psychology, Florida State
University, Tallahassee, FL 32306-1270. E-mail: email@example.com or
Journal of Personality and Social Psychology Copyright 2007 by the American Psychological Association
2007, Vol. 92, No. 2, 325–336 0022-3514/07/$12.00 DOI: 10.1037/0022-35184.108.40.2065
press; Stucke & Baumeister, 2006), and managing one’s intake of
food and alcohol (Kahan et al., 2003; Muraven, Collins, & Nien-
haus, 2002; Muraven, Collins, Shiffman, & Paty, 2005; Vohs &
Thus, there is ample evidence that self-control processes operate
as if they depend on some kind of limited energy resource. But
what might that energy resource actually be? Glucose may be one
facet of the energy dynamics of self-control.
Self-Control and Glucose
Glucose is one vital fuel for the brain. The brain’s activities rely
heavily on glucose for energy (e.g., Laughlin, 2004; Siesjo, 1978;
Weiss, 1986). The metabolization of glucose from the bloodstream
allows each brain region to carry out its given functions (e.g.,
McNay, McCarty, & Gold, 2001; Reivich & Alavi, 1983).
Even though nearly all of the brain’s activities consume some
glucose, most cognitive processes are relatively unaffected by
subtle or minor fluctuations in glucose levels within the normal or
healthy range. Controlled, effortful processes that rely on execu-
tive function, however, are unlike most other cognitive processes
in that they seem highly susceptible to normal fluctuations in
glucose. For instance, low glucose has been linked with impaired
performance on difficult (incongruent) but not easy (congruent)
trials of the Stroop color word interference task (Benton, Owens,
& Parker, 1994) and on complex but not simple reaction time tasks
(Owens & Benton, 1994). One study found that low glucose was
associated with poor performance on a driving simulation task, but
only toward the end of the task, when participants were fatigued
and the task was most demanding (as cited in Benton, 1990). Low
glucose, therefore, seems to impair controlled or effortful pro-
cesses but not the simpler or automatic processes, most likely
because controlled processes require more glucose than automatic
processes (Fairclough & Houston, 2004).
Self-control relies on controlled or executive processes in that
the self must effortfully override urges, thoughts, emotions, and
habitual or automatic response tendencies. Self-control, therefore,
may be highly susceptible to fluctuations in glucose. Indeed,
indirect evidence suggests that self-control failure may be more
likely when glucose is low or when glucose is not transported
effectively from the body to the brain. For instance, poor self-
control is one of the leading causes of criminal behavior (Gottfred-
son & Hirschi, 1990; Pratt & Cullen, 2000), and several studies
have linked criminal behavior to decrements in the processing of
glucose (e.g., Bolton, 1979; Virkkunen & Huttunen, 1982). Prob-
lems with glucose have been associated with increases in aggres-
sion and impulsivity (Donohoe & Benton, 1999; Lustman, Frank,
& McGill, 1991) and with decrements in concentration and emo-
tion regulation (Benton & Owens, 1993; Benton et al., 1994).
Alcohol impairs many forms of self-control (Baumeister, Heath-
erton, & Tice, 1994), and likewise, alcohol reduces levels of
glucose in the brain and body (Altura, Altura, Zhang, & Zakhari,
1996). Failures at self-control are more likely later in the evening
than during the day (Baumeister et al., 1994), and glucose is used
less effectively later in the evening than during the day (Van
Cauter, Polonsky, & Scheen, 1997). Glucose has also been found
to facilitate coping with stress (Simpson, Cox, & Rothschild, 1974)
and quitting smoking (West, 2001). These links between self-
control and glucose suggest that glucose may be an important
component of the energy source on which self-control relies.
Overview and Hypotheses of the Present Work
We used nine studies to test the hypothesis that decrements in
self-control are caused in part by low glucose. This relatively large
number of studies allowed us to provide converging multimethod
evidence that would demonstrate the effects of glucose on a broad
range of self-control behaviors and rule out potential alternative
explanations. We hypothesized that completing a self-control task
would use up a relatively large amount of glucose, compared with
completing a cognitive task that does not require self-control.
Because there exists an equilibrium between glucose in the blood-
stream and the brain (Lund-Anderson, 1979), low blood glucose
levels after an initial self-control task were then predicted to impair
performance on subsequent self-control tasks, insofar as available
quantities of glucose are too low for self-control to function
unimpaired, and possibly because the self starts to avoid effortful
activities in order to conserve its reduced remaining stock of
glucose (Muraven, Shmueli, & Burkley, 2006). Restoring glucose
to higher and optimal levels should replenish the ability to exert
The first step in the present investigation was to show that acts
of self-control reduce the level of glucose in the bloodstream. In
Studies 1 and 2, we examined whether completing a task that
required self-control, as compared with a task that did not require
self-control, would cause a drop in levels of glucose in the blood.
Next, we sought to link reduced glucose to the behavioral signs of
prior self-control exertion. In Studies 3– 6, we tested the hypoth-
esis that low glucose after an initial self-control task would be
associated with greater self-control impairments, in the form of
poorer performance on a subsequent self-control task. Having
established that glucose correlated with greater impairments to
self-control, we then turned to showing a causal role for glucose.
More precisely, we aimed to show that experimental manipulations
of glucose levels (administration of glucose drinks) would coun-
teract the effects of prior exertions of self-control (see Studies
7–9). We predicted that participants who performed the initial
self-control task would exhibit the typical pattern of performing
worse than others on the second task but that receiving a glucose
drink would reduce this effect.
Two of the major goals of this investigation were (a) to establish
that blood glucose levels are reduced from before to after perform-
ing an initial self-control task and (b) to show that low levels of
glucose after a first self-control task predict behavioral deficits on
a second self-control task. In theory, we may have been able to do
both of these in the same study. In practice, methodological
complications rendered the two goals somewhat incompatible.
Fasting facilitates (a) but interferes with (b). If participants have
eaten recently, then some glucose may be entering the bloodstream
at unpredictable intervals (possibly even when we might be as-
sessing their glucose levels after the first self-control task), so the
level may be rising when our hypothesis would predict a decrease,
even if our hypothesis were completely correct. Hence, in Studies
1 and 2, we had participants fast prior to the experiment so their
blood glucose would be stable apart from the impact of our
manipulations. However, there is some evidence that hungry par-
GAILLIOT ET AL.
ticipants already perform poorly on self-control tasks,
made it much harder to find behavioral effects of a laboratory
manipulation of self-control depletion. Therefore, we did not re-
quire fasting in Studies 3– 6 in order to obtain better behavioral
Study 1 provided an initial test of the hypothesis that exerting
self-control uses up a relatively large amount of glucose. Partici-
pants completed a task that either did or did not require self-
control. At the end of this task, we assessed glucose levels. We
predicted that the self-control task would diminish glucose relative
to the task that did not require self-control.
Participants. Participants were 110 undergraduates (69
women and 1 unreported) enrolled in an introductory psychology
course. Seven participants indicated having medical conditions
related to glucose (e.g., diabetes), and so their data were discarded
from all analyses. This left a final sample of 103 participants (64
women and 1 unreported). Participants in this and all subsequent
studies received credit toward fulfilling a course requirement.
Participants were instructed not to eat for 3 hr prior to arriving
at the experiment. Glucose levels fluctuate regularly throughout
the day as a result of eating (and for rather long and variable
intervals after eating). Requiring participants to refrain from eating
thus allowed glucose levels to stabilize, which greatly reduced
extraneous variance in glucose measurement.
Procedure. Participants were run individually and were told
the study was investigating physiological measures and task per-
formance. First, the experimenter assessed baseline blood glucose
levels. Blood samples were taken with single-use blood sampling
lancets. Blood glucose levels were measured (mg/dL) using an
Accu-Chek compact meter.
Next, participants completed the video task that served as the
manipulation of self-control exertion. Participants watched a 6-min
video (without sound) of a woman talking (modified from Gilbert,
Krull, & Pelham, 1988). In the bottom corner of the screen,
common one-syllable words (e.g., hair, hat, pulse) appeared indi-
vidually for 10 s. Participants randomly assigned to the attention
control condition were instructed to focus their attention only on
the woman’s face and to refrain from looking at the words. If they
happened to look at the words, then they were to refocus their
attention on the woman as quickly as possible. Participants ran-
domly assigned to the watch normally condition were instructed to
watch the video as they would normally. After completing this
task, the experimenter assessed blood glucose levels a second time.
In Studies 1– 8, we assessed current mood valence and arousal
after participants completed the initial self-control task or prior to
any subsequent self-control task. Participants completed either the
Brief Mood Introspection Scale (Mayer & Gaschke, 1988), the
Positive and Negative Affect Schedule (Watson, Clark, & Telle-
gen, 1988), or the UWIST Mood Adjective Checklist (Matthews,
Jones, & Chamberlain, 1990). None of the mood measures (in-
cluding the Valence and Arousal subscales) had a significant effect
on the dependent measure in any study. To save space, we do not
report all these null findings individually. In each study, partici-
pants were thanked and debriefed at the end of the experiment.
Results and Discussion
Analyses confirmed that the self-control task used up a rela-
tively large amount of glucose. A 2 ⫻ 2 mixed model analysis of
variance (ANOVA) indicated a significant interaction between
attention control condition and time of measurement, F(1, 100) ⫽
6.08, p ⬍ .05. Among participants in the attention control condi-
tion, glucose was lower after the video task (M ⫽ 101.22, SD ⫽
18.34) than before (M ⫽ 107.10, SD ⫽ 21.02), t(50) ⫽⫺2.57, p ⬍
.05. Among participants in the watch normally condition, glucose
levels did not differ from before (M ⫽ 102.24, SD ⫽ 21.20) to
after (M ⫽ 103.24, SD ⫽ 18.71) the video task (t ⬍ 1). Thus, all
participants watched the same video, but glucose levels dropped
only among participants who had to exert self-control while watch-
These results are consistent with the notion that exerting self-
control uses up a relatively large amount of glucose. Blood glucose
levels were lower after participants regulated their attention while
watching a video—lower than their own levels before the video
and lower than those of participants who had just watched the
same video without controlling attention.
The purpose of Study 2 was to provide additional evidence that
self-control uses enough glucose to partially deplete the supply in
the bloodstream. Suppressing stereotypes or prejudice, such as
during the course of an interracial interaction, has been shown to
deplete self-control strength (Gordijn et al., 2004; Richeson &
Shelton, 2003; Richeson & Trawalter, 2005; Richeson et al., 2005;
see also Richeson et al., 2003). Therefore, we examined in Study
2 whether an interracial interaction would deplete glucose.
Restraining prejudice may be more difficult for some people
than for others. Therefore, Study 2 measured individual differ-
ences in internal motivation to respond without prejudice (Plant &
In support of the notion that hunger obscures the effects of self-control
exertion on subsequent behavior, a pilot study found that hunger impaired
future self-control to the same extent as did prior self-control exertion.
Specifically, participants (N ⫽ 27) were instructed to arrive hungry to the
laboratory and completed the same video task used in Study 1. They were
randomly assigned to one of three conditions: hunger/attention control,
hunger/watch normally, or no hunger/watch normally. Participants in the
no-hunger/watch normally condition received an orange juice drink and a
muffin bar to eat after the video task. Participants in the other two
conditions proceeded immediately to the next part of the experiment.
Next, participants completed the Stroop task for 3 min. The number of
trials completed (i.e., speed) on this task served as the dependent measure
of Stroop performance.
A one-way ANOVA indicated a main effect of condition, F(1, 24) ⫽
4.22, p ⬍ .05. Participants in the no-hunger/watch normally condition
(M ⫽ 198.88, SD ⫽ 8.72) completed more Stroop trials than participants
in the hunger/watch normally (M ⫽ 168.60, SD ⫽ 7.80) and no-hunger/
attention control (M ⫽ 168.78, SD ⫽ 8.22) conditions. The difference
between the hunger/watch normally and no-hunger/attention control con-
ditions was not significant (t ⬍ 1). Thus, the Stroop measure suggested that
self-regulatory performance was impaired to about the same degree by
prior self-regulation as by hunger. These results are also consistent with the
hypothesis that poor self-control is caused by low glucose, insofar as
hunger is associated with low glucose (e.g., Cox, Eickhoff, Gonder-
Frederick, & Clarke, 1993).
SELF-CONTROL AND GLUCOSE
Devine, 1998), which were predicted to moderate the consumption
of glucose during an interracial interaction. Internal motivation
reflects the desire to respond without prejudice because of the
personal importance of endorsing nonprejudiced beliefs. Internally
motivated people are less likely than others to respond with racial
bias across all sorts of situations. For them, suppressing prejudice
should be a well practiced and hence presumably automatic way of
acting, and therefore it should be easier for them to avoid express-
ing bias in our laboratory (Plant, 2004). In contrast, people who are
low in internal motivation to avoid prejudice prefer to avoid
interracial interactions than to struggle to suppress their views
while in them (Plant, 2004), and so when they find themselves
having to suppress prejudice in a sensitive situation, they may have
to exert considerable self-control to speak and act appropriately.
For them, unlike the others, stifling prejudicial thoughts may
require effortful self-control rather than falling back on a habitual
pattern. We therefore expected that an interracial interaction might
reduce glucose primarily among participants low in internal mo-
tivation to respond without prejudice.
Participants. The final sample included 38 White college un-
dergraduates (29 women). We excluded from all analyses 1 par-
ticipant because he was diabetic. Participants were randomly as-
signed to interact with either a Black or a White experimenter.
Participants were instructed not to eat for 3 hr prior to arriving at
Procedure. Upon arrival at the laboratory, participants were
greeted by a Hispanic female experimenter. Participants were told
that the study was examining factors related to different tasks and
First, we assessed initial glucose levels. Next, participants had a
5-min conversation with either a Black or a White female exper-
imenter. After introducing themselves, participants were asked to
state their opinions on affirmative action and criminal profiling (in
counterbalanced order) and were given 2 min to discuss each topic.
These topics were chosen because they involve unequal treatment
of individuals on the basis of race and would therefore be likely to
activate racial stereotypes.
Glucose levels then were assessed a second time. Last, partici-
pants completed a questionnaire on which they indicated the extent
to which they felt like they exerted effort during the interaction so
as to avoid saying anything negative, using a 9-point scale ranging
from 1 (very little)to9(a lot). The questionnaire also contained
the Internal Motivation to Respond Without Prejudice scale (IMS;
Plant & Devine, 1998), which contains five items (e.g., “Because
of my personal values, I believe that using stereotypes about
Blacks is wrong”).
Results and Discussion
Glucose levels after interaction. The hypothesis was that dis-
cussing racially sensitive material with a member of another race
would require self-control and therefore deplete glucose—mainly
for people who do not habitually stifle prejudicial thoughts and
feelings (i.e., for people low in IMS). A regression analysis that
regressed standardized IMS, condition (same-race vs. interracial
interaction), and their interaction upon postinteraction glucose
levels yielded the predicted significant interaction between IMS
and experimental condition (b ⫽ 4.41), t(1, 33) ⫽ 2.20, p ⬍ .05
(see Figure 1). The postinteraction glucose levels were controlled
for baseline (preinteraction) glucose levels, which did not differ by
condition or IMS (ts ⬍ 1). To interpret the interaction, we assessed
the simple effect of condition among participants who were rela-
tively high versus low in IMS (one standard deviation above and
below the mean on IMS, respectively; Aiken & West, 1991).
Results indicated that the effect of condition was significant for
low-IMS participants (b ⫽⫺3.28), t(1, 33) ⫽⫺2.33, p ⬍ .05, but
not for high-IMS participants (b ⫽ 1.16), t(1, 33) ⫽ 0.92, p ⫽ .36.
Thus, discussing a sensitive topic with a member of a different race
used up a significant amount of glucose among people with low
IMS. Glucose was not depleted in people who discussed the same
topics with members of their own race or among people who are
dispositionally motivated to stifle prejudicial thoughts and feel-
ings. This pattern is consistent with the view that acts of self-
control deplete blood glucose.
Effort. Interracial interactions require self-control because one
often exerts effort to avoid expressing negative attitudes or opin-
ions (Richeson & Trawalter, 2005). In support of this, in the
interracial interaction condition, IMS scores predicted self-
reported effort, r(21) ⫽⫺.48, p ⬍ .05; effort predicted postinter-
action glucose levels (controlling for preinteraction glucose levels)
r(18) ⫽⫺.54, p ⬍ .05; and a Sobel test for mediation pointed
toward the conclusion that effort mediated the effect of IMS on
postinteraction glucose levels, although it fell short of two-tailed
significance (z ⫽ 1.79, p ⫽ .07). Glucose dropped primarily
among low-IMS participants because they found the interracial
interaction more effortful than did high-IMS participants. Effort
did not appear to mediate the effect of IMS in the same-race
condition (z ⫽⫺.20, ns), and the preconditions for mediation were
not met either. These findings suggest that the self-regulatory
effort needed to avoid negative responses during an interracial
interaction depletes an energy source on which self-control relies
Internal Motivation to Respond Without Prejudice
Figure 1. Glucose levels after an interaction (controlling for glucose
levels before the interaction) as a function of interaction condition and
internal motivation to respond without prejudice (see Study 2).
GAILLIOT ET AL.
With Study 3, we began testing the hypothesis that low levels of
blood glucose following a self-control task would predict poor
performance on behavioral measures of self-control. In Studies
3– 6, all participants were subjected to an initial depleting task, and
we assumed that these would be more depleting to some than to
others. The prediction for Studies 3– 6 was that the participants
with the lowest levels of blood glucose would perform worst at
self-control. After measuring blood glucose, we used the attention
control task from Study 1 to create depletion. Following a second
glucose measurement, participants completed the Stroop task,
which is one of the most frequently used measures of self-control
(e.g., Richeson & Shelton, 2003; von Hippel & Gonsalkorale,
2005). The Stroop task requires the participant to override an
incipient response (i.e., to read aloud the name of the word) in
order to say instead the color in which the word is printed, and in
that sense it requires self-regulation. Lower blood glucose should
impair Stroop performance in the sense of causing the person to
take longer to get the right answer and in terms of making more
errors along the way.
Participants. Participants were 16 college undergraduates (12
women). We excluded from this sample 1 additional participant
because of equipment malfunction.
Procedure. Participants were told that the study was examin-
ing the relationship between physiological factors and task perfor-
mance. First, baseline blood glucose levels were assessed. Next,
participants completed the attention control task used in Study 1.
All participants were instructed to exert self-control by refraining
from looking at the words in the video. At the end of the task,
glucose levels were assessed a second time.
Last, participants completed the Stroop task. They were shown
words (i.e., red, green, and blue) that appeared in a font color (i.e.,
red, green, or blue) that diverged from the meaning of the word
(e.g., red appeared in blue ink). Participants completed 80 trials for
which they were to state aloud the color ink that each word
appeared in and to refrain from reading the word. The amount of
time participants took to complete the Stroop task (i.e., speed) and
the number of errors (i.e., accuracy) constituted the dependent
measures of Stroop performance. The assumption behind that
measure is that effective self-control enables the person to override
the initial response to say the word so as to be able to state the
color of the ink. When self-control is weak or ineffective, the
person takes longer to override the initial (wrong) response, or the
person makes more errors.
Results and Discussion
Glucose levels at the start of the experiment did not predict
Stroop performance (i.e., neither time to complete the Stroop nor
errors; ⫺.20 ⬍ rs ⱕ ⫺.13, ps ⬎ .48). In contrast, lower glucose
after having watched the video was associated with poorer Stroop
performance (i.e., taking more time to complete the Stroop task),
r(14) ⫽⫺.62, p ⬍ .05, and this relationship remained significant
after controlling for baseline glucose levels, r(12) ⫽⫺.66, p ⬍
.05. Errors on the Stroop task showed a similar though nonsignif-
icant pattern, such that lower glucose was associated with making
more errors, r(14) ⫽⫺.23, ns.
These findings are consistent with the idea that self-control
impairments stemming from a prior act of self-control are attrib-
utable to low glucose. Participants with less glucose after an initial
self-control task performed worse at a subsequent self-control task.
One alternative explanation for these findings is that low glu-
cose simply made participants slow to respond rather than impair-
ing their self-control. In Studies 4 – 6, we addressed this limitation
by using a dependent measure that was not directly related to speed
Studies 4 – 6
Studies 4 – 6 provided conceptual replications of Study 3. In
each study, we had participants engage in a self-control task
designed to impair their self-control afterwards. Participants either
controlled their attention (see Study 4), completed the Stroop task
(see Study 5), or regulated their emotions (see Study 6). Regarding
the last of these, emotion regulation is a common, well-recognized
form of self-control and likewise has been shown to impair later
attempts at self-control (e.g., Muraven, Tice, & Baumeister, 1998).
In each study, blood glucose was measured before and after the
initial self-control task.
Last, participants were assigned to perform a figure-tracing task.
Unbeknownst to participants, the task was unsolvable, and we
timed how long they persisted. Persistence at the unsolvable task
requires self-control because the discouraging, frustrating failures
at the task give rise to impulses to quit, which the person must
override in order to continue striving on the task (see Baumeister
et al., 1998). We predicted that low glucose after the initial
self-control task would be associated with less effortful persis-
Participants. Study 4 had a final sample of 12 participants (10
women). Two additional participants were excluded, 1 because a
second glucose reading could not be obtained and the other be-
cause the participant erroneously claimed to have solved the
figure-tracing task. Study 5 had a final sample of 23 participants
(15 women). Four additional participants were excluded, 1 because
a second glucose reading could not be obtained and 3 others who
either suspected the figure-tracing task was unsolvable or errone-
ously claimed to have solved it. The final sample in Study 6 was
17 participants (7 women). Two additional participants were ex-
cluded from analyses. One of the excluded participants did not
complete the emotion regulation task, and the other was an outlier
(three standard deviations above the mean) on the dependent
Procedure. Participants were told that the study was investi-
gating the relationship between physiological factors and task
performance. Participants first provided a blood sample to assess
glucose level. Next, they either completed the same attention
control task used in the previous studies (see Study 4), the Stroop
task used in Study 3 for 4 min (see Study 5), or an emotion
regulation task (see Study 6). For the emotion regulation task,
participants watched a 2-min video clip of animals in a slaughter-
house and a 2-min video clip of the comedy show Jay Leno.
SELF-CONTROL AND GLUCOSE
Participants in Study 6 were instructed to suppress any emotional
feelings or reactions they had while watching the clips. After their
respective task, participants provided a second sample of blood to
Last, all participants completed the figure-tracing task, adapted
from Glass, Singer, and Friedman (1969) and Feather (1961).
Participants were given a figure and asked to trace the lines of the
figure without lifting their pen. Participants were told that they
would be given five figures in total and were to solve each of them
as quickly as possible. They were to notify the experimenter as
soon as they solved each figure or decided that they could not and
wanted to give up. The first figure given to participants, however,
was unsolvable. Participants were surreptitiously timed. Partici-
pants were given a maximum of 20 min for the task, at which time
any participants still working were asked to stop.
Results and Discussion
Glucose levels at the start of each experiment were not signif-
icantly related to persistence (all rs ⬍ .27, ps ⬎ .24). Low glucose
after the initial self-control task, however, was associated with
persisting less on the figure-tracing task in all three studies; Study
4 r(10) ⫽ .56, p ⬍ .05; Study 5 r(21) ⫽ .45, p ⬍ .05; Study 6
r(15) ⫽ .43, p ⬍ .05. (Because these studies were replications, we
used one-tailed tests to examine our hypothesis.) Furthermore, the
correlation between glucose levels after the initial task and persis-
tence remained significant when controlling for glucose levels at
the start of each experiment ( ps ⬍ .05). The link between posttest
glucose levels and persistence, therefore, did not appear attribut-
able to initial glucose levels.
Although the samples in these three studies were small, the
magnitude of the correlations was consistently quite robust. The
consistency of findings across the three different samples and
methods lends confidence to the conclusion that low levels of
glucose are associated with decrements in self-control.
Moreover, these results help to refute the alternative explanation
for Study 3 that low glucose simply makes people generally slower
at tasks rather than impairing their self-control. If low glucose only
made people slow at various tasks, then one might have expected
that low glucose would have predicted greater persistence (i.e.,
being slower to give up).
The results thus far support the hypothesis that impairments to
self-control following an initial self-control task are linked to low
glucose in the bloodstream. Studies 1 and 2 showed that exerting
self-control reduced blood glucose levels. Studies 3– 6 showed that
lower glucose after an initial exertion of self-control predicted
poorer self-control on the second task (e.g., quitting sooner on the
frustrating puzzle). The remaining studies sought to provide evi-
dence of a causal relationship between glucose and self-control by
using direct manipulations of blood glucose, namely, drinking a
glucose beverage. Glucose is absorbed into the bloodstream at a
rate of 30 calories per minute and after about 10 min can be
metabolized to the brain (Donohoe & Benton, 1999). Hence,
glucose drinks are a viable means of increasing the amount of
glucose available for self-regulatory tasks, provided that one al-
lows some time for the glucose to reach the bloodstream. Partic-
ipants in Study 7 first completed either a task that required self-
control (attention control as in the previous studies) or a task that
did not require self-control.
Participants then drank Kool-Aid lemonade that had been sweet-
ened either with sugar (and hence glucose) or Splenda (a good-
tasting sugar substitute that does not increase blood glucose). The
sugar lemonade shake should restore glucose and therefore would
be expected to replenish self-control strength after the initial
Hence, we predicted that participants who controlled their at-
tention during the video task would perform worse on a subsequent
Stroop task than participants who did not control their attention,
but that a glucose drink would attenuate this effect.
Participants. Participants were 62 college undergraduates (49
women and 1 unreported). We excluded from this final sample 1
additional participant who chose not to consume the drink. Partic-
ipants were randomly assigned to depletion and glucose condi-
tions. They were randomly assigned among the conditions. The
two variables (attention control and lemonade content) were ma-
Procedure. Participants were told that the study was examin-
ing factors related to different tasks and food. Participants first
completed 20 Stroop trials (as in Study 3) as a baseline measure of
Stroop ability. They then were administered the attention control
manipulation used in Study 1, with half simply watching the video
and the other half being instructed to keep their attention focused
on the woman and not on the words while watching it. Study 7
added a manipulation check. After watching the tape, participants
rated how often they had looked at the woman and the words,
Next, participants were given 14 ounces of lemonade sweetened
with either sugar (glucose condition) or a sugar substitute (placebo
condition). The glucose drink contained approximately 140 calo-
ries, whereas the placebo contained 0 calories. Participants and the
experimenter were blind to condition. Participants consumed the
glucose drink and then completed three measures of liking for the
drink (e.g., “How pleasant was it for you while drinking the
beverage?”; ␣⫽.63) that were embedded among other filler
measures intended to bolster the cover story regarding the taste and
appearance of the drink. Participants then completed filler ques-
tionnaires for 12 min to allow the glucose from the drink (if they
had any) to be metabolized (Donohoe & Benton, 1999). Last,
participants completed 80 Stroop trials separated into four blocks.
Speed and errors on the Stroop task constituted the dependent
measures of self-control performance.
Results and Discussion
Manipulation check. Participants appeared to be successful in
following the video task instructions. Participants in the attention
control condition reported having looked at the words (M ⫽ 2.03,
SD ⫽ 0.68) less often and at the woman (M ⫽ 6.72, SD ⫽ 0.45)
more often than participants in the watch normally condition
reported having looked at the words (M ⫽ 4.76, SD ⫽ 1.95) and
woman (M ⫽ 5.12, SD ⫽ 1.44), respectively (ts ⬎ 4.30, ps ⬍
GAILLIOT ET AL.
Stroop performance. A 2 (self-control condition) ⫻ 2 (glucose
condition) between-subjects analysis of covariance (ANCOVA) on
the number of errors made on the final Stroop task (controlling for
errors at baseline) indicated that the main effects of self-control
condition and glucose condition were not significant (Fs ⬍ 2.23,
ps ⬎ .14). Their interaction, however, was significant, F(1, 57) ⫽
5.04, p ⬍ .05, and consistent with predictions (see Figure 2 for
adjusted means). In the placebo condition, attention control par-
ticipants made more errors than watch normally participants, F(1,
58) ⫽ 5.14, p ⬍ .05, d ⫽ 0.73. This, however, was not the case in
the glucose condition (F ⫽ 1.12, ns). A glucose drink thus elim-
inated the tendency for an initial self-control task to impair Stroop
performance, consistent with the hypothesis that glucose replen-
ishes what has been depleted.
A 2 (self-control condition) ⫻ 2 (glucose condition) ANOVA
on speed on the Stroop indicated no significant effects (Fs ⬍ 1.76,
ps ⱖ .19). To be sure, there was no speed–accuracy trade-off, such
as if attention control participants in the placebo condition made
more errors because they responded faster. The above interaction
between self-control condition and glucose condition on the num-
ber of errors on the Stroop remained significant even when con-
trolling for speed, F(1, 56) ⫽ 6.18, p ⬍ .05.
This pattern of results diverges from Study 3, in which speed on
the Stroop rather than errors was more sensitive to glucose. This
difference is probably attributable to the different methods used in
Studies 3 and 8. In Study 3, participants completed 80 Stroop trials
without a break, and so performing quickly (rather than making
errors) was probably more onerous and hence more sensitive to
glucose. In Study 7, the Stroop trials were divided into four blocks.
The delay between blocks probably allowed participants to main-
tain their speed, and so avoiding errors was probably more oner-
Liking for the drinks. The effects of depletion and glucose did
not appear to be the result of liking for the drinks. In particular, the
interaction between self-control condition and glucose condition
on the number of errors on the Stroop approached significance
when controlling for liking (F ⫽ 3.47, p ⫽ .07).
Study 8 sought additional evidence that increasing glucose
levels can counteract the detrimental effects of prior efforts at
self-control. More precisely, in Study 8, we examined whether
glucose would improve self-control among people who have coped
with thoughts and fears of dying. The thought of death can be
aversive and threatening, and people therefore seem motivated to
avoid thinking about death (e.g., Aries, 1981; Becker, 1973). Past
research has indicated that avoiding the thought of death is effort-
ful and demanding (Arndt, Greenberg, Solomon, Pyszczynski, &
Simon, 1997; Greenberg, Arndt, Schimel, Pyszczynski, & So-
lomon, 2001) and requires self-control (Gailliot et al., 2006).
Likewise, thinking about death (i.e., mortality salience) has been
found to impair subsequent performance on tasks that require
self-control (e.g., the Stroop task), presumably because the self-
control expended to cope with the thought of death depletes the
energy source on which self-control relies (Gailliot et al., 2006).
To assess whether glucose underlies self-control, we examined
whether a glucose drink would benefit self-control after a mortality
First, participants received either a glucose drink or a placebo.
We administered it first on the assumption that the glucose would
start to reach the bloodstream around the time participants com-
pleted the next task, at which point those who had coped with
thoughts of death would otherwise be suffering from depleted
glucose. They then wrote about either death or a control topic
(dental pain). For the subsequent self-control task, participants
were given a sheet filled with word fragments and were asked to
solve them all. The word fragments were relatively easy, and with
enough persistence (self-control), one could eventually solve each
one by substituting into the word fragment different letters in the
alphabet. Consistent with the idea that coping with thoughts of
death impairs self-control afterwards, Gailliot et al. (2006) found
that mortality salience caused participants to leave more word
fragments unsolved, suggesting that mortality salience activated
thoughts of death, and the person had to use his or her resources in
order to defend against that threat, resulting in less resources
available for making oneself persist on the word fragment task.
The present experiment was designed to extend this finding by
showing that a glucose drink attenuates this effect.
Participants. Participants were 73 undergraduates (51
women). One additional participant was excluded from all analy-
ses because this participant declined the drink. Participants were
randomly assigned to mortality salience versus dental pain and
glucose drink versus placebo conditions.
Procedure. Participants were told that the study was investi-
gating food and personality. First, participants consumed either a
glucose drink or a placebo and completed measures of taste and
liking for the drink. The drinks and measures were the same as
those used in Study 7.
Next, participants in the mortality salience condition were asked
to describe the emotions that the thought of their own death
Errors on Stroop
Figure 2. Stroop performance as a function of self-control and glucose
conditions (see Study 7).
SELF-CONTROL AND GLUCOSE
aroused in them and to write about what would happen to their
bodies as they physically die (see Rosenblatt, Greenberg, So-
lomon, Pyszczynski, & Lyon, 1989). Participants in the dental pain
condition answered parallel questions about dental pain.
Participants next worked on a crossword puzzle and completed
a filler questionnaire. These tasks took approximately 6 min. This
delay was intended to allow participants to cope with any thoughts
of death they might have had and hence consume the energy on
which self-control relies (see Gailliot et al., 2006).
Participants then received a sheet that contained 20 word frag-
ments (e.g., —WER) and were asked to solve all of them. The
number of word fragments left unsolved constituted the dependent
measure of self-control. The word fragments were relatively easy
to solve, and with enough persistence (self-control), participants
could have solved them all. Therefore, we used the number of
unsolved fragments as the measure of poor self-control.
Last, participants completed the 33-item Marlowe-Crowne So-
cial Desirability Scale (Crowne & Marlowe, 1960), which allowed
us to assess whether any differences in the number of unsolved
word fragments may have been attributable to response bias rather
than to poor self-control. Participants then estimated the number of
calories they thought that their drink had contained, as a check of
whether there was any discernible difference in taste between the
glucose and placebo drinks that may have influenced persistence at
solving the word fragments.
Results and Discussion
Unsolved word fragments. We predicted that mortality sa-
lience would cause participants to leave more word fragments
unsolved but that a glucose drink would eliminate this effect, and
hence that mortality salience participants who consumed the pla-
cebo drink would leave more word fragments unsolved than any
other group of participants. A 2 (mortality salience condition) ⫻ 2
(glucose condition) ANOVA on the number of unsolved word
fragments produced no significant effects (Fs ⬍ 2.02, ns). To test
our specific prediction, which constituted a theoretical replication
of Study 7, we conducted a planned comparison, which confirmed
our prediction, F(1, 69) ⫽ 5.45, p ⬍ .05 (see Figure 3).
In the placebo condition, mortality salience participants left
unsolved more word fragments than did dental pain participants,
F(1, 69) ⫽ 3.91, p ⫽ .05, d ⫽ 0.33. Thus, mortality salience
impaired self-control. In the glucose drink condition, mortality
salience participants did not differ from dental pain participants in
the number of word fragments left unsolved (F ⬍ 1). A glucose
drink thus eliminated the detrimental effects of mortality salience
on self-control. These results suggest that the self-regulatory im-
pairments following mortality salience may be at least partly
attributable to low glucose.
Social desirability, calories, and ratings of liking. Analyses
indicated that the obtained pattern of results was likely not attrib-
utable to differential liking for the drinks, social desirability, or
estimations of the caloric content of the drinks. For instance, the
number of word fragments left unsolved was not significantly
related to any of these factors ( ps ⬎ .53; p ⫽ .21 for social
desirability), and the planned comparison reported above remained
significant or approached significance when controlling for each of
these factors ( ps ⱕ .06). Moreover, estimations of the caloric
content of the drinks did not differ between those who consumed
the glucose drink and those who consumed the placebo (F ⬍ 1).
In the final study in this investigation, we sought to apply the
link between glucose and self-control to an important form of
interpersonal behavior, namely, helping. Helping is oftentimes
effortful (see, e.g., Dawkins, 1976) and hence requires some
amount of self-control. Indeed, previous studies show that an
initial act of self-control reduces helpfulness later on (Gailliot,
Maner, DeWall, & Baumeister, 2005) as well as other forms of
prosocial behavior (Finkel & Campbell, 2001; Vohs, 2004).
Study 9 was a field experiment. We had participants exert
self-control by taking an actual examination in their psychology
course. Previous work has confirmed that taking exams or engag-
ing in logical reasoning both require self-control (Gailliot et al.,
2006; Schmeichel, Vohs, & Baumeister, 2003), and other work has
shown that taking a difficult exam reduces glucose (Hall & Brown,
1979). Study 9 did not have a no-self-control (control) condition,
insofar as everyone took the exam, but we were able to assess
individual differences in the extent of self-control exertion. Spe-
cifically, the experimenter recorded the sequential order in which
participants finished the exam. Gailliot et al. (2005) found that
taking longer to complete an exam was associated with higher
scores on a questionnaire measure of impaired self-control (see
Twenge, Muraven, & Tice, 2004).
After finishing the exam, participants were randomly assigned
to receive a glass of lemonade that was either high or low in
glucose. Then their willingness to help others was measured with
two items: one involved donating money to charity, and the other
involved helping a fellow student with a housing problem. If
glucose underlies self-control and self-control facilitates helping,
then participants who consume the glucose rather than the placebo
drink should be more willing to help.
Glucose Drink Condition
Unsolved Word Fragments
Figure 3. Effortful persistence as a function of mortality salience and
glucose conditions (see Study 8).
GAILLIOT ET AL.
Participants. Participants were 18 undergraduates (10 men).
We excluded from all analyses 1 additional participant who was an
outlier (nearly three standard deviations above the grand mean) on
the dependent measure. Participants were randomly assigned to a
glucose or placebo drink condition.
Procedure. Participants were told the study was investigating
the relationship between their attitudes and factors related to food.
First, participants completed a 26-item multiple-choice exam that
was an actual examination that counted toward the course grade.
The experimenter recorded the order in which participants finished
(e.g., first, second, third), as a measure of the duration of self-
The actual experiment began after participants finished their
exam. First, when participants turned in their exam, they were
handed either a glucose or placebo drink. Participants and the
experimenter were blind to condition. They then completed mea-
sures of liking for the drink (␣⫽.63) that were embedded among
other filler measures regarding the taste and appearance of the
drink intended to bolster the cover story. Participants then sat
quietly for approximately 14 min to allow the glucose to be
metabolized. (The reason for the delay was not explained to
Last, participants responded to two hypothetical scenarios that
assessed their willingness to help strangers. For the first scenario,
participants were to indicate how much money they would donate
to a charity. For the second scenario, participants were told to
imagine that a stranger from their class had been evicted from his
or her apartment. Participants were to indicate, on a 7-point scale,
what they would do to help him or her, with higher points on the
scale representing more labor-intensive forms of helping. Specif-
ically, participants’ options were to: (1) do nothing, (2) give him or
her an apartment guide, (3) help him or her find a new place to live
by driving him or her around for a few hours, (4) offer to have him
or her come to stay with the participant for a couple of days
(provided the participant had space), (5) offer to have him or her
come stay with the participant for a week (provided the participant
had space), (6) offer to have him or her come stay with the
participant until he or she found a new place (provided the partic-
ipant had space), or (7) offer to have him or her come live with the
participant rent free (provided the participant had space). Re-
sponses to the two scenarios correlated moderately strong, r(16) ⫽
.63, p ⬍ .05, and so they were standardized and averaged to form
the final dependent measure of helping that ranged from ⫺.90 to
Results and Discussion
Likelihood of helping. Participants in the glucose condition
(M ⫽ 0.48, SD ⫽ 0.95) indicated that they were willing to give
significantly more help than those in the placebo condition (M ⫽
⫺0.61, SD ⫽ 0.27), t(16) ⫽ 3.13, p ⬍ .01, d ⫽ 1.57. Thus, after
exerting self-control, consuming a glucose drink increased the
likelihood and extent of helping.
Exam effort and likelihood of helping. We used the order in
which participants completed the exam as a rough index of self-
regulatory effort expended on the exam. Exam completion order
does not appear to be related to intelligence or exam performance
(e.g., Bridges, 1985), and it therefore seemed reasonable that
taking longer on the exam reflected a greater expenditure of effort.
In other words, independent of ability, the students who spent
longer on the exam would have exerted effort for a longer period
of time and therefore were likely to have used more self-control
than the others.
Among those participants who received no glucose, likelihood
of helping correlated negatively and quite strongly with exam
completion order, r(6) ⫽⫺.83, p ⫽ .01. Those who took longer
on the exam offered less help. This fits the hypothesis that working
long and hard on the exam depletes some resource that could
otherwise be used to increase helpfulness. In the glucose condition,
however, the relationship between helping and completion order
was not significant, r(8) ⫽ .29, ns, and if anything, the relationship
was in the opposite direction. Moreover, the strength of this
relationship was significantly different than in the placebo condi-
tion (z ⫽ 2.52, p ⬍ .05). To the extent that taking longer on the
exam reflected expending more self-control effort, then it appears
that glucose eliminated the effect of self-control exertion (taking
an exam) on helping.
Ratings of liking. Ratings of liking for the drinks did not differ
between the glucose and placebo conditions (t ⬍ 1). Furthermore,
these ratings were not significantly related to helping (r ⬍ .22, ns).
It therefore does not appear that the difference in helping between
the two conditions was attributable to liking for the drinks. The
most parsimonious conclusion seems to be that glucose increased
helping by counteracting the effects of prior self-control exertion.
Because of the small sample size, however, we do present these
results with some caution.
Self-control is one of the most important and central processes
in personality and self-theory. Past work has shown that self-
control operates as if a limited resource is consumed or expended
in the process. The nature of such a resource, however, has
remained a matter of conjecture and metaphor. The goal of this
investigation was to flesh out that model by moving from meta-
phor to at least one plausible physiological process. Specifically,
we investigated the hypothesis that some patterns of poor self-
control are attributable to drops in glucose.
Three main sets of findings supported the hypothesis that self-
control depends on glucose. First, measurements of blood glucose
showed significant drops following acts of self-control (e.g., dur-
ing an interracial interaction), primarily among participants who
worked hardest. Second, low glucose after an initial self-control
task (e.g., emotion regulation) was linked to poor self-control on a
subsequent task. Third, experimental manipulations of glucose
reduced or eliminated self-control decrements stemming from an
initial self-control task.
By testing hypotheses using a variety of methods, the present
findings sought to rule out a variety of alternate explanations. The
improvements and impairments of self-control were not linked to
mood, arousal, or other emotional states. Neither the taste nor the
pleasure of the glucose snacks appeared to be a factor. The effects
were not limited to any particular sort of task or measure. In fact,
the procedures of these nine studies were deliberately set up to
encompass a variety of behaviors, including well-validated and
artificial laboratory measures (e.g., the Stroop task) and complex
social behaviors (e.g., willingness to help a stranger).
SELF-CONTROL AND GLUCOSE
Implications, Qualifications, and Future Directions
The present results suggest that self-control involves blood
glucose and that the effects of an initial self-control task stem
partly from reduced levels of glucose. As our title indicates,
willpower is more than a metaphor. The nature of the resource that
gets depleted has been a puzzle since the earliest findings on
self-control depletion emerged (Baumeister et al., 1998). Some
researchers have thought that self-control depletion was primarily
a motivational deficit. This view may have been (somewhat un-
wittingly) encouraged by Muraven and Slessareva’s (2003) dem-
onstration that offering motivational incentives could counteract
depletion effects. However, the present results suggest that self-
control depletion involves a shortage of fuel for brain activities. As
Muraven and Slessareva showed, behavior during that state can
still respond to motivational incentives, but the state of self-control
depletion is not primarily a deficit in motivation. By analogy,
money may induce a person to continue working despite physical
exhaustion, but that does not mean that physical exhaustion is
essentially the result of a lack of money. Insofar as self-control
depletion is a deficit in fuel, it reflects an impairment in capacity
to perform not in desire (motivation) to perform, even though
stimulating desire can sometimes offset a reduced capacity.
Another implication of the present work is that individual dif-
ferences in glucose processing may contribute to different out-
comes in self-control. Consistent with that suggestion, juvenile
delinquents lack self-control (Gottfredson & Hirschi, 1990; Pratt &
Cullen, 2000), and there is some evidence that juvenile delinquents
process and tolerate glucose less effectively than comparable ad-
olescents who do not have legal troubles (Gans et al., 1990;
Matykiewicz, La Grange, Vance, Mu, & Reyes, 1997).
There is no reason to assume, however, that self-control is
unique in its reliance on blood glucose. As far as we understand,
all brain processes use some glucose, though some use more than
others and are therefore more susceptible to fluctuations of the
supply in the bloodstream. We would certainly expect that the
patterns we have observed for self-control would generalize to
other executive functioning or controlled processing. Evidence is
accumulating that these other executive processes rely on the same
energy source as self-control. Engaging in effortful choice and
decision making has been shown to lead to impairments in subse-
quent self-control (Baumeister et al., 1998; Vohs, Baumeister,
Twenge, Schmeichel, Tice, & Crocker, 2004). Along the same
lines, controlled thinking processes, such as logical reasoning and
extrapolation, are impaired among people who have performed
recent acts of self-control (Schmeichel et al., 2003). We regard the
self as the controller of controlled processes, and so its effortful
activities are likely to involve glucose dynamics just like self-
control. It is also plausible that there are some additional brain
activities that would not involve the self but would still depend on
the glucose in the same way, though we do not know just what
those might be. For now, the most reasonable and parsimonious
formulation would seem to be that the agentic, effortful activities
of the self use significant amounts of glucose and are impaired
when the supply of glucose has been depleted.
Still, it seems likely that relatively few psychological processes
are as expensive as self-control in terms of requiring large amounts
of glucose. There are two reasons for thinking that self-control is
unusual, even if not unique. First, what it accomplishes is rather
advanced and difficult. We assume that the psychological system
evolved to want and do things. To interrupt and override these
well-established responses, especially after they are already in
process, seems quite difficult. In plain terms, an animal may have
evolved to feel and act on strong desires, such as for food or sex,
and so self-control requires an inner mechanism strong enough to
counter those powerful responses. Second, the widespread occur-
rence of self-control failures is evidence that self-control is not
easy, and high metabolic cost would be one likely explanation for
Not all self-control processes will be equally expensive, either.
The present results showed ample evidence of individual differ-
ences. Studies 2–6 specifically showed that the same task depleted
some people more than others. These differences may reflect
differences in personality and values (as in Study 2) or in physi-
ological capabilities and response patterns. We also assume, but
have not shown, that different self-control tasks would be differ-
entially depleting even to the same person in the same circum-
stances. At its core, self-regulatory change involves overriding one
response in order to enable a different response. The stronger the
initial response or impulse, the more difficult the self-control task
will be—and, we would assume, the greater amount of energy in
the form of blood glucose the system would have to expend in
order to succeed.
Another implication is that self-control tasks that have a direct
impact on blood glucose may raise particular problems for self-
control. Most obviously, dieting essentially involves restricting
one’s caloric intake, and there may be an ironic conflict in which
the dietary restriction produces lower glucose, which, in turn,
undermines the willpower needed to refrain from eating. Further
research in self-control may explore how efforts to control some
behaviors paradoxically undercut the capacity for control by in-
terfering with the body’s glucose processes.
We do not wish to overstate the importance of glucose to
self-control. Self-control may have multiple physiological and
psychological links and processes. Sleep and rest, for example,
seem beneficial to self-control independently of caloric processes
(see Baumeister et al., 1994, for a review). Although it is possible
that their effects could be mediated by glucose, they may signal
that other factors can also be decisive. The present results none-
theless do point to glucose as important for self-control.
Moreover, despite our manipulations, we do not intend to ad-
vocate consuming large quantities of sugar as an ideal strategy for
improving self-control. Eating several candy bars, for instance,
may give one a boost of energy and better self-control, but these
benefits are likely to disappear when glucose levels eventually
drop. Protein or complex carbohydrates may be more effective for
sustained self-control. We used sugar in our studies because it is
fast-acting and convenient.
It has long been known that action consumes energy. More
recent evidence has indicated that some brain and cognitive pro-
cesses likewise consume substantial amounts of energy—indeed,
some far more than others. The “last-in, first-out rule” states that
cognitive abilities that developed last ontogenetically are the first
to become impaired when cognitive and physiological resources
are compromised. Self-control, as a relatively advanced human
GAILLIOT ET AL.
capacity, was probably one of the last to develop and hence may
be one of the first to suffer impairments when resources are
inadequate. The present findings suggest that relatively small acts
of self-control are sufficient to deplete the available supply of
glucose, thereby impairing the control of thought and behavior, at
least until the body can retrieve more glucose from its stores or
ingest more calories. More generally, the body’s variable ability to
mobilize glucose may be an important determinant of people’s
capacity to live up to their ideals, pursue their goals, and realize
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interaction. Newbury Park, CA: Sage.
Altura, B., Altura, B., Zhang, A., & Zakhari, S. (1996). Effects of alcohol
on overall brain metabolism. In H. Begleiter & B. Kissin (Eds.), The
pharmacology of alcohol and alcohol dependence: Alcohol and alco-
holism (pp. 145–180). New York: Oxford University Press.
Aries, P. (1981). The hour of our death. New York: Oxford University
Arndt, J., Greenberg, J., Solomon, S., Pyszczynski, T., & Simon, L. (1997).
Suppression, accessibility of death-related thoughts, and cultural world-
view defense: Exploring the psychodynamics of terror management.
Journal of Personality and Social Psychology, 73, 5–18.
Baumeister, R. F. (2005). The cultural animal: Human nature, meaning,
and social life. New York: Oxford University Press.
Baumeister, R. F., Bratslavsky, E., Muraven, M., & Tice, D. M. (1998).
Self-control depletion: Is the active self a limited resource? Journal of
Personality and Social Psychology, 74, 1252–1265.
Baumeister, R. F., Gailliot, M. T., DeWall, C. N., & Oaten, M. (in press).
Self-regulation and personality: Strength-boosting interventions and trait
moderators of ego depletion. Journal of Personality.
Baumeister, R. F., Heatherton, T. F., & Tice, D. M. (1994). Losing control:
How and why people fail at self-regulation. San Diego, CA: Academic
Becker, E. (1973). The denial of death. New York: Academic Press.
Benton, D. (1990). The impact of increasing blood glucose on psycholog-
ical functioning. Biological Psychology, 30, 13–19.
Benton, D., & Owens, D. (1993). Is raised blood glucose associated with
the relief of tension? Journal of Psychosomatic Research, 37, 1–13.
Benton, D., Owens, D. S., & Parker, P. Y. (1994). Blood glucose influences
memory and attention in young adults. Neuropsychologia, 32, 595–607.
Bolton, R. (1979). Hostility in fantasy: A further test of the hypoglycemia-
aggression hypothesis. Aggressive Behavior, 2, 257–274.
Bridges, K. R. (1985). Test-completion speed: Its relationship to perfor-
mance on three course-based objective examinations. Educational and
Psychological Measurement, 45, 29 –35.
Cox, D., Eickhoff, K., Gonder-Frederick, L., & Clarke, W. (1993). Hunger:
A sensitive but nonspecific symptom of hypoglycemia. Diabetes, 16,
Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability
independent of psychopathology. Journal of Consulting Psychology, 24,
Dawkins, R. (1976). The selfish gene. London: Oxford University Press.
DeWall, C. N., Baumeister, R. F., Stillman, T. F., & Gailliot, M. T. (in
press). Violence restrained: Effects of self-regulatory capacity and its
depletion on aggressive behavior. Journal of Experimental Social Psy-
Donohoe, R. T., & Benton, D. (1999). Blood glucose control and aggres-
siveness in females. Personality and Individual Differences, 26, 905–
Duckworth, A. L., & Seligman, M. E. P. (2005). Self-discipline outdoes IQ
in predicting academic performance of adolescents. Psychological Sci-
ence, 16, 939 –944.
Dunbar, R. I. M. (1998). The social brain hypothesis. Evolutionary An-
thropology, 6, 178 –190.
Fairclough, S. H., & Houston, K. (2004). A metabolic measure of mental
effort. Biological Psychology, 66, 177–190.
Feather, N. T. (1961). The relationship of persistence at a task to expec-
tation of success and achievement related motives. Journal of Abnormal
and Social Psychology, 63, 552–561.
Finkel, E. J., & Campbell, W. K. (2001). Self-control and accommodation
in close relationships: An interdependence analysis. Journal of Person-
ality and Social Psychology, 81, 263–277.
Freud, S. (1961a). The ego and the id. In J. Strachey (Ed. and Trans.), The
standard edition of the complete psychological works of Sigmund Freud
(Vol. 19, pp. 12– 66). London: Hogarth Press. (Original work published
Freud, S. (1961b). New introductory lectures on psycho-analysis. In J.
Strachey (Ed. and Trans.), The standard edition of the complete psycho-
logical works of Sigmund Freud (Vol. 22, pp. 7–182). London: Hogarth
Press. (Original work published 1933)
Gailliot, M. T., Maner, J. K., DeWall, C. N., & Baumeister, R. F. (2005).
The selfishness heuristic. Manuscript in preparation.
Gailliot, M. T., Schmeichel, B. J., & Baumeister, R. F. (2006). Self-
regulatory processes defend against the threat of death: Effects of
self-control depletion and trait self-control on thoughts and fears of
dying. Journal of Personality and Social Psychology, 91, 49 –62.
Gans, D., Harper, A., Bachorowski, J., Newman, J., Shrago, E., & Taylor,
S. (1990). Sucrose and delinquency: Oral sucrose tolerance test and
nutritional assessment. Pediatrics, 86, 254 –262.
Gilbert, D. T., Krull, D. S., & Pelham, B. W. (1988). Of thoughts unspo-
ken: Social inference and the self-regulation of behavior. Journal of
Personality and Social Psychology, 55, 685–694.
Glass, D. C., Singer, J. E., & Friedman, L. N. (1969). Psychic cost of
adaptation to an environmental stressor. Journal of Personality and
Social Psychology, 12, 200 –210.
Gordijn, E. H., Hindriks, I., Koomen, W., Dijksterhuis, A., & Van Knip-
penberg, A. (2004). Consequences of stereotype suppression and internal
suppression motivation: A self-regulation approach. Personality and
Social Psychology Bulletin, 30, 212–224.
Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime.
Stanford, CA: Stanford University Press.
Greenberg, J., Arndt, J., Schimel, J., Pyszczynski, T., & Solomon, S.
(2001). Clarifying the function of mortality salience-induced worldview
defense: Renewed suppression or reduced accessibility of death-related
thoughts? Journal of Experimental Social Psychology, 37, 70 –76.
Hall, J. B., & Brown, D. A. (1979). Plasma glucose and lactic acid
alterations in response to a stressful exam. Biological Psychology, 8,
Kahan, D., Polivy, J., & Herman, C. P. (2003). Conformity and dietary
disinhibition: A test of the ego-strength model of self-regulation. Inter-
national Journal of Eating Disorders, 32, 165–171.
Laughlin, S. B. (2004). The implications of metabolic energy requirements
for the representation of information in neurons. In M. S. Gazzaniga
(Ed.), The cognitive neurosciences (3rd ed., pp. 187–196). Cambridge,
MA: MIT Press.
Lund-Anderson, H. (1979). Transport of glucose from blood to brain.
Physiological Review, 59, 305–352.
Lustman, P. J., Frank, B. L., & McGill, J. B. (1991). Relationship of
personality characteristics to glucose regulation in adults with diabetes.
Psychosomatic Medicine, 53, 305–312.
Matthews, G., Jones, D. M., & Chamberlain, A. G. (1990). Refining the
measurement of mood: The UWIST Mood Adjective Checklist. British
Journal of Psychology, 81, 17– 42.
Matykiewicz, L., La Grange, L., Vance, P., Mu, W., & Reyes, E. (1997).
SELF-CONTROL AND GLUCOSE
Adjudicated adolescent males: Measures of urinary
5-hydroxyindoleacetic acid and reactive hypoglycemia. Personality and
Individual Differences, 22, 327–332.
Mayer, J. D., & Gaschke, Y. N. (1988). The experience and meta-
experience of mood. Journal of Personality and Social Psychology, 55,
McNay, E. C., McCarty, R. C., & Gold, P. E. (2001). Fluctuations in brain
glucose concentration during behavioral testing: Dissociations between
brain areas and between brain and blood. Neurobiology of Learning and
Memory, 75, 325–327.
Muraven, M., & Baumeister, R. F. (2000). Self-regulation and depletion of
limited resources: Does self-control resemble a muscle? Psychological
Bulletin, 126, 247–259.
Muraven, M., Collins, R. L., & Neinhaus, K. (2002). Self-control and
alcohol restraint: An initial application of the self-control strength
model. Psychology of Addictive Behaviors, 16, 113–120.
Muraven, M., Collins, R. L., Shiffman, S., & Paty, J. A. (2005). Daily
fluctuations in self-control demands and alcohol intake. Psychology of
Addictive Behaviors, 19, 140 –147.
Muraven, M., Shmueli, D., & Burkley, E. (2006). Conserving self-control
strength. Journal of Personality and Social Psychology, 91, 524 –537.
Muraven, M., & Slessareva, E. (2003). Mechanisms of self-control failure:
Motivation and limited resources. Personality and Social Psychology
Bulletin, 29, 894 –906.
Muraven, M., Tice, D. M., & Baumeister, R. F. (1998). Self-control as a
limited resource: Regulatory depletion patterns. Journal of Personality
and Social Psychology, 74, 774 –789.
Owens, D. S., & Benton, D. (1994). The impact of raising blood glucose
on reaction times. Neuropsychobiology, 30, 106–113.
Plant, E. A. (2004). Responses to interracial interactions over time. Per-
sonality and Social Psychology Bulletin, 30, 1458 –1471.
Plant, E. A., & Devine, P. G. (1998). Internal and external motivation to
respond without prejudice. Journal of Personality and Social Psychol-
ogy, 75, 811– 832.
Pratt, T. C., & Cullen, F. T. (2000). The empirical status of Gottfredson and
Hirschi’s general theory of crime: A meta-analysis. Criminology, 38,
Reivich, M., & Alavi, A. (1983). Positron emission tomographic studies of
local cerebral glucose metabolism in humans in physiological and patho-
logical conditions. Advances in Metabolic Disorders, 10, 135–176.
Richeson, J. A., Baird, A. A., Gordon, H. I., Heatherton, T. F., Wyland,
C. L., Trawalter, S., & Shelton, J. N. (2003). An fMRI investigation of
the impact of interracial contact on executive function. Nature Neuro-
science, 6, 1323–1328.
Richeson, J. A., & Shelton, J. N. (2003). When prejudice does not pay:
Effects of interracial contact on executive function. Psychological Sci-
ence, 14, 287–290.
Richeson, J. A., & Trawalter, S. (2005). Why do interracial interactions
impair executive function? A resource depletion account. Journal of
Personality and Social Psychology, 88, 934–947.
Richeson, J. A., Trawalter, S., & Shelton, J. N. (2005). African Americans’
implicit racial attitudes and the depletion of executive function after
interracial interactions. Social Cognition, 23, 336–352.
Rosenblatt, A., Greenberg, J., Solomon, S., Pyszczynski, T., & Lyon, D.
(1989). Evidence for terror management theory I: The effects of mor-
tality salience on reactions to those who violate or uphold cultural
values. Journal of Personality and Social Psychology, 57, 681– 690.
Schmeichel, B. J., Vohs, K. D., & Baumeister, R. F. (2003). Intellectual
performance and ego depletion: Role of the self in logical reasoning and
other information processing. Journal of Personality and Social Psy-
chology, 85, 33– 46.
Shoda, Y., Mischel, W., & Peake, P. K. (1990). Predicting adolescent
cognitive and self-regulatory competencies from preschool delay of
gratification: Identifying diagnostic conditions. Developmental Psychol-
ogy, 26, 978 –986.
Siesjo, B. K. (1978). Brain energy metabolism. Chichester, England:
Simpson, G. C., Cox, T., & Rothschild, D. R. (1974). The effects of noise
stress on blood glucose level and skilled performance. Ergonomics, 17,
Stucke, T. S., & Baumeister, R. F. (2006). Ego depletion and aggressive
behavior: Is the inhibition of aggression a limited resource? European
Journal of Social Psychology, 36, 1–13.
Tangney, J. P., Baumeister, R. F., & Boone, A. L. (2004). High self-control
predicts good adjustment, less pathology, better grades, and interper-
sonal success. Journal of Personality, 72, 271–322.
Twenge, J. M., Muraven, M., & Tice, D. M. (2004). Measuring state
self-control: Reliability, validity, and correlations with physical and
psychological stress. Unpublished manuscript, San Diego State Univer-
Van Cauter, E., Polonsky, K. S., & Scheen, A. J. (1997). Roles of circadian
and rhythmicity and sleep in human glucose regulation. Endocrine
Reviews, 18, 716 –738.
Virkkunen, M., & Huttunen, M. O. (1982). Evidence for abnormal glucose
tolerance test among violent offenders. Neuropsychobiology, 8, 30 –34.
Vohs, K. (2004). The health of romantic relationships relies on self-
regulation. Paper presented at the Society for Personality and Social
Psychology, New Orleans, LA.
Vohs, K. D., Baumeister, R. F., & Ciarocco, N. J. (2005). Self-regulation
and self-presentation: Regulatory resource depletion impairs impression
management and effortful self-presentation depletes regulatory re-
sources. Journal of Personality and Social Psychology, 88, 632– 657.
Vohs, K. D., Baumeister, R. F., Twenge, J. M., Schmeichel, B. J., Tice,
D. M., & Crocker, J. (2004). Decision fatigue: Making multiple deci-
sions depletes the self. Manuscript in preparation.
Vohs, K. D., & Faber, R. J. (2004). To buy or not to buy? Self-control and
self-regulatory failure in purchase behavior. In R. F. Baumeister & K. D.
Vohs (Eds.), Handbook of self-regulation: Research, theory, and appli-
cations (pp. 509 –524). New York: Guilford Press.
Vohs, K. D., & Heatherton, T. F. (2000). Self-regulatory failure: A
resource-depletion approach. Psychological Science, 11, 249–254.
von Hippel, W., & Gonsalkorale, K. (2005). “That is bloody revolting!”
Inhibitory control of thoughts better left unsaid. Psychological Science,
Watson, D., Clark, L., & Tellegen, A. (1988). Development and validation
of brief measures of positive and negative affect: The PANAS scales.
Journal of Personality and Social Psychology, 54, 1063–1070.
Weiss, V. (1986). From memory span to the quantum mechanics of
intelligence. Personality and Individual Differences, 7, 737–749.
West, R. (2001). Glucose for smoking cessation: Does it have a role? CNS
Drags, 15, 261–265.
Received February 2, 2006
Revision received September 12, 2006
Accepted September 18, 2006 䡲
GAILLIOT ET AL.