Content uploaded by Miguel A. Vadillo
Author content
All content in this area was uploaded by Miguel A. Vadillo on Jul 12, 2016
Content may be subject to copyright.
Psychological Science
1 –8
© The Author(s) 2016
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797616654911
pss.sagepub.com
Research Article
According to the law of conservation of energy, the total
amount of energy of an isolated system can never
increase. In the domain of psychology, the idea that
energy is a limited resource originated with Freud
(1923/1961). Energy models have been little used in psy-
chology since Freud, though, with the rare exception
of the ego-depletion model developed by Baumeister,
Bratslavsky, Muraven, and Tice (1998). According to this
research, “the self’s acts of volition draw on some limited
resource, akin to strength or energy. . . , therefore, one act
of volition will have a detrimental impact on subsequent
volition” (Baumeister et al., 1998, p. 1252). Research on
ego depletion has substantial implications. It has been
claimed that reliably exerting self-control, either actively
doing something “good” or avoiding the temptation to act
on “bad” impulses, can greatly reduce many of the major
ills that affect society and people’s personal lives, such as
“crime, violence, unwanted pregnancy, drug addiction,
venereal diseases, bankruptcy, and premature deaths”
(Baumeister, Muraven, & Tice, 2000, p. 130). It is not sur-
prising that the work of Baumeister et al. (1998) has had
an impact on a number of disciplines, including advertis-
ing, behavioral economics, business, consumerism, law,
management, marketing, and medicine. In fact, it is fair to
say that this seminal article by Baumeister et al. has
become a classic: At the time of this writing, it has more
than 1,250 citations in the Web of Science.
When ego depletion was first proposed, the idea of a
limited resource was a convenient metaphor. Given how
fundamental exerting self-control is thought to be, it is
important to establish the energy source that is depleted
and to provide a mechanism by which ego depletion
occurs. The most popular explanation found in the litera-
ture involves glucose. Gailliot et al. (2007) presented
nine studies supporting three main findings: (a) Blood
glucose levels are reduced after performing a self-control
654911PSSXXX10.1177/0956797616654911Vadillo et al.The Glucose Model of Ego Depletion
research-article2016
Corresponding Author:
Magda Osman, Biological and Experimental Psychology Group,
School of Biological and Chemical Sciences, Queen Mary University
of London, Fogg Building, Mile End Rd., London E1 4NS, United
Kingdom
E-mail: m.osman@qmul.ac.uk
The Bitter Truth About Sugar and
Willpower: The Limited Evidential Value
of the Glucose Model of Ego Depletion
Miguel A. Vadillo1, Natalie Gold2, and Magda Osman3
1Department of Primary Care and Public Health Sciences, King’s College London; 2Department of Philosophy,
King’s College London; and 3Biological and Experimental Psychology Group, School of Biological and Chemical Sciences,
Queen Mary University of London
Abstract
The idea behind ego depletion is that willpower draws on a limited mental resource, so that engaging in an act of self-
control impairs self-control in subsequent tasks. To present ego depletion as more than a convenient metaphor, some
researchers have proposed that glucose is the limited resource that becomes depleted with self-control. However, there
have been theoretical challenges to the proposed glucose mechanism, and the experiments that have tested it have
found mixed results. We used a new meta-analytic tool, p-curve analysis, to examine the reliability of the evidence
from these experiments. We found that the effect sizes reported in this literature are possibly influenced by publication
or reporting bias and that, even within studies yielding significant results, the evidential value of this research is weak.
In light of these results, and pending further evidence, researchers and policymakers should refrain from drawing any
conclusions about the role of glucose in self-control.
Keywords
ego depletion, glucose, meta-analysis, p-curve, self-control, open data
Received 11/20/15; Revision accepted 5/25/16
Psychological Science OnlineFirst, published on July 11, 2016 as doi:10.1177/0956797616654911
by guest on July 11, 2016pss.sagepub.comDownloaded from
2 Vadillo et al.
task but not after performing a comparable cognitive task
that does not require self-control; (b) low levels of blood
glucose after a first self-control task predict behavioral
deficits on a second self-control task; and (c) participants
whose glucose levels are restored by ingesting a glucose
drink after a self-control task perform better on a subse-
quent task than do participants who are given a diet
drink between tasks. They concluded that self-control
depletes blood glucose, which leads to decreased self-
control on subsequent tasks, and restoring glucose levels
replenishes the ability to exert self-control.
Although the conclusions drawn by Gailliot et al.
(2007) have been extraordinarily influential, their glucose
hypothesis remains controversial. The mechanism they
propose has been challenged, and the reliability of their
results has been disputed. Kurzban (2010) argued that
the glucose mechanism, as presented by Gailliot et al., is
biologically implausible. The mechanism invokes the
idea that self-control tasks deplete glucose because of
energy consumption by the brain, but the supporting
evidence shows changes only in blood glucose levels.
Kurzban cited evidence that the sort of self-control tasks
used in the literature have little effect on brain metabo-
lism and that changes in blood glucose are unlikely to
reflect blood glucose uptake by the brain.
There are also concerns about the empirical evidence
for the glucose mechanism. For instance, Schimmack
(2012) showed that the number of significant results
reported by Gailliot et al. (2007) is too large, given their
average power. In other words, these results are likely to
be influenced by publication bias or p-hacking (see
Francis, 2012; Simmons, Nelson, & Simonsohn, 2011).
Kurzban’s (2010) concerns were supported by a reanaly-
sis of the data from Gailliot et al., in which he found that
self-control does not decrease blood glucose levels, and
by recent failures to replicate the effect of glucose on
self-control (Job, Walton, Bernecker, & Dweck, 2013;
Kelly, Sünram-Lea, & Crawford, 2015; Lange & Eggert,
2014; Lange, Seer, Rapior, Rose, & Eggert, 2014).
The effect of glucose on self-control has been exam-
ined in three broad categories of studies:
Correlational studies measure effects on blood
glucose levels before and after self-control is
exerted (Dvorak & Simons, 2009; Gaillot, 2012;
Gailliot et al., 2007; Molden et al., 2012).
Glucose-ingestion studies manipulate blood glu-
cose levels by leaving enough time between inges-
tion and a control task for glucose to be absorbed
into the bloodstream (Birnie, Smallwood, Reay, &
Riby, 2015; Denson, von Hippel, Kemp, & Teo,
2010; DeWall, Baumeister, Gailliot, & Maner, 2008;
Dickinson, McElroy, & Stroh, 2014; Gailliot et al.,
2007; Gailliot, Peruche, Plant, & Baumeister, 2009;
Howard & Marczinski, 2010; Job et al., 2013; Kelly
et al., 2015; Lange & Eggert, 2014; Lange et al.,
2014; Masicampo & Baumeister, 2008; Wang &
Dvorak, 2010). Although the findings from these
studies provide mixed support for the glucose
hypothesis, the methods used imply a mechanism
that is consistent with the proposals of Gailliot
et al. (2007).
Glucose-rinsing studies examine the impact of sim-
ply rinsing one’s mouth with a glucose solution
before exerting self-control (Hagger & Chatzisarantis,
2013; Lange & Eggert, 2014; Molden et al., 2012;
Sanders, Shirk, Burgin, & Martin, 2012). Results
from these studies suggest that the signal of glu-
cose from the mouth to the brain is sufficient to
neutralize the ego-depletion effect. This mecha-
nism is consistent with the results of the ingestion
manipulations but suggests that the effect does not
depend on a metabolic explanation.
To help settle the growing concerns in the academic
community regarding the reliability of the glucose mecha-
nism, which in turn has implications for the ego-depletion
hypothesis that it underpins, we sought to use a new
meta-analytic tool, p-curve analysis, to investigate the pres-
ence of publication and reporting biases in this literature.
Method
Literature-search strategy
We looked for studies supporting the idea that sugar con-
sumption is related to ego depletion and self-regulation.
Specifically, we considered any study exploring the
hypothesis that glucose ingestion or rinsing improves
performance (e.g., overcoming an impulse, inhibiting an
aggressive reaction, or controlling a cognitive process) or
ameliorates the effect of an ego-depleting experience in
these laboratory self-regulation tasks. We also included
studies testing whether performance in laboratory tasks
(again, specifically those that explicitly require self-
regulation) is correlated with pre- or posttesting sugar
levels. Studies in which participants were not asked to
drink a sugary beverage but simply to rinse their mouths
with it were also included in the present analyses; this
literature also supports the idea that sugar consumption
improves self-regulation (even if it challenges the specific
hypothesis that such improvement is achieved through
metabolic processes).
Given these criteria, we excluded experiments show-
ing a relation between sugar consumption and cognitive
processes (e.g., short-term memory or general cognition
function) that prima facie do not seem to pose demands
on self-regulation (e.g., Carter & McCullough, 2013;
by guest on July 11, 2016pss.sagepub.comDownloaded from
The Glucose Model of Ego Depletion 3
Owen, Scholey, Finnegan, Hu, & Sünram-Lea, 2012). In
addition, we also excluded studies that investigated the
correlation between general glucose levels (or regular
glucose ingestion) and self-regulated behavior in natural-
istic settings over many days. This included, for instance,
studies on the relationships between glucose ingestion
and smoking cessation and studies on the relationship
between diabetes and various psychological processes
(see Gailliot & Baumeister, 2007). These studies rely on
measures that differ substantially from the dependent
variables gathered in laboratory-based ego-depletion
tasks, and the lack of experimental control makes the
results amenable to alternative explanations that bear lit-
tle or no relation to ego depletion and self-regulation.
We began our search by inspecting a small set of stud-
ies that had included an exhaustive literature review.
These included a meta-analysis by Hagger, Wood, Stiff,
and Chatzisarantis (2010) on the general ego-depletion
literature, a study by Job et al. (2013) exploring individ-
ual differences in the impact of glucose on self-control,
and Lange and Eggert’s (2014) recent attempt to replicate
the effect of sugar consumption or rinsing on ego deple-
tion. Then, to make sure that we included all relevant
studies, we conducted a systematic search in Web of Sci-
ence and Google Scholar for the term “glucose” along
with “ego depletion,” “self-control,” or “self-regulation.”
This strategy allowed us to identify 18 articles with one
or more eligible studies. All these studies are listed in
Table 1 and are also marked with asterisks in the refer-
ence list. Furthermore, we found out that one of our
selected studies (Masicampo & Baumeister, 2008) had
been included in the famous project on the reproducibil-
ity of psychological science (Open Science Collabora-
tion, 2015). This replication was also included in our
analyses, which resulted in a total of 19 articles.
p-curve analysis
To assess the reliability of this set of studies, we used
p-curve analysis, a recently designed meta-analytic tool
that allows for the exploration of various biases solely by
examining the distribution of significant p values
(Simonsohn, Nelson, & Simmons, 2014). Imagine a set of
studies exploring an effect that does not exist. Occasion-
ally, these studies will yield a significant result (i.e., a
p value lower than .05) just by chance. In this scenario, all
p values will be equally likely: 5% of studies will have
p values lower than .05, 4% of studies will have p values
lower than .04, and so on. Consequently, the p values of
a set of studies exploring a nonexistent effect should typi-
cally follow a flat distribution. Note that this is not the
case if the studies are exploring a true effect: In that case,
significant p values should follow a right-skewed distribu-
tion in which small p values (e.g., p < .01) are more
prevalent than larger p values (e.g., p between .04 and
.05). As explained by Simonsohn et al. (2014), this can be
easily understood if one imagines an experimenter explor-
ing a very large effect with a large sample of participants.
Most likely, the experimenter will observe a very low
p value. Experiments with smaller effect sizes and smaller
samples are simply less extreme versions of this ideal sce-
nario. Even for low-powered studies, the distribution of
p values should be right skewed. This implies that, in
principle, one can know whether a set of experiments is
exploring true effects or null effects simply by checking
whether their p values follow a right-skewed distribution
or a rather flat distribution. An interesting feature of this
approach is that it focuses exclusively on significant p values
(i.e., studies for which p < .05); consequently, its results
are unaffected by publication bias.
Simonsohn et al. (2014) designed an online appli-
cation (available at http://www.p-curve.com) that allows
researchers to test whether an observed distribution of
p values is significantly right skewed or suspiciously flat,
which could suggest that the significant results are false
positives. A simple way to test whether the distribution of
p values is significantly right skewed is to compare the
number of significant p values lower than .025 with the
number of p values between .025 and .05 by using a bino-
mial test. A potential shortcoming of this approach is that
this binomial test gives the same weight to exceptionally
small p values (e.g., .00001) as to p values barely smaller
than .025 (e.g., .024). To overcome this limitation, the lat-
est versions of the p-curve application conduct not only a
binomial test but also an alternative analysis, known as a
continuous test, that is sensitive to the exact p values.
If the distribution of p values is not significantly right
skewed, this might mean that the studies lack any eviden-
tial value or, in other words, that the significant results
could be false positives. However, failure to find a signifi-
cant right-skewed distribution might also be due to a lack
of statistical power (e.g., if the analysis includes a very
small number of studies). Simonsohn et al. (2014) sug-
gested that in order to determine whether the distribution
of p values is too flat, one should test whether the p-curve
is flatter than the theoretical distribution that one would
observe in a set of studies with 33% statistical power. If
the p-curve is significantly flatter than this very flat stan-
dard, a common conclusion is that the set of studies might
lack evidential value and that they might be the product
of publication bias, selective reporting, or p-hacking.
Selection of statistical contrasts
We selected the key statistical contrasts of each study
following the guidelines offered by Simonsohn et al.
(2014). In the case of correlational studies or experiments
with just two groups, we registered the target correlation
by guest on July 11, 2016pss.sagepub.comDownloaded from
4 Vadillo et al.
coefficient, the statistic testing the regression slope,
or the statistic testing the difference of means. In com-
plex factorial designs, if researchers expected the ego-
depletion effect to disappear in a specific condition, then
we registered the statistic testing the interaction. In con-
trast, if they expected to find a complete cross-over inter-
action, we registered the statistics for the two simple
effects. A total of 38 statistical contrasts were included in
the analyses. In accordance with the recommendations of
Simonsohn et al. (2014), when two statistics were equally
valid, we used one of them in the main analysis and the
other one in a second analysis that we refer to as a
robustness test. In most cases (four out of five), we
adopted the general rule of selecting the first one to
Table 1. Studies Included in the Analyses and Their Key Statistical Contrasts
Study Sugar rinsing? Key statistical contrast p
Birnie, Smallwood, Reay, and Riby (2015) No t(15) = 2.469a.02605
Denson, von Hippel, Kemp, and Teo (2010), Study 1 No t(67) = −2.19 .03201
Denson et al. (2010), Study 2 No t(151) = 2.24 .02655
DeWall, Baumeister, Gailliot, and Maner (2008), Study 2 No F(1, 55) = 6.64 .01268
Dickinson, McElroy, and Stroh (2014) No z = 1.88 .06011
Dvorak and Simons (2009) No F(1, 177) = 5.63 .01873
Gailliot (2012) No r(50) = –.30 .03071
Gailliot et al. (2007), Study 1 No F(1, 100) = 6.08 .01537
Gailliot et al. (2007), Study 2 No t(33) = 2.20 .03492
Gailliot et al. (2007), Study 3 No r(14) = –.62a.01041
Gailliot et al. (2007), Study 4 No r(10) = .56 .05828
Gailliot et al. (2007), Study 5 No r(21) = .45 .03120
Gailliot et al. (2007), Study 6 No r(15) = .43 .08493
Gailliot et al. (2007), Study 7 No F(1, 57) = 5.04 .02866
Gailliot et al. (2007), Study 8 No F(1, 69) = 5.45 .02249
Gailliot et al. (2007), Study 9 No t(16) = 3.13 .00646
Gailliot, Peruche, Plant, and Baumeister (2009) No t(47) = 2.21a.03201
Hagger and Chatzisarantis (2013), Study 1 Yes F(1, 24) = 8.42 .00783
Hagger and Chatzisarantis (2013), Study 2 Yes F(1, 30) = 6.12 .01925
Hagger and Chatzisarantis (2013), Study 3 Yes F(1, 32) = 4.06 .05238
Hagger and Chatzisarantis (2013), Study 4 Yes F(1, 40) = 10.32 .00260
Hagger and Chatzisarantis (2013), Study 5 Yes F(1, 36) = 7.28 .01055
Howard and Marczinski (2010) No F(4, 75) = 2.95 .02544
Job, Walton, Bernecker, and Dweck (2013), Study 1 No t(78) = 2.10 .03896
Job et al. (2013), Study 2 No F(1, 58) = 5.16 .02684
Job et al. (2013), Study 3 No F(1, 139) = 5.28 .02306
Kelly, Sünram-Lea, and Crawford (2015) No F(1, 67) = 0.80 .37430
Lange and Eggert (2014), Study 1 No F(1, 68) = 1.12 .29366
Lange and Eggert (2014), Study 2 No F(1, 110) = 0.01a.92053
Lange, Seer, Rapior, Rose, and Eggert (2014) No t(68) = 0.05a.96027
Masicampo and Baumeister (2008) No F(1, 111) = 5.311 .02305
Molden et al. (2012), Study 1 No F(1, 83) = 2.05 .15596
Molden et al. (2012), Study 2 Yes F(1, 39) = 4.54 .03947
Molden et al. (2012), Study 3 Yes F(1, 28) = 5.02 .03317
Open Science Collaboration (2015); replication of
Masicampo and Baumeister (2008)
No F(1, 158) = 0.379 .53902
Sanders, Shirk, Burgin, and Martin (2012) Yes t(49) = −2.129 .03831
Wang and Dvorak (2010), Simple Effect 1 No t(31) = 2.55 .01593
Wang and Dvorak (2010), Simple Effect 2 No t(32) = 3.12 .00381
Note: The statistical contrasts were selected according to the guidelines of Simonsohn, Nelson, and Simmons (2014). A complete
p-curve disclosure table justifying the selection of each statistical contrast is available at the Open Science Framework (https://osf
.io/yf8p3/).
aThese statistical contrasts were replaced by alternative contrasts in the robustness test (for further details, see the p-curve
disclosure table at the Open Science Framework).
by guest on July 11, 2016pss.sagepub.comDownloaded from
The Glucose Model of Ego Depletion 5
appear in the text for the main analysis and the second
one to appear for the robustness test. However, on one
occasion (Birnie et al., 2015), we broke this rule because
the conclusions of the authors relied more heavily on
one of the statistics than on the other. In this particular
case, we selected the more appropriate statistic for the
main analysis and the other one for the robustness test.
We found no studies in which three or more statistics
were equally valid for p-curve analysis. A p-curve disclo-
sure table with all the selected statistics and the justifica-
tion for our choices are available at Open Science
Framework (https://osf.io/yf8p3/).
Results
The key statistical contrasts of the studies included in our
analysis are shown in Table 1. Figure 1 plots the fre-
quency of each range of p values among these studies. As
the figure shows, the main p-curve did not fit with the
right-skewed distribution that one would expect if these
studies were exploring a true effect. Although there were
no significant results immediately below .05, p values in
the interval between 0 and .04 show, if anything, a left-
skewed distribution. Not surprisingly, the statistical con-
trast testing the right skewness of the p-curve was
nonsignificant (binomial test comparing the proportions
of contrasts with p values < .025 and those with p values
between .025 and .05: p = .575; continuous test: z = −0.83,
p = .204). In other words, the distribution of p values was
not significantly different from what would be expected
if the null hypothesis (in this case, that the average effect
size is zero) were true. Furthermore, the observed distri-
bution is flatter than would be expected if the studies
were simply underpowered. Statistical analyses con-
firmed that the p-curve was significantly flatter than
would be expected if there were an effect but the studies
had only 33% power on average (binomial test: p = .019;
continuous test: z = −2.08, p = .019). Thus, we can reject
the hypothesis that although there was an effect, the
studies had an average power of only 33% to detect it.
The results were very similar for the robustness test,
which also failed to find significant evidence of right
skewness (binomial test: p = .500; continuous test:
z = −0.88, p = .190) and, in fact, detected that the p-curve
was significantly flatter than would be expected if there
were an effect but the studies had only 33% power on
average (binomial test: p = .033; continuous test:
z = −1.99, p = .023).
Note that when we removed from the analyses all the
studies that explored the effects of glucose rinsing
(because these have a somewhat different theoretical
background), p-curve results did not improve. After the
glucose-rinsing experiments were removed, neither the
continuous test (z = −0.57, p = .285) nor the binomial test
(p = .668) suggested that the remaining set of studies had
any evidential value. Furthermore, both tests (continu-
ous: z = −1.94, p = .026; binomial: p = .024) suggested
that the p-curve was significantly flatter than would be
expected if there were an effect but the studies had only
33% power on average. In other words, the poor results
of the previous tests cannot be attributed to the inclusion
of glucose-rinsing studies in the analyses.
Discussion
The results of our analyses suggest that the relationship
between glucose levels and self-control behaviors might
be unreliable. Figure 1 shows that the key p values of the
19 studies included in the present analyses follow a sur-
prisingly flat distribution. This is exactly the pattern of
results that one would expect to find if those results were
false positives. These results remain unchanged regard-
less of whether glucose-rinsing studies are included or
excluded from the sample.
These results may not come as a complete surprise
given the empirical challenges to the glucose hypothesis
suggested by failed replications (Job et al., 2013; Kelly
et al., 2015; Lange & Eggert, 2014; Lange et al., 2014) and
theoretical critiques regarding its biological plausibility
(Kurzban, 2010; Osman, 2014). Furthermore, a detailed
analysis of the seminal article suggesting the glucose
hypothesis showed that the number of significant find-
ings reported in that article was too large, given the low
power of each study (Schimmack, 2012). In other words,
the results were too good to be true (Francis, 2012).
Nevertheless, the findings from the present study are a
surprise in the context of the wide acceptance of the
glucose hypothesis in general scientific research and its
popularity, as evidenced by the number of citations of
0
10
20
30
40
50
.01 .02 .03 .04 .05
Percentage of Test Results
p Value
Main Analysis
Robustness Test
Zero Effect
True Effect, 33% Power
Fig. 1. Distribution of observed p values for both the main analysis and
the robustness test, along with the expected distribution of p values if the
null hypothesis is true (zero effect) and if the alternative hypothesis is
true but the experiments lack sufficient power (true effect, 33% power).
by guest on July 11, 2016pss.sagepub.comDownloaded from
6 Vadillo et al.
Gailliot et al. (2007) in the literature and the continued
influence of this hypothesis in recent reviews on ego
depletion (e.g., Baumeister, 2014; Baumeister & Alghamdi,
2015). Moreover, the hypothesis has intuitive and seem-
ingly practical appeal. If one accepts that a failure of self-
control in regulating actions contributes to the many
personal and societal problems that people face (Bau-
meister et al., 2000), then glucose supplements would
provide a simple means to enhance willpower and ame-
liorate these problems (Baumeister & Tierny, 2011). In
light of our results, it is doubtful that such a recommen-
dation will work in the real world. This conclusion con-
verges with recent evidence that glucose might have little
or no impact on domain-general decision-making tasks
(Orquin & Kurzban, 2016) and with an intriguing series
of meta-analyses and preregistered replications suggest-
ing that the ego-depletion effect itself might be less
robust than previously thought (Carter, Kofler, Forster, &
McCullough, 2015; Hagger et al., in press).
Previous criticisms of the glucose model of ego deple-
tion have typically focused on individual articles (e.g.,
Kurzban, 2010; Schimmack, 2012). Article-level analyses
such as those by Francis (2012) are ideal in some respects
because they ensure that all the studies under scrutiny
are grounded in the same theoretical view and rely on
very similar research methods. Unfortunately, only a cou-
ple of the articles included in the present review contain
a sufficiently large number of studies to allow this type
of analysis (Gailliot et al., 2007; possibly Hagger &
Chatzisarantis, 2013). An examination of the wider litera-
ture, such as the one offered in the present article, must
necessarily collate studies with heterogeneous methods
and theoretical backgrounds. In exchange, this approach
allows researchers to check for publication and reporting
biases in areas of research in which articles with a small
number of studies are prevalent. In this sense, our study
adds to the conclusions of article-level analyses by sug-
gesting that the kinds of biases that have been detected
in isolated studies might be representative of the wider
area of research on the glucose model of ego depletion.
In any case, the rest of the experiments included in the
present analyses, with the possible exception of glucose-
rinsing studies, share a common theoretical background.
It is worth noting that, as with any other statistical test,
p-curve analysis is not a flawless indicator of bias (Bishop
& Thompson, 2016; Bruns & Ioannidis, 2016; Lakens,
2015). Our results suggest that, on average, these studies
have little or no evidential value, but they do not allow
us to determine whether the significant results are due to
publication bias, selective reporting of outcomes or anal-
yses, p-hacking, or all of these. It is not impossible that
some of these studies are exploring small but true effects
and that their evidential value may be diluted by the
biases that pervade the rest of the studies. Perhaps future
research will show that glucose does play a role in ego-
depletion effects, but our conclusions are based on the
analysis of the extant literature in this area. Thus, our
contribution must be seen as an additional piece of infor-
mation in the wider context of attempts to verify the reli-
ability of the glucose model of ego depletion. Note that
the kind of biases explored in the present study are prev-
alent in other (but not all) areas of psychological research
(e.g., Bakker, van Dijk, & Wicherts, 2012) and that low
reproducibility is not exclusively a problem of psycho-
logical research (Camerer et al., 2016; Errington et al.,
2014). In fact, it is fair to say that psychology is taking a
leading role in the dissemination of open research prac-
tices (Open Science Collaboration, 2015). We hope that
this new trend in psychological research will soon render
meta-analytic studies like this one unnecessary.
Action Editor
D. Stephen Lindsay served as action editor for this article.
Author Contributions
All the authors developed the study concept. The literature
search was conducted by M. A. Vadillo and N. Gold. M. A.
Vadillo performed the data analysis and interpretation. M.
Osman drafted the manuscript, and M. A. Vadillo and N. Gold
provided critical revisions. All the authors approved the final
version of the manuscript for submission.
Acknowledgments
We are indebted to Greg Francis, David R. Shanks, Eric-Jan
Wagenmakers, and two anonymous reviewers for their valu-
able comments on earlier versions of the manuscript.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Funding
This work was supported by European Research Council Grant
283849 (to N. Gold) under the European Union’s Seventh
Framework Programme (FP/2007–2013).
Open Practices
All data have been made publicly available via the Open Sci-
ence Framework and can be accessed at https://osf.io/4a6jk/.
The complete Open Practices Disclosure for this article can
be found at http://pss.sagepub.com/content/by/supplemental-
data. This article has received the badge for Open Data. More
information about the Open Practices badges can be found
at https://osf.io/tvyxz/wiki/1.%20View%20the%20Badges/ and
http://pss.sagepub.com/content/25/1/3.full.
by guest on July 11, 2016pss.sagepub.comDownloaded from
The Glucose Model of Ego Depletion 7
References
References marked with an asterisk indicate studies included in
the meta-analysis.
Bakker, M., van Dijk, A., & Wicherts, J. M. (2012). The rules
of the game called psychological science. Perspectives on
Psychological Science, 7, 543–554.
Baumeister, R. F. (2014). Self-regulation, ego depletion, and
inhibition. Neuropsychologia, 65, 313–319.
Baumeister, R. F., & Alghamdi, N. G. (2015). Role of self-control
failure in immoral and unethical actions. Current Opinion
in Psychology, 6, 66–69.
Baumeister, R. F., Bratslavsky, E., Muraven, M., & Tice, D. M.
(1998). Ego depletion: Is the active self a limited resource?
Journal of Personality and Social Psychology, 74, 1252–1265.
Baumeister, R. F., Muraven, M., & Tice, D. M. (2000). Ego
depletion: A resource model of volition, self-regulation,
and controlled processing. Social Cognition, 18, 130–150.
Baumeister, R. F., & Tierny, J. (2011). Willpower: Rediscovering
the greatest human strength. New York, NY: Penguin Press.
*Birnie, L. H. W., Smallwood, J., Reay, J., & Riby, L. M. (2015).
Glucose and the wandering mind: Not paying attention or
simply out of fuel? Psychopharmacology, 232, 2903–2910.
Bishop, D. V. M., & Thompson, P. A. (2016). Problems in using
p-curve analysis and text-mining to detect rate of p-hacking
and evidential value. PeerJ, 4, Article e1715. doi:10.7717/
peerj.1715
Bruns, S. B., & Ioannidis, J. P. A. (2016). p-curve and p-hacking
in observational research. PLoS ONE, 11, Article e0149144.
doi:10.1371/journal.pone.0149144
Camerer, C. F., Dreber, A., Forsell, E., Ho, T. H., Huber, J.,
Johannesson, M., . . . Wu, H. (2016). Evaluating replicabil-
ity of laboratory experiments in economics. Science, 351,
1433–1436.
Carter, E. C., Kofler, L. M., Forster, D. E., & McCullough, M. E.
(2015). A series of meta-analytic tests of the depletion effect:
Self-control does not seem to rely on a limited resource.
Journal of Experimental Psychology: General, 144, 796–815.
Carter, E. C., & McCullough, M. E. (2013). After a pair of self-
control-intensive tasks, sucrose swishing improves subse-
quent working memory performance. BMC Psychology, 1,
Article 22. doi:10.1186/2050-7283-1-22
*Denson, T. F., von Hippel, W., Kemp, R. I., & Teo, L. S. (2010).
Glucose consumption decreases impulsive aggression in
response to provocation in aggressive individuals. Journal
of Experimental Social Psychology, 46, 1023–1028.
*DeWall, C. N., Baumeister, R. F., Gailliot, M. T., & Maner, J. K.
(2008). Depletion makes the heart grow less helpful: Helping
as a function of self-regulatory energy and genetic relatedness.
Personality and Social Psychology Bulletin, 34, 1653–1662.
*Dickinson, D. L., McElroy, T., & Stroh, N. (2014). Impact of glucose
on Bayesian versus heuristic-based decision making. Journal
of Neuroscience, Psychology, and Economics, 7, 237–247.
*Dvorak, R. D., & Simons, J. S. (2009). Moderation of resource
depletion in the self-control strength model: Differing
effects of two modes of self-control. Personality and Social
Psychology Bulletin, 35, 572–583.
Errington, T. M., Iorns, E., Gunn, W., Tan, F. E., Lomax, J., & Nosek,
B. A. (2014). An open investigation of the reproducibility of
cancer biology research. eLife, 3, Article e04333. doi:10.7554/
eLife.04333
Francis, G. (2012). Too good to be true: Publication bias in
two prominent studies from experimental psychology.
Psychonomic Bulletin & Review, 19, 151–156.
Freud, S. (1961). The ego and the id. In J. Strachey (Ed. &
Trans.), The standard edition of the complete psychologi-
cal works of Sigmund Freud (Vol. 19, pp. 1–66). London,
England: Hogarth Press. (Original work published 1923)
*Gailliot, M. T. (2012). Improved self-control associated
with using relatively large amounts of glucose: Learning
self-control is metabolically expensive. Psychology, 3,
987–990.
Gailliot, M. T., & Baumeister, R. F. (2007). The physiol-
ogy of willpower: Linking blood glucose to self-control.
Personality and Social Psychology Review, 11, 303–327.
*Gailliot, M. T., Baumeister, R. F., DeWall, C. N., Maner, J. K.,
Plant, E. A., Tice, D., . . . Schmeichel, B. J. (2007). Self-
control relies on glucose as a limited energy source:
Willpower is more than a metaphor. Journal of Personality
and Social Psychology, 92, 325–336.
*Gailliot, M. T., Peruche, B. M., Plant, E. A., & Baumeister, R. F.
(2009). Stereotypes and prejudice in the blood: Sucrose
drinks reduce prejudice and stereotyping. Journal of
Experimental Social Psychology, 45, 288–290.
*Hagger, M. S., & Chatzisarantis, N. L. D. (2013). The sweet taste
of success: The presence of glucose in the oral cavity mod-
erates the depletion of self-control resources. Personality
and Social Psychology Bulletin, 39, 28–42.
Hagger, M. S., Chatzisarantis, N. L. D., Alberts, H., Anggono, C. O.,
Batailler, C., Birt, A., . . . Zwienenberg, M. (in press). A mul-
tilab preregistered replication of the ego-depletion effect.
Perspectives on Psychological Science.
Hagger, M. S., Wood, C., Stiff, C., & Chatzisarantis, N. L. (2010).
Ego depletion and the strength model of self-control: A
meta-analysis. Psychological Bulletin, 136, 495–525.
*Howard, M. A., & Marczinski, C. A. (2010). Acute effects of a
glucose energy drink on behavioral control. Experimental
and Clinical Psychopharmacology, 18, 553–561.
*Job, V., Walton, G. M., Bernecker, K., & Dweck, C. S. (2013).
Beliefs about willpower determine the impact of glucose
on self-control. Proceedings of the National Academy of
Sciences, USA, 110, 14837–14842.
*Kelly, C. L., Sünram-Lea, S. I., & Crawford, T. J. (2015). The role
of motivation, glucose and self-control in the antisaccade
task. PLoS ONE, 10, Article e0122218. doi:10.1371/journal
.pone.0122218
Kurzban, R. (2010). Does the brain consume additional glu-
cose during self-control tasks? Evolutionary Psychology, 8,
244–259.
Lakens, D. (2015). What p-hacking really looks like: A comment
on Masicampo and LaLande (2012). Quarterly Journal of
Experimental Psychology, 68, 829–832.
*Lange, F., & Eggert, F. (2014). Sweet delusion. Glucose drinks
fail to counteract ego depletion. Appetite, 75, 54–63.
*Lange, F., Seer, C., Rapior, M., Rose, J., & Eggert, F. (2014).
Turn it all you want: Still no effect of sugar consumption on
ego depletion. Journal of European Psychology Students,
5(3), 1–8. doi:10.5334/jeps.cc
by guest on July 11, 2016pss.sagepub.comDownloaded from
8 Vadillo et al.
*Masicampo, E. J., & Baumeister, R. F. (2008). Toward a physi-
ology of dual-process reasoning and judgment: Lemonade,
willpower, and expensive rule-based analysis. Psychological
Science, 19, 255–260.
*Molden, D. C., Hui, C. M., Scholer, A. A., Meier, B. P., Noreen,
E. E., D’Agostino, P. R., & Martin, V. (2012). Motivational
versus metabolic effects of carbohydrates on self-control.
Psychological Science, 23, 1137–1144.
*Open Science Collaboration. (2015). Estimating the repro-
ducibility of psychological science. Science, 349, 943.
doi:10.1126/science.aac4716
Orquin, J. L., & Kurzban, R. (2016). A meta-analysis of blood
glucose effects on human decision making. Psychological
Bulletin, 142, 546–567.
Osman, M. (2014). Future-minded: The psychology of agency
and control. Basingstoke, England: Palgrave Macmillan.
Owen, L., Scholey, A. B., Finnegan, Y., Hu, H., & Sünram-Lea,
S. I. (2012). The effect of glucose dose and fasting interval on
cognitive function: A double-blind, placebo-controlled, six-
way crossover study. Psychopharmacology, 220, 577–589.
*Sanders, M. A., Shirk, S. D., Burgin, C. J., & Martin, L. L. (2012).
The gargle effect: Rinsing the mouth with glucose enhances
self-control. Psychological Science, 23, 1470–1472.
Schimmack, U. (2012). The ironic effect of significant results
on the credibility of multiple-study articles. Psychological
Methods, 17, 551–566.
Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-
positive psychology: Undisclosed flexibility in data collec-
tion and analysis allows presenting anything as significant.
Psychological Science, 11, 1359–1366.
Simonsohn, U., Nelson, L. D., & Simmons, J. P. (2014).
P-curve: A key to the file-drawer. Journal of Experimental
Psychology: General, 143, 534–547.
*Wang, X. T., & Dvorak, R. D. (2010). Sweet future: Fluctuating
blood glucose levels affect future discounting. Psychological
Science, 21, 183–188.
by guest on July 11, 2016pss.sagepub.comDownloaded from