Content uploaded by Wouter Boendermaker
Author content
All content in this area was uploaded by Wouter Boendermaker on Jul 30, 2018
Content may be subject to copyright.
Executive Functions and Motivation as Moderators of the
Relationship Between Automatic Associations and Alcohol
Use in Problem Drinkers Seeking Online Help
Denise S. van Deursen, Elske Salemink, Wouter J. Boendermaker, Thomas Pronk,
Wilhelm Hofmann, and Reinout W. Wiers
Background: Dual process models posit that problem drinking is maintained by an imbalance
between relatively strong automatic processes and weak controlled processes, a combination of execu-
tive functions and motivation. Few studies have examined how the interplay between automatic pro-
cesses and executive functions is affected by motivation to change. This study examined this
relationship in problem drinkers seeking online help to change their alcohol use. It was expected that
executive functions (i.e., working memory, response inhibition) would moderate the relationship
between automatic (valence and approach) associations and alcohol use and that this effect would be
stronger in individuals with strong motivation to change.
Methods: A sample of 302 problem drinkers (mean age: 51.7 years) participated in this study as part
of the baseline assessment before an Internet intervention. Participants completed an online version of
the brief Implicit Association Test (valence and approach associations), the self-ordered pointing task
(working memory), the Stroop task (response inhibition), the Readiness to Change Questionnaire (mo-
tivation to change), and the Timeline Follow-Back Questionnaire (alcohol use). Hierarchical moderated
regression analysis was used to test the 4 hypothesized 3-way interactions.
Results: As expected, the interaction between valence associations and working memory only pre-
dicted alcohol use among individuals with strong motivation. This pattern was neither found for
response inhibition nor for approach associations.
Conclusions: Results provide partial support for the moderating role of motivation in the interplay
between automatic processes and executive functions. Future studies should investigate this relationship
in participants with the full range of motivation and alcohol use.
Key Words: Problem Drinking, Automatic Associations, Executive Functions, Motivation to
Change, Brief Implicit Association Test.
ONE OF THE defining characteristics of alcohol depen-
dence is the continuation of excessive alcohol use,
despite the motivation to restrain or reduce (American Psy-
chiatric Association, 2013). Dual process models of addic-
tion (Gladwin et al., 2011; Wiers et al., 2007) attribute this
discrepancy between motivation and behavior to an imbal-
ance between automatic and controlled cognitive determi-
nants of alcohol use. Automatic processes are relatively fast,
require lower levels of cognitive resources, and are less
amendable to conscious effort. Extended periods of heavy
drinking are thought to increase the strength of automatic
processes, resulting in a selective attention for alcohol-related
cues, automatic positive and approach associations with
alcohol, and automatic action tendencies toward alcohol.
The extent to which these automatic biases influence alcohol
use is thought to depend on controlled processes: a combina-
tion of the motivation to restrain and executive functions.
Several executive functions can been distinguished, including
set shifting, updating and monitoring working memory, and
inhibiting a prepotent response (Miyake et al., 2000).
Although separable, these functions also show considerable
overlap, which is likely caused by shared task demands and
the fact that they all require suppression of irrelevant infor-
mation (Miyake et al., 2000). In the context of dual process
models, working memory and response inhibition have been
studied most.
Several studies in (at-risk) adolescents and heavy-drinking
students have found evidence that the influence of automatic
processes on alcohol use is moderated by working memory
or response inhibition. In line with the predictions derived
from dual process theories, automatic biases were shown to
be more strongly related to alcohol use in individuals with
From the Addiction Development and Psychopathology (ADAPT)
Lab (DSD, ES, WJB, TP, RWW),Department of Psychology,
University of Amsterdam, Amsterdam, the Netherlands; and Social Cog-
nition Center Cologne (WH), University of Cologne, Cologne, Germany.
Received for publication September 19, 2014; accepted June 16, 2015.
Reprint requests: Denise van Deursen, MSc, Addiction Development
and Psychopathology (ADAPT) Lab, Department of Psychology,
University of Amsterdam, Weesperplein 4, 1018 XA Amsterdam, the
Netherlands. Tel.: +31-(0)20-5256212; Fax: +31-(0)20-639-0279;
E-mail: d.s.vandeursen@uva.nl
Clinical trial registration details: Netherlands Trial Register:
NTR3875.
Copyright ©2015 by the Research Society on Alcoholism.
DOI: 10.1111/acer.12822
1788 Alcohol Clin Exp Res, Vol 39, No 9, 2015: pp 1788–1796
ALCOHOLISM:CLINICAL AND EXPERIMENTAL RESEARCH Vol. 39, No. 9
September 2015
low compared to high working memory (Grenard et al.,
2008; Salemink and Wiers, 2014; Thush et al., 2008) and
response inhibition (Houben and Wiers, 2009; Peeters et al.,
2012, 2013), although this pattern was not always replicated
(van Hemel-Ruiter et al., 2011; Pieters et al., 2012). As these
studies have all focused on either working memory or
response inhibition, there is little empirical evidence to argue
why one or the other should be more important in the inter-
play with automatic processes in alcohol use. In the domain
of eating behavior, it was shown that both working memory
and response inhibition independently moderate the relation
between automatic valence associations and candy consump-
tion (Hofmann et al., 2009), but this finding has yet to be
replicated in context of alcohol use. Similarly, different auto-
matic cognitive biases have been studied (i.e., positive alco-
hol associations, alcohol approach associations, and
approach action tendencies), but rarely in conjunction. While
theoretically these constructs are expected to be related
(Wiers et al., 2007), this is not consistently supported by
research (e.g., van Hemel-Ruiter et al., 2011; Lindgren et al.,
2013; Thush et al., 2007; van den Wildenberg et al., 2006),
possibly due to the limited reliability of implicit (reaction
time) measures. In sum, there is evidence that the relation-
ship between automatic processes and alcohol use is moder-
ated by executive functions, but it is unknown to what extent
different indicators of automatic biases and executive func-
tions uniquely contribute to the prediction of alcohol use.
The interplay between executive functions and controlled
processes has also been studied, although research in this
area has largely focused on alcohol expectancies and less on
the role of motivation to restrain. The aforementioned study
by Thush and colleagues (2008) found a moderating effect of
working memory capacity in the expected direction: Explicit
alcoholexpectancieswerepredictiveofalcoholuseinat-risk
adolescents with higher but not in those with lower levels of
working memory capacity. In contrast, 2 recent studies in
large student samples have found little to no evidence for the
expected interactions between alcohol expectancies and cog-
nitive functions (Lavigne, 2013; Littlefield et al., 2011),
thereby countering the predictions made by dual process
models. It has to be noted that these studies, as well as those
studying the moderation of automatic processes, have been
conducted in (at-risk) adolescents and (heavy drinking) stu-
dents, who are likely to have experienced relatively few nega-
tive consequences of their alcohol use and are therefore
expected to have less negative expectancies regarding the
effects of drinking alcohol and less motivation to restrain
compared to individuals who are already experiencing alco-
hol use problems (Jones and McMahon, 1998).
Only a few studies have attempted to test the joint contri-
bution of automatic biases, executive functions, and motiva-
tion to restrain (Ostafin et al., 2008; Sharbanee et al., 2013;
Tahaney et al., 2014). An experimental study in heavy-drink-
ing adults found that positive alcohol associations were more
predictive of alcohol use during a taste test after performing
a task that depleted cognitive recourses, than after a control
task requiring little cognitive effort (Ostafin et al., 2008).
Importantly, all participants were given a situational incen-
tive to restrain their alcohol use, consistent with the assump-
tion that the interplay between automatic processes and
executive control only becomes apparent when individuals
are motivated to exert control over their impulses. Recently,
this assumption was explicitly tested in a heavy-drinking, lar-
gely student sample (Tahaney et al., 2014). As expected, it
was found that the interaction between appetitive responses
and executive functions predicted real-life alcohol use in indi-
viduals with high, but not in those with low levels of
restraint. However, instead of using automatic biases as a
measure of appetitive reactions to alcohol, this study
assessed participants’ explicit response to being exposed to
alcohol (i.e., the urge to drink, and the anticipated stimulant
effects of drinking). It is unclear whether these findings gen-
eralize to automatic processes, as the relation between auto-
matic biases and explicit craving or expectancies tends to be
small (Reich et al., 2010), although likely strengthened by
cue exposure (Field et al., 2009). Furthermore, it remains to
be seen whether the same pattern would be observed in prob-
lem drinkers who are motivated to change. As the level of
alcohol-related problems increases, motivation to restrain is
thought to increase (Wiers et al., 2007). At the same time,
executive functions are expected to weaken, while automatic
biases are thought to strengthen after prolonged heavy alco-
hol use. It would therefore be highly relevant—both from a
theoretical and from a clinical perspective—to investigate
whether the moderating role of executive functions and moti-
vation also holds for problem drinkers.
This study aimed to answer this question by examining
these interactions in a large sample of adult problem drinkers
seeking online help to change their alcohol use. We hypothe-
sized that executive functions would moderate the relation-
ship between automatic associations and drinking and that
this effect would be stronger in individuals with strong moti-
vation to change. More specifically, we expected that alcohol
use would be predicted by the interactions between (i) auto-
matic valence associations, working memory, and motiva-
tion; (ii) automatic valence associations, response inhibition,
and motivation; (iii) automatic approach associations, work-
ing memory, and motivation; and (iv) automatic approach
associations, response inhibition, and motivation.
MATERIALS AND METHODS
Participants
Participants were seeking online help in the form of cognitive bias
modification and were recruited through newspaper articles, a tele-
vision documentary on cognitive bias modification, the Addiction
Development and Psychopathology (ADAPT) Lab website, the
DrinkingLess website (Riper et al., 2008), and word of mouth com-
munication with existing participants. In total, 689 participants
completed the consent form (see Fig. 1 for the flow of participants
through the study). Participants had to meet the following inclusion
criteria to be included in the study: (i) an Alcohol Use Disorders
Identification Test (AUDIT; Babor et al., 2001) score >7; (ii)
AUTOMATIC ASSOCIATIONS AND PROBLEM DRINKING 1789
alcohol consumption of >21 (men) or >14 (women) standard drinks
(containing 10 g of ethanol) a week in the past 2 weeks, as assessed
with an adapted version of the Timeline Follow-Back (TLFB;
Sobell and Sobell, 1992) method; (iii) age between 18 and 64 years;
(iv) Internet access; and (v) no professional treatment for problem
drinking at the start of the study. A total of 427 problem drinkers
who wanted to reduce their alcohol use met the inclusion criteria.
The study was approved by the Ethics Committee of the psychology
department of the University of Amsterdam and registered at the
Netherlands Trial Register (NTR3875).
Measures
This study used the baseline measurement of a larger research
project on the efficacy of cognitive bias modification (for an over-
view of all measures, see van Deursen et al., 2013).
Self-Ordered Pointing Task. A computerized version of the self-
ordered pointing task (SOPT; Petrides and Milner, 1982) was used
to assess working memory capacity. Participants were presented
with a grid of pictures of concrete objects and were instructed to
click on each picture just once. After every click, the location of the
pictures changed, and participants could not click on the picture
that appeared at the location of the previous picture they selected.
The task consisted of a practice block of 4 pictures, and 5 test blocks
with 6, 8, 10, 12, and 12 pictures. The outcome measure was the
total number of correctly selected pictures during the test blocks,
ranging from 5 to 48, with higher scores indicating better working
memory capacity. This measure has been shown to have good test–
retest reliability, r=0.82 (Ross et al., 2007).
Stroop Task. The Stroop task was used to assess response inhi-
bition (MacLeod, 1991). Participants had to indicate the color of a
cue presented in the middle of the screen as quickly as possible by
pressing the e (red), f (yellow), j (green), or i (blue) key on the key-
board. The task started with a practice block of 40 trials. The next
block was an alcohol Stroop of 56 trials, which was not used in this
study. The final block consisted of the classical Stroop of 56 trials,
with 8 congruent trials (e.g., red in red), 24 incongruent trials (e.g.,
red in green), and 24 neutral trials (e.g., %%% in green). The out-
come measure was the difference score between mean response times
on incongruent minus neutral trials, with higher scores indicating
more Stroop interference and poorer response inhibition. The cur-
rent version of the task was found to have acceptable test–retest reli-
ability, r=0.73 (Peeters et al., 2013).
Brief Implicit Association Test. Two versions of the brief Impli-
cit Association Test (bIAT; Sriram and Greenwald, 2009) were used
to measure valence and approach associations with alcohol. Partici-
pants had to indicate whether the word presented in the middle of
the screen belonged to 1 or 2 categories at the top of the screen, by
pressing the designated “yes” or “no” buttons on the keyboard. As
689 signed the consent form
622 completed the screening
for inclusion criteria 185 excluded, because (>1 reason possible):
•16 AUDIT <8
•38 TLFB < 22 drinks (men) or <15 drinks (women) a
week in the past two weeks
•36 Age >64 years
•19 did not have (near) daily internet access
•64 currently received professional treatment for
problem drinking
10 removed from the analyses, because they were not
unique parcipants (created a new account aer being
excluded)
437 included in the study
302 included in the analycal
sample
36 parcipants were removed from one or more
regression analyses, because they were mulvariate
outliers, as indicated by Mahalanobis distances
exceeding the crical chi-square value for 10
predictors: 29.59 (regression 1: 17, regression 2: 17) or
11 predictors: 31.26 (regression 3: 12, regression 4: 11)
67 did not complete the screening for inclusion criteria
17 removed from the analysis, because they had
excessive error scores (>35%) on the valence bIAT (6),
the approach bIAT (5) or both (6)
108 did not complete at least one of the the following
measures : valence bIAT, approach bIAT, Stroop task,
SOPT, RCQ
427 valid parcipants
included in the study
Fig. 1. Participant flowchart.
1790 VAN DEURSEN ET AL.
in the study of Menatti and colleagues (in press), the 7-block struc-
ture of the IAT was maintained, but with fewer trials per block. This
bIAT version was found to have good internal consistency, with
Cronbach’s alphas ranging from 0.73 to 0.79. In the first block of
the valence bIAT (12 trials), alcoholic (wine, beer, vodka) or nonal-
coholic drinks (coke, water, juice) had to be classified as either
belonging to the category “alcohol” or not. In the second block (12
trials), words had to be classified as either pleasant (funny, cheerful,
sociable) or not (nauseous, sad, tired). In the third (12 trials) and
fourth (24 trials) block, the target (alcohol) and attribute (pleas-
ant) categories were combined. In the fifth block (12 trials), the
opposite attribute category, unpleasant, was practiced and then
combined with the target category (alcohol) in the sixth (12 trials)
and seventh (24 trials) block. The approach bIAT (Palfai and
Ostafin, 2003) followed the same structure, with the attributes
“approach” (grab, touch, take) and “avoid” (let go, avert,
ignore). The order of the 2 bIATs was randomized over partici-
pants, as well as the order of the combined blocks (but if the
valence bIAT started with the alcohol-pleasant block, the
approach bIAT always started with the alcohol-approach block
and the other way around), and correct response–key combina-
tion (yes =iandno=e or vice versa). The outcome measure
was the standardized difference in latencies (D600 score)between
the different combined blocks (Greenwald et al., 2003).
Self-Report Measures. The Readiness to Change Questionnaire
(RCQ; Defuentes-Merillas et al., 2002) was used to assess partici-
pants’ motivation to change their alcohol use. Participants indicated
to what extent they agreed with 12 statements, reflecting the precon-
templation (e.g., “It’s a waste of time thinking about my drinking”),
contemplation (e.g., “My drinking is a problem sometimes”), or
action stage (e.g., “I am trying to drink less than I used to”). In addi-
tion to assigning participants to a stage of change, a total score
ranging from 12 to 60 was computed by reversing scores on the pre-
contemplation items, which has been shown to have good internal
consistency (a=0.85; Budd and Rollnick, 1996). Note that the
RCQ was administered after a set of questionnaires on which partic-
ipants received personalized feedback, with the aim of motivating
participants for the intervention (i.e., a second assessment of the
TLFB and AUDIT, and a questionnaire on the perceived pros and
cons of drinking alcohol. These questionnaires were not used in this
study). This may have increased participants’ motivation to change.
The TLFB (Sobell and Sobell, 1992) method was used to measure
participants’ alcohol use in the past 14 days. Participants indicated
the number of standard alcoholic drinks they drank on each of the
specified days, using drop-down menu ranging from 0 to 40. The
number of alcoholic drinks per day was then added to compute the
total alcohol use over the past 2 weeks.
The AUDIT (Babor et al., 2001) was used to measure partici-
pants’ alcohol-related problems in the past 12 months. The ques-
tionnaire consisted of 10 items and scores ranged from 0 to 40.
Procedure
The study was entirely completed online. After creating an
account with their email address, participants were first informed
about the goals and the contents of the study. A copy of this infor-
mation letter was automatically emailed to participants and
included information on how to contact the researchers by email or
telephone in case they experienced technical difficulties or if they
had any questions regarding the study. Contact details were also
included in an overview of frequently asked questions, a link to
which was shown before the start of every new measure. As the reac-
tion time tasks were not suitable for use on smart phones, partici-
pants could only complete the study on a PC or laptop. After giving
online informed consent, participants first filled out a brief auto-
mated screening for eligibility, including the TLFB and the AUDIT.
Participants who were included in the study then completed the 2
versions of the bIAT, the Stroop task, and the SOPT, in random
order. The RCQ was administered last. In total, the assessment took
approximately 75 minutes to complete.
RESULTS
Sample
Due to missing data (n=108) and excessive errors
(n=17), the analytical sample consisted of 302 participants
(see Fig. 1 for participant flowchart). Characteristics of the
sample are shown in Table 1. Participants had a mean age of
51.7 years (SD =8.4), and about two-thirds were female
(64.9%). On average, participants drank 83.6 (SD =43.0)
alcoholic drinks during the past 14 days, which equals almost
6 standard units per day. The mean AUDIT score was 23.1
(SD =6.2), suggesting likely alcohol dependence (Babor
et al., 2001). Men drank more than women t(143) =5.51,
p<0.001. Based on the RCQ, all participants were in the
contemplation (84.8%) or action (15.2%) stages of change,
which corresponds with the fact that they were seeking help
to change their alcohol use. At group level, participants’
Stroop scores significantly differed from 0, confirming Stroop
interference, t(301) =20.66, p<0.001. Participants’ D600
scores on the valence bIAT were not significantly different
from 0, t(301) =1.83, p=0.069, indicating no difference in
the strength of participants’ positive and negative valence
associations with alcohol. Participants did on average show
relatively strong approach compared to avoidance associa-
tions with alcohol: D600 scores on the approach bIAT were
significantly greater than 0, t(301) =12.62, p<0.001. As the
D600 scores of both the valence, t(300) =0.866, p<0.001,
and the approach, t(300) =0.688, p=0.008, bIAT depended
on the order of the blocks within the task (alcohol-pleasant/
approach first, or alcohol-unpleasant/avoid first), block order
was included in the analyses. The order in which the 2 bIAT
versions were administered only significantly influenced the
approach bIAT, t(300) =5.485, p=0.027, and was only
included in analyses involving this task. Neither of the bIAT
scores was affected by the key–response combination (yes =i
and no =eorviceversa).
Internal Consistencies and Correlations
The 2 versions of the bIAT both showed acceptable relia-
bility: The Spearman–Brown corrected correlation between
the D600 score of the practice and test blocks was r=0.68
for the valence bIAT and r=0.60 for the approach bIAT.
The Stroop interference score had poor (r=0.45) even–odd
reliability. The (5 blocks of the) SOPT showed good reliabil-
ity (a=0.76), the reliability of the RCQ was good
(a=0.79), and the TLFB had very high reliability
(a=0.93).
Correlations between all study variables are shown in
Table 1. There was a strong positive correlation between
AUTOMATIC ASSOCIATIONS AND PROBLEM DRINKING 1791
alcohol use (TLFB) and alcohol problems (AUDIT). Alco-
hol problems were also positively related to motivation
(RCQ), indicating that those with more alcohol problems
were more motivated to change their alcohol use. Stronger
positive alcohol associations were related to higher alcohol
use (TLFB) and poorer working memory (SOPT).
Regression Analyses
The 4 hypothesized 3-way interactions were tested in 4
hierarchical moderated regression analyses. The dependent
variable was alcohol use in the past 2 weeks (TLFB), which
was positively skewed and therefore log-transformed to
enhance normality of the data.
1
To arrive at the correct stan-
dardized beta weights, all continuous predictor variables
were z-standardized and interaction terms were created with
these z-standardized variables. Multivariate outliers were
determined separately for each of the 4 hierarchical moder-
ated regression analyses and were removed only from the rel-
evant analysis (see Fig. 1 for details). At Step 1, age and
gender were entered. The relevant individual predictors were
added at Step 2: the order of the bIAT blocks and tasks,
valence or approach associations (bIAT), motivation to
change (RCQ), and inhibitory control (Stoop) or working
memory capacity (SOPT). All relevant 2-way interactions
between automatic associations, executive functions, and
motivation were added at Step 3. At Step 4, the hypothesized
3-way interactions between automatic associations, executive
functions, and motivation were added.
Table 1. Means, Standard Deviations, and Pearson Correlations
123456 78
M51.68 0.06 0.31 49.65 40.28 171.75 83.57 23.06
SD 08.36 0.55 0.42 05.50 04.24 144.49 42.97 06.18
Median 53 0.10 0.30 50 41 166.54 75 24
Minimum 25 1.52 1.11 32 11 416.00 29 9
Maximum 64 1.61 1.38 60 47 661.13 340 40
1. Age —
2. Valence bIAT 0.03 —
3. Approach bIAT 0.04 0.02 —
4. RCQ 0.03 0.04 0.02 —
5. SOPT 0.09 0.12* 0.01 0.04 —
6. Stroop 0.01 0.09 0.04 0.09 0.02 —
7. TLFB 0.00 0.15* 0.01 0.05 0.01 0.03 —
8. AUDIT 0.06 0.05 0.02 0.23*** 0.10 0.03 0.44*** —
*p<0.05, ***p<0.001.
bIAT, brief Implicit Association Test. Positive numbers on the valence or approach bIAT reflect relatively strong positive or approach associations with
alcohol. RCQ, Readiness to Change Questionnaire; SOPT, self-ordered pointing task. A positive SOPT score indicates larger working memory capacity.
A positive Stroop score reflects more problems with inhibition. TLFB, TimelineFollow-Back; AUDIT, Alcohol Use Disorders Identification Test.
Table 2. Hypothesis 1: Predicting Alcohol Use with the Valence
bIAT 9SOPT 9RCQ Interaction
Step 1 Step 2 Step 3 Step 4
DR²0.12 0.04 0.01 0.03
DF18.57 2.98 1.37 8.92
p<0.001 0.019 0.251 0.003
Variable (b)
Sex 0.329*** 0.311*** 0.315*** 0.309***
Age 0.010 0.014 0.016 0.016
Valence bIAT block order 0.060 0.065 0.065
Valence bIAT D-score 0.075** 0.081** 0.083**
RCQ 0.021 0.024 0.040
SOPT 0.053* 0.053* 0.053*
Valence bIAT 9SOPT 0.040 0.055*
Valence bIAT 9RCQ 0.024 0.031
RCQ 9SOPT 0.025 0.020
Valence bIAT 9RCQ 9SOPT 0.084**
*p<0.05, **p<0.01, ***p<0.001.
For abbreviations, see notes for Table 1. Sex was coded as 0 =male
and 1 =female. 4
4,1
4,2
4,3
4,4
4,5
4,6
4,7
4,8
4,9
5
Negative (-1 SD) Positive (+1 SD)
ALCOHOLUSE
(log transformed)
Low WMC (-1 SD)
High WMC (+ 1 SD)
ALCOHOL ASSOCIATIONS
Fig. 2. Relationship between automatic valence associations and alco-
hol use as a function of working memory capacity, for participants with
strong (+1 SD) motivation to change.
1
The K-S test indicated that TLFB scores differed significantly from normal
before, D(302) =0.112, p<0.001, but not after log-transformation, D
(302) =0.031, p=0.200.
1792 VAN DEURSEN ET AL.
The first regression analysis tested the interaction
between valence bIAT, SOPT, and RCQ (see Table 2).
The final model explained 19.2% of the total variance,
adjusted R
2
=0.16, F(10, 274) =6.49, p<0.001. Higher
alcohol use was related to male sex, positive alcohol asso-
ciations (valence bIAT), stronger working memory
(SOPT), and the 2-way interaction between SOPT and
valence bIAT. As hypothesized, alcohol use was signifi-
cantly predicted by the interaction between valence associ-
ations, working memory capacity, and motivation to
change. To decompose this 3-way interaction (Aiken and
West, 1991), the 2-way interaction between valence associ-
ations and working memory capacity was tested separately
for individuals with relatively strong and weak motivation
to change. Among those with weak motivation (1SD),
the interaction between valence associations and working
memory capacity was not significant (b=0.028,
p=0.429). Furthermore, an explorative simple slope
analysis among those with weak motivation showed no
significant relation between valence associations and alco-
hol use (b=0.065, p=0.206). As predicted, however, in
those with strong motivation, the interaction between
valence associations and working memory capacity was
significant (b=0.139, p=0.001; see Fig. 2). Simple
slope analyses confirmed that in individuals with a combi-
nation of strong motivation (+1 SD) and high working
memory capacity (+1 SD), valence associations were unre-
lated to alcohol use (b=0.025, p=0.633), whereas in
participants with strong motivation (+1 SD) but low work-
ing memory capacity (1 SD), positive alcohol associa-
tions were related to higher levels of alcohol use
(b=0.253, p<0.001).
2
In none of the 3 other hierarchical moderated regression
analyses that were conducted did the addition of the hypoth-
esized 3-way interaction between automatic associations,
executive functions, and motivation add to the prediction of
alcohol use (see Tables 3–5).
DISCUSSION
The results of this study partially support the expected
interplay between automatic associations, executive func-
tions, and motivation to change. Main findings can be sum-
marized as follows: First, as hypothesized, motivation to
Table 3. Hypothesis 2: Predicting Alcohol Use with the Valence
bIAT 9Stroop 9RCQ Interaction
Step 1 Step 2 Step 3 Step 4
DR²0.12 0.03 0.02 0.01
DF19.902.132.042.24
p<0.001 0.077 0.109 0.136
Variable (b)
Sex 0.339*** 0.327*** 0.320*** 0.320***
Age 0.004 0.004 0.001 0.002
Valence bIATblock order 0.062 0.056 0.052
Valence bIAT D-score 0.070** 0.071** 0.074**
RCQ 0.011 0.014 0.018
Stroop 0.015 0.005 0.008
Valence bIAT 9Stroop 0.002 0.001
Valence bIAT 9RCQ 0.014 0.010
RCQ 9Stroop 0.067* 0.062*
Valence bIAT 9RC Q 9Stroop 0.057
*p<0.05, **p<0.01, ***p<0.001.
For abbreviations, see notes for Table 1. Sex was coded as 0 =male
and 1 =female.
Table 4. Hypothesis 3: Predicting Alcohol Use with the Approach
bIAT 9SOPT 9RCQ Interaction
Step 1 Step 2 Step 3 Step 4
DR²0.12 0.01 0.02 0.00
DF19.830.841.990.211
p<0.001 0.523 0.116 0.639
Variable (b)
Sex 0.334*** 0.319*** 0.327*** 0.329***
Age 0.011 0.019 0.017 0.018
Approach bIAT block order 0.017 0.022 0.022
Approach bIAT task order 0.035 0.024 0.025
Approach bIAT D-score 0.004 0.005 0.006
RCQ 0.022 0.016 0.015
SOPT 0.045 0.040 0.041
Approach bIAT 9SOPT 0.037 0.037
Approach bIAT 9RCQ 0.044 0.043
RCQ 9SOPT 0.029 0.030
Approach bIAT 9RCQ
9SOPT
0.015
***p<0.001.
For abbreviations, see notes for Table 1. Sex was coded as 0 =male
and 1 =female.
Table 5. Hypothesis 4: Predicting Alcohol Use with the Approach
bIAT 9Stroop 9RCQ Interaction
Step 1 Step 2 Step 3 Step 4
DR²0.12 0.01 0.02 0.00
DF19.810.311.870.02
p<0.001 0.904 0.135 0.876
Variable (b)
Sex 0.336*** 0.330*** 0.331*** 0.331***
Age 0.009 0.010 0.006 0.006
Approach bIAT block order 0.025 0.014 0.014
Approach bIAT task order 0.035 0.028 0.028
Approach bIAT D-score 0.003 0.003 0.003
RCQ 0.007 0.001 0.001
Stroop 0.022 0.014 0.014
Approach bIAT 9Stroop 0.003 0.002
Approach bIAT 9RCQ 0.034 0.034
RCQ 9Stroop 0.058* 0.058
Approach bIAT 9RCQ
9Stroop
0.006
*p<0.05, ***p<0.001.
For abbreviations, see notes for Table 1. Sex was coded as 0 =male
and 1 =female.
2
The hypothesized 3-way interaction between valence bIAT, SOPT, and
RCQ was also found when Bonferroni correction was applied to account for
multiple testing, when the expected 3-way interactions were tested in 1 hierar-
chical moderated regression analysis instead of in 4 separate analyses, and
when the TLFB data were Winsorized instead of log-transformed.
AUTOMATIC ASSOCIATIONS AND PROBLEM DRINKING 1793
change moderated the interaction between valence associa-
tions and working memory capacity in the prediction of
(heavy) alcohol use. Decomposition of the interaction indi-
cated that working memory capacity moderated the relation-
ship between valence associations and drinking in
individuals with strong motivation, but not in those with
relatively weak motivation. In those with strong motivation
but low working memory capacity, positive associations were
associated with higher alcohol use. In individuals with both
strong motivation and high working memory capacity, posi-
tive associations were unrelated to alcohol use. Second, con-
trary to our expectations, this pattern was not found for
other indices of executive functions (i.e., response inhibition)
or automatic associations (i.e., approach associations). We
discuss these main findings in turn.
The finding that working memory capacity only moder-
ated the relationship between valence associations and alco-
hol use in those who were strongly motivated is consistent
with dual process models (Gladwin et al., 2011; Wiers et al.,
2007) and previous research (Tahaney et al., 2014) and sug-
gests that the ability to control one’s automatic impulse to
drink only becomes relevant when one is motivated to do so.
In other words, the combination of both strong motivation
and high working memory capacity appears to be required
to successfully inhibit the effect of automatic associations on
drinking. However, explorative testing in those with rela-
tively weak motivation revealed no significant relationship
between valence associations and alcohol use. If both high
working memory capacity and strong motivation are
required to control the influence of associations on drinking,
associations should be related to alcohol use in those with
weak motivation.
In contrast to our hypotheses, motivation only moderated
the interaction between working memory capacity and
valence associations; this effect did not hold for response
inhibition and approach associations. A possible explana-
tion for these discrepant results is the fact that the reliability
of the measure of response inhibition (Stroop) was poor
compared to that of the working memory capacity task
(SOPT). As the product of the reliabilities of the individual
predictors determines the reliability of an interaction term
(Busemeyer and Jones, 1983), this could explain why the
only interaction effect that was found included the executive
function measure with relatively higher reliability. Further-
more, the limited reliability of the reaction time measures
might play a role in the fact that neither the executive func-
tion tasks nor the 2 implicit measures were correlated, and
limits the conclusions that can be drawn from the (null)
results. Aside from the reliability problems, there are 2 other
possible explanations for why we did not find most of the
expected interactions, the first of which lies in the selection
of participants. In contrast to previous research, which
investigated the interaction between automatic processes,
executive functions, and motivation in participants with rel-
atively low motivation to restrain, the current study set out
to investigate this interaction in a sample of problem drin-
kers with relatively high motivation to change. As excessive
alcohol use was also an inclusion criterion, participants were
by definition dyscontrolled drinkers: motivated to restrain
their alcohol use, yet failing to do so. Naturally, this restricts
the possible variability in the studied variables, although the
ranges of both motivation and alcohol use were still ade-
quate to examine the hypothesized interactions. Alterna-
tively, it is possible that the factors that contribute to the
initiation of (heavy) alcohol use differ from those that main-
tain problematic alcohol use in the long term, and that the
interplay between automatic processes, executive functions,
and motivation becomes less pronounced as alcohol use
becomes more habitual (Everitt and Robbins, 2005;
Vollst€
adt-Klein et al., 2010). Ideally, future research should
include participants with the full range of motivation to
change and alcohol use, so that these explanations can be
formally tested.
A number of limitations are important to consider
when interpreting the results. First, although lack of relia-
bility is a general problem of many reaction time mea-
sures (e.g., Schmukle, 2005), several aspects of the design
of this study may have negatively influenced the reliability
and validity of the measures, specifically (i) the web-based
assessment, with little control over how the measures were
completed; (ii) the large number of measures, which may
have resulted in fatigue; (iii) the fact that the response
inhibition measure (classical Stroop task) was preceded by
an addiction Stroop task (not used in the current study),
which may have caused practice effects and decreased
Stroop interference; and (iv) the fact that motivation was
assessed after participants received personalized feedback
on questionnaires regarding their alcohol use, which is
likely to have increased their motivation to change.
Notwithstanding these potential concerns, there were also
advantages to these aspects of the design. For example,
the web-based assessment enabled us to include a large
group of problem drinkers, which would have been hard
to achieve in a more controlled, laboratory-based setting.
A different approach to reduce measurement error could
be the use of a latent variable approach, instead of using
single indicators of different components of executive
functions and automatic associations (Miyake et al.,
2000). Another limitation of this study is that about 25%
of all eligible participants did not complete all of the
measures and were thus removed from the analyses.
Although not uncommon during baseline assessments of
web-based interventions for problem drinking (Blankers
et al., 2011), this may have biased the data in the sense
that the participants who dropped out might have been
less motivated to change their alcohol use. Other factors
that may have contributed to dropout include technical
problems and difficulties understanding the task
instructions. Participants were repeatedly reminded that
they could contact the researchers by email or phone if
they had questions or experienced any problems, but it
cannot be ruled out that some simply discontinued their
1794 VAN DEURSEN ET AL.
participation. A final limitation is that the cross-sectional
nature of this study prevents causal conclusions. It would
be highly relevant to test these interactions in a longitudi-
nal and/or experimental design, for example, by investi-
gating whether retraining strong automatic processes
results in larger decreases in alcohol use in individuals
with weak executive functions and strong motivation,
compared to those with both strong executive functions
and strong motivation. Some steps in this direction have
recently been made (Eberl et al., 2013; Houben et al.,
2011), for example, in a study on approach bias modifica-
tion in alcoholic patients, which found larger training
effects in older patients, who presumably have lower levels
of executive functions.
To the best of our knowledge, this is the first study to
examine the joint interplay between automatic associa-
tions, executive functions, and motivation in a large sam-
ple of problem drinkers. Although the results were not
consistent across measures, the finding that motivation
interacted with working memory capacity and valence
associations suggests that future studies should not over-
look the role of motivation when studying the interaction
between automatic associations and executive functions.
More knowledge about this interplay would enable empir-
ically supported tailoring of interventions for problem
drinking.
ACKNOWLEDGMENTS
We thank Filip Smit for valuable comments on an earlier
version of this manuscript, Bruno Boutin for his technical
assistance, and Katherina Martin Abello, Brigitte Boon,
Filip Smit, and Heleen Riper for their help in enabling the
collaboration between the University of Amsterdam and
Trimbos Institute.
SOURCES OF FUNDING
DSD, WJB, TP, and RWW were supported by a VICI
grant (453-08-001), and ES was supported by a VENI grant
(451-10-029), both from the Dutch National Science Foun-
dation, N.W.O. TP received additional funding from the
European Union’s Seventh Framework Programme (FP7/
2007-2013), under Grant Agreement no 266813. The views
expressed here reflect those of the authors only, and the
European Union is not liable for any use that may be made
of the information contained therein.
CONFLICT OF INTEREST
None.
REFERENCES
Aiken LS, West SG (1991) Multiple Regression: Testing and Interpreting
Interactions. Sage, Thousand Oaks, CA.
American Psychiatric Association (2013) Diagnostic and Statistical Manual
of Mental Disorders. 5th ed. American Psychiatric Association, Washing-
ton, DC.
Babor T, Higgins-Biddle JC, Saunders J, Monteiro MG (2001) The Alcohol
Use Disorders Identification Test: Guidelines for Use in Primary Care.
2nded.WorldHealthOrganization,Geneva.
Blankers M, Koeter MWJ, Schippers GM (2011) Internet therapy versus
internet self-help versus no treatment for problematic alcohol use: a ran-
domized controlled trial. J Consult Clin Psychol 79:330–341.
Budd RJ, Rollnick S (1996) The structure of the readiness to change ques-
tionnaire: a test of Prochaska & DiClemente’s transtheoretical model. Br J
Health Psychol 1:365–376.
Busemeyer JR, Jones LE (1983) Analysis of multiplicative combination rules
when the causal variables are measured with error. Psychol Bull 93:549–
562.
Defuentes-Merillas L, Dejong C, Schippers G (2002) Reliability and validity
of the Dutch version of the readiness to change questionnaire. Alcohol
Alcohol 37:93–99.
van Deursen DS, Salemink E, Smit F, Kramer J, Wiers RW (2013) Web-
based cognitive bias modification for problem drinkers: protocol of a ran-
domised controlled trial with a 2x2x2 factorial design. BMC Public Health
13:674.
Eberl C, Wiers RW, Pawelczack S, Rinck M, Becker ES, Lindenmeyer J
(2013) Approach bias modification in alcohol dependence: do clinical
effects replicate and for whom does it work best? Dev Cogn Neurosci
4:38–51.
Everitt BJ, Robbins TW (2005) Neural systems of reinforcement for drug
addiction: from actions to habits to compulsion. Nat Neurosci 8:1481–
1489.
Field M, Munaf
o MR, Franken IH (2009) A meta-analytic investigation of
the relationship between attentional bias and subjective craving in sub-
stance abuse. Psychol Bull 135:589–607.
Gladwin TE, Figner B, Crone EA, Wiers RW (2011) Addiction, adolescence,
and the integration of control and motivation. Dev Cogn Neurosci 1:364–
376.
Greenwald AG, Nosek BA, Banaji MR (2003) Understanding and using the
implicit association test: I. An improved scoring algorithm. J Pers Soc Psy-
chol 85:197–216.
Grenard JL, Ames SL, Wiers RW, Thush C, Sussman S, Stacy AW (2008)
Working memory capacity moderates the predictive effects of drug-related
associations on substance use. Psychol Addict Behav 22:426–432.
van Hemel-Ruiter ME, de Jong PJ, Wiers RW (2011) Appetitive and regula-
tory processes in young adolescent drinkers. Addict Behav 36:18–26.
Hofmann W, Friese M, Roefs A (2009) Three ways to resist temptation: the
independent contributions of executive attention, inhibitory control, and
affect regulation to the impulse control of eating behavior. J Exp Soc Psy-
chol 45:431–435.
Houben K, Wiers RW (2009) Response inhibition moderates the relationship
between implicit associations and drinking behavior. Alcohol Clin Exp
Res 33:626–633.
Houben K, Wiers RW, Jansen A (2011) Getting a grip on drinking behavior
training working memory to reduce alcohol abuse. Psychol Sci 22:968–
975.
Jones BT, McMahon J (1998) Alcohol motivations as outcome expectancies,
in Treating Addictive Behaviors (Miller WR, Heather N eds), pp 75–91.
Plenum Press, New York, NY.
Lavigne AM (2013) Alcohol Cognitions and College Student Drinking: The
Moderating Effect of Executive Abilities. Open Access Dissertations, paper
49. Available at: http://digitalcommons.uri.edu/oa_diss/49. Accessed Jan-
uary 27, 2014.
Lindgren KP, Neighbors C, Teachman BA, Wiers RW, Westgate E, Green-
wald AG (2013) I drink therefore I am: validating alcohol-related implicit
association tests. Psychol Addict Behav 27:1–13.
Littlefield AK, Verg
es A, McCarthy DM, Sher KJ (2011) Interactions
between self-reported alcohol outcome expectancies and cognitive
functioning in the prediction of alcohol use and associated problems: a fur-
ther examination. Psychol Addict Behav 25:542–546.
AUTOMATIC ASSOCIATIONS AND PROBLEM DRINKING 1795
MacLeod CM (1991) Half a century of research on the Stroop effect: an inte-
grative review. Psychol Bull 109:163–203.
Menatti AR, Smyth FL, Teachman BA, Nosek BA (in press) Reducing
stigma toward individuals with mental illnesses: a brief online intervention.
Stigma Res Action.
Miyake A, Friedman NP, Emerson MJ, Witzki AH, Howerter A, Wager TD
(2000) The unity and diversity of executive functions and their contribu-
tions to complex “frontal lobe” tasks: a latent variable analysis. Cogn Psy-
chol 41:49–100.
Ostafin BD, Marlatt GA, Greenwald AG (2008) Drinking without thinking:
an implicit measure of alcohol motivation predicts failure to control alco-
hol use. Behav Res Ther 46:1210–1219.
Palfai TP, Ostafin BD (2003) Alcohol-related motivational tendencies in haz-
ardous drinkers: assessing implicit response tendencies using the modified-
IAT. Behav Res Ther 41:1149–1162.
Peeters M, Monshouwer K, Schoot RA, Janssen T, Vollebergh WA, Wiers
RW (2013) Automatic processes and the drinking behavior in early adoles-
cence: a prospective study. Alcohol Clin Exp Res 37:1737–1744.
Peeters M, Wiers RW, Monshouwer K, Schoot R, Janssen T, Vollebergh
WA (2012) Automatic processes in at-risk adolescents: the role of alcohol-
approach tendencies and response inhibition in drinking behavior. Addic-
tion 107:1939–1946.
Petrides M, Milner B (1982) Deficits on subject-ordered tasks after frontal-
and temporal-lobe lesions in man. Neuropsychologia 20:249–262.
Pieters S, Burk WJ, Van der Vorst H, Wiers RW, Engels RC (2012) The
moderating role of working memory capacity and alcohol-specific rule-set-
ting on the relation between approach tendencies and alcohol use in young
adolescents. Alcohol Clin Exp Res 36:915–922.
Reich RR, Below MC, Goldman MS (2010) Explicit and implicit measures
of expectancy and related alcohol cognitions: a meta-analytic comparison.
Psychol Addict Behav 24:13–25.
Riper H, Kramer J, Smit F, Conijn B, Schippers G, Cuijpers P (2008) Web-
based self-help for problem drinkers: a pragmatic randomized trial. Addic-
tion 103:218–227.
Ross TP, Hanouskova E, Giarla K, Calhoun E, Tucker M (2007) The relia-
bility and validity of the self-ordered pointing task. Arch Clin Neuropsy-
chol 22:449–458.
Salemink E, Wiers RW (2014) Alcohol-related memory associations in posi-
tive and negative affect situations: drinking motives, working memory
capacity, and prospective drinking. Psychol Addict Behav 28:105–113.
Schmukle SC (2005) Unreliability of the dot probe task. Eur J Pers 19:595–
605.
Sharbanee JM, Stritzke WG, Wiers RW, Young P, Rinck M, MacLeod C
(2013) The interaction of approach-alcohol action tendencies, working
memory capacity, and current task goals predicts the inability to regulate
drinking behavior. Psychol Addict Behav 27:649–661.
Sobell LC, Sobell MB (1992) Timeline Follow-back: a technique for assess-
ing self-reported ethanol consumption, in Measuring Alcohol Consump-
tion: Psychosocial and Biological Methods (Allen J, Litten RZ, Totowa NJ
eds), pp 41–72. Humana Press, Totowa, NJ.
Sriram N, Greenwald AG (2009) The brief implicit association test. Exp Psy-
chol 56:283–294.
Tahaney KD, Kantner CW, Palfai TP (2014) Executive function and appeti-
tive processes in the self-control of alcohol use: the moderational role of
drinking restraint. Drug Alcohol Depend 138:251–254.
Thush C, Wiers RW, Ames SL, Grenard JL, Sussman S, Stacy AW (2007)
Apples and oranges? Comparing indirect measures of alcohol-related cog-
nition predicting alcohol use in at-risk adolescents. Psychol Addict Behav
21:587–591.
Thush C, Wiers RW, Ames SL, Grenard JL, Sussman S, Stacy AW (2008)
Interactions between implicit and explicit cognition and working memory
capacity in the prediction of alcohol use in at-risk adolescents. Drug Alco-
hol Depend 94:116–124.
Vollst€
adt-Klein S, Wichert S, Rabinstein J, B€
uhler M, Klein O, Ende G, Her-
mann D, Mann K (2010) Initial, habitual and compulsive alcohol use is
characterized by a shift of cue processing from ventral to dorsal striatum.
Addiction 105:1741–
1749.
Wiers RW, Bartholow BD, van Den Wildenberg E, Thush C, Engels RCME,
Sher KJ, Grenard J, Ames SL, Stacy AW (2007) Automatic and controlled
processes and the development of addictive behaviors in adolescents: a
review and a model. Pharmacol Biochem Behav 86:263–283.
van den Wildenberg E, Beckers M, Van Lambaart F, Conrod PJ, Wiers RW
(2006) Is the strength of implicit alcohol associations correlated with alco-
hol-induced heart-rate acceleration? Alcohol Clin Exp Res 30:1336–1348.
1796 VAN DEURSEN ET AL.
