Content uploaded by Morten Moshagen
Author content
All content in this area was uploaded by Morten Moshagen on Oct 06, 2017
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=iart20
Download by: [217.251.79.52] Date: 03 June 2017, At: 11:10
Addiction Research & Theory
ISSN: 1606-6359 (Print) 1476-7392 (Online) Journal homepage: http://www.tandfonline.com/loi/iart20
A test of expectancy-value theory in predicting
alcohol consumption*
Jennifer Nicolai, Morten Moshagen & Ralf Demmel
To cite this article: Jennifer Nicolai, Morten Moshagen & Ralf Demmel (2017): A test of
expectancy-value theory in predicting alcohol consumption*, Addiction Research & Theory, DOI:
10.1080/16066359.2017.1334201
To link to this article: http://dx.doi.org/10.1080/16066359.2017.1334201
Published online: 03 Jun 2017.
Submit your article to this journal
View related articles
View Crossmark data
ORIGINAL ARTICLE
A test of expectancy-value theory in predicting alcohol consumption
Jennifer Nicolai
a
, Morten Moshagen
b
and Ralf Demmel
c
a
Cognition and Individual Differences, School of Social Sciences, Psychology, University of Mannheim, Mannheim, Germany;
b
Ulm University,
Ulm, Germany;
c
Bodelschwinghsche Stiftungen Bethel, University of M€
unster, Muenster, Germany
ABSTRACT
Background: Research on alcohol-related outcome expectancies has primarily focused on the likelihood
of the anticipated effects, while comparatively little attention has been paid to their subjective evalu-
ation. However, according to expectancy-value theory, the expectation that alcohol use will produce
certain consequences and the evaluation of those consequences jointly and interactively determine an
individual's decision to consume alcohol. Previous research on this issue was hampered by multiple
regression strategies that are plagued by measurement error and low statistical power.
Method: To overcome this limitation, we investigated expectancy-value interactions in predicting drink-
ing variables by drawing on latent variable methodology using the five expectancy-value dimensions
from the Comprehensive Alcohol Expectancy Questionnaire. Expectancy-value models were tested in a
sample of college students (N¼1053) and a sample of alcohol-dependent inpatients (N¼699).
Results: Significant expectancy-value interactions emerged concerning social assertiveness among stu-
dents as well as for aggression and tension reduction among alcohol-dependent inpatients. The rela-
tionship between expectancy and drinking was strongest for pronounced (either positive or negative)
valuations of the effect. Effect sizes were small, however.
Conclusions: The results are in partial agreement with basic premises of expectancy-value theory.
However, this study also identifies limits to the universal validity of expectancy-value theory, given that
prediction of alcohol use depends on the effect domains, alcohol outcome measures, and study
populations.
ARTICLE HISTORY
Received 28 July 2016
Revised 13 February 2017
Accepted 21 May 2017
KEYWORDS
Alcohol expectancies;
outcome valuations;
evaluations of drinking
outcomes; expectancy-value
theory; CAEQ; alcohol use
A test of expectancy-value theory in predicting
alcohol consumption
Beliefs about the likelihood of certain affective, cognitive,
and behavioral effects of alcohol use (alcohol-related out-
come expectancies, AE) have been shown to be powerful
predictors of alcohol-related behaviors (Settles et al. 2014),
such as the initiation and maintenance of drinking (Aas
et al. 1998; Jester et al. 2015), the transition from alcohol use
to alcohol abuse (Patrick et al. 2010), and the outcome fol-
lowing treatment (Jones & McMahon 1994; McKellar et al.
2006; Sawayama et al. 2012). Indeed, the mere belief that
alcohol has been consumed can alter behavior regardless of
actual alcohol consumption (Goldman et al. 1999) and both
subjective and behavioral effects of alcohol strongly depend
on the anticipated outcomes of drinking (Jones et al. 2001).
According to both subjective expected utility (Edwards
1954; Atkinson 1957) and social-cognitive theories (Fishbein
& Ajzen 1975; Bandura 1986; Kuther 2002), an individual's
tendency to perform a certain behavior is a function of an
individual’s expected probability of a certain outcome and
the incentive value (or valence) of this outcome. More spe-
cifically, expectancy-value theories (EVT) predict that the
more likely one perceives that a specific behavior will pro-
duce a specific positive outcome, and the more highly one
values the outcome, the more likely it is that the individual
will engage in the specific behavior. Thus, a particular AE
will affect drinking to the extent that the individual (posi-
tively or negatively) values the outcome. More specifically,
strong AE are likely to increase drinking if the expected out-
come is valued as highly positive and likely to decrease
drinking if the expected outcome is valued as highly nega-
tive. However, if an outcome is valued neutrally (neither
positively nor negatively), endorsing a particular expectancy
should play little role in adopting and maintaining the asso-
ciated behavior.
Although EVT thus inherently stresses the importance of
an individual’s evaluation of an anticipated effect of alcohol,
research on AE has primarily focused on the likelihood of
the anticipated effects, while comparatively little attention
has been paid to their subjective evaluation (Fromme et al.
1993). At the level of measurement instruments, the desir-
ability of an effect (‘positive’versus ‘negative’) is typically
determined a-priori by the researchers, whereas individuals
are not explicitly asked for their personal judgments of this
CONTACT Jennifer Nicolai nicolai@uni-mannheim.de Cognition and Individual Differences, School of Social Sciences, University of Mannheim, Schloss EO
254, 68161, Mannheim, Germany
This paper is based on part of a thesis submitted by the first author to the University of Muenster, Germany, in partial fulfillment of the requirements of the
Doctor of Philosophy degree.
ß2017 Informa UK Limited, trading as Taylor & Francis Group
ADDICTION RESEARCH & THEORY, 2017
https://doi.org/10.1080/16066359.2017.1334201
effect. However, ‘positive’and ‘negative’AE are not always
viewed as such by individuals (Fromme et al. 1993; Demmel
& Hagen 2003a,b; Ham et al. 2005; Valdivia & Stewart
2005). For example, whereas most individuals negatively
value cognitive-behavioral impairments following alcohol
consumption, some college students regard these effects as
positive motivators for drinking (e.g. Fromme et al. 1994).
Similarly, ‘negative’expectancies, which are theoretically
assumed to act as protective factors for drinking, have some-
times been shown to be related to more (rather than less)
drinking (e.g. Mann et al. 1987; McMahon et al. 1994). This
might reflect that heavier drinkers have more experience
with the negative effects of alcohol (Lee et al. 1999; Hasking
& Oei 2002), but may also suggest that ‘negative’expectan-
cies are not valued negatively by all individuals. In light of
the considerable variability in the evaluation of the effects of
alcohol, it is necessary to assess rather than to assume the
valence individuals attach to specific alcohol-related out-
comes (Fromme et al. 1993; Valdivia & Stewart 2005; Mallett
et al. 2008; Kachadourian et al. 2014).
Mixed results were obtained in studies that investigated
the role of an individual’s subjective evaluation of specific
AE (Leigh 1987; Fromme et al. 1993; Grube et al. 1995;
Burden & Maisto 2000; Fromme & D’Amico 2000;
Zamboanga 2006; Zamboanga et al. 2009). These discrepan-
cies might be due to differences in the populations studied
(adolescents, high school students, college students, commu-
nity samples, alcohol-dependent individuals), the background
variables included (age, gender, ethnicity, attitudes, etc.), the
assessment instruments used (e.g. AEQ, Brown et al. 1987;
CEOA, Fromme et al. 1993; CAEQ, Demmel & Hagen
2003a,b; B-CEOA, Ham et al. 2005), and the measure of
alcohol consumption (e.g. frequency, quantity, quantity/fre-
quency index, lifetime alcohol use, abstinence survivorship).
Moreover, the majority of studies failed to test for interactive
effects of expectancies and values. According to EVT, how-
ever, the effects of either expectancy or value on alcohol use
are interrelated: As evaluations increase, the effect of high
versus low expectancies on alcohol use should also increase,
and vice versa. Consequently, EVT explicitly predicts inter-
active effects of expectancies and values rather than isolated
(additive) effects of either or both variables (Nagengast et al.
2011).
Some studies examined interactive expectancy-value
effects by multiplying expectancies with values and regress-
ing drinking on this multiplicative composite without includ-
ing the main effects of expectancy and value (Bauman et al.
1985; Critchlow 1987; Stacy et al. 1990). However, such com-
posite scores represent a blend of interactive and additive
effects, because they also carry the information of the main
effects (Evans 1991; Cohen et al. 2013). A multiplicative
composite only represents a genuine interaction effect if the
components are also included in the regression equation.
Analyses of a product-term-only model can therefore not
inform the relationship of expectancies and values on
drinking.
Until now, only two studies assessed both additive and
interactive effects of expectancies and valuations on alcohol
consumption using an appropriate data analytic strategy.
Using a sample of 1751 adolescents, Grube and colleagues
(Grube et al. 1995) investigated the utility of outcome
expectancies, evaluations, and expectancy-value interactions
in predicting a drinking index derived from factor analysis.
Hierarchical regression analyzes revealed that that the inclu-
sion of values and expectancy-value interactions significantly
improved the prediction of drinking over-and-above expect-
ancies alone. The association between positive expectancies
and alcohol use was highest when they were valued as desir-
able, whereas negative expectancies were better predictors
when they were valued less desirable. In contrast, Jones and
McMahon (1996a) failed to find support for an increase in
the variance explained in post-treatment abstinence survivor-
ship of 128 alcohol-dependent individuals by the expectancy-
value interactions. However, the failure to find significant
contributions of the interaction terms in the latter study may
be due to lack of sufficient statistical power resulting from
the rather small sample. Indeed, interaction effects are typic-
ally quite small in magnitude, so that large sample sizes are
necessary to detect these effects reliably (McClelland & Judd
1993; Shieh 2010). This is a particular problem for the trad-
itional approach of investigating interaction effects in a mul-
tiple regression framework, as the measurement error
inherent in the main effects leads to a strong underestima-
tion of the interaction effect (McClelland & Judd 1993;
Nagengast et al. 2011).
In order to investigate the interactive effects of alcohol
outcome expectancies and subjective evaluations of alco-
hol's effects in predicting quantity and frequency of drink-
ing, the present study evaluated an interactive model using
latent variable methodology (e.g. Bollen 1989; Lomax &
Schumacker 2012; Kline 2015). Latent variable models
determine a latent construct of interest (e.g. a certain class
of outcome expectancies) indirectly by considering the
commonalities between observed indicator variables (e.g. a
set of items designed to measure the said construct).
These models assume that each observed indicator has
two causes: the common latent factor representing the
shared variance among all relevant indicator variables and
a specific factor representing variance that is unique to
the indicator. Thus, each unique factor captures measure-
ment error associated with a particular item, whereas the
common latent variable can be considered an error-free
measure of the latent variable of interest. As such, resort-
ing to latent variable models remedy the measurement
error problems associated with multiple regression tech-
nique (e.g. Kenny & Judd 1984). To go further beyond
previous research, five expectancy-value dimensions (rather
than just negative-positive dimensions) were considered.
Finally, we investigated two highly diverse samples (stu-
dents and inpatients), because relevant expectancies associ-
ated with the decision to drink have been shown to differ
substantially for casual drinkers and alcohol-dependent
individuals (Jones et al. 2001; Nicolai et al. 2010; Monk &
Heim 2013; Tibboel et al. 2015).
In accordance with EVT, it was hypothesized that expect-
ancy-value interactions significantly predict drinking meas-
ures over and beyond additive effects of outcome
expectancies and valuations alone. Specifically, given that
2 J. NICOLAI ET AL.
positive expectancies, in particular those related to social
assertiveness or social enhancement, are important correlates
of alcohol use in college student samples (e.g. Cable &
Sacker 2008; Ham & Hope 2003; Patrick et al. 2010), it
was expected that expectancy-value interactions most likely
occur in the prediction of drinking among students in the
social domain. Concerning alcohol dependent inpatients,
evidence suggests that relaxation and tension reduction
expectancies clearly distinguish problem drinkers from non-
problem drinkers (e.g. Goldman et al. 1991; Schmidt et al.
2013). Heavier drinkers also tend to view the negative
effects of alcohol as more benign (e.g. Williams &
Ricciardelli 1996) and endorse stronger aggression expect-
ancies (e.g. Connors et al. 1986; Townshend & Duka 2007).
Consequently, expectancy-value interactions were expected
to incrementally predict drinking among alcohol dependent
inpatients in particular in the tension reduction and aggres-
sion domains.
Methods
Participants and procedures
The study was a cross-sectional study based on two different
samples. First, data were collected from students of a large
German university during class time. In addition to this, a
‘snowball’strategy was also used, i.e. students were asked to
distribute the package of questionnaires to their peer-group.
All participants gave written informed consent and did not
receive any compensation for completing the study. Of the
N¼1053 participants, 59%were female. Mean age was 23.84
(SD ¼7.86) years. As the sample was predominantly com-
posed of college or university students, we refer to this sam-
ple as student sample. Participants reported drinking alcohol
on an average of 6.73 (SD ¼5.55) days during the past
30 days. The average grams of pure alcohol consumed on a
typical drinking occasion was 75.40 (SD ¼65.10), the mean
quantity-frequency index was 16.98 (SD ¼20.14), and the
number of standard drinks (defined as 14 grams of pure
alcohol) consumed during the past 30 days was M¼36.40
(SD ¼43.17).
In addition to this, data were collected from a clinical
sample of 699 alcohol-dependent inpatients (74%male). A
research assistant, at the place of recruitment, dispensed a
packet of questionnaires (containing a consent form, the
measures described below, and several other questionnaires
that are not pertinent to the present study). Participants
were free to complete the questionnaires in their own time.
At the time of questionnaire completion, all participants
were completely detoxified and free from clinically signifi-
cant symptoms of acute alcohol withdrawal. Participation
was voluntary and anonymous. Participants did not receive
any compensation for completing the study. Mean age was
43.59 (SD ¼8.00) years. Patients reported drinking alcohol
on an average of 19.97 days (SD ¼10.12) within the last
30 days before hospitalization with a mean of 217.99
(SD ¼150.47) grams of pure alcohol consumed on a typical
drinking occasion. The quantity-frequency index was
M¼152.60 (SD ¼140.07), and the number of standard
drinks consumed during the past 30 days before hospitaliza-
tion was M¼326.99 (SD ¼300.16).
The study was designed in accordance with the
Declaration of Helsinki and its current amendments. The
Institutional Review Board of the University of M€
unster,
Germany, approved the study and written informed consent
was obtained from all participants prior to their participation.
Measures
Drinking behavior
Alcohol use was assessed by the following measures: (1) the
number of drinking days during the past 30 days (frequency)
and (2) the number of drinks (beer, wine, or distilled spirits)
on an average drinking day during the past 30 days (quan-
tity). Responses to the drinking items were transformed into
grams of pure alcohol following the guidelines outlined in
B€
uhringer et al. (2002).
Alcohol expectancy and value ratings
A revised form of the Comprehensive Alcohol Expectancy
Questionnaire (CAEQ) (Demmel & Hagen 2003a,b; Nicolai
et al. 2010) was used to assess outcome expectancies and
valuations. The revised CAEQ comprises 41 items and five
subscales: (1) Social Assertiveness And Positive Affect (15
items), (2) Tension Reduction (7 items), (3) Cognitive
Impairment And Physical Discomfort (10 items), (4)
Aggression (4 items), and (5) Sexual Enhancement (5 items).
The test format elicits two types of responses for each item:
(1) the subjective probability of the described alcohol effect
(outcome expectancy) and (2) the subjective evaluation of
this effect (outcome value). For assessing outcome expectan-
cies, items take the form of short phrases prefaced by
‘When I drink alcohol …’ (e.g. ‘I forget about worries’)
rated on a five-point scale with the endpoints 1 ¼‘not at all’
and 5 ¼‘definitely’. For assessing outcome values, items take
the form of short phrases prefaced by ‘This effect of alcohol
is/would be …’ (1 ¼‘very unpleasant’to 5 ¼‘very pleasant’).
Internal consistencies of the expectancy and evaluation
scales ranged from a¼.71 to a¼.93 and from a¼.67 to
a¼.90, respectively (see Table 1). Note that only the scale
assessing evaluations of sexual enhancement exhibited an
internal consistency below .70 in the student sample. Given
that this scale comprises five items only, the comparatively
low internal consistency can still be considered acceptable.
The CAEQ was shown to possess adequate temporal stability
over a test-retest interval of seven (r
tt
¼.78 –.89) and four-
teen days (r
tt
¼.82 –.89), respectively. Evidence for the valid-
ity of the CAEQ has been demonstrated (Demmel & Hagen
2003a; Nicolai et al. 2010).
Data analysis
We used structural equation modeling to investigate our
hypotheses. Specifically, two latent variables were specified
for each subscale of the CAEQ. One latent variable indicated
the degree to which the effect was expected (as measured by
the responses to the expectancy items) while another latent
ADDICTION RESEARCH & THEORY 3
variable indicated the degree to which the effect was valued
(as measured through the responses to the evaluation items).
Thus, the full model comprised five latent variables indicat-
ing expectancies, five latent variables indicating evaluations,
and another five latent variables indicating the latent expect-
ancy-value interactions.
In order to model interactions between the latent expect-
ancy and the latent evaluation factors, we used the uncon-
strained product-indicator approach (Little et al. 2006; Marsh
et al. 2004,2006). This approach has been shown to yield
unbiased parameter estimates and exhibits reasonable power
in comparison to other approaches (e.g. Brandt et al. 2014).
In the product indicator approach, a latent interaction factor
is formed by considering the products of the indicators of the
latent predictors using a matched pair-strategy. For instance,
to identify the latent interaction between expectancies and
valuations of impairment, each item of the impairment
expectancy scale (e.g. expectancy to 'forget about worries')
was multiplied with its respective counterpart of the impair-
ment valuation scale (e.g. value of 'forget about worries'). The
resulting product variables were then used as indicators of the
latent interaction factor. As recommended by Marsh et al.
(2004), the expectancy and value items were mean-centered,
and product indicators were calculated by multiplying the
mean-centered expectancy score with the mean-centered
value score for each item. Missing data were imputed by
means of two-way imputation (Sijtsma & van der Ark 2003;
van Ginkel & van der Ark 2005) prior to calculating the prod-
uct indicators.
1
To identify the latent interaction model, it is
further necessary to set the mean of the latent product varia-
bles equal to the covariance of the latent composites (Marsh
et al. 2004). The unstandardized loading of one item (the one
exhibiting the strongest standardized loading on the respect-
ive factor) was set equal to one to set a scale for the associated
factor (reference indicator approach; see Kline 2015). As is
the case for interaction terms of observed variables in a mul-
tiple regression framework, appropriately specified latent-
interaction models must include both the main-effects and
the product term.
A complication in the present context was that the full
model comprised a total of 15 latent variables (including five
latent interactions) and 123 observed items, so that it did not
prove feasible to estimate all effects jointly. Therefore, we
applied a stepwise strategy to identify significant interaction
effects in the prediction of frequency and quantity. First, we
regressed the dependent variable (frequency or quantity) on
expectancy, value, and the expectancy-value interaction separ-
ately for each of the five CAEQ dimensions. Age and gender
(dummy-coded with females as reference category) were
included to control for effects of demographic background
variables, as both drinking behavior and expectancies have
been shown to remarkably differ by age and gender (Nicolai
et al. 2012). Next, we computed a model including all signifi-
cant predictor terms (and main effects where appropriate) as
identified in the first step. The final model was then com-
puted by dropping all non-significant predictors from the
model estimated in the second step (except for main effects
when the interaction effect was significant as well as age and
gender as background variables).
Since the product-indicator approach inherently violates
the multivariate normality assumption (Brandt et al. 2014),
we used robust standard errors and test-statistics (Satorra &
Bentler 2001). Because the chi-square model test has very
large power to detect even trivial misspecifications (Moshagen
& Erdfelder 2016) and is severely inflated for models involv-
ing many variables (Herzog et al. 2007; Moshagen 2012),
model fit was evaluated by means of descriptive indices of
model fit (RMSEA and SRMR, in line with recent recommen-
dations by Moshagen & Auerswald 2017). Analyses were per-
formed using SPSS version 20 and Mplus version 7.11.
Results
Student sample
The model predicting drinking quantity exhibited an accept-
able fit to the data, SB-v
2
(1585) ¼4833.01, p<.01;
Table 1. Descriptive statistics.
M (SD) Internal consistency
Scale # Items Student sample Clinical sample Student sample Clinical sample
Social assertiveness and positive affect
Expectancies 15 3.35 (0.68) 3.41 (0.81) .92 .93
Values 15 3.55 (0.49) 3.43 (0.66) .85 .90
Tension reduction
Expectancies 7 2.88 (0.70) 3.64 (0.73) .78 .78
Values 7 3.46 (0.58) 3.68 (0.69) .73 .80
Cognitive impairment and physical discomfort
Expectancies 10 2.97 (0.59) 3.18 (0.75) .80 .87
Values 10 2.00 (0.49) 2.23 (0.65) .76 .83
Aggression
Expectancies 4 1.82 (0.78) 2.63 (1.22) .84 .91
Values 4 1.68 (0.79) 2.14 (0.92) .80 .79
Sexual enhancement
Expectancies 5 2.64 (0.74) 2.83 (0.92) .71 .82
Values 5 3.37 (0.69) 3.18 (0.87) .67 .80
Cronbach’s alpha estimate of internal consistency. Responses were made on a scale ranging from 1 to 5.
1
Two-way imputation handles missing data by estimating the missing scores
and then imputing these estimates, which is an advantage over other
approaches (such as Full Information Maximum Likelihood) in the context of
the product-indicator approach used in this study. Two-way imputation
improves upon simple mean or person-mean imputation by correcting for
score differences between respondents and for score differences between
items and has been shown to perform on par with other approaches for
multidimensional Likert-type data (Bernaards & Sijtsma, 2000).
4 J. NICOLAI ET AL.
RMSEA ¼.044 (90%-CI: .043–.046); SRMR ¼.053. Gender,
valuations of tension reduction, expectancies of social assert-
iveness, and the expectancy-value interaction for social
assertiveness were significant positive predictors, whereas age
was significantly negatively related to quantity (overall
R
2
¼.06; Table 2). We used a simple-slopes plot (Preacher
et al. 2006) to facilitate the interpretation of the interaction
effect. Figure 1 shows the effect of social assertiveness
expectancies on drinking quantity at the mean of social
assertiveness valuations and at one and two standard devia-
tions below and above the mean, respectively. In interpreting
this plot, it is important to note that the mean evaluation of
social assertiveness was not equal to a neutral evaluation:
The mean on the five-point scale ranging from 1–5 was 3.55
(SD ¼0.49), so that social assertiveness was rated rather neu-
trally even for those scoring 1 SD below the mean.
Endorsement of social assertiveness expectancies was associ-
ated with drinking increasingly larger amounts of alcohol as
this effect was valued more positively. The strongest relation-
ship was evident for individuals valuing social assertiveness
most positively. However, if social assertiveness was valued
negatively (2SD), the degree to which this effect was
expected had a small effect on drinking.
Frequency was significantly positively predicted by age
(b¼.23; p<.01), gender (b¼.20; p<.01), and tension
reduction expectancies (b¼.22; p<.01), and significantly
negatively predicted by impairment expectancies (b¼.12;
p<.01). Neither valuations nor expectancy-value interactions
significantly contributed to the prediction of frequency. The
model accounted for 15%of the variance. Model fit was
adequate, SB-v
2
(167) ¼782.33, p<.01; RMSEA ¼.059 (90%-
CI: .055–.063); SRMR ¼.050.
Clinical sample
The model predicting quantity yielded an acceptable fit to
the data, SB-v
2
(692) ¼1873.60, p<.01; RMSEA ¼.049 (90%-
CI: .047–.052); SRMR ¼.063. Quantity was significantly posi-
tively predicted by gender (b¼.18; p<.01), tension reduc-
tion expectancies (b¼.23; p<.01), and aggression
expectancies (b¼.09; p<.05), and significantly negatively
predicted by age (b¼.16; p<.01), social assertiveness
expectancies (b¼.09; p¼.051), and impairment expectan-
cies (b¼.15; p<.01). Note that the effect of social assert-
iveness expectancies only approached statistical significance,
so it might not be reliable. The model accounted for 10%of
the variance in quantity. Neither valuations nor expectancy-
value interactions significantly contributed to the prediction
of quantity.
The model predicting frequency exhibited a mediocre fit,
SB-v
2
(1218) ¼6203.07, p<.01; RMSEA ¼.07 (90%-CI:
.075–.078); SRMR ¼.147. In this model, frequency was sig-
nificantly predicted by the expectancy-value interaction for
tension reduction and the expectancy-value interaction for
aggression (R
2
¼.03; Table 3). Again, the interaction effects
are in line with EVT, as shown by the simple-slopes plots
displayed in Figure 2. Inpatients positively valued the tension
reducing effects of alcohol (M¼3.68; SD ¼0.69), so that
individuals scoring 1 SD below the mean are best character-
ized as holding neutral evaluations. From Figure 2(A),itis
evident that expecting tension reduction only increased
drinking frequency if it was valued positively, while it had
Table 2. Latent regression results for the prediction of drinking quantity in the student sample.
Predictor bSE p
Age .12 .02 <.01
Gender .09 .03 <.01
Tension reduction evaluation .10 .03 <.01
Social assertiveness expectancy .17 .05 <.01
Social assertiveness evaluation .02 .05 .70
Social assertiveness expectancy X evaluation .09 .03 <.01
Note. Standardized regression coefficients with associated standard errors and p-values.
Figure 1. Simple-slopes plot showing the combined effects of social assertive-
ness expectancy and values (both in standard deviations) in predicting drinking
quantity in the student sample. The mean value of social assertiveness on a
scale ranging from 1–5 was 3.55 (SD ¼0.49).
Table 3. Latent regression results for the prediction of drinking frequency
(prior to hospitalization) in the clinical sample.
Predictor bSE p
Age .04 .03 .27
Gender .05 .04 .13
Tension reduction expectancy .06 .06 .38
Tension reduction evaluation .03 .05 .55
Tension reduction expectancy X evaluation .11 .04 .01
Aggression expectancy .00 .04 .99
Aggression evaluation .03 .05 .51
Aggression expectancy X evaluation .09 .05 .04
Standardized latent regression coefficients with associated standard errors and
p-values.
ADDICTION RESEARCH & THEORY 5
no impact for neutral (1SD) and a slightly negative impact
for negative evaluations (-2 SD). Becoming aggressive after
drinking was negatively valued on average (M¼2.14;
SD ¼0.92), so that individuals scoring 1 SD above the mean
neutrally evaluated becoming aggressive. As can be seen
from Figure 2(B), expecting aggression was associated with
less frequent drinking only when this effect was rather
strongly negatively valued (1SD). For moderately negative
(M) evaluations, expecting becoming aggressive was not
related to frequency. However, for participants neutrally (þ1
SD) or positively valuing aggression (þ2SD), expecting
aggression increased the likelihood of drinking.
Discussion
Alcohol-related outcome expectancies have been studied
extensively and can be considered as one of the core con-
cepts in the psychology of alcohol use. The concept of out-
come expectancies is grounded in EVT (Edwards 1954;
Fishbein & Ajzen 1975; Bandura 1986), which predicts an
individual's tendency to perform a certain behavior as a
function of both the expectation that the behavior will pro-
duce certain consequences and the evaluation of those conse-
quences. Of central importance for this class of theories is
the proposition that the effect of either expectancy or value
on consumption depends on the other variable. The more
(positively or negatively) one values an alcohol effect, the
larger should be the influence of expectancies on alcohol
use. This basic premise has rarely been tested adequately,
however, and the few studies on this issue yielded mixed
results (Grube et al. 1995; Jones & McMahon 1996a), prob-
ably due to applying statistical models that are plagued by
measurement error (Nagengast et al. 2011). The present
study is the first that examined expectancy-value interactions
in predicting quantity and frequency of alcohol use by five
diverse expectancy-value domains drawing on latent variable
methodology.
In line with the basic tenet of EVT, the results show that
the expectancy-value interaction of social assertiveness sig-
nificantly predicted quantity of drinking among students.
The prediction of drinking by expectancies thus was moder-
ated by the specific evaluation of the effect such that positive
evaluations increased the association, whereas negative evalu-
ations led to a weak association between expectancy and
quantity. This finding adds to the growing body of evidence
on the importance of social enhancement expectancies
among students (Ham & Hope 2003; Gilles et al. 2006; Ham
2009; Monk & Heim 2013). Although this interaction pattern
is consistent with expectancy-value theory, we did not find
compelling evidence for a reversed (negative) association
between expectancy and drinking when social assertiveness
was valued negatively. However, it must be kept in mind
that social assertiveness was generally valued positively in
our sample, so the lack of a reversed association may be due
to the absence of individuals giving a pronounced negative
evaluation to these effects.The interaction effect between
expectancies and evaluations promoting aggression observed
in the prediction of drinking frequency in the clinical sample
was more closely aligned with EVT. Expecting aggression
decreased the frequency of drinking when aggression was
valued strongly negative, had no impact on frequency when
aggression was evaluated slightly negatively, and increased
the frequency when aggression was valued neutrally to posi-
tively. This suggests that expecting an increased likelihood of
aggression can be both a risk (Ham & Hope 2003) and a
protective factor, depending on its evaluation (Fromme et al.
1993; Quigley & Leonard 2006). A similar, although less pro-
nounced, interaction pattern was observed for the prediction
of frequency by tension reduction in the clinical sample.
Expecting tension reduction increased drinking when it was
valued positively, was only weakly related to drinking when
Figure 2. Simple-slopes plots showing the combined effects of tension reduction (left) and aggression (right) expectancies and values (both in standard deviations)
in predicting drinking frequency in the clinical sample. The mean value of tension reduction was 3.68 (SD ¼0.69) and that of aggression 2.14 (SD ¼0.92), both on
scales ranging from 1–5.
6 J. NICOLAI ET AL.
it was valued neutrally, and slightly decreased drinking fre-
quency when it was valued negatively, in turn highlighting
the importance of the tension reduction effects for the
development and maintenance of alcohol use disorders
(Brown 1985; Ham & Hope 2003; Schmidt et al. 2013).
Interestingly, the endorsement of tension reduction may act
as a protective factor against frequent drinking when this
effect was valued negatively. One explanation may be that
some individuals are primarily motivated to drink alcohol
for its stimulating and activating effects, and may therefore
consider the sedative effects of alcohol as less desirable or
even aversive. For these individuals, strong tension reduc-
tion expectancies thus may in fact reduce the likelihood of
drinking.
It is important to note that in the clinical sample neither
the main effect of expectancies nor values reached signifi-
cance for tension reduction and aggression. Thus, ignoring
the interaction effect and proceeding by considering the
main effects in isolation would have led to the erroneous
conclusion that tension reduction and aggression are unre-
lated to alcohol consumption in this sample.
Interactions between expectancies and values emerged for
different effect domains in the prediction of different drink-
ing outcomes in the two samples. Among students, expect-
ancy-value interactions were evident for social assertiveness
in the prediction of drinking quantity, whereas interactions
for tension reduction and aggression occurred in the predic-
tion of drinking frequency in the clinical sample. While the
differences in the relevant effect domain arguably reflect dif-
ferent drinking styles among students versus inpatients, it is
less clear why the interactions occurred in the prediction of
different drinking outcomes. One might speculate that the
latter is a consequence of the relevant effect domain. For
example, unlike inpatients, students typically consume alco-
hol in social settings (Ham & Hope 2003). Individuals who
expect and positively value the social enhancing effects of
alcohol might be more responsive to peer pressure to drink
larger amounts, but may not experience more drinking
opportunities. By contrast, when individuals both expect and
positively value the tension relieving effects of alcohol, it
seems likely that alcohol is used to dampen responses to
stressful stimuli encountered in the environment. Clearly,
such stressors occur regularly, in turn resulting in more fre-
quent drinking, but not necessarily in drinking larger
amounts. Similarly, individuals who expect to become
aggressive after consuming alcohol, but negatively value this
effect, might attempt to cut-down the frequency of drinking
episodes in order to avoid aggression. However, once a
drinking episode has been initiated (for reasons different
from aggression), the quantity of alcohol consumed on such
an occasion might not be affected.
Despite the fact that the results discussed above support
central predictions from EVT, other parts of the results
question the universal validity of EVT in the domain of alco-
hol consumption. The results indicate that expectancy-value
interactions do not always occur, as there were no inter-
action effects when predicting drinking frequency among
students or drinking quantity among inpatients, where the
only significant predictors were the main effects of various
expectancy dimensions. Thus, the proposed interactive effect
of expectancies and values does not seem to be as universal
and powerful as implied by EVT. Importantly, showing the
influence of mere main effects of expectancies cannot be
considered as sufficient evidence for the validity of EVT,
because the central preposition states that the effect of a par-
ticular expectancy depends on its valuation. In addition, the
observed interaction effects were –although not atypical –
small in magnitude, further casting doubt on the universality
of EVT in explaining the relationship between expectancies
and drinking.
Some limitations should also be considered when inter-
preting the results of the present study. First, due to non-
probability sampling procedures neither sample can be con-
sidered representative of the general population of students
or inpatients. A related issue pertains to the large proportion
of males in the clinical sample, so that these results should
not be readily generalized to females. Note, however, that
males are more likely to develop alcohol related problems
(Wilsnack et al. 2000; Nolen-Hoeksema 2004) and that
research on alcohol expectancies suggests gender may not be
as important as age or drinking patterns (Lundahl et al.
1997; Leigh & Stacy 2004; Nicolai et al. 2012). Second, as the
study was correlational in nature, only associations can be
tested and causality cannot be assumed. Third, the instruc-
tional set used in the present study did not specify the dose
of alcohol consumed or the particular context of alcohol
consumption, both of which have been shown to influence
alcohol expectancies (e.g. Sher 1985; Read & O'Connor
2006). Although we are not aware of any study investigating
this issue, it seems highly likely that these factors also affect
the evaluation of the effects of alcohol. For example, the
expectation that alcohol enhances interpersonal effectiveness
is likely to be more positively valued in a social compared to
a solitary drinking context, whereas the reverse may be true
concerning the sedative effects of alcohol. Investigating
expectancy-value interactions in different drinking contexts
thus seems to be a promising avenue for future research.
Fourth, consumption variables were assessed relying on
retrospective self-reports. As with any self-reported data,
there are issues of social desirability (Moshagen et al. 2010)
and other distorting effects (e.g. inaccuracy or recall errors),
so the validity cannot be guaranteed (Ekholm 2004).
Concerning possible social desirability biases regarding the
anticipated effects of alcohol, a viable alternative may be to
resort to implicit measures, as these are less vulnerable to
demand characteristics (e.g. Wiers et al. 2002). Finally, reli-
ance on self-reported data might also be a problem concern-
ing the measurement of expectancies and valuations in the
clinical sample, as past drinking behavior (before hospitaliza-
tion) was predicted by expectancies and valuations that
might have been amended during treatment. Considering
this, the finding of a substantial proportion of variance
explained in the quantity of drinking is a rather surprising
one. One explanation may be that alcohol dependent indi-
viduals tend to endorse rather stable outcome expectancies
that are difficult to modify during treatment (see e.g.
Connors et al. 1993; Jones & McMahon 1996b, for similar
results).
ADDICTION RESEARCH & THEORY 7
Conclusions
In summary, using latent variable methodology and two
diverse samples, the present study supports basic premises of
EVT by showing interactive effects of certain expectancies
and valuations in predicting drinking outcomes. Thus, future
studies should take potential interaction effects between
expectancies and valuations into account in order to avoid
drawing incorrect conclusions about the influence of a par-
ticular anticipated effect on alcohol consumption. However,
our study also identifies the boundaries of EVT, as there was
little support for the universal validity of expectancy-value
interactions in the prediction alcohol use. Most importantly,
expectancies, values, and expectancy-value interactions of
various dimensions have very different validities in the pre-
diction of either the quantity or the frequency of alcohol use
in college students or alcohol-dependent inpatients, which is
in line with previous studies (Jones et al. 2001; Patrick &
Maggs 2011; Monk & Heim 2013). Future studies should
therefore compare different theoretical frameworks in the
prediction of alcohol use. For example, Borders and col-
leagues (Borders, et al. 2004) argue that extending EVT to
multiple expectancies for alternative behaviors is essential,
because behavioral decisions are complex and rarely limited
to single target behaviors and their accompanying expectan-
cies. To test this assumption, future studies have to include
alternative behaviors to drinking so that individuals can
choose from alternative behaviors and compare various
expectancies. Such comparative studies might help to clarify
the limits of EVT. Finally, as we have shown herein, the val-
idity of EVT may vary by expectancy domain, alcohol out-
come, and population, so a fine-grained approach seems
mandatory when evaluating and further refining EVT in the
domain of alcohol use.
Disclosure statement
The authors report no conflicts of interest. The authors alone are
responsible for the content and writing of the article.
References
Aas HN, Leigh BC, Anderssen N, Jakobsen R. 1998. Two-year longitu-
dinal study of alcohol expectancies and drinking among Norwegian
adolescents. Addiction. 93:373–384.
Atkinson JW. 1957. Motivational determinants of risk-taking behavior.
Psychol Rev. 64:359–372.
Bandura A. 1986. Social foundations of thought and action: a social
cognitive theory. Upper Saddle River (NJ): Prentice Hall.
Bauman KE, Fisher LA, Bryan ES, Chenoweth RL. 1985.
Relationship between subjective expected utility and behavior: a
longitudinal study of adolescent drinking behavior. J Stud
Alcohol. 46:32–38.
Bernaards CA, Sijtsma K. 2000. Influence of imputation and EM meth-
ods on factor analysis when item nonresponse in questionnaire data
is nonignorable. Multivariate Behav Res. 35:321–364.
Bollen KA. 1989. Structural equations with latent variables. New York:
Wiley.
Borders A, Earleywine M, Huey S. 2004. Predicting problem behaviors
with multiple expectancies: Expanding expectancy-value theory.
Adolescence. 39:539–550.
Brandt H, Kelava A, Klein A. 2014. A simulation study comparing
recent approaches for the estimation of nonlinear effects in
SEM under the condition of nonnormality. Struct Eq Model.
21:181–195.
Brown SA. 1985. Reinforcement expectancies and alcoholism treatment
outcome after a one-year follow-up. J Stud Alcohol Drugs. 46:304.
Brown SA, Christiansen BA, Goldman MS. 1987. The Alcohol
Expectancy Questionnaire: an instrument for the assessment of ado-
lescent and adult alcohol expectancies. J Stud Alcohol Drugs.
48:483–491.
B€
uhringer G, Augustin R, Bergmann E, Bloomfield K, Funk W, Kraus
L, Merfert-Diete C, Rumpf H-J, Simon R, Toeppich J. 2002. Alcohol
consumption and alcohol-related problems in Germany. Munich
(Germany): Hogrefe Publishing.
Burden JL, Maisto SA. 2000. Expectancies, evaluations and attitudes:
prediction of college student drinking behavior. J Stud Alcohol.
61:323–331.
Cable N, Sacker A. 2008. Typologies of alcohol consumption in adoles-
cence: predictors and adult outcomes. Alcohol Alcohol. 43:81–90.
Cohen J, Cohen P, West SG, Aiken LS. 2013. Applied multiple regres-
sion/correlation analysis for the behavioral sciences. Mahwah:
Erlbaum.
Connors GJ, O'Farrell TJ, Cutter HSG, Thompson DL. 1986. Alcohol
expectancies among male alcoholics, problem drinkers, and nonpro-
blem drinkers. Alcoholism Clin Exp Res. 10:667–671.
Connors GJ, Tarbox AR, Faillace LA. 1993. Changes in alcohol expect-
ancies and drinking behavior among treated problem drinkers.
J Stud Alcohol. 54:676–683.
Critchlow B. 1987. A utility analysis of drinking. Addict Behav.
12:269–273.
Demmel R, Hagen J. 2003a. The comprehensive alcohol expectancy
questionnaire: II. prediction of alcohol use and clinical utility. Sucht.
49:300–305.
Demmel R, Hagen J. 2003b. The comprehensive alcohol expectancy
questionnaire: I. scale development. Sucht. 49:292–299.
Edwards W. 1954. The theory of decision making. Psychol Bull.
51:380–417.
Ekholm O. 2004. Influence of the recall period on self-reported alcohol
intake. Eur J Clin Nutr. 58:60–63.
Evans MG. 1991. The problem of analyzing multiplicative composites:
Interactions revisited. Am Psychol. 46:6–15.
Fishbein M, Ajzen I. 1975. Belief, attitude, intention and behavior: an
introduction to theory and research. Reading (MA): Addison-Wesley
[Database].
Fromme K, D’Amico EJ. 2000. Measuring adolescent alcohol outcome
expectancies. Psychol Addict Behav. 14:206–212.
Fromme K, Stroot EA, Kaplan D. 1993. Comprehensive effects of alco-
hol: development and psychometric assessment of a new expectancy
questionnaire. Psychol Assess. 5:19–26.
Fromme K, Marlatt GA, Baer JS, Kivlahan DR. 1994. The alcohol skills
training program: a group intervention for young adult drinkers.
J Substance Abuse Treat. 11:143–154.
Gilles DM, Turk CL, Fresco DM. 2006. Social anxiety, alcohol expect-
ancies, and self-efficacy as predictors of heavy drinking in college
students. Addict Behav. 31:388–398.
Goldman MS, Brown SA, Christiansen BA, Smith GT. 1991.
Alcoholism and memory: broadening the scope of alcohol-expect-
ancy research. Psychol Bull. 110:137–146.
Goldman MS, Del Boca FK, Darkes J. 1999. Alcohol expectancy theory:
the application of cognitive neuroscience. In: Psychological theories
of drinking and alcoholism. 2nd ed. New York: Guilford Press;
p. 203–246.
Grube JW, Chen M-J, Madden P, Morgan M. 1995. Predicting adoles-
cent drinking from alcohol expectancy values: a comparison of addi-
tive, interactive, and nonlinear models1. J Appl Social Pyschol.
25:839–857.
Ham LS. 2009. Positive social alcohol outcome expectancies, social anx-
iety, and hazardous drinking in college students. Cogn Ther
Res.33:615–623.
8 J. NICOLAI ET AL.
Ham LS, Hope DA. 2003. College students and problematic drinking: a
review of the literature. Clin Psychol Rev. 23:719–759.
Ham LS, Stewart SH, Norton PJ, Hope DA. 2005. Psychometric assess-
ment of the comprehensive effects of alcohol questionnaire: compar-
ing a brief version to the original full scale. J Psychopathol Behav
Assess. 27:141–158.
Hasking PA, Oei TPS. 2002. The differential role of alcohol expectan-
cies, drinking refusal self-efficacy and coping resources in predicting
alcohol consumption in community and clinical samples. Addict Res
Theory. 10:465–494.
Herzog W, Boomsma A, Reinecke S. 2007. The model-size effect on
traditional and modified tests of covariance structures. Struct Eq
Model. 14:361–390.
Jester JM, Wong MM, Cranford JA, Buu A, Fitzgerald HE, Zucker RA.
2015. Alcohol expectancies in childhood: change with the onset of
drinking and ability to predict adolescent drunkenness and binge
drinking. Addiction. 110:71–79.
Jones BT, Corbin W, Fromme K. 2001. A review of expectancy theory
and alcohol consumption. Addiction. 96:57–72.
Jones BT, McMahon J. 1994. Negative and positive alcohol expectancies
as predictors of abstinence after discharge from a residential treat-
ment program: a one-month and three-month follow-up study in
men. J Stud Alcohol. 55:543–548.
Jones BT, McMahon J. 1996a. A comparison of positive and negative
alcohol expectancy and value and their multiplicative composite as
predictors of post-treatment abstinence survivorship. Addiction.
91:89–99.
Jones BT, McMahon J. 1996b. Changes in alcohol expectancies during
treatment relate to subsequent abstinence survivorship. Br J Clin
Psychol. 35:221–234.
Kachadourian LK, Quigley BM, Leonard KE. 2014. Alcohol expectancies
and evaluations of aggression in alcohol-related intimate-partner ver-
bal and physical aggression. J Stud Alcohol Drugs. 75:744–752.
Kenny DA, Judd CM. 1984. Estimating the nonlinear and interactive
effects of latent variables. Psychol Bull. 96:201–210.
Kline RB. 2015. Principles and practice of structural equation modeling.
New York: Guilford.
Kuther TL. 2002. Rational decision perspectives on alcohol consump-
tion by youth: Revising the theory of planned behavior. Addict
Behav. 27:35–47.
Lee NK, Greely J, Oei TPS. 1999. The relationship of positive and nega-
tive alcohol expectancies to patterns of consumption of alcohol in
social drinkers. Addict Behav. 24:359–369.
Leigh BC. 1987. Evaluations of alcohol expectancies: do they add to
prediction of drinking patterns? Psychol Addict Behav. 1:135–139.
Leigh BC, Stacy AW. 2004. Alcohol expectancies and drinking in differ-
ent age groups. Addiction. 99:215–227.
Little TD, Bovaird JA, Widaman KF. 2006. On the merits of orthogon-
alizing powered and product terms: implications for modeling inter-
actions among latent variables. Struct Eq Model. 13:497–519.
Lomax RG, Schumacker RE. 2012. A beginner's guide to structural
equation modeling. New York: Routledge.
Lundahl LH, Davis TM, Adesso VJ, Lukas SE. 1997. Alcohol expectan-
cies: effects of gender, age, and family history of alcoholism. Addict
Behav. 22:115–125.
Mallett KA, Bachrach RL, Turrisi R. 2008. Are all negative consequences
truly negative? Assessing variations among college students’percep-
tions of alcohol related consequences. Addict Behav. 33:1375–1381.
Mann LM, Chassin L, Sher KJ. 1987. Alcohol expectancies and the risk
for alcoholism. J Consult Clin Psychol. 55:411–417.
Marsh HW, Wen Z, Hau K-T. 2004. Structural equation models of
latent interactions: evaluation of alternative estimation strategies and
indicator construction. Psychol Methods. 9:275–300.
Marsh HW, Wen Z, Hau K-T. 2006. Structural equation models of
latent interaction and quadratic effects. In: Hancock GR, Mueller
RO, editor. Structural equation modeling: a second course.
Greenwich: IAP; p. 225–265.
McClelland GH, Judd CM. 1993. Statistical difficulties of detecting
interactions and moderator effects. Psychol Bull. 114:376–390.
McKellar JD, Harris AH, Moos RH. 2006. Predictors of outcome for
patients with substance-use disorders five years after treatment drop-
out. J Stud Alcohol. 67:685–693.
McMahon J, Jones BT, O'Donnell P. 1994. Comparing positive and
negative alcohol expectancies in male and female social drinkers.
Addict Res. 1:349–365.
Monk RL, Heim D. 2013. A critical systematic review of alcohol-related
outcome expectancies. Subst Use Misuse. 48:539–557.
Moshagen M. 2012. The model size effect in SEM: inflated goodness-
of-fit statistics are due to the size of the covariance matrix. Struct Eq
Model. 19:86–98.
Moshagen M, Auerswald M. 2017. On congruence and incongruence of
measures of fit in structural equation modeling. Psychol Methods.
(Epub ahead of print).
Moshagen M, Erdfelder E. 2016. A new strategy for testing structural
equation models. Struct Eq Model. 23:54–60.
Moshagen M, Musch J, Ostapczuk M, Zhao Z. 2010. Reducing
socially desirable responses in epidemiologic surveys: an
extension of the randomized-response technique. Epidemiology.
21:379–382.
Nagengast B, Marsh HW, Scalas LF, Xu MK, Hau K-T, Trautwein U.
2011. Who took the'x' out of expectancy-value theory? A psycho-
logical mystery, a substantive-methodological synergy, and a cross-
national generalization. Psychol Sci. 22:1058–1066.
Nicolai J, Demmel R, Moshagen M. 2010. The comprehensive alcohol
expectancy questionnaire: confirmatory factor analysis, scale refine-
ment, and further validation. J Personal Assess. 92:400–409.
Nicolai J, Moshagen M, Demmel R. 2012. Patterns of alcohol expectan-
cies and alcohol use across age and gender. Drug Alcohol Depend.
126:347–353.
Nolen-Hoeksema S. 2004. Gender differences in risk factors and conse-
quences for alcohol use and problems. Clin Psychol Rev. 24:981–1010.
Patrick ME, Wray-Lake L, Finlay AK, Maggs JL. 2010. The long arm of
expectancies: adolescent alcohol expectancies predict adult alcohol
use. Alcohol Alcohol. 45:17–24.
Patrick ME, Maggs JL. 2011. College students’evaluations of
alcohol consequences as positive and negative. Addict Behav.
36:1148–1153.
Preacher KJ, Curran PJ, Bauer DJ. 2006. Computational tools for prob-
ing interactions in multiple linear regression, multilevel modeling,
and latent curve analysis. J Educ Behav Stat.31:437–448.
Quigley BM, Leonard KE. 2006. Alcohol expectancies and intoxicated
aggression. Aggress Violent Behav. 11:484–496.
Read JP, O'Connor RM. 2006. High- and low-dose expectancies as
mediators of personality dimensions and alcohol involvement. J Stud
Alcohol. 67:204–214.
Satorra A, Bentler PM. 2001. A scaled difference chi-square test statistic
for moment structure analysis. Psychometrika. 66:507–514.
Sawayama T, Yoneda J, Tanaka K, Shirakawa N, Sawayama E, Ikeda T,
Higuchi S, Miyaoka H. 2012. The predictive validity of the Drinking-
Related Cognitions Scale in alcohol-dependent patients under abstin-
ence-oriented treatment. Subst Abuse Treat Prev Policy. 7:17.
Schmidt AF, Eulenbruch T, Langer C, Banger M. 2013. Interoceptive
awareness, tension reduction expectancies and self-reported drinking
behavior. Alcohol Alcohol. 48:472–477.
Settles RE, Zapolski TCB, Smith GT. 2014. Longitudinal test of a devel-
opmental model of the transition to early drinking. J Abnorm
Psychol. 123:141–151.
Sher KJ. 1985. Subjective effects of alcohol: the influence of setting and
individual differences in alcohol expectancies. J Stud Alcohol.
46:137–146.
Shieh G. 2010. Sample size determination for confidence intervals of
interaction effects in moderated multiple regression with continuous
predictor and moderator variables. Behav Res Methods. 42:824–835.
Sijtsma K, van der Ark LA. 2003. Investigation and treatment of miss-
ing item scores in test and questionnaire data. Multivariate Behav
Res. 38:505–528.
Stacy AW, Widaman KF, Marlatt GA. 1990. Expectancy models of alco-
hol use. J Pers Soc Psychol. 58:918.
ADDICTION RESEARCH & THEORY 9
Tibboel H, De Houwer J, Spruyt A, Brevers D, Roy E, No€
el X. 2015.
Heavy social drinkers score higher on implicit wanting and liking
for alcohol than alcohol-dependent patients and light social drinkers.
J Behav Ther Exp Psychiatry. 48:185–191.
Townshend JM, Duka T. 2007. Avoidance of alcohol-related stimuli in
alcoholdependent inpatients. Alcoholism Clin Exp Res. 31:1349–1357.
Valdivia I, Stewart SH. 2005. Further examination of the psychometric
properties of the comprehensive effects of alcohol questionnaire.
Cogn Behav Ther. 34:22–33.
van Ginkel JR, van der Ark LA. 2005. SPSS syntax for missing value imput-
ation in test and questionnaire data. Appl Psychol Measure. 29:152–153.
Wiers RW, Van Woerden N, Smulders FT, De Jong PJ. 2002. Implicit
and explicit alcohol-related cognitions in heavy and light drinkers.
J Abnorm Psychol. 111:648–658.
Wilsnack RW, Vogeltanz ND, Wilsnack SC, Harris TR. 2000. Gender dif-
ferences in alcohol consumption and adverse drinking consequences:
cross-cultural patterns. Addiction. 95:251–265.
Williams RJ, Ricciardelli LA. 1996. Expectancies elate to
symptoms of alcohol dependence in young adults. Addiction.
91:1031–1039.
Zamboanga BL. 2006. From the eyes of the beholder: alcohol expectan-
cies and valuations as predictors of hazardous drinking behaviors
among female college students. Am J Drug Alcohol Abuse.
32:599–605.
Zamboanga BL, Schwartz SJ, Ham LS, Hernandez Jarvis L, Olthuis JV.
2009. Do alcohol expectancy outcomes and valuations mediate peer
influences and lifetime alcohol use among early adolescents? J Genet
Psychol. 170:359–376.
10 J. NICOLAI ET AL.
