Professional Documents
Culture Documents
Copyright1988bythe AmericanPsychologicalAssociation,Inc.
0021-843X/88/$00.75
lacking. The model proposed here postulates that alcohol abuse can be predicted from a causal chain
that includes alcohol consumption and "drinking to cope" as proximal determinants and general
coping skills and positive alcohol expectancies as more distal determinants. To evaluate this model
in a way that permits simultaneous consideration of its multiple determinants and control for demographic influences, path analytic techniques were applied to data from problem and nonproblem
drinkers drawn from a general population sample. The hypothesized model accounted for significant
variance in abuse status. Drinking to cope emerged as the most powerful predictor, exerting influence
via direct and indirect pathways. Coping styles indicative of avoidance of emotion emerged as more
important predictors of abuse than problem-focused coping. The predictive value of coping was
moderated by alcohol expectancies such that avoidant styles of coping with emotion were predictive
of abuse status only among drinkers expressing greater beliefin alcohol's positive reinforcing properties. These findings both support and refine the social learning perspective on alcohol abuse. Theoretical and treatment implications are considered.
Drinking to Cope
Fundamental to the social learning perspective are the dual
premises that alcohol may be used as a generalized coping
mechanism and that the use of alcohol to cope will promote
heavier drinking and alcohol abuse. Considerable empirical evidence exists to support both notions.
Drinking to cope is defined as the tendency to use alcohol to
escape, avoid, or otherwise regulate unpleasant emotions. Correlational research examining motives for drinking consistently
reveals that a substantial percentage of drinkers, typically ranging from 10% to 25%, report drinking to regulate negative emo-
Alcohol Expectancies
According to the social learning perspective, beliefs about the
effects of alcohol, referred to as alcohol expectancies, should
influence the likelihood that alcohol will be used to cope with
negative emotions. Presumably one must first believe that alcohol will in some way ameliorate unpleasant emotions before alcohol would be used instrumentally to regulate or reduce negative affect.
Expectancy research has focused on mapping the domain of
alcohol-related expectancies and their relation to alcohol use
and abuse. Early research identified six dimensions of positive
expectancies (Brown, Goldman, Inn, & Anderson, 1980). Two
of these dimensions were highly general, indicating the belief
that alcohol is capable of "magically" transforming or enhancing a broad range of physical and social experiences. The remaining four dimensions tapped expectancies for sexual enhancement, increased power and aggression, increased social
219
assertiveness, and tension reduction. Rohsenow (1983) subsequently elaborated this typology to include two additional dimensions reflecting expectancies for the negative effects of alcohol, in particular for performance impairment and irresponsibility.
Expectancy patterns have successfully predicted drinking behavior at all points along the continuum, from the onset of
drinking during adolescence to alcoholism. Expectancies assessed among 12- to 14-year-olds, prior to the onset of drinking,
have predicted subsequent patterns and levels of consumption
as well as the onset of problem drinking at 1- and 2-year followup intervals (Roehling, Smith, Goldman, & Christiansen, 1987;
Smith, Roehling, Christiansen, & Goldman, 1986). Among adolescent, college, and adult populations, the strength and pattern of alcohol expectancies have discriminated between light
and heavy drinkers, at-risk and control groups, and problem
and nonproblem drinkers (see Goldman, Brown, & Christiansen, 1987, for a review). Finally, expectancies have prospectively
predicted relapse among groups of treated alcoholics (Brown,
1985).
Experimental studies using the balanced placebo design provide further evidence that expectancies may significantly influence alcohol consumption (see Hull & Bond, 1986, for a recent
meta-analytic review of balanced placebo studies). Thus, experimental and correlational evidence provide convergent support
for the role of expectancies in determining actual levels of consumption as well as the behavioral consequences of consumption. The prospective studies recently reported by Smith et al.
(1986) and Roehling et al. (1987) provide compelling evidence
that expectancies precede the onset of drinking and drinking
problems. Collectively, these data strongly suggest that expectancies may play a causal role in the development of alcohol
abuse. It is not known, however, whether all or part of the relationship between expectancies and alcohol abuse is mediated
through reliance on alcohol as a coping mechanism.
220
(f Positive"~
/~-IP" ~ ExpectanciesJ~_
Coping / -
"
"
Method
Sample
Subjects in this study were 119 adults meeting DSM-III criteria for
current alcohol abuse or dependence and a comparison group of 948
drinkers, all of whom drank within the past year and have no history of
alcohol abuse or dependence. Subjects meeting these criteria were
drawn from a recently completed random sample survey of 1933 household residents in Erie County, New York. Designated respondents were
identified in a three-stage probability sample, stratified on race (Black
vs. non-Black) and education (less than high school, high school, some
college). The overall sample completion rate was 78.3%, and the majority (84.5%) of noncompletions were refusals.
Subjects in the present study ranged in age from 19 to 91 (M = 40
years, SD = 15.7), had completed on average one year of college (M =
13.0 years of schooling, SD = 2.4) and reported a median income of
$8,500 (M = $10,503, SD = $9,695). Fifty-seven percent were female
and 51% were Black. All subjects had consumed alcohol within the past
year, reporting an average of one drink per day, containing approximately .5 ounces absolute alcohol.
Procedures
Data were collected by a corp of 27 interviewers in the summer and
fall of 1986. Interviewers received five days of intensive training on general interviewing techniques, administration of the survey instrument
and study-specific procedures. Interviews were conducted in respondents' homes using a highly structured interview schedule that contained both interviewer- and self-administered portions. The complete
interview required approximately 90 minutes to administer. Respondents were compensated $25 for their time.
Routine steps were taken to ensure the standardization of interview
procedures (e.g., careful interviewer training, the development and use
of field manuals, etc.), which served to minimize the introduction of
both random and systematic error. In addition, standard procedures
were followed to assure respondents of the anonymity and confidentiality of the data, including conducting interviews in private whenever possible. Of the 1057 respondents included in the present study, however,
259 (24.5%) were interviewed in the presence of older children or adults.
Subsequent analyses of these data showed that the presence of an adult
or older child did not significantly influence the reporting of alcohol
consumption or alcohol problems. Thus it would appear that social desirability biases were not invoked by the presence of others with regard
to the reporting of these key and potentially sensitive data.
Measures
Measures were administered in a fixed order. Except for the expectancies and anger coping measures, which were self-administered, all measures included in the present study were interviewer-administered.
Problem drinking status. Subjects were administered questions from
the National Institute of Mental Health Diagnostic Interview Schedule
(DIS, Robins, Helzer, Croughan, Williams, & Spitzer, 1981) to approximate a psychiatric diagnosis of alcohol abuse or dependence according
to DSM-III criteria. Subjects were first asked if they had ever experienced any of 17 symptoms (e.g., needing a drink before breakfast, having trouble on the job or at school because of drinking, having the
"shakes" after stopping or cutting down drinking). Affirmative responses were followed by a question regarding the number of times that
particular symptom had been experienced within the past year. Following DSM.III criteria, symptoms were divided into three categories: (A)
a pattern of pathological alcohol use, (B) impairment in social or occupational functioning due to alcohol use, and (C) evidence of tolerance
or withdrawal. Subjects had to report the occurrence of at least one
221
222
Results
3 It has been suggested that low internal reliability may reflect an inherent property of the construct rather than inadequate measurement.
Specifically, Lazarus and Folkman (1984) have argued that the successful use of one strategy within a particular domain of strategies may obviate the need to use other strategies within that domain. Thus, one would
not necessarily expect to obtain high estimates of internal consistency
reliability.
4 The only exceptions to this pattern are the significant, though modest (. 16 ~< r < .25), positive correlations between avoidance coping, in
the first subset, and the remaining subscales of the HDL Coping Response Index, Active Cognitive and Active Behavioral Coping, in the
second subset. Presumably at least some portion of the relationship between these subseales reflects shared method variance. In addition,
there may also be a spurious component due to perceived stressfulness
of the event or situation with which the individual was coping.
,',',',,,', 9
"0
( (()"
....
~. . . .
....
m~m
i ~
i...
. . . . . .
0~
3~
223
t~
"' . . . .
I .....
Hierarchical multiple regression analyses were used to estimate the model depicted in Figure 1. The full model, consisting
of three equations, was estimated twice: once using the emotion-focused coping indexes and a second time using the problem-focused coping indexes. 5 In Equation 1, drinking to cope
was regressed on all hypothesized predictors (i.e., positive alcohol-related expectancies, the coping indexes and, in a subsequent step, the appropriate Expectancies Coping interaction
terms). In Equation 2, alcohol consumption (the average number of drinks consumed per day over the past year) was regressed
on all variables to its left in the model (drinking to cope, expectancies, and coping followed by Expectancies X Coping). Finally, in Equation 3, problem drinking status was regressed on
all variables to its left (consumption, drinking to cope, expectancies, and coping, again followed by Expectancies Coping).
This analytic strategy is highly similar to path analysis with
variables postulated as effects regressed simultaneously on all
variables postulated as causes. Thus, the contribution of each
term in the model is evaluated, controlling for all other terms
in the model. As such, the present analytic strategy is highly
conservative, attributing only nonoverlapping variance to each
factor. In addition, this strategy enables the estimation of direct
and indirect effects among variables.
Strictly speaking, however, the present analytic strategy
differs from traditional path analysis in two respects9 First, an
<
-d
.~
~ - - ~ - - ~
. .
. . . . . .
AI
5 Individual coping scales are used throughout these analyses. Although use of composite coping measures would simplify subsequent
analyses and potentially facilitate their interpretation, the composites
formed on the basis oftbe reported factor analysis were not highly reliable. For the anger-in, anger-out, anger-reflect (reverse scored), avoidance coping composite, coefficientalpha was 9 for the three problemfocused coping indexes, coefficient alpha was 9 These estimates are
not only lower than six oftbe seven reliability estimates obtained for the
individual subscales, but also suggest that unique information carried
by the individual scales might be lost by compositing.
224
Table 2
R2
change
Standardized
beta weight
.245***
.389***
.170***
.123**
-.023
-.010
.015***
.073*
.084*
-.006
.033
.272***
Problem-focused coping
(N= 951)
Variables entered on
R2
change
Step 2
Expectancies
Coping indexes
JH Active
HDL Active Cognitive
HDL Active Behavioral
.207***
Step 3
Expectancies Coping
JH Active
HDL Active Cognitive
HDL Active Behavioral
.007*
Total R 2
Standardized
beta weight
.438***
-.053
.069
.018
-.088**
.049
-.012
.226***
Note. JH Active = "John Henryism" active coping style; HDL = Health and Daily Living Coping Response Index. Both analyses were conducted
forcing age and sex into the equation on Step 1. R 2 for Step 1 was .012 (p < .01) across both equations.
*p < .05. **p < .01. ***p < .001.
interaction term has been incorporated in the model and, second, multiple indicators for one of the components in the model
(coping) are included, among which no causal relationships
have been hypothesized. However, both modifications to standard path analytic procedures have been explicitly discussed in
the literature and recommended procedures have been followed
here (see Cohen & Cohen, 1983). 6
Predicting drinking to cope. Table 2 presents the results of
the first pair of parallel multiple regression analyses in which
drinking to cope is regressed on positive expectancies and two
sets of coping indexes. Results are presented for emotionfocused coping and expectancies in Columns 1 and 2 and for
problem-focused coping and expectancies in Columns 3 and
4. Beta weights are reported throughout in order to facilitate
comparison of the relative importance of variables in the
model.
Summarizing across both equations, Table 2 shows that the
block of expectancy and coping variables was significant, accounting for more than 20% of the variance in drinking to cope
after controlling for sex and age. Examination of the beta
weights reveals that expectancies made a significant positive
contribution across both equations. Among the coping indexes,
only two of the emotion-focused coping variables, anger-in and
avoidance coping, made significant independent contributions.
The direction of their beta weights suggests that individuals who
suppress their anger and use avoidance coping are more likely
to drink to cope. Comparison of the magnitude of the betas for
coping and expectancies shows that expectancies contributed
relatively more to the prediction of drinking to cope.
Blocks of Coping Expectancy interaction terms were entered on Step 3. The interactions of both Avoidance Coping and
Anger-In Positive Expectancies were significant and together
accounted for 1.5% additional variance. The JH Active Coping
225
Variables entered on
change
Step 2
Drinking to cope
Expectancies
Coping indexes
HDL Avoidance
Anger-In
Anger-Out
Anger-Reflect
Step 3
Expectancies Coping
HDL Avoidance
Anger-In
Anger-Out
Anger-Reflect
Total R E
.157**
Problem-focused coping
(/7=951)
Standardized
beta weight
Variables entered on
Step 2
Drinking to cope
Expectancies
Coping indexes
JH Active
HDL Active Cognitive
HDL Active Behavioral
.163"*
.365**
.070*
.015
-.053
.022
-.019
Step 3
Expectancies Coping
JH Active
HDL Active Cognitive
HDL Active Behavioral
.000
-.014
-.026
.041
.035
.002
.275**
TotalR 2
R2
change
Standardized
beta weight
.369**
.076*
.024
-.014
-.076"
-.017
.013
.008
.283**
Note. JH Active = "John Henryism" active coping style; HDL = Health and Daily Living Coping Response Index. Both analyses were conducted
forcing age and sex into the equation on Step 1. R2 for Step I ranged from. 115 to. 119 ( ps < .001) across both equations.
*p < .05. **p < .01.
cies, the correlation was only - . 0 7 (p < . 10) compared with a
correlation o f - . 0 1 (p > . 10) among individuals low in expectancies.
Results of these analyses provide clear support for the proposed model. Positive alcohol expectancies and an avoidant
style of coping with emotion accounted for nearly 25% of the
variance in the self-reported use of alcohol to cope. Conversely,
more active, problem-focused coping strategies did not predict
drinking to cope] In addition, expectancies and coping interacted, as hypothesized, to predict the use of alcohol to cope.
Examination of the form of the interaction showed that drinking to cope is most likely among individuals who rely on an
avoidant style of coping with emotion and hold strong positive
expectancies for the effects of alcohol.
Predicting alcohol consumption. Table 3 presents the results of the second pair of parallel regression analyses in which
alcohol consumption was regressed on drinking to cope, positive expectancies, and the two sets of coping indexes. Because
alcohol consumption was highly skewed (skewness = 5.5), these
analyses were conducted using a log-transformation of the
dependent variable. Skewness of the transformed variable
was -.023.
Analyses for emotion-focused and problem-focused coping
revealed few differences. Summarizing across both equations,
it can be seen that drinking to cope, positive expectancies, and
the general coping indexes accounted for approximately 16% of
the variance in alcohol consumption after controlling for sex
and age. As predicted, drinking to cope and positive expectancies made significant independent contributions. Active behavioral coping also made a significant, though unpredicted, contribution. Comparison of the relative magnitude of these effects
reveals, however, that drinking to cope is a substantially more
important determinant of alcohol consumption than either ac-
226
Table 4
Hierarchical Multiple Regression Analyses Predicting Problem Drinking Status From Consumption,
Drinking to Cope, Positive Expectancies, and Two Sets of Coping Indexes
Emotion-focused coping
(N = 947)
Variables entered on
Step 2
Alcohol consumption
Drinking to cope
Expectancies
Coping indexes
HDL Avoidance
Anger-In
Anger-Out
Anger-Reflect
Step 3
Expectancies Coping
HDL Avoidance
Anger-In
Anger-Out
Anger-Reflect
Total R 2
R2
Change
Problem-focused coping
(N= 951)
Standardized
beta weight
.207***
.279***
.220***
.067*
.028
-.033
.104**
.015
.002
.018
.044
-.033
-.005
.271 ***
R2
change
Step 2
Alcohol consumption
Drinking to cope
Expectancies
Coping indexes
JH Active
HDL ActiveCognitive
HDL ActiveBehavioral
.204***
Step 3
Expectancies Coping
JH Active
HDL ActiveCognitive
HDL ActiveBehavioral
.002
Total R 2
Standardized
beta weight
.283***
.235***
.082*
.016
-.013
-.012
-.003
-.051
.045
.267***
Note. JH Active = "John Henryism" active coping style; HDL = Health and Daily Living Coping Response Index. Both analyses were conducted
forcing age and sex into the equation on Step 1. R 2 for Step 1 was .062 ( ps < .001) across both equations.
*p < .05. **p < .01. ***p < .001.
Summarizing across both equations, it can be seen that quantity of alcohol consumed, drinking to cope, and positive expectancies made significant independent contributions to the prediction of problem drinking status after controlling for age and
sex. Jointly these variables accounted for more than 20% of the
variance in abuse status. Calculation of a squared multiple correlation eliminating consumption from the equation showed
that drinking to cope and expectancies alone account for nearly
14% of the variance in abuse status.
Examination of the beta weights shows that all relationships
were positive and that drinking to cope and alcohol consumption contributed nearly equally to the prediction of abuse. In
contrast, the expectancy effect was substantially weaker. Interestingly, anger-out also significantly predicted problem drinking status, although the magnitude of the effect was quite small
(sr2 = .006). The direction of the beta weight suggests that individuals who express their anger outwardly are more likely to
abuse alcohol. Overall, these results are consistent with predictions. They suggest that, at every level of alcohol consumption,
individuals who report using alcohol to cope and who hold
strong positive expectancies for the effects of alcohol are more
likely to abuse alcohol.
227
w,"I
Positive I-
9......
.17"**
Avoidance
"c:::.;c' ). ;
. , J
.32"**
\.,
'".
'"_
"~f
***
Alcohol
,-,
~ Dependence)
Drinking"~ ~
<su
Figure2. Estimated model relating drinking to cope, expectancies, and emotion-focused
coping to alcohol use and abuse. (* p < .05; ** p < .01; *** p < .001 .)
hen & Cohen, 1983). Following this logic, indirect effects are
computed only if the Bs for all paths are significant atp < .05.
In order to estimate the indirect effects for anger-in and avoidance coping on abuse status, supplementary analyses were required. Because of the significant Coping X Expectancies interaction terms, two subgroup regressions were conducted in
which drinking to cope was regressed on avoidance coping and
anger-in simultaneously for individuals one standard deviation
above or below the mean on positive expectancies. Results of
these analyses showed that, for individuals holding strong positive expectancies (n = 174), both anger-in and avoidance coping
made significant independent contributions to the prediction of
drinking to cope (Bs = .231 and .284 for anger-in and avoidance
Table 5
Direct
Indirect
Total
Alcohol consumption
Drinking to cope
Expectancies
Emotion-focused coping
(at + 1SD above mean)
Avoidance
Anger-in
.279**
.220**
.067*
-.102
.137
.279
.320
.204
.000
.000
.000
.166
.113
.053
.166
.113
.053
Note. All indirect effects are statistically significant (except for alcohol
consumption) according to the criterion suggestedby Cohen and Cohen
(1983).
*p < .05. **p < .001.
228
the results of this study provide strong support for the proposed
model, linking coping styles, alcohol expectancies, and drinking
to cope to alcohol use and abuse. However, these data suggest
that the relevant coping domain is restricted to avoidant styles
of coping with emotion. Expectancies were also shown to moderate the relationship between avoidant styles of emotion-focused coping and drinking to cope, such that individuals who
hold strong positive expectancies and also use avoidant styles of
coping with emotion were most likely to drink to cope. In turn,
individuals who hold strong positive expectancies and who
drink to cope not only drink more, but are also more likely to
experience problems as a result of their drinking. Collectively,
the direct effects of consumption, drinking to cope, and expectancies account for approximately 20% of the variance in alcohol abuse, with drinking to cope and expectancies alone accounting for almost 14% of that variance. Although magnitude
estimates for indirect effects cannot be directly converted into
percentage of variance estimates, comparison of the relative
magnitude of the combined indirect effects for drinking to cope,
expectancies and general coping styles (. 10 + . 14 + . 17 = .41)
with the combined direct effects of drinking to cope and expectancies (.23 + .07 -- .30) reveals that the total indirect effects
are approximately 1.4 times greater than the total direct effects.
Given that the direct effects alone account for 14% of the variance in abuse status, these data suggest that the total contribution of drinking to cope, expectancies, and emotion coping to
alcohol abuse are substantively important.
Reliance on drinking as a coping strategy emerges as the most
powerful explanatory variable in the model, contributing to alcohol abuse via direct and indirect pathways. Importantly, these
data reveal that individuals who drink to cope are more likely
to experience problems indicative of abuse syndromes regardless of their level of consumption. One implication of this finding is that reasons for drinking are important determinants of
the consequences of drinking and, in particular, that drinking
to cope may be intrinsically maladaptive. It seems reasonable
to speculate that individuals who rely on alcohol to cope with
dysphoric feelings may become more psychologically dependent on alcohol. Increased psychological dependence, in turn,
may promote continued drinking despite the experience of negative consequences indicative of abuse syndromes.
To the extent that this interpretation is valid, nonpathological
reasons for drinking (e.g., social reasons) should not directly
predict abuse after controlling for consumption. In fact, when
the model depicted in Figure 1 was reestimated by using a threeitem Drinking for Social Reasons scale (Polich & Orvis, 1979;
= .73) in lieu of drinking to cope, the Drinking for Social
Reasons scale did not predict problem drinking status directly,
although it made a small significant indirect contribution via
increased consumption. Additional evidence of discriminant
validity is provided by the set of regressions in which drinking
for social reasons served as dependent variable. These analyses
revealed that expectancies, but not coping, significantly predicted drinking for social reasons, thereby suggesting that
drinking for social reasons is unrelated to general levels of coping ability.
Collectively, these secondary analyses provide additional support for the notion that motivations for drinking differentially
influence the outcomes of alcohol use. Specifically, these analy-
229
230
tional Institute of Mental Health Diagnostic Interview Schedule. Archives in General Psychiatry, 38, 381-389.
Robins, L. N., Helzer, J. E., Croughan, J., Williams, J. B. W., & Spitzer,
R. L. (1981). NIMH Diagnostic Interview Schedule: Version IlL
(Contract No. MH 278-79-00). Rockville, MD: National Institute of
Mental Health.
Roehling, P. V., Smith, G. T., Goldman, M. S., & Christiansen, B. A.
(1987, August). Alcohol expectancies predict adolescent drinldng: A
three year longitudinal study, Paper presented at the 95th Annual
Convention of the American Psychological Association, New York
City.
Rohsenow, D. J. (1983). Drinking habits and expectancies about alcohol's effects for self versus others. Journal of Consulting and Clinical
Psycholog3z, 51, 752-756.
Smith,, G. T., Roehling, P. V., Christiansen, B. A., & Goldman, M. S.
(1986, August). Alcohol expectancies predict early adolescent drinking: A longitudinal study Paper presented at the 94th Annual Convention of the American Psychological Association, Washington, DC.
Spielberger, C. D., Johnson, E. H., Russell, S. E, Crane, R. J., Jacobs,
G. A., & Worden, T. J. (1985). In M. A. Chesney & R. H. Rosenman
(Eds.), Anger and hostility in cardiovascular and behavioral disorders
(pp. 5-30). Washington, DC: Hemisphere Publishing.
Stone, A. A., Lennox, S., & Neale, J. M. (1985). Dally coping and alcohol use in a sample of community adults. In S. Shiffman & T. A.
Wills (Eds.), Coping and substance use (pp. 199-220). Orlando, FL:
Academic Press.
Timmer, S. G., Veroff, J., & Colten, M. E. (1985). Life stress, helplessness, and the use of alcohol and drugs to cope: An analysis of national
survey data. In S. Shiffman & T. A. Wills (Eds.), Coping and substance use (pp. 171-198). Orlando, FL: Academic Press.
Wilsnack, S. C., Klassen, A. D., & Wilsnack, R. W. (1984). Drinking
and reproductive dysfunction among women in a 1981 national survey.Alcoholism: Clinical and Experimental Research, 8, 451-458.
Zajonc, R. B. (1984). On the primacy of affect. American Psychologist,
39, 117-123.
Received August 26, 1987
Revision received November 20, 1987
Accepted November 30, 1987 9