University of Southern California Dissertations and Theses

Expectancies for alternative behaviors predict drinking problems: A test of a cognitive reformulation of the matching law

(USC Thesis Other)

Expectancies for alternative behaviors predict drinking problems: A test of a cognitive reformulation of the matching law

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORM ATION TO USERS
This manuscript has been reproduced from the microfilm master. U M I films
the text directly from the original or copy submitted. Thus, some thesis and
dissertation copies are in typewriter face, while others may be from any type of
computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality illustrations
and photographs, print bteedthrough, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send UM I a complete manuscript
and there are missing pages, these w ill be noted. Also, if unauthorized
copyright material had to be removed, a note w ill indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and continuing
from left to right in equal sections with small overlaps.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations appearing
in this copy for an additional charge. Contact UMI directly to order.
ProQuest Information and Learning
300 North Zeeb Road. Ann Arbor, M l 48106-1346 USA
800-521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
EXPECTANCIES FOR ALTERNATIVE BEHAVIORS PREDICT DRINKING
PROBLEMS: A TEST OF A COGNITIVE REFORMULATION OF THE
MATCHING LAW
by
Boaz Levy
A Thesis Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF ARTS
(Psychology)
Copyright December 1999 Boaz Levy
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 1409641
___ __®
UMI
UMI Microform 1409641
Copyright 2002 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor. Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY O F SOUTHERN CALIFORNIA
THC GRADUATE SCHOOL.
UNIVERSITY PARK
LO S ANGELES. CALIFORNIA SOOOT
This thesis, written by
& 0/x-t LEM'J______________
under the direction of hcS. Thesis Committee,
and approved by all its members, has been pre
sented to and accepted by the Dean of The
Graduate School, in partial fulfillment of the
requirements for the degree of
Master of A rts
THESIS COMMITTEE
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Contents
Abstract..............................................................2
Preface............................................................... 3
Introduction....................................................... 5
Method................................................................15
Results.................................................................17
Discussion..........................................................40
References..........................................................56
Index.................................................................. 59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract
This study explored the role expectancies for alternative behaviors play in
problem drinking. The matching law predicts that when behavior is free to vary
(reinforcement is available for more than one behavior), animals and humans will
allocate their responses across behaviors in exact proportion to the value derived from
each. Extending the theoretical framework of the matching law from an operant
conditioning to a cognitive mediational model, we hypothesized that expectancies for
alternative behaviors in a given setting would influence the link between alcohol
expectancies and drinking behavior. Under this model, the alternatives serve as
competing behaviors to drinking. We thus expected high alternative expectancies to
buffer drinking problems for individuals holding high alcohol expectancies. Specifically,
this model predicted that among subjects reporting high levels of alcohol expectancies,
high scorers on alternative expectancies scales will drink less and develop fewer
problems than low scorers. The setting explored in this study was the university campus
and the participants were undergraduate students. The alternative expectancies assessed
were: Studying, Working, Hobbies, Sports and Family. Analyses suggested a buffering
effect for the “studying” and “working” expectancies, but no effect for “hobbies”,
“sports” and “family”. In addition, these scales accounted for unique variance in problem
drinking. These results suggest that enhancing expectancies for alternative behaviors may
decrease problem drinking.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Preface
This study approaches problem drinking from a perspective of decision making and
choice. Emphasizing the role that choice plays in problem drinking seems particularly
important given the general growing predominance of genetic, neurological, biological and
pharmacological paradigms in psychopathology treatment and research. While the
proliferation of research exploring the physiological basis of abnormal behavior has generally
made major contributions to the development of effective interventions, it has also impressed
the notion that psychological maladjustment principally results from neuro-chemical
impairments.
Taken to the extreme, the focus on the neuro - biological approach may significantly
minimize the appreciation of the control clients have over dysfunctional behavior and the
responsibility or shame associated with it. Social and moral questions regarding the
responsibility a client has over problem drinking are, of course, beyond the scope of
empirical study. However, these issues are pertinent here because they constitute important
yet potentially controversial aspects of the context that shapes the different ways that
researchers approach the study of problem drinking. Accentuating choice in problem drinking
may raise an opposition against the accusatory implications this approach may carry for
alcohol abusers. Such accusations by no means motivate this line of research. On the
contrary, applying the psychology of choice to the study of problem drinking may lead to
productive research that will prove valuable, especially for prevention.
The question of choice may, indeed, seem less relevant for some disorders (such as
the schizophrenias); however, it probably plays a more significant role in explaining
substance abuse. In essence, the pathological state in alcohol abuse/dependence is
progressive. After all, no client had ever reported irresistible craving for alcohol before she
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ever had her first drink. Drinking alcohoi involves volition. Thus, in addition to neuro -
biological predisposition for drinking, cognitive factors of choice and decision making
probably play an important role in guiding one’s drinking behavior before one develops
problems.
The assumption underlying the general line of research proposed in this study holds
that, before reaching a pathological state, alcohol abusers make a lot of choices that guide
their drinking behavior; and, apparently, a lot of dysfunctional ones. The proposed approach
thus suggests that the application of psychological models of decision making and choice to
problem drinking research may help in characterizing differences between the ways alcohol
abusers and non - problem drinkers make choices. This knowledge may contribute to the
understanding of problem drinking behavior, and may also prove useful for practical
purposes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
Introduction
In an effort to decrease alcohol - related problems, researchers have enhanced the
accuracy of assessment of the cognitive and behavioral factors that influence alcohol
consumption. Major measurement advances have focused on evaluating alcohol
expectancies: beliefs about subjective effects of alcohol (Brown, Goldman, Inn &
Anderson, 1980). Expectancies covary with drinking patterns in adolescence
(Christiansen, Goldman, & Inn, 1982; Fromme, Stroot and Kaplan, 1993); and beliefs
about the reinforcing effects of drinking influence the quantity and frequency o f alcohol
consumption (Goldman, Brown & Christiansen, 1987). Adolescent problem drinkers
expect more positive effects o f alcohol on social, cognitive and motor performance
(Christiansen and Goldman, 1983) than non - problem drinkers. Beyond discriminating
heavy from light drinkers (Southwick, Steele, Marlatt, & LindelL, 1981) and problem
from nonproblem drinkers (Brown, Golden and Cristiansen, 1985), alcohol outcome
expectancies predict problem drinking among college students and adolescents (Brown,
1985b). Similarly, expectancies predicted longitudinal transition from nonproblem to
problem drinking status among adolescents (Christiansen, Smith, Roehling & Goldman,
1989). These data suggest that alcohol expectancies may mediate the development of
problematic drinking patterns. (Brown et al. 1985; Cox & Klinger, 1990). In other words,
individual who expect to derive considerably more positive outcome from alcohol may be
at greater risk for developing alcohol - related problems (Fromme, 1994).
The research on alcohol expectancies emphasizes the important role cognitive factors
play in problem drinking. Extending the basic model of operant conditioning to a cognitive
mediational framework enabled to account for individual differences in drinking behavior.
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The cognitive mediational model predicts which individuals in a certain social group (i.e.:
college students) will drink more than others when the availability of alcohol and the social
reinforcement for drinking do not vary significantly among group members. The theoretical
proposition of the cognitive behavioral model is, therefore, that behavior is determined not
only by the schedule of the reinforcement but also by its quality. In addition, this model
maintains that the quality of reinforcement is subjective. The practical advantage of this
perspective is that it can help in making predictions on an individual level in an uncontrolled
environment (outside the laboratory) because, as opposed to schedules of reinforcements, the
subjective quality of a reinforcement should not vary considerably across situations'. Thus,
theoretically, it is the cognitive factors (alcohol expectancies in this case) that should help us
in targeting problem drinkers outside the laboratory - the ultimate goal of this line of
research. Following the same line of reasoning, to further enhance the accuracy of problem
drinking predictions in an uncontrolled environment, the present study attempts to extend
Hermstein’s matching theory (1970), which constitutes an operant model of choice, into a
cognitive mediational framework of reinforcement expectancies.
Hermstein’s matching law is a theory of choice. It was devised as a mathematical
model for predicting response allocation across concurrent schedules of reinforcement,
offering a theoretical framework of applied behavior analysis for studying choice (Baum &
Rachlin, 1969; McDowell, 1988; Rachlin, 1989; Hermstein, 1997). Within this framework,
any behavior and its controlling contingency serve as one option among an array of
concurrently available alternative responses with differing schedules of reinforcement (Neef,
Mace, Shea, & Shade, 1992). The matching law does not predict whether an organism will
1 To qualify the latter assertion, it is clear that alcohol expectancies vary from one situation to the next (i.e.: the
degree to which one expects alcohol to make one more sexually aroused depends also on one company while
drinking). The assumption here, though, is that there are general and relatively stable expectations of alcohol
effects which people bring into the situation (like a personality trait) and on which people tend to differ.
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
eat, drink, sleep, work or play in the next minute or ten minutes. Instead, the law predicts
what fraction of hours per day the organism will spend in eating, sleeping, playing, working,
or, relevant to the purpose of this study, drinking alcohol. In Hermstein words, the matching
law predicts “not whether the pitcher will throw a fastball, curve, slider or changeup at his
next throw, but the proportion of each type of pitch he will throw to a given batter”
(Herrnstein, 1997). Under this theoretical framework, the dependent variable is the relative
frequency of responding.
The matching law views life in terms of choice. It assumes that when behavior is free
to vary (that is, reinforcement is available for more than one behavior), animals or humans
will allocate their responses across behaviors in exact proportion to the value derived from
each. More generally, the distribution of behavior across response alternatives tends to be
proportional to the reinforcement received from those alternatives. This equality between the
relative rate of responding and the relative rate of reinforcement is called matching.
Under highly controlled laboratory conditions, using non-human subjects, the
matching law appears to offer relatively accurate predictions of the proportions of behavior
distribution across available alternatives (Davison & McCarthy, 1988). Attempts to replicate
parallel experiments in the operant laboratory with humans suggest that the matching law can
also describe human behaviors under the highly controlled settings of experimental
conditions (Pierce & Epling, 1983). A few studies have investigated the applicability of
matching theory to socially important human behavior. McDowell (1981) found that natural
covariations between self-injurious behavior and parental attention were predicted by
Hermstein’s (1970) single - alternative formula of the matching law (for description of this
methodology, see McDowell, 1988). Likewise, Martens and Houk (1989) reported such
covariations between disruptive behavior and teacher attention. In a study of social behavior
under concurrent VIVI schedules, Conger and Killeen (1974) had each of their subjects
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
discuss drug use with three experimenters. Two of the experimenters provided
complimentary statements contingent on conversation directed at the experimenters
according to independent VI schedules, while the third experimenter moderated the
discussion. As predicted by the matching law, subjects divided their conversation between
the two experimenter in approximate proportion to the rate of attentive comments supplied by
the experimenters.
These results extend matching theory to applied situations that closely resemble the
controlled settings of the operant laboratory. However, the translation of these robust
findings to applied clinical settings may be hampered by important differences between
laboratory conditions and the complex human environment (Baer, 1981; Faqua, 1984). One
fundamental difference between laboratory conditions and natural contingencies for human
behavior is the variation in reinforcements for alternative behaviors. In particular, laboratory
conditions typically hold the quality (same kind of attention or food) of reinforcements
constant and merely vary their rate. For humans in natural settings, however, the quality of
reinforcement varies across response alternatives (praise vs. money, etc.). In an important
experiment, Neef et al (1992) demonstrated how matching relation is disrupted when, as
occurs in most natural choice situations, the quality of the reinforcement differs across the
response options. In this experiment, students completed math problems from two alternative
sets of concurrent VI schedules of reinforcement under equai-quality and non equal quality
conditions. Results indicated a bias toward the higher quality reinforcement that is
independent of reinforcement rate. Thus, people vary their response rate with the quality of a
reinforcement as well as with its schedule.
Assessing the quality of reinforcement under the operant paradigm is highly accurate
because the researcher can observe behavioral preference directly. However, in an
uncontrolled environment, inferences about observed behaviors become problematic.
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Furthermore, the observational method, in general, is more difficult to apply, especially when
assessing a large number of subjects. To complicate the picture even further, the notion that
the quality of reinforcement varies across individuals is totally absent in the operant
paradigm. Consequently, the operant model lacks the ability to account for individual
differences in choice behavior. Since the necessity for obtaining comparative information is
intrinsic to targeting problem drinking, this limitation impedes the application of the
matching law theory to this line of research. The current research proposes that inserting
cognitive mediational factors, such as expectancies of reinforcement, may help to overcome
some of these limitations.
From a cognitive behavioral perspective, behavior patterns should vary across
individuals even when the schedule of reinforcements for all the available alternatives is held
constant for all subjects. To illustrate, in a vacation resort where all the enjoyable activities
are equally available for everyone, while one person eats a two hour lunch, then sleeps for
another two hours and afterwards plays ping pong for one hour, another person takes only
one hour for lunch, then swims for two hours and then sits at the bar for forty minutes. The
underlying assumption here is that people differ in their expectations for how reinforcing
each alternative is; or, in turn, the quality of their alternative expectancies is subjective.
The cognitive perspective holds that expectancies should vary across individuals even
if the history of reiforcement is controlled. It is assumed here that expectancies are shaped by
the unique characteristics of the individual. For example, a highly intelligent person will
probably experience more positive reinforcement from engaging in challenging intellectual
activities than a dull person. Likewise, individuals possessing athletic physical qualities will
probably enjoy competitive sports more than others. In addition, the cognitive approach
assumes that, in addition to schedules of reinforcements, learning is influenced by processes
that are primarily cognitive: memory, mental representations and attributions. As Edward
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Tolman (1932) demonstrated, learning occurs also without reinforcement; and, as exhibited
by Albert Bandura (1977), human behavior can be acquired vicariously through imitation. It
is thus assumed that expectations also develop through cognitive processes. Finally, research
on cognitive therapies provide robust evidence that dysfunctional behavior can be altered by
cognitive restructuring. Hence, if alternative expectancies play a casual role in problem
drinking, changing these expectancies may lead to attenuation in the drinking behavior.
The proposed study examines the link between alcohol expectancies and drinking
behavior within a cognitive mediational framework of the matching theory. The basic
assumption in this study is that at any given moment in which individuals drink alcohol they
have the option to pursue a different potentially rewarding engagement2. The matching law
tells us that when reinforcement is available for more than one behavior, the occurrence of a
particular behavior (alcohol drinking in this case) is no longer determined merely by its
contingent reinforcement. Instead, it becomes a function of the constellation of
reinforcements concurrently available for alternative behaviors. In a similar manner, we
hypothesize that in addition to alcohol expectancies, drinking behavior is also influenced by
expectancies an individual possesses for alternative reinforcing activities available under the
circumstances. Matching Theory suggests that alcohol drinking also depends on the situation
and should thus be studied in context.
More specifically, Hermstein’s formula predicts that when the schedule of
reinforcement for a particular behavior is kept constant, the engagement in this activity will
decrease as new, competing alternatives are introduced into the situation. Hence, we
hypothesize that under some given circumstances3 , when two individuals possess similar
levels of alcohol expectancies, the person who holds lower expectancies for the alternative
2 This assumption excludes subjects who meet the criteria for alcohol dependence.
3 Circumstances in a general sense. I.E: undergraduate students who live on campus.
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
behaviors available will drink more than the person higher on those expectancies. More
generally, this ideographic approach suggests that possessing high expectancies for available
alternatives that compete with drinking may lower drinking behavior and buffer drinking
problems.
A major concern in this study regards the practical limitations of assessing the entire
scope of expectancies for alternative behaviors in an uncontrolled setting. The focus should
thus be narrowed to those alternative behaviors deemed common and most influential across
the vast majority of individuals in a particular setting. The setting explored in this study is the
university campus and the subjects are undergraduate students. The alternative activities
deemed to capture the largest common ground undergraduate students share are: studying,
working, physical training and sports, intellectual and recreational activities that do not
involve alcohol (going to the movies or reading a book, etc) and social interactions that do
not involve alcohol. Participants will complete four questionnaires that assess drinking
habits, drinking problems, alcohol expectancies and alternative expectancies4.
The current study will attempt to replicate previous findings showing that alcohol
expectancies covary with drinking problems. We thus hypothesize that analysis will show a
significant positive relation between beliefs about the reinforcing effects of alcohol drinking
and drinking problems. In addition, using alcohol expectancies as a predictor, we will try to
obtain the probability estimate and a confidence interval for detecting problem drinkers. This
analysis is consistent both with the practical goal of targeting problem drinkers and the major
theoretical advantage the cognitive mediational model - namely, the ability to account for
individual differences in drinking,
4 The alternative expectancies questionnaires were devised for this study.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The major aim of the study is to explore the role expectancies of and actual engagement
in alternative reinforcing behaviors play in drinking habits and drinking problems. We
hypothesize that alternative expectancies and activities will account for unique variance in
drinking problems, over and above the variance alcohol expectancies account for alone.
The second hypothesis is that among students reporting the same level of alcohol
expectancies and activities, high scorers on alternative expectancies scales will drink less
and develop fewer alcohol related - problems than low scorers. Similar to the one predictor
analysis, the multiple predictors analysis will attempt to obtain a probability estimate and a
confidence interval for the individual case. That is to say, the analysis will try to answer the
question, “what is the probability that, among students reporting similar levels of alcohol
expectancies, a randomly sampled student with high alternative expectancies will develop
fewer drinking problems than a randomly sampled student with low alternative
expectancies?’.
Analysis on the individual level is important for theoretical purposes because it helps
specify the conditions under which problem drinking develops. The ultimate goal is to be
able to predict whether an individual client will engage in problem drinking. To accomplish
this goal, we need to characterize the differences between individuals who develop drinking
problems and those who do not under hypothesized high - risk conditions. Understanding the
differences between these types of individuals should lead to further refinement in high - risk
conditions for problem drinking, and enhance the probability for a successful prediction. For
example, we know that, on average, alcohol abusers tend to possess higher alcohol
expectancies than non - problem drinkers. On the other hand, many individuals possess high
alcohol expectancies but do not develop drinking problems. We believe these individuals
differ in their alternative expectancies. If the analysis supports this hypothesis, we have
succeeded in further specifying the high - risk conditions for the development of problem
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
drinking. The next step should be to explore the difference between the anomalies that
emerge and the successful predictions.
Estimating the probability of problem drinking on an individual level provides us with a
measure of how well we understand the conditions under which drinking problems develop.
The higher the probability for making a successful, individual prediction, the more anomalies
we can account for; and, in turn, the more insight we have into the problem - drinking
phenomenon. Differentiating between successful predictions and anomalies should increase
the probability of making future successful predictions.
From a practical perspective, the analysis on the individual level helps us assess the
confidence we have in predictions about drinking behaviors. Much is gained if we can
provide an instrument (or instruments) that will actually specify the probability of drinking
problems for individual clients. In forensic psychology, for instance, a common goal is to
help the Dependency Court decide about custody cases. Substance abuse problems are a
major cause for denying custody from biological parents. Thus, in many cases, the judge
wishes to evaluate the risk that the parent will relapse to abusing alcohol. Providing the judge
with an actual probability estimate may be of considerable assistance. The term “high - risk”
may mean that the parent is five times more likely to exhibit problem drinking behavior in
the future in comparison to the risk - level in the general population. However, if the risk in
the general population is 1% this means that the probability that the parent will actually
relapse to problem drinking is 5%. Five percent chance may not be sufficient to deny custody
from a biological parent. On the other hand, if research shows that the conditions which the
parent meet suggest that there is 70% chance of relapse, the judge knows that if she decides
not to deny custody under such circumstances, out of 100 cases, she will err 70 times. This is
practical knowledge, necessary for making crucial decisions. For this reason, a 0.3
probability for a successful prediction on an individual level is so much more informative
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
than a 0.3 Pearson correlation. The Pearson estimate will not help the judge reach a good
decision.
The analyses in this study do not yet provide this practical information; however they are
designed to support the proposed model directly. The analysis on the individual level
provides us with an estimate of how well the theory predicts individual events under the
condition specified. Namely: “what is the probability that, among students reporting similar
levels of alcohol expectancies, a randomly sampled student with high alternative
expectancies will develop fewer drinking problems than a randomly sampled student with
low alternative expectancies?’ Theoretically, we expect that the probability for a successful
prediction will increase (approach 1) toward the higher quantiles (of alcohol expectancies)
and drop toward the lower quantiles (approach 0.5 = random chance for successful
prediction) of the empirical distribution of alcohol expectancies. This effect should arise
more conspicuously on the higher levels of alcohol expectancies because drinking
behavior is expected to be more intense on these levels. Likewise, students who score on
the lower quantiles of alcohol expectancies should drink very little, or not at all, because
they do not expect a positive experience from drinking; thus, the effect in these quantiles
should disappear.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
METHOD
Subjects
Ninety-three undergraduate students, ages 19 - 30 (mean=20.7, S.D=3.11),
participated in the study for extra credit in an undergraduate psychology class at the
University of Southern California. The sample was ethnically diverse (European
Americans, African Americans, Asian American, Latino-Americans); and the gender
distribution was 25% males and 75% females.
Instruments
A. Drinking Habits. Participants completed a Quantity/ Frequency/ Maximum Index.
They reported the average number of standard drinking (i.e. 12 oz. Beer, 4 oz. glass
of wine, 12 oz. wine cooler, mixed drink, or 1.5 oz. liquor) they consumed per
drinking occasion in the past three months, the average number of drinking occasion
per week and per month in the past three months and the maximum number of drinks
consumed on one occasion in the last three months. A similar measure has been used
in previous work and appears to be a valid index of alcohol consumption (Earleywine
& Martin, 1993).
B. Drinking Problems. Participants completed the Rutgers Alcohol Problem Index
(White, Labouvie, 1989). The RAPI assesses alcohol-related problems in adolescent
and young adult populations. In the current study, the test score was used as a
continuous variable that indicates the frequency of experiencing negative
consequences due to alcohol use. The questionnaire consist of 23 items (i.e. went to
school high or drunk), each containing 5 possible frequency ranges during the last
three years (i.e. 0 = never, 1 = 1-2 times, 2 = 3-5 times, 3 = 6-10 times, 4 = more than
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10 times). For the purposes of this study, we have used a 0 to 10 scale instead, in
order to get a more refined range of scores.
C. Alcohol Expectancies. Participants completed the Alcohol Effects Questionnaire -
AEFQ (Brown, Goldman and Andersen, 1980). This questionaire assesses
undesirable effects of alcohol as well as positive reinforcing effects. It assesses only
personal beliefs (beliefs about the effect of alcohol on oneself as opposed to beliefs
about the effects of alcohol on people in general). The questionnaire consists of 40
items, and has a response format of 6 point rating scale (from strongly agree to
strongly disagree). For the purposes of this study, we extended the response format to
a 10 point rating scale.
D. Alternative Expectancies. Participants completed an Alternative Expectancies
questionnaire devised for the current study. The questionnaire assesses personal
beliefs about the potential pleasure or gain inherent in activities deemed
incompatible with drinking. The questionnaire assesses expectancies of five
different domains o f activities: studying, sports and physical training, social
interaction, hobies and intellectual interests, and work. The response format
consists of a 10 point rating scale (from strongly agree to strongly disagree). In
addition, participants rated in descending order the four activities that take up
most o f their time, and the four activities they enjoy the most.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Results
For each subject, two alcohol expectancies scores were calculated, positive and
negative, by summing across all items of each category separately. Likewise, each subject
received a RAPI score and five separate scores for each of the alternative expectancies. All
scores were calculated by summing across items.
From 93 respondents, 82 completed the alcohol expectancies questionnaire. The
11 participants who failed to fill out this section were non-drinkers who, instead of
responding to the alcohol expectancies items, simply stated, “I never drink”. Although
irrelevant for most analyses performed in the current study, these subjects are important
for discussing the possible implications the empirical distribution in this study suggests
for the way problem drinking is distributed in the population (because they all scored 0
on the RAPI). 30 subjects reported 0 drinks per month, and 19 subjects reported 0 drinks
per drinking occasion.
The Cronbach’s alpha coefficients for the scales of alternative expectancies were
calculated and appear to be satisfactory. The working expectancy scale was the only
exception. Its initial alpha estimate was 0.S3; however, after one item was dropped, the
estimate increased to 0.73. This result seems satisfactory, taking into account that after
dropping the item, the scale remains with only 4 items. The Cronbach Alpha estimates of
the alternative expectancies scales are summarized in the table below (Table 1):
Tabic l:_Croabach’s Alpha estimates for the alternative expectancies, alcohol expectancies and R API scales
Studying Sports Hobbies Family Work Positive Negative RAPI
0.83 0.78 0.80 0.86 0.73 0.95 0.85 0.95
To observe the inter - dependencies among the variables in the study, the Pearson correlation
was calculated for all variables (see table 2 in the next page).
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2. S iau u ry of the oae predictor correlatioa aaalyscs.
RAPI Al.Ex
Pos.
Al.Ex
Neg.
Freq.
week
Freq.
month
Aver. Max Study Work Sport Hob. Fam.
RAPI
1 .44*
-.26*
.49* .52* .27* .46*
-.29*
.17 .05 .08 -.10
Al.Ex
Pos.
1 .64* .34* .33* .24* .25* -.10 .16 .03 .12 .01
Al.Ex
Neg.
1 .05 .02 .06 .09 -.17 .18* .03 .167 -.14
Freq.
week
1 .95* .36* .52* .06 .17 .1 2 .03 .05
Freq.
month
1 .34* .54* -.08 .11 .12 .01 .02
Aver.
1 .74* -.12 .15 .01 .05 .02
Max
1
1
K )
si
•
-.1 2 .1 0 .06 -.12
Study
1 .18* .2 2 * .19* .32*
Work
1 .31* .27* .23*
Sport
1 .41* .32*
Hob.
1 .42*
Fam.
1
To test whether alternative expectancies account for unique variance, over and above the
variance alcohol expectancies account for alone, a two predictor analysis was performed. The
unique variance was significant only for the studying and working expectancies, and only
when the dependent variables were problem drinking and maximum number of drinks per
one occasion (See tables 3 and 4).
Table 3. Summary of the two predictor correlatioa aaalysis:
Depeadeat variable - Average
First predictor R square Second
Predictor
Cumulative R
square
Delta R square
Positive Exp. 0.06 Studying 0.07 0.01
Negative Exp. 0.003 Studying 0.01 0.01
Positive Exp. 0.06 Sports 0.06 0
Negative Exp. 0.003 Sports 0.003 0
Positive Exp. 0.06 Hobbies 0.07 0.01
Negative Exp. 0.003 Hobbies 0.003 0
Positive Exp. 0.06 Family 0.05 0
Negative Exp. 0.003 Family 0.003 0
Positive Exp. 0.06 Work 0.06 0
Negative Exp. 0.003 Work 0.02 0.01
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Dependent variable - freqaeacy/week
First predictor R square Second Predictor Cumulative R
square
Delta R square
Positive Exp. 0.12 Studying 0.12 0
Negative Exp. 0.002 Studying 0.002 0
Positive Exp. 0.12 Sports 0.13 0.01
Negative Exp. 0.002 Sports 0.01 0.01
Positive Exp. 0.12 Hobbies 0.12 0
Negative Exp. 0.002 Hobbies 0.002 0
Positive Exp. 0.12 Family 0.14 0.02
Negative Exp. 0.002 Family 0.01 0.01
Positive Exp. 0.12 Work 0.14 0.02
Negative Exp. 0.002 Work 0.02 0.01
Depeadeat variable - F reqaeacy/moath
First predictor R square Second Predictor Cumulative R square Delta R square
Positive Exp. 0.1 Studying 0.11 0.01
Negative Exp. 0.00 Studying 0.06 0.06
Positive Exp. 0.1 Sports 0.11 0.01
Negative Exp. 0.00 Sports 0.01 0.01
Positive Exp. 0.1 Hobbies 0.11 0.01
Negative Exp. 0.00 Hobbies 0.00 0
Positive Exp. 0.1 Family 0.11 0.01
Negative Exp. 0.00 Family 0.00 0
Positive Exp. 0.1 Work 0.1 0
Negative Exp. 0.00 Work 0.01 0.01
Depeadeat variable • M ax
First predictor R square Second Predictor Cumulative R
square
Delta R square
Positive Exp. 0.06 Studying 0.12 0.06*
Negative Exp. 0.008 Studying 0.07 0.06*
Positive Exp. 0.06 Sports 0.07 0.01
Negative Exp. 0.008 Sports 0.01 0.009
Positive Exp. 0.06 Hobbies 0.07 0.01
Negative Exp. 0.008 Hobbies 0.01 0.006
Positive Exp. 0.06 Family 0.07 0.01
Negative Exp. 0.008 Family 0.01 0.006
Positive Exp. 0.06 Work 0.05 0
Negative Exp. 0.008 Work 0.01 0.006
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 4. Depeadeat variable - Rapi
First predictor R square Second Predictor Cumulative R
square
Delta R square
Positive Exp. 0.20 Studying 0.26 0.06*
Negative Exp. 0.06 Studying 0.13 0.07*
Positive Exp. 0.20 Sports 0.20 0
Negative Exp. 0.06 Sports 0.07 0.01
Positive Exp. 0.20 Hobbies 0.20 0
Negative Exp. 0.06 Hobbies 0.07 0.01
Positive Exp. 0.20 Family 0.20 0
Negative Exp. 0.06 Family 0.06 0
Positive Exp. 0.20 Work 0.21 0.01*
Negative Exp. 0.06 Work 0.11 0.05*
To test whether alternative expectancies moderate the association between drinking
habits and drinking problems as well as between alcohol expectancies and drinking
problem, a three steps analysis was performed, adhering to the format suggested by
Baron and Kenny (1986). The first two steps are summarized in table 5 and the third step
appears in table 6. The interaction variable was calculated by multiplying the
standardized scores of predictor and moderator. Only the interaction between “drinking
occasions per month” and working expectancies reached significance level.
Table S. S a u u ry of the two predictor correlatioo aoalysis: First predictor - Drioltiog habit measore (Rows);
second predictor - alteroative expectancies (Cotaaus); depeadeat variable - R A P L “* " = sigailicaat at the 0.05
level.
Studying Sports Hobbies Family Work
Average .376** .28 .29 .28 .325
Frequency/w .56** .5 .5 .588 .55**
Frequency/m .57* .53 .53 .612 .55*
Max .49 .48 .48 .49 .49
Table 6. Summary of the moderation analysis. Independent variable: Rapi
Predictor Moderator Two predictor R R - interaction Significance
Average Studying 0.37 0.38 0.3
Frequency/w Studying 0.56 0.56 0.34
Frequency/m Studying 0.57 0.58 0.4
Max Studying 0.275 0.50 0.26
Positive. Exp Studying 0.51 0.51 0.8
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 6. C oatiaac
Predictor Moderator Two predictor R R - interaction Significance
Average Working 0.32 0.38 0.06
Frequency/w Working 0.55 0.57 0.09
Frequency/m Working 0.55 0.6 0.03*
Max Working 0.45 0.5 0.4
Positive. Exp Working 0.46 0.48 0.213
The preliminary correlation analysis yielded significant correlation estimates for both the
positive (r = 0.448) and negative (r = 0.27) factors o f alcohol expectancies. These factors
were correlated with the problem drinking measure (RAPI). The analysis summary
appears in the table below (Table 7):
Tabic 7: Samaury for the correbtioa analyse: alcohol eipectaacies factors predict drinking problem.
Predictor R R Square T Sig.
Positive 0.448 0.201 4.48 0.00
Negative 0.27 0.072 2.48 0.015
Although negative alcohol expectancies appear to be significantly correlated with problem
drinking, when this factor was used as a second predictor, it did not seem to account for
unique variance (Table 8):
Table 8: Snmmary of the two predictor aaalysb: alcohol eipectaacies factors predict driakiag problems
Predictor R R square t Sig.
1. Positive
2.Negative
0.449 0.201
Positive 3.58 0.01
Negative -0.29 0.77
Likewise, there seems to be little difference between the predictions of one versus two
predictors of a runmean smoother5. The smoother of the one predictor (positive expectancies)
predicts values close to S O when the positive expectancies score approaches and exceeds 200
(see graph 1); whereas the smoother of the two predictors (three dimensional graph) indicates
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 1. Positive alcohol expectancies (X axis) predict problem drinking (Y axis).
| g w — i — w » h m . I k * * * B _____________ I K — _______________| | * * # u i s
a similar pattern (see graph 2 in the next page). The two predictors graph tends to peak when
the values of the negative expectancies are around 40 (and the values of the positive
expectancies exceed 200), but the peak does not exceed the value of SO . Thus, the negative
expectancies scale does not enhance our ability to target high scorers on the RAP16. For this
reason, subsequent analyses in this study, in which alcohol expectancies were used as a first
predictor and alternative expectancies as a second predictor, were based only on the positive
factor. To neutralize the effect of bad leverage point (outliers), we obtained a correlation
9 The run mean smoother is designed to estimate the trimmed mean o f y corresponding to a given interval of x
values. fr=l, tr=0.2
4 Although the two predictor graph does suggest that very high scorers (around 80) on the negative scale will
tend lower RAPI scores for high scorers on the positive scale.
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 2. Positive (X axis) and negative (Y axis) alcohol expectancies predict
problem drinking (Y axis).
«| a 'n' w | * > « • * < » ________ | K ■ * — _________ |f*Tmjs
estimate for the bulk of the observations, using a more robust measure of association7 . The
Percentage Bend Correlation suggests a higher and more significant correlation estimate for
the bulk of the observation (See table 9 below):
Table 9: Summary of the Percentage Bead Correlation analysis.
Model Correlation Test Statistic Sig.
Percentage Bend C 0.59 6.5S 0.00000000446
To further characterize the ways in which alcohol expectancies may distinguish problem
from non problem drinkers, the data were classified into two groups: 1) Top 20% of RAPI
7 Although using T to test Ho: p = 0 provides good control over type I error some exceptions occur. (Wilcox,
1993, 1994).
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
scores8 (problem drinkers). 2) Bottom 80% of RAPI scores (non problem drinkers). The
RAPI cutoff score for the top 20% was x > SO . Because the groups differ substantially in
variance and sample size, the t test was not used. When the sample sizes are unequal and the
assumption of common variance is violated, t test may be highly unsatisfactory even when
sampling from normal distributions (Wilcox, 1994). In particular, under such circumstance,
control over Type I error is poor and the estimate of the confidence interval becomes highly
inaccurate. Hence, instead of comparing means, the trimmed means of the positive alcohol
expectancies of these groups were compared. A Yuen test suggests that the problem drinkers
group has a higher trimmed mean of alcohol expectancies (See table 10).
Table 10: Summary of the Yuen teat for the difference in alcohol expectancies between problem and
non - problem drinkers.
Model C.I. Dif. Sig. s.e. Test.stat Crit d.f.
Yuen
20.6-63.13
41.87 0.00047 10.27 4.07 2.06 22.89
The scatterplot (graph 1) shows that it is probably the problem drinkers who carry the largest
residuals, suggesting that these observations (problem drinkers) deflate the correlation
estimate between positive alcohol expectancies and RAPI. To observe this effect, we
calculated the correlation, trimming single observations (without replacement) at each step,
starting in descending order with the highest scorer on the RAPI. Almost with each trimming
the correlation increased (see table 1 1 below):
Tabic 11. Correlation estimates alter trimming the problem drinkers in descending order.
Observation
Trimmed
0
1 2 3 4 5 6 7 8 9 10 11
Cofrdation
Estimate
0.44 0.44 0.45 0.47 0.48 0.48 0.54 0.54 0.52 0.56 0.56 0.6
8 If the theory is supported for the 20% cutoff score (SO occasions on which the subject experience
negative consequences as a result from drinking alcohol), it will provide reasons to believe that the same
principles may apply for the mare extreme end of the distribution as well. Evidence based on the 20%
cutoff will justify the collection of a much larger sample size required for evaluating the theory at the top
S% cutoff.
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 3. Group differences in alcohol expectancies between problem (left figure) and
non - problem (right figure) drinkers.
The scatterplot further suggests that most of the variance in problem drinking is accounted by
the lower end of the alcohol expectancies distribution. After trimming the lower third of the
distribution, the correlation estimate dropped to 0.2. The proportion of variance accounted for
thus decreased from 20% to 4%. The Boxplot (see graph 3 above) and scatterplot (see graph
1) both suggest that the problem drinking group has a smaller variance but a higher trimmed
mean (inferred from the Yuen test) and median. As the boxplot indicates, the median of the
problem drinking group approaches the third quadrant of the non problem group. The
empirical distribution (of alcohol expectancies scores) of the problem drinking group thus
appears to be condensed into the upper portion of the distribution of the non problem
drinkers. Nevertheless, the groups seem to overlap: the expectancies values of the higher
quantiles of the problem drinking group do not exceed those of the non problem drinking
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 4. Differences in alcohol expectancies between problem and
non - problem drinkers for corresponding quantiles.
» « « — ■ ■ ■ I * X * K B _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ I * —■ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ j|* IA U I
group. That is, the highest expectancies scores of the problem drinkers do not exceed the
highest expectancies scores of the non problem drinkers. This is apparent in the shift
function graph, which shows the difference between expectancies values for corresponding
quantiles in the two distributions. As the graph shows (see graph 4 above), when
expectancies values exceed 200 (the higher quantiles of the distribution), the difference
between the groups rapidly approaches 0.
The second phase of analysis focused on estimating the contribution of alternative
expectancies as second predictors to targeting problem drinking. The preliminary two
predictors analysis suggests that only studying expectancies and working expectancies may
account for unique variance. None of the other scales approached significance level.
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The correlation estimates increased when studying and working expectancies served as
second predictors; and the P values for the t tests were relatively small (P<0.28 and P<0.32,
see table 12 in the next page):
Table 12. Swnmary of the two predictors aaalysu: accooatiag for u iq o e variance.
Predictor R R square t Sig.
1. Alcohol Exp.
2.Study Exp.
0.521 0.272
Alcohol Exp. 4.55 0.00
Study Exp. -2.32 0.028
Predictor R R square t Sig.
1. Alcohol Exp.
2.Work Exp.
0.46 0.212
Alcohol Exp. 3.9 0.00
Work Exp. -2.19 0.032
To further explore the role of Studying Expectancies and Working Expectancies as
second predictors, a two predictors runmean smoother was fitted to the data for each of these
expectancies scales. The graph of Studying Expectancies (see graph 5 in the next page)
approaches 70 on the RAPI scale for the lower values of studying expectancies (y < 130) and
the higher values of alcohol expectancies (x > 190). Likewise, the graph of Working
Expectancies (see graph 6 in the next page) approaches 60 on the RAPI scale when Working
Expectancies approach 0 and Alcohol Expectancies are above 190. These predictions peak
higher than the predictions based on alcohol expectancies alone: the runmean smoother for
both one (positive) and two (positive and negative) alcohol expectancies predictors did not
exceed 50 on the RAPI (see graph 2).
To perform the binomial analysis, we formed groups that were homogeneous in
alcohol expectancies. The observations were rearranged in ascending order, and divided
into two groups: above the median and below the median.
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 5. Alcohol expectancies (X axis) and studying expectancies (Y axis)
predicting drinking problem (Z axis).
Graph 6. Alcohol expectancies (X axis) and working expectancies (Y axis)
predicting drinking problems (Z axis).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Members of the same group were considered as scoring on the same level of alcohol
expectancies. For each of the alternative expectancies, each observation was classified
into one of three levels: high scorers, middle scorers and low scorers. The cutoff criteria
for high and low scorers was set to 30% (high = top 30% and low = bottom 30%). This
criterion varied slightly across the different alternative expectancies because o f ties
among the observation. Within each category that contained homogeneous alcohol
expectancies scorers, the RAPI score (the scale of problem drinking) of every high scorer
(on a given alternative expectancy scale) was compared with the score of every low
scorer (on the same scale). A successful prediction was counted when a low scorer on a
given alternative expectancy scale scored higher on the Rapi than the high scorer on the
same alternative scale.
The results of the binomial procedure were consistent with the two predictors
regression analyses: only Studying Expectancies and Working Expectancies showed a
significant effect. Table 13 specifies the probability estimates for a randomly sampled subject
low on a given alternative expectancies scale to score higher on the RAPI than a randomly
sampled subject high on the same scale, given that the two subjects score on the same level
on the alcohol expectancies scale. The confidence intervals were calculated using Mee’s
(1990) method.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
Table 13. Sammary o f the binom ial analysts: the raoderatioa effect o f alteraative expectancies.
Scale
Prob.
above
median
C.I.
Prob.
Below
median
C.I.
Studying
0.93
0.77 - 0.98*
0.64
0.48-0.80
Working
0.74
0.54 - 0.88*
0.60
0.47-0.74
Sport
0.61
0.38-0.80
0.31
0.15-0.54
Family
0.53
0.31 -0.75
0.63
0.48-0.79
Hobbies
0.39
0.20-0.63
0.33
0.24-0.58
Table 14. The cutoff alternative expectancies scale scores for “high” and “fcw" groups
Scale Cutoff score for “high” n Cutoff score for “low” N
Studying X >= 150 23 X <= 130 29
Working X >= 53 27 X <=41 22
Sports X>= 60 25 X <49 26
Family X>= 132 29 X <= 116 25
Hobbies X >= 182 25 X <= 150 25
Following these results, a few items of interest in the Studying and Working scales were
analyzed in the same manner. The numbering of the items below correspond to the item
numbers in the table.
1. How many hours per week do you usually study (not including class attendance)?
2. How would you rate your dedication to school work relative to other students?
3. Studying is really important to me
4. How significant is your job to your graduating from college?
Table IS. Sammary of the binomial aaalysis: the moderation effect of core items.
Item
Prob .
above
median
C.I.
Prob.
below
median
C.I.
1 0.86
0.67 - 0.95*
0.58
0.43 -0.75
2 0.85
0.62 - 0.95*
0.56
0.34-0.80
3 0.66
0.38-0.81
0.65
0.5-0.81
4 0.80
0.58 - 0.99*
0.43
0.58-0.28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 16. The cutoff item scores for “high” a id “low” groups
Item Cutoff score for “high” n Cutoff score for “low” N
1 X>=10 28 X<=6 30
2 X > 8 22 X <=6 25
3 X > 8 25 X<=6 30
4 X >= 7 31 X<=2 28
5 X>=20 32 X<=10 28
To complement the binomial method, a group comparison analysis was performed.
While the binomial method provides information about the probability that one subject will
drink more than another, group comparison measures by how much. Whereas the ultimate
goal is to devise a sensitive metric that will predict the magnitude of difference in drinking
problems on the individual level, the following preliminary analysis is based on a group
design. To use the cutoff scores, inductively derived from the data in this study, to perform
the analysis on the individual level, a new, independent sample is required. Following the
results of the binomial analysis, the current analysis attempts to characterize group
differences in problem drinking between “high” and “low” scorers on the following scale and
items: Studying Expectancies, item 1 and item 4. These scales and items were chosen
because they showed the largest effect in the binomial analysis. The reason for choosing only
those items and scales showing large effects was to assure sufficient power for rejecting
multiple comparisons of medians9 between relatively small groups (12 -18 subjects per
group). We chose to compare medians because medians provide information that is relevant
to the purpose of the study: making predictions on an individual level. Thus, if the median of
a certain group is estimated above the problem drinkers cutoff predicting problem drinking
for subjects who fall into this category will be more often successful than unsuccessful. All
members of the groups compared here scored above the median on the alcohol expectancies
scale. The results were significant for all three comparisons, suggesting
9 Although medians are robust, the efficiency of their measure of scale is only fair.
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 7. Differences in RAPI scores between “high” and “low” groups of studying
expectancies.
| I m —»___________||x i-mx « ■
that the “low” groups had a higher median then the “high” groups. Rom’s procedure (1990)
suggests that experiment-wise type I error did not exceed 0.05 level. The variances of the
sampling distributions of Ml and M2 were estimated by the Maritz-Jarret procedure (1978).
Table 17. Summary o f the MJ test for difference ia medians of RAPI scores between the “low” and
“high” groups of studying expectancies.
Model
VuiiMe
Dif. C.I Sig.
s.e.
Tca.Sut
CriL
MJ
Studying
Expectancies
45
71.6-18.35
0.0003
13.6 3.3 1.64
The boxplot (see graph 7) indicates that the score of the first quadrant of the “low” group is
above the third quadrant of the high group, suggesting that the “low” group tends to score
significantly higher on the RAPI on most quantiles. In consistency with the boxplot, the test
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 8. Differences in RAPI scores between “high” and “low” groups of item i : the
number of studying hours per week.
suggests that there is a large difference between the medians of the groups. The median of the
“high” group is estimated at well below 12, whereas the median estimate of the low “group”
approaches 60.60 is the 0.84 quantile of the empirical distribution of problem drinking;
hence, almost half of the members of the “low” group fall into the top 14% of the problem
drinking scores in this data.
Table 18. Summary of the MJ teat for difference in medians of RAPI scores between the “low” and
“high” groups of item 1.
Model
Variable
Dif. Sig.
s.e.
Ten. Sut
C rit. n High n Low
MJ
Item 1
23 0.04 13.49 1.7 1.64 13 15
The one tailed test for item 1 (hours of studying per week) suggests that the median of the
“low” group is significantly higher than the median of the “high” group. The observed
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
difference in medians of the groups (23) is substantially smaller than the difference observed
for the entire studying expectancies scale (45). There is also a drop in the significance level
(from 0.0003 to 0.04). These results suggest that most members of the “low” group (low on
studying hours) generally score higher on the RAPI than the members of the “high” group;
however the entire studying expectancies scale seems to be superior in differentiating
problem from non problem drinkers. Thus, while the binomial analysis suggested that, in
individual comparisons, the behavioral assessment (item 1) and the entire studying
expectancy scale can equally predict who will score higher or lower on the RAPI, the group
analysis suggests that the entire scale can predict larger group differences, and target more
subjects who score above 50 (top 20% of the data) on the RAPI.
Table 19. Sammary o f the MJ test for dlflereace in mediae* of RAPI scores between the “tow” and
“Ugh" groups of item 4.
Model
Variable
Dif. Sig.
s.e.
TesLStit
Crit n High n Low
MJ
Item 4
41 0.04 24.4 1.68 1.64 19 8
The one tailed test for medians (item 4) was significant. The boxplot (see graph 9) suggests
that the groups overlap considerably but seem to differ in skewness. If the skewness is heavy,
as suggested by the boxplot, the groups should differ more in medians than in trimmed
means. To confirm this hypothesis, we also tested for trimmed means, using the Yuen model,
and the significance level indeed dropped significantly (from 0.04 to 0.19). The difference in
medians, however, is more important for the purposes of targeting problem drinkers on an
individual level. The median of the “low” group is estimated at 65 (the 0.88 quantile of the
empirical distribution of problem drinking); whereas the median of the “high” group is
estimated at 29 (the 0.64 quantile). This finding suggest that many more observations in the
higher end of the RAPI scale are members of the “low” rather than the “high” group.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 9. Differences in RAPI scores between “high” and “low” groups of item 4: the
financial significance of students’ job to their graduation from college.
_ _ _ _ _ _ _ I * — » _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ||« c-p u is
The final analysis compared the drinking habits of the “low” and “high” groups of
Studying Expectancies. All members of both groups scored above the median on alcohol
expectancies. A Yuen test found no difference in the trimmed means of the drinking habits
measures of these groups. However, the boxplot (see graph 10) suggests that the groups differ
in the medians for the maximum number of drinks consumed in one occasion. The MJ test
for medians found a highly significant difference (P < 0.008). The median of the “low” group
is estimated at 12; whereas the median of the “high” group is estimated at 6 and its third
quadrant is estimated at less than 10. This suggests that many more members in the “low”
group have had more then 10 drinks in one occasion. The Yuen tests was not significant
probably due to the heavy skewedness of the “low” distribution that pulled the trimmed mean
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 10. Differences in the maximum number of drinks in one occasion between
“high” and “low” groups of studying expectancies.
«■<■»— I l « — ■ ll« » H U «
downwards. The boxplots of the other measures suggest that there is very little difference in
medians between the groups (see graphs 11,12,13).
Table 20. The sammary of the Yuen and MJ tests for the differences in maximum number o f drinks
consumed on one occasion between the “low” and “high” groups of studying expectancies.
Model
Variable
Dif. C.I Sig.
s.e.
Test. Sut
Crit. df
Yuen
Studying
Expectancies
-3.4
©
1
0.054
1.8 2.12 2.17 12.04
Model
Variable
Dif. C.I Sig.
s.e.
Teat. Slat
Crit
MJ
Studying
Expectancies
6
1.1 - 10.9 0.008
2.5 2.4 1.64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 11. Differences in the number of drinks per week between “high” and “low”
groups of studying expectancies.
i..| » x w b I n — lln a - w m
Table 21. The summary of the Yuen test for the differences in number of drinks
Consumed per week between the “km ” and “high” groups of studying expectancies.
Model
Variable
Dif. C.I Sig.
s.e.
Test. Slat
Crit. df
Yuen
Studying
Expectancies
0.05
-1 .2 -1 .3 0.9
0.59 0.09 2.13 14.63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
Graph 12. Differences in the number of drinks per month between “high” and “low”
groups of studying expectancies.
| I X — ■ lla »*ui»
Table 22. The sammary of the Yuen teat for the differences in number of drinks
Consumed per month between the “low" and ‘ 'Ugh” groups of studying expectancies.
Model
Variable
Dif. C.I Sig.
s.e.
TestStat
Crit df
Yuen
Studying
Expectancies
1.4
-3.15-6.04 0.5
2.15 0.66 2.12 15.26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Graph 13. Differences in the average number of drinks per occasion between “high”
and “low” groups of studying expectancies.
Table 23. The summary of the Yuen test for the differences in average number of drinks
Consumed per occasion between the “low” and “high” groups of studying expectancies.
Model
Variable
Dif. C.I Sig.
s.e.
TeS.Stat
Crit df
Yuen
Studying
Expectancies
0.72
-1.02-2.4 0.39
0.81 0.88 2.14 14.11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
Discussion
The first hypothesis of the study was fully supported. The preliminary analysis
estimates the correlation between positive alcohol expectancies and problem drinking at
0.448, and this result is highly significant. This finding replicates past results and further
supports the notion that alcohol expectancies covary with problem drinking and that the
association between these variables tends to be positive. Analysis further indicated that the
trimmed mean of alcohol expectancies scores of those who scored at the top 20% on the
RAPI (in this data) was significantly higher than the trimmed mean of the rest of the
observations. Together these analyses support the general notion that, on average, problem
drinking tends to increase as positive alcohol expectancies become high.
To further explore the role alcohol expectancies play in problem drinking, subsequent
analyses focused on the individual rather than the group level. Analysis on an individual level
is of theoretical importance because it assists in specifying the conditions under which, as
well as the mechanisms through which, problem drinking may develop; and it is of practical
importance because information pertaining to the individual level directly may contribute to
preventive and therapeutic endeavors. As discussed in length in the introduction section, in
clinical psychology clinicians and assessment experts deal with individual clients, so they
need assessment instruments that will provide them with practical and accurate information
about the individuals they work with. Much can be gained if clinicians could estimate the
prognosis probability for certain clients as well as the probability they will exhibit specific
behaviors of interest (as in forensic cases). This is not to say that psychologists will reach
perfect accuracy in their predictions. Rather, they should be able to obtain an accurate
estimate for the probability of making an error. From a theoretical perspective, studying the
phenomenon on the individual level enables us to examine anomalies: cases that meet the
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
conditions specified by the theory for developing drinking problems yet do not manifest the
problem, and vice versa. Exploring the differences between anomalies and cases which the
theory accounts for may help in further refining the conditions under which problem drinking
develops.
One major purpose of this study was to obtain a probability estimate of successfully
predicting problem drinking on the basis of alcohol expectancies. In other words, the goal is
to search for a score value or a range of score values on the alcohol expectancies scale that
predicts problem drinking; and to obtain a probability estimate and a confidence interval for a
successful prediction. More generally, this line of analysis attempts to evaluate what kind of
information we can derive about the drinking behavior of an individual client or subject from
his or her score on an alcohol expectancies scale.
The unique pattern in which the observations are distributed across the X and Y axes
appears to be informative. As the scatterplot reveals, the variance of the error term of the
predicted value (of problem drinking) increases considerably as the values of alcohol
expectancies get large. This specific pattern of hetroscedastic error term suggests that the
confidence interval for problem drinking scores increases significantly for the higher score
values of alcohol expectancies. For alcohol expectancies score values under 120, confidence
intervals for RAPI scores may not exceed 25; whereas it may exceed 100 for alcohol
expectancies score values above 124. In this data set, 100 is a range of scores that contains
almost the entire distribution of problem drinking. This pattern suggests that in the higher
score values of alcohol expectancies predictions of problem drinking become less accurate
and less or not at all informative. The implication is that, based on alcohol expectancies, we
can make relatively accurate predictions for some of the people that drink very little or not at
all (those who score below 120); however, we are unable to target problem drinkers with a
reasonable probability of success or accuracy. In fact, every alcohol expectancies (or range of
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
scores) value scored by problem drinkers is also shared by many more non - problem
drinkers. Hence, if we predict drinking problems for those expectancies scores, our
predictions will be wrong much more often than right. In this data set, only about 27% of
those scoring above 124 on the alcohol expectancies scale also scored at the top 20% on the
problem drinking scale. This basic problem exists regardless of the criterion used to define
the problem - drinking zone; and even gets worse as the cutoff score chosen for problem
drinking moves higher on the RAPI scale.
The scatterplot offers a more specific characterization to the positive association the
correlation analysis indicates. The correlation analysis estimates the correlation between
alcohol expectancies and drinking problems at 0.448. A common interpretation of this
finding would hold that alcohol expectancies account for a little more than 20% of the
variance in problem drinking (R square). The rationale for this line of interpretation is that
the proportion of variance that we can account for provides us with a measure of how well
we can predict Y from knowing its X value; so the general goal is to maximize R square.
This line of reasoning, however, may be misleading because there are many different data
patterns that can mathematically produce the same number (r = 0.448 in this case), but each
of them may carry different theoretical implications. Mathematically, R square increases as
the sum of squared residuals decreases. This means that the observations that are closest to
the regression line are the ones that inflate the value of R square, while the observations most
distanced from the line deflate it. In order to make a theoretical inference from the data
analysis, it is important to characterize which observations are the closest to the line and
which are the most distanced. The particular pattern that emerges may carry theoretical
implications.
In this particular data set, the observations most distanced from the regression line are
the observations of interest: the problem drinkers. Hence, it is the non - problem drinkers
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
who inflate the correlation estimate, while it is the problem drinkers who deflate it. The value
r = 0.448 thus represents an association between alcohol expectancies and problem drinking
mainly in the non - problem drinking range. R square is relatively large not because but
despite the problem drinkers. Hence, knowing that the correlation between alcohol
expectancies and drinking problems is as high as 0.448 does not allow us to maintain that
alcohol expectancies play a role in the development of drinking problems. Although the Yuen
test provided strong evidence that the problem drinking group tends to hold significantly
higher alcohol expectancies, this finding is not a necessary sequel of the high correlation. The
data can be easily rearranged so the differences between problem and non-problem drinkers
in alcohol expectancies are eliminated even though the correlation estimate will remain
unchanged. Hence, the theoretical implications may not be apparent from knowing the
correlation alone.
Furthermore, in order to calculate R square, we aggregate residuals according to a
mathematical rather than a theoretical principle. In other words, by summing across all
squared residuals, we attempt to characterize the association between alcohol expectancies
and problem drinking with just one number. In this case, summing up all the information of
the group into one number prevents us from observing theoretical implications. As
demonstrated earlier, the correlation estimate in this data set is high not because but despite
the problem drinkers. To further illustrate this point, when we used a robust measure of
association, the correlation estimate increased to 0.59; however, in this case, we have
trimmed the phenomenon of interest. Likewise, when we trimmed observations in a
theoretically meaningful way (trimming the problem drinkers in descending order, one at a
time), the correlation estimate was inflated to 0.6.
Rather than focusing on the magnitude of R square, it may be more revealing to
interpret possible theoretical implications this number may carry for the particular pattern of
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the data manifested in the scatterplot. This particular data pattern is telling us to look for a
theoretical explanation for why alcohol expectancies account for a lot of variance at the low
levels but for almost none of the variance at high levels. The cognitive mediational form of
the matching theory provides precisely such an account. According to this model, possessing
low reinforcement expectancies predicts that behavior will be exhibited either rarely or not at
all; so behavior should not vary considerably across individuals on the low end of the
expectancies distribution. On the other hand, when expectancies are high, the degree to which
the behavior will be manifested heavily depends on the expectancies for alternative behaviors
available under the circumstances. Since the alternative expectancies are deemed to be
subjective, people with high expectancies for the behavior of interest (alcohol drinking) will
tend to vary considerably in the degree to which they engage in it.
Additionally, characterizing the shape of the problem drinking distribution may also
contribute to making theoretical inferences from the data. The empirical distribution of
problem drinking in this study peaks at 0; it is heavily skewed to the right and has a heavy
right tail1 0 . It is the tail that should capture the focus of attention in this line of research
because it contains the problem drinkers. There are relatively few people in the tail, for
problem drinking is a rare phenomenon (because it is by definition an abnormal behavior); so
the challenge is to differentiate this rather small number of people from the bulk of the
observation. Since the regression line is determined mostly by the bulk of the observations, it
should not be surprising that the problem drinkers would be the most distanced from it, while
the non - problem drinkers (the bulk of the observations) are the closest observations to the
line. On the other hand, this should not necessarily be the case. In theory, the problem
drinkers, although very distanced from the bulk of the observations on their Y values (RAPI
1 0 25% of the observations scored 0 on the RAPI; almost 20% scored under 10, while the top 20% of the
scores range from 50 to 186.
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
scores), may form a good continuation to the regression line and become good rather than
bad leverage points. If this were the case, strong arguments could be made for a simple
mediational mechanism of alcohol expectancies. The residuals of the problem drinkers would
drop dramatically and, also due to their large distance from the bulk of the observations on
both axes, inflate the correlation estimate. This time, however, the inflation in the correlation
estimate would be accompanied by informative confidence intervals that can target problem
drinkers. Under this scenario, the residuals could be interpreted as errors in measurement.
But this scenario is not the case in this data set; in fact, in order to bring the problem
drinkers to form a good continuation to the regression line, their alcohol expectancies scores
need to reach the maximum possible score and, in many cases, even exceed it. This, of
course, is not possible. Hence, there seem to be compelling reasons to be skeptical as to the
simple mechanism possibility.
As a group, on the other hand, problem drinkers score higher than non - problem
drinkers on the alcohol expectancies scale. The boxplot (graph 3) and scatterplot (graph 1)
suggest that the distribution of alcohol expectancies scores of problem drinking is congested
into the top 65% of the general distribution, but it does not exceed the alcohol expectancies
range of non - problem drinkers. Similarly, as the shift function graph (graph 4) indicates, the
difference between the corresponding quantiles of the two groups rapidly approaches 0 for
alcohol expectancies values above 200 (the top quantiles). Thus, the alcohol expectancies of
problem drinkers generally tend to be higher than a certain portion of the non - problem
drinkers but not higher than all of them. Hence, the data strongly suggest that the cognitive
mediational mechanism underlying problem drinking is complex. To reformulate the initial
mediational hypothesis more precisely, as the data suggest, alcohol expectancies may play a
mediational role in a more complex mediational mechanism. The complex mechanism should
explain why some individuals who possess high alcohol expectancies develop drinking
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
problems, while others do not; and, again, this is exactly what the cognitive form of the
matching law framework attempts to account for.
All the two predictors analyses found an effect for studying and working expectancies
but not for “hobbies”, “sports” and “family”. These results suggest that, in the college setting,
possessing high working and studying expectancies may buffer drinking problems for
students who have high alcohol expectancies. The binomial analysis provides support to the
theoretical model offered in this study more directly, for it estimates the probability that an
individual event will occur under the conditions the theory specifies. The model predicted
that when comparing two students who hold high yet similar levels of alcohol
expectancies, the student who is low on the alternative expectancies would develop more
drinking problems than the one holding high alternative expectancies. The probabilities
of making successful predictions for such individual comparisons were highly significant
and carried a large effect for both the studying and working scales; and, as the model
predicted, this effect disappeared for the low levels of alcohol expectancies.
The results from the regression-based analyses are more difficult to interpret. As
demonstrated in the discussion on the one-predictor analysis, under heteroscedastic
conditions and due to the particular aberrant shape of the problem drinking distribution, R
square does not provide sufficient information for interpreting the data. In the least, a
three dimensional scatterplot is needed in order to understand why R square increased or
remained unchanged when an additional predictor is introduced; and what possible
theoretical implications may consequently be considered. As illustrated earlier, under
heteroscedastic conditions, the mathematical operation of aggregating residuals may be
unsatisfactory for making theoretical inferences from the data. The three dimensional
smoother o f the runmean conveys a general pattern that supports the hypotheses: the
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
trimmed mean of the predicted value of problem drinking peaks above the top 20% o f the
problem drinking distribution at the expected score zones of alcohol expectancies and
alternative expectancies (studying and working only). The variance of the error term,
however, remains a problem, suggesting that further factors need to be specified in order
to target problem drinkers more reliably. A third predictor would add a fourth dimension
to the scatterpolt and thus render it impossible to present visually or interpret. As earlier
demonstrated, an increase in R square should not be interpreted without a scatterplot;
hence, at this point, since a four dimensional scatterpolt is unavailable, using a three
predictor analyses is not recommended for exploring the role additional variables may
play in the development of problem drinking.
Following the same line of reasoning, it is very difficult to interpret the
moderation analysis. In this analysis, there are three predictors, or, in turn, four
dimensions. As argued above, a four-dimensional scatterplot, necessary for interpreting
the increase in R square, is yet to be devised. In addition, the fourth dimension in this
case is defined by the product of standardized scores, complication the picture even
further.
Only one of the moderation analyses turned out significant; this result is difficult
to interpret. One of the problems is that the moderators were correlated with the
predictors as well as with the dependent variable. As Baron and Kenny (1990) suggest,
“it is desirable that the moderator variable be uncorrelated with both predictor and
criterion to provide a clear interpretable interaction term. In addition, as Baron and
Kenny suggest, error of measurement in either the predictor or moderator results in low
power. It is thus possible that the error of measurement inherent in the self - report
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
instruments used in this study attenuated power considerably. At any rate, even if the
moderation effect were significant, it would still be difficult to interpret its meaning for
the reasons elaborated above. Finally, it may be that alternative expectancies do not play
a moderating role at all. Rather, as suggested earlier, alternative expectancies may play a
causal role in a complex mediational mechanism. If this theoretical interpretation is
correct, it may carry an important practical implication: manipulating studying and
working expectancies may decrease problem drinking for students. This line of
interpretation is highly consistent with Hermstein’s operant model o f matching theory:
changing the schedules of reinforcement for alternative behaviors should change the
proportion of response allocation; or, in turn, change the amount o f time the organism
spends on the behavior of interest. As discussed in length in the introduction section, this
principle carries over to the cognitive mediational framework proposed in this study.
The group based analyses suggest that “low” and “high” groups of studying
expectancies and of at least one crucial element of studying expectancies'1 differ
substantially in medians. The estimated difference between these groups exceeded 40
points on the RAPI. This difference seems relatively large because only the top 22%
percent o f subjects in our data exceeded the RAPI score o f 40. Hence, the group based
analysis suggests that relatively high number o f subjects in the “low’' groups of these
scales may fall into the top portion of the problem drinking distribution. At this point,
however, we do not yet have sufficient evidence that the difference in medians is indeed
so impressive because the confidence interval (for the difference in medians) was too
large. One major goal o f the next study would thus be to shorten the confidence interval
in order to gain a more tenable estimate of the difference and its possible implications.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
On the other hand, the actual difference between the medians may be even larger
than the estimate obtained here. In order to obtain a sufficient number o f subjects to
allow these group comparisons, we had to use a more heterogeneous level o f alcohol
expectancies: instead o f comparing the same deciles, as was the case in the binomial
analysis, the group comparisons were made between subjects who scored above the
median. More homogeneous comparisons are expected to yield larger group differences
and further enhance the accuracy o f prediction for the individual case.
An interesting finding in this study was that, although “low” and “high” groups
tend to differ substantially in their problem drinking scores, they do not seem to differ in
the frequency in which they drink. This result suggests that students with high alcohol
expectancies who are also high on studying and working expectancies do not drink less
often than comparable students who are low on those scales; but, rather, these students
tend to better control the consequences and the amount of their drinking. Consistent with
this hypothesis, the groups did differ markedly in the maximum number of drinks they
reported drinking on one occasion. The median for the “low” group (on the studying
scale) was as high as 12 standard drinks, whereas the median for the “high” group was
estimated at 6. This finding suggests that many more of those low on studying
expectancies and high on alcohol expectancies may have a greater tendency to drink to
the point in which the risk for problematic behaviors as a result of drinking increases
significantly. To illustrate, it may be that a non problem drinker who reports drinking
alcohol twice a week refers to a beer he drinks in front of the T.V in the middle o f the
week and a glass of wine he drinks at dinner with his parents on a Friday night. By
contrast, a problem drinker reporting drinking twice a week may refer to spending 5
1 1 The degree to which the work is financially significant to the student’s graduating from college.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
hours at a bar drinking to incapacitating intoxication. This description, however, is
inconsistent with the analysis that found no difference between the groups in the average
number o f drinks consumed per occasion. To further explore the issue, future research
may need to refine the measurement of drinking behavior.
While the “studying” and “working” scales support the proposed model, the
“hobbies”, “sports” and “family” scales showed no effect. It may be the case that
behaviors related to family interaction do not constitute competing activities to drinking
while students are on campus. On the other hand, the activities the “hobbies” and “sports”
scales tap into are certainly incompatible with drinking in the sense that engaging in these
activities is impossible or highly improbable while or after drinking. It is possible that
errors of measurement due to flaws in these scales masked the effect; however, this does
not seem to be the case. Even when the difference between the “low” and “high” scales
was made more distinct (up to top/bottom 10%), P value did not markedly decline toward
significance level. In addition, since these scales appear to possess at least face validity, it
seems plausible to suggest that, although behavior associated with these scales compete
with drinking, high expectancies on these scales do not buffer drinking problems.
Following this line of reasoning, the alternative activities that show an effect
seem to be qualitatively different from those that do not. Studying and working may be
described as behaviors associated with duty, responsibility, demands and long term
commitments; whereas, activities related to hobbies and sports may be described as
behaviors associated with leisure, recreation and “time off’. These two types of activities
may be considered as oppositions, by the same rationale work and leisure would be
considered antonyms. From this perspective, the results of this study suggest students
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
may develop drinking problems not because they possess a narrower repertoire of leisure
time activities, but because they expect less reinforcement from studying and working.
For these students, drinking may come at the expense o f the number of hours they study
or attend class, but they will not play less basketball or see fewer movies. It is interesting
that the working expectancies scale generally showed a moderate effect, but item 4 -
“working is financially significant to my graduating from college” - showed an
exceptionally strong effect (P = 0.94). This result suggest that working expectancies may
buffer problem drinking especially when the students expect his/her job to carry long
term consequences.
Following this line of interpretation, in light of the specific pattern of results
emerging in this study, extending Hermstein’s notion of melioration (Hermstein, 1997)
may offer an interesting hypothesis. By melioration, Hermstein refers to an uneconomical
behavior exhibited in choice situations in which an individual pursues those behaviors
that offer the highest level of reinforcement for the present moment or very near future,
but fail to engage in behaviors that carry long term payoffs. Hermstein showed that, like
pigeons, many people tend to behave in such uneconomical manner. From a cognitive
perspective, melioration may involve cognitive structures that underlie planning and
decision making; and it is thus possible that individuals tend to differ in these cognitive
factors. Perhaps problem and non - problem drinkers differ in cognitive aspects which
play a role in long term planning and decision making. This may be an interesting
hypothesis to follow in future research.
Following more directly from the current study, the next study should attempt to
obtain the probability estimate of success in making an individual prediction of problem
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
drinking based both on alcohol expectancies and alternative expectancies. In other words,
the question that the next study will attempt to answer would be as follows: when we
predict that an individual subject is a problem drinker, based on his/her alcohol
expectancies and alternative expectancies scores, what is the probability that we will be
correct? The expectancies cutoff scores and ranges of score values that will set the rules
necessary for making such predictions will be derived from the data in the current study,
while the subjects in the new study will be used to assess how well these rules perform.
One of the main interests in this study would be to estimate the contribution of alternative
expectancies to the prediction accuracy for the individual client, over and above the level
o f accuracy alcohol expectancies possess alone.
In a broader perspective, the theory applied in this study and its encouraging
results propose a future line of research that will approach drinking problems and
drinking behavior in general in a context o f setting. As the matching law research
demonstrates, when behavior is free to vary, the amount of time and energy one spends
on engaging in a particular activity heavily depends on the alternative reinforcement
available in one’s environment. Thus, a change of setting and circumstances in one’s life
may lead to a change in one’s drinking behavior. The current study was not designed to
measure how a change of setting influences drinking behavior; however, it did provide
interesting evidence that may accentuate the role circumstances may play in drinking
problem. The analysis o f working expectancies, for example, demonstrated that students
who believed their job is highly significant to their graduating from college tended to
score much lower on the RAPI, even if they possessed high alcohol expectancies. This
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
finding implies that circumstances or situational factors may be important aspects to
study in order to enhance the understanding of drinking problems.
From a practical perspective, changing one’s beliefs about the possible gains one
can derive from alternative behaviors in the environment may help to decrease drinking
problems. Strengthening alternative expectancies may work by the same principle of
reinforcing incompatible behaviors. For example, enhancing studying expectancies in
college students may buffer drinking problems for those possessing high alcohol
expectancies.
The current study focused on the college setting; and, thus, the alternative
reinforcements we assessed were tailored to this specific context. These alternative
reinforcements, of course, cannot be generalized to other settings directly; however, the
principle of alternative expectancies may apply to other situations as well. The setting we
chose was one of convenience; but we used it to demonstrate a principle. At this point,
there seem to be no particular connection between the setting and principle in this study
that would limit the generalizability of the principle across settings.
Moreover, it may also be difficult to apply the same cutoff scores for alcohol
expectancies and alternative expectancies we used in this study even across colleges. It
may be the case that the distributions of problem drinking, alcohol expectancies and
alternative expectancies are in themselves context dependent. In other words, colleges
may differ in the degree to which their students abuse alcohol because they admit
different types of students and constitute different environments that provide different
reinforcements for drinking. The cutoff scores are likely to change if these distributions
change; but the principle should stay the same.
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Thus, the norms of drinking problems may vary across colleges. If this is the case,
the same RAPI score would correspond to different quantiles in the problem drinking
distribution of different colleges. Since the RAPI was not normed on a nation wide
sample, it is difficult to define the problem drinking range by a cutoff score that would
correspond to some extreme quantile (top 5%, for example) in the same way giftedness
and mental retardation are defined in intelligence measures. Perhaps, in addition to the
RAPI, future studies should incorporate a more conceptually based definition for problem
drinking like the criterion based definition the DSMIV specifies for alcohol abuse. In
this way, the sensitivity of measurement will not be lost, while a more conceptually based
definition for alcohol abuse will be available to assess how well the instruments we use
predict who foils into this category. This seems to be a requisite step because the RAPI
contains many items that differ in their conceptual but not in the numerical weight. For
instance, the item, “tried to cut down on drinking” probably indicates a higher level of
problem drinking than the item, “had a bad time”; however, these items do not carry
different weights in computing the scale score. Since there are multiple ways to elevate
the RAPI score, the same score may carry different conceptual implications for different
people. Hence, conceptually based criteria may help to guide our future research as to
which factors appear to be more important for predicting the actual phenomenon we are
interested in.
Several steps may be taken to improve the assessment accuracy of alternative
expectancies. Adding more items to the scales will probably improve their internal
consistency. Adding items may contribute especially to the working scale, which has very
few items. In addition, in order to increase the foce validity o f measurement, it may be
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
valuable to ask subjects to describe their expectations from the different behaviors and
incorporate these into the scales in case they add innovative aspects. Finally, future
research should attempt to estimate the temporal stability o f the scales in order support
attempts to base longitudinal predictions on these scales.
The results of this preliminary study await replication. This line o f research is,
evidently, yet premature to allow a discussion about the applicability of alternative
expectancies. And yet, the results of this study do call attention to the possibility that a
change of setting or a cognitive restructuring o f reinforcement expectancies for behaviors
within one’s setting may constitute additional strategies to abate or prevent problem
drinking behavior more effectively.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
References
Bandura, A. (1977). Social learning theory. New Jersey: Prentice-HalL, Inc.
Baum, M. (1974). On two types of deviation from the matching law: Bias and
undermatching. Journal of Experimental Analysis o f Behavior. 22.231 - 242.
Baum, M., Rachlin, H. C. (1969). Choice as time allocation. Journal o f Experimental
and Analysis of Behavior. 12. 861 - 874.
Baron, M. R., & Kenny, A. D. (1986) The Moderator - Mediator Variable Distinction in
Social Psychological Research: Conceptual, Strategic, and Statistical
Considerations. Journal of Personality and Social Psychology. 53. 1173 - 1182.
Brown, S. A. (1985b). Expectancies versus background in the prediction o f college
drinking practices. Journal o f Consulting and Clinical Psychology: 53. 123 -
130.
Brown, S. A., Goldman, M. S., Inn. A., & Anderson, L. (1980). Expectation of
reinforcement from alcohol: Their domain and relation to drinking patterns.
Journal of Consulting and Clinical Psychology: 4 8 .419 - 426.
Brown, S. A., Goldman, M. S., & Christiansen, B. A. (1985). Do alcohol
expectancies mediate drinking patterns in adults? Journal of Consulting and
Clinical Psychology. 53. 512 - 519.
Christiansen, B.A., Goldman. M. S., & Inn. A. (1982). The development of alcohol-
related expectancies in adolescents: Separating pharmacological from social
learning influences. Journal of Consulting and Clinical Psychology: 50. 336 -
344.
Christiansen, B. A., & Goldman, M. S. (1983). Alcohol related expectancies versus
demographic/background variables in the prediction of adolescents drinking.
Journal of Consulting and Clinical Psychology: 51. 249 - 257.
Christiancen, B.A., Smith G. T., Roehling, P.V., & Goldman, M. S. (1989). Using
alcohol expectancies to predict adolescent drinking behavior after one year.
Journal of Consulting and clinical Psychology. 57.93 -99.
Cogner, R., & Killeen, P. (1974). Use of Concurrent Option Small Group Research: A
Demonstration. Pacific Sociological Review. 17. 399 - 416.
Cohen J (1988). Statistical Power Analysis for the Behavioral Sciences (2n d ed.)
Hillsdale, NJ: Lawrence Earlbaum.
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Davison., and McCarthy, D. (1988). The matching Law: A research review. Hillsdale,
N.J.: Earlbaum.
Earleywine, M., & Martin, C. S. (1993). Anticipated stimulant and sedative effects of
ethanol vary with limb of the blood alcohol curve and dosage. Alcoholism:
Clinical and Experimental Research. 17. 135 - 139.
Faqua, W. (1984). Comments on the applied relevance of the matching law. Journal of
Applied Behavior Analysis. 7,381 - 386.
Fromme, K. (1994). An experimental test of affect, subjective craving, and alcohol
outcome expectancies as motivators of young adult drinking. Addictive
Behaviors. 19 631 -641.
Fromme, K., Stroot, E., & Kaplan, D. (1993). Comprehensive effects of Alcohol:
Development and Psychometric Assessment of a New Expectancy Questionnaire.
Psychological Assessment. 5 .19 - 26
Goldman, M. S., Brown, S.A., & Christiansen, B.A. (1987). Expectancy theory:
Thinking about drinking. In H. T. Blane and K. E. Leonard (Eds.),
Psychological theories of drinking and alcoholism: (pp. 181 - 226). New
York: Guilford Press.
Hermstein, R. J., (1961). Relative and absolute strength of response as a function of
frequency of reinforcement. Journal of Experimental Analysis o f Behavior. 4.
267 -272.
Hermstein, R. J. (1970). On the law of effect. Journal of Experimental Analysis of
Behavior. 13. 243 - 266.
Hermstein, R. J. (1997) The Matching Law. NY: Harvard University Press.
Maritz. J. S., and Jarrett, R. G. (1978). A note on estimating the variance of the sample
median. Journal o f American Statistical Association. 73. 194 - 196.
Martens, B. K., & Houk, J. L. (1989). The application of Hermstein’s law of effect to
disruptive and on-task behavior of a related adolescent girl. Journal of
Experimental Analysis of Behavior. 5 1 .17-28.
McDowell, J. J. (1981). On the validity and utility of Herrenstein’s hyperbula in
applied behavior analysis. In C. M. Bradshaw, E. Szabadi, & C. F. Lowe
(Eds.), Quantification of steadv-state operant behavior, (pp. 311 - 324).
Amsterdam: Elsevier/ North Holland.
McDowell, J. J. (1988). Matching theory in natural human environments. The
Behavior Analyst. 11.95 - 109.
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Nlee. R- W., (1990). Confidence intervals for probabilities and tolerance regions based on
a generalization of the Man-Whitney statistic. Journal o f the American Statistical
Association. 85. 793 - 800.
Mooney. D.. Fromme. K.. Kivlahan. D., & Marian. G. (1986). Sex. age. and
expectancies relate differentially to quantity and frequency. Addictive
Behaviors. 14. S31 - 536.
National Institute on Alcohol Abuse and Alcoholism (1995). Assessing alcohol
Problems: A Guide for Clinicians and Researchers. Allen. P. J.. & Columbus M.
(Eds.). NIH publication No. 95-3745.
Neef A. N.. Mace. F. C.. Shea. C. M.. & Shade. D. (1992). Effects of reinforcer rate
and reinforcer quality on time allocation: Extent ion o f matching theory to
educational settings. Journal of Applied Behavior Analysis. 25. 691 - 699.
Rachlin. H. (1989). Judgement, decision, and choice: A cognitive behavioral
synthesis. New York: W. H. Freeman.
Rom. D. M. (1990). A sequentially rejective test procedure based on a modified
Befferoni inequality. Biometrica. 77.663-666
Pierce. W. D.. & Epling, W. F. (1983). Choice matching and human behavior: A
review of the literature. The Behavior Analyst. 6. 57 - 76.
Southwiek. L.L.. Steele, C.M.. Marlatt. G. A.. & LindeL M. (1981). Alcohol related
expectancies: defined by phase of intoxication and drinking experience.
Journal of Consulting and Clinical Psychology: 49. 713-721.
Tolman. E. C. (1932). Purposive behavior in animals and men. New York: Century.
Wilcox, R. R. (1996). Statistics for the Social Sciences. New York: Academic Press.INC.
Wilcox. R.R. (1997). Introduction to Robust Statistics and Hypothesis Testing. New
York: Academic Press
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
Studying n m r tiB tiB
Please rate on a scale from 1 to 10 the extent to which you agree with the statements
below.
I = absolutely do not agree
10 = absolutely agree
1) Earning a college degree will allow me employment opportunities that may
significantly enhance my future economical possibilities. 1..2..3..4..5..6..7..8..9..10
2) Earning a college degree will allow me employment opportunities that are
significantly more interesting/challenging than the ones that will otherwise
be available to me. 1 ..2..3..4..5..6..7..8..9.. 10
3)(The degree to which I will materialize these opportunities
depends on my GPA. 1 ..2..3..4..5..6..7..8..9.. 10
4) Earning a college degree will enhance my social status in the long run. (Relative to the
social status I would otherwise have). 1 ..2..3..4..5..6..7..8..9.. 10
5) A college degree will earn me more respect from others. (Relative to the respect others
would pay me if 1 would not have studied in college). 1..2..3..4..5..6..7..8..9.. 10
6) Earning a college degree is really important to my parents. I ..2..3..4..5..6..7..8..9.. 10
7) Earning a college degree is really important to me. 1..2..3..4..5..6..7..8..9..10
8) Maintaining a high GPA is really important to me. 1 ..2..3..4. J..6..7..8..9.. 10
9) The more I study the higher my grades are. 1..2..3..4..5..6..7..8..9..10
10) How many hours per week do you usually study (not including class
attendance)?__________.
11) How would you rate your class attendance relative to other students (the percentiles
represent the portion of students that who's attendance is lower than yours) ?
10%........ 20%... .30%......40%........50%......60%......70%......80%.....90%..... 100%
12) On how many of your next ten exams/assignments do you expect to get at least the
grade that you are aiming at? 1 ..2..3..4.J..6..7..8..9.. 10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13) How would you rate your dedication to school work relative to other student (the
percentiles represent the portion of students who’s dedication to school work is lower
than yours)?
10%...20%....30%...... 40%........50%...60%...70%.....80%......90%.....100%
14) It is important to me to do well on every exam or paper. 1 ..2..3..4..5..6..7..8..9.. 10
15) It is important to me to feel well prepared before
I take an exam. 1..2..3..4..5..6..7..8..9.. 10
16) It is important to me to read as much as possible before
coming to class, or on a regular basis. 1..2..3..4..5..6..7..8..9..10
17) Studying is really important to me. 1..2..3..4..5..6..7..8..9.. 10
Snorts and physical training.
1) Exercising (jogging, swimming, biking, working out in the gym, aerobics)
is healthy for me. 1 ..2..3..4..5..6..7..8..9.. 10
2) Exercising makes me feel good. 1 ..2..3..4..5..6..7..8..9.. 10
3) Exercising make my body look better. 1. J..3..4..5..6..7..8..9.. 10
4) Exercising reduces my level of stress. 1..2..3..4..5..6..7..8..9..10
5) How many kinds of sport games (Tennis, Ballet. Football,
Basketball Surfing, etc.) do you enjoy playing? 1..2..3..4..5..6..7..8..9..10
6) How many hours per week do you devote to sport games? 1..2..3..4..5..6..7..8..9..10
7) How many hours a week do you exercise? I ..2..3..4..5..6..7..8..9.. 10
8) Sport is really important to me. 1 „2..3..4..5..6..7..8..9.. 10
Intellectual Interests and hobbies.
1) How much pleasure do you expect from
reading a book o f your choice? 1..2..3..4..5..6..7..8..9..10
2) Of the next ten books that you will read for your
own pleasure, how many do you think will get a score
higher than 5 (from the previous question). 1..2..3..4..5..6..7..8..9..10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3) How many books do you read for your own
pleasure during one semester?
4) How much pleasure do you expect from
going to the theater to see a play?
3) Of the next ten plays that you will see. how many do
you think will get a score higher than S
(from the previous question)?
6) How many plays do you usually go to in one year?
7) How much pleasure do you expect from
going to art exhibitions or museums?
8) Of the next ten an exhibitions or museums
that you will visit, how many do you think will
get a score higher than S (from the previous question)?
9) How many an exhibitions or museums
do you usually go to in one year?
10) How much pleasure do you expect from going
to a concerts o f classical music or jazz?
11) Of the next ten concerts that you will attend
in the future, how many do you think will get
a score higher than S (from the previous question).
12) How many concerts o f jazz or classical music do
you go to in one year?
13) How much pleasure do you expect from
going to a movie of your choice?
14) Of the next ten movies that you will see,
how many do you think will get a score higher
than 3 (from the previous question)?
13) How many movies do you usually watch
in one semester?
16) How much pleasure do you expect
from watching T.V. at home?
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2.J..4..5..6..7..8..9..I0
1..2..3..4..5..6..7..8..9..10
1..2.J..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17) Ofthe next ten T.V shows that you will watch,
how many do you think will get a score higher
than 5 (from the previous question)? 1 ..2..3..4..5..6..7..8..9.. 10
18) How many hours per week do you watch T.V? 1..2.J..4..5..6..7..8..9.. 10
19) How much pleasure do you expect from
playing a musical instrument? 1 ..2~3..4..5..6..7..8..9.. 10
20) How many hours per week do you play? 1 ~2..3..4..5..6..7..8..9.. 10
21) How much pleasure do you expect from
listening to music at home. 1..2..3..4..5..6..7..8..9..10
22) How many hours per week do you
spend listening to music at home? 1 ..2..3..4..5..6..7..8..9.. 10
23) How much pleasure do you expect from
going out hiking or camping? 1 ..2..3..4..5..6..7..8..9.. 10
24) How many times per year do you
go out camping or hiking? 1 ..2..3..4..5..6..7..8..9.. 10
25) How much do you enjoy staying
at home by yourself? 1..2..3..4..5..6..7..8..9..10
26) Do you have a hobby that you especially enjoy (Le. photography, cooking..). Y /N
If yes. please specify_________________ (sports and music excluded). How
much do you enjoy it? 1..2..3..4..5..6..7..8..9.. 10
27) How many hours per week do you spend
engaging in this hobby? 1..2..3..4..5..6..7..8..9..10
28) How much pleasure do you expect from reading
a journal of a specific topic (i.e. National Geographic.
computers, fashion, sport)? Please specify_____________ . 1..2..3..4..5..6..7..8..9..10
29) How much do you enjoy going to molls? 1 ..2..3..4..5..6..7..8..9.. 10
Family and Social interactioa«
1) How much do you like talking to your father?
(If father not available, circle N.A) 1..2..3..4..5..6..7..8..9..10 N.A
2) How many times a month do you call him? 1 ..2..3..4..5..6..7..8..9.. 10
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
3) How much do you like talking to your mother?
(If mother not available, circle N.A.) 1 ..2..3..4..5..6..7..8..9.. 10.N.A
4) How many times a month to you call her? 1.2. J..4..5..6..7..8..9.. 10
5) How much do you like talking to your brothers/sisters?
(In not available, circle N.A) 1 ..2..3..4„5..6..7..8..9.. 10.N.A
6) How many times a month do you talk to them? 1..2..3..4..5..6..7..8..9.. 10
7) How much do you enjoy spending time with your family? 1 ..2..3..4..5..6..7..8..9.. 10
8) How involved do you think you are in the
lives o f your family members? 1 ..2..3..4..5..6..7..8..9.. 10
9) How much would you enjoy spending time with
your girlfriend/boyfriend/spouse alone
(If you don’t currently have a partner, answer
according to what you think would like)? 1..2..3..4..5..6..7..8..9..10
10) How much would you enjoy talking with your
gvlfriend/boyfriend/spo use/partner about things that
bother them? (If you don’t currently have a partner, answer
according to what you think you would like). 1 ..2..3..4..S..6..7..8..9.. 10
11) How rewarding is it for you to listen to your
friends when they talk about their difficulties? 1 ..2..3..4..5..6..7..8..9.. 10
12) How rewarding is it for you to talk to your
friend about your own difficulties? 1..2..3..4..5..6..7..8..9.. 10
13) How much do you enjoy inviting or going to
friends over for a quite evening at home. 1..2..3..4..5..6..7..8..9..10
14) How often do you do that during a month period? 1 ..2..3..4..S..6..7..8..9.. 10
13) List the three social activities that you enjoy the most.
(A = the most preferable social activity).
A )_____________________
B )__________________
C )___________________
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
16) List the three social activities that you engage in the most. (The difference
between this question and the previous one lies in the possibility that the activities
which your social group engage in the most do not accurately correspond to your
personal priorities o f social activities).
A )_________________________. Hours per week_________________
B ) . Hours per_week_________________
C )_________________________ . Hours per week_________________
17) How important to you is staying in touch with the family? 1 ..2..3..4..5..6..7..8..9.. 10
18) How important to you is having social (peer) interaction? 1..2..3..4..5..6..7..8..9..10
Work.
1) Do you currently have a job (also part time)? Y /N
2) If not. are you currently looking for a job? Y /N
If you are. answer the following questions camming you have found the type o f job you
are looking for.
3) How financially significant is your job to your
graduating from college? 1 ..2..3..4..S..6..7..8..9.. 10
4) How many hours per week do you work?___________
5) How important is it for you to earn your own money
at this point in your life? 1 ..2..3..4..S..6..7..8..9.. 10
6) How much do you enjoy the social interactions at work? 1..2..3..4..5..6..7..8..9..10
7) How important to uou is your job? 1 ..2.J..4..5..6..7..8..9.. 10
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
Please rate in descending order the four activities that take up the most o f your
time (exclude eating and sleeping, unless you devote more than 9 hours for sleeping and
two hours for meals per day). Then rate on a scale from 1 to 10 how much you enjoy
engaging in those activities.
I ) _____________________ Hours per week__________
How much do you enjoy engaging in this activity? 1 ..2..3..4..5..6..7..8..9.. 10
How much do you think engaging in this
activity is worthwhile for you in the long run? 1 ..2..3..4..5..6..7..8..9.. 10
II ) _____________________ Hours per week__________
How much do you enjoy engaging in this activity? 1 ~2..3~4..5..6..7..8..9.. 10
How much do you think engaging in this
activity is worthwhile for you in the long run? 1..2..3..4..5..6..7..8..9..10
HI) __________________ Hours per week___
How much do you enjoy engaging in this activity? 1..2..3..4..5..6..7..8..9..10
How much do you think engaging in this
activity is worthwhile for you in the long run? I ..2..3..4..5..6..7..8..9.. 10
D )____________________ Hours per week__________
How much do you enjoy engaging in this activity? I..2..3..4..5..6..7..8..9..10
How much do you think engaging in this
activity is worthwhile for you in the long run? 1..2..3..4..5..6..7..8..9..10
Please rate in descending order the four activities you enjoy the most.
A )________________ Hours per week
B )________________ Hours per week,
C )______________ Hours per week.
D )__________________ Hours per week.
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
Drinking Habits
Now we would like to ask about your drinking habits over the last
year. Please write the appropriate number in the blank on your
answer sheet.
7. Approximately how many times per week do you drink alcohol
(beer, wine, liquor, mixed drinks, wine coolers, etc.)? 0-7.
8. Approximately how many times per month do you drink alcohol
(beer, wine, liquor, mixed drinks, wine coolers, etc.)?
0-30.
In the next questions, assume 1 drink = 1 beer, 4 oz. wine, 1 wine
cooler, 1 mixed drink, or 1« oz. liquor.
9. On the average, how many drinks did you usually have when you
drank alcohol over the last year?
10. What's the largest number of drinks you ever drank on one
occasion in the last year?
Please rate from 0 (not at all) to 10 (extremely) how much you
think alcohol would produce the following feelings IMMEDIATELY
AFTER FINISHING 4 drinks that you drank in one hour. If you have
never consumed 4 drinks in an hour, give your best guess of how you
would feel. Please circle one number for each feeling.
1 . DIFFICULTY CONCENTRATING 0 1 2 3 4 5 6 7 8 9 10
2. DOWN 0 1 2 3 4 5 6 7 8 9 10
3. ELATED 0 1 2 3 4 5 6 7 8 9 10
4. ENERGIZED 0 1 2 3 4 5 6 7 8 9 10
5. EXCITED 0 1 2 3 4 5 6 7 8 9 10
6. HEAVY HEAD 0 1 2 3 4 5 6 7 8 9 10
7 . INACTIVE 0 1 2 3 4 5 6 7 8 9 10
8. SEDATED 0 1 2 3 4 5 6 7 8 9 10
9. SLOW THOUGHTS 0 1 2 3 4 5 6 7 8 9 10
10. SLUGGISH 0 1 2 3 4 5 6 7 8 9 10
11. STIMULATED 0 1 2 3 4 5 6 7 8 9 10
12. TALKATIVE 0 1 2 3 4 5 6 7 8 9 10
13. UP 0 1 2 3 4 5 6 7 8 9 10
14. VIGOROUS 0 1 2 3 4 5 6 7 8 9 10
66
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
Please rate how much you think alcohol would produce the following
feelings ONE AND A HALF HOURS AFTER FINISHING 4 drinks that you
drank in an hour. In other words, imagine that after drinking 4
drinks in an hour you did not drink any more alcohol. Please rate
how you would feel one and a half hours later.
1 . DIFFICULTY CONCENTRATING 0 1 2 3 4 5 6 7 8 9 10
2. DOWN 0 1 2 3 4 5 6 7 8 9 10
3. ELATED 0 1 2 3 4 5 6 7 8 9 10
4. ENERGIZED 0 1 2 3 4 5 6 7 8 9 10
5. EXCITED 0 1 2 3 4 5 6 7 8 9 10
6. HEAVY HEAD 0 1 2 3 4 5 6 7 8 9 10
7. INACTIVE 0 1 2 3 4 5 6 7 8 9 10
8. SEDATED 0 1 2 3 4 5 6 7 8 9 10
9. SLOW THOUGHTS 0 1 2 3 4 5 6 7
8 9 10
10. SLUGGISH 0 1 2 3 4 5 6 7 8 9 10
11. STIMULATED 0 1 2 3 4 5 6 7 8 9 10
12. TALKATIVE 0 1 2 3 4 5 6 7 8 9 10
13. UP 0 1 2 3 4 5 6 7 8 9 10
14. VIGOROUS 0 1 2 3 4 5 6 7 8 9 10
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
67
Alcohol Effects Questionnaire
Directions: We would like to find out what you personally experience after you have
had a f e w a l c o h o l i c drinks. Please indicate on a scale from 1 to 10 the degree to which
the following possible experiences match your own personal experience after a few
drinks. (1 “ least matching. 10 = most notching).
1 . Drinking makes me feel flushed. 1 -2..3..4..5..6..7..S..9.. 10
2. Alcohol decreases muscular tension in my body. 1..2..3..4..5..6..7..8..9..10
3. A few drinks make me feel less shy. 1 ..2..3..4..5..6..7..8..9.. 10
4. alcohol enables me to fall asleep more easily. I ~2..3..4..5..6..7..8..9.. 1 0
3. I feel powerful when I drink, as if I can really
influence other people to do what I want. 1 J..4..5..6..7..8..9.. 10
6. I'm more clumsy after a few drinks. I..2. J..4..5..6..7..8..9.. 10
7. I’m more romamic when I drink. 1 .J..3..4..5..6..7..8..9.. 10
8. Drinking makes the future seem brighter to me. 1 ~2..3..4..5..6..7..8..9.. 10
9. If I have had a couple of drinks it is easier for me
to tell someone off. 1 „2..3..4. J..6..7-8..9.. 1 0
10. I can’t act as quickly when I've been drinking. 1..2..3..4..5..6..7..8..9..I0
11. Alcohol can act as an anesthetic for me. that is,
it can deaden pain.
1 2. I often feel sexier after I’ve had a few drinks.
13. Drinking makes me feel good.
1 4. Alcohol makes me careless about my actions.
1 3. Some alcohol has a pleasant, cleansing, tmgly,
taste to me.
1 6. Drinking increases my aggressiveness.
1 7. Alcohol seems like magic to me.
1 8. Alcohol makes it hard for me to concentrate.
19. I am a better lover after a few drinks.
20. W hen I’m drinking, it is easier to open up
and express my feelings.
1..2..3..4..5..6..7..8..9..10
1..1.3..4.J..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1..2..3..4..5..6..7..8..9..10
1 ..2..3..4..S..6..7..8..9.. 1 0
1..2..3..4..5..6..7..8..9..10
I ..2. J..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9..10
I ~2..3..4..5..6..7..8..9.. 1 0
1 ~2-3..4„5..6..7..8..9.. 10
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
21. Drinking adds a certain warmth to social
occasions.
22. If I'm feeling restricted in any way, a few
drinks make me feel better.
2 3 .1 can't think as quickly after I drink.
24. Having a few drinks is a nice way for
me to celebrate special occasions.
25. Alcohol makes me worry less.
26. Drinking makes me inefficient.
27. Drinking is pleasurable because it is enjoyable
for me to join in with people who are enjoying
themselves.
28. After a few drinks, I am more sexually responsive.
2 9 .1 feel more coordinated after I drink.
30. I'm more likely to say embarrassing things
after drinking.
31. 1 enjoy having sex more if I've had some alcohol.
32. I'm more likely to get into an argument if I've
had some alcohol.
33. Alcohol makes me less concerned about doing
things well.
34. Alcohol makes me sleep better.
35. Drinking gives me more confidence in myself.
36. Alcohol makes me more irresponsible.
37. After a few drinks it is easier for me to
pick a fight.
38. A few drinks make it easier for me to talk
to people.
39. If I have a couple of drinks it is easier to
express my feelings.
40. Alcohol makes me more interesting.
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
R.A.P.I
Different things happen to people when they drink ALCOHOL, or as a result o f their
ALCOHOL use. Some o f these things are listed below. Please indicate how many
time* each has happened to you during the last 3 yean while you were drinking
alcohol or as a result o f your alcohol use. (0 K never, 1 ■ once, 2 * twice..... 10* ten
times).
How many times have the following things happen to you while you were drinking
alcohol or because o f your alcohol »«* during the last three veers?
1. Not able to do your homework or study for a test. 1..2~3..4..5..6..7..8..9..10
2. Got into fights, acted bad. or did mean things. 1 ..2..3..4..5..6..7..8..9.. 10
3. Missed out in other things because you spent too
much money on alcohol 1 ..2..3..4..5..6..7..8..9.. 10
4. Went to work or school high or drunk. 1 ..2..3..4..5..6..7..8..9.. 10
5. Caused shame or embarrassment to someone. 1 ~2..3..4..5..6..7..8..9.. 10
6. Neglected your responsibilities. 1 ..2..3..4..5..6..7..8..9.. 10
7. Relatives avoided you. 1 ..2..3..4..5..6..7..8..9.. 10
8. Felt that you needed more alcohol than you used
to use in order to get the same effect. 1..2..3..4..5..6..7..8..9.. 10
9. Tried to control your drinking by trying to drink
only at certain times o f day or certain places. 1..2..3..4..5..6..7..8..9..10
10. Had withdrawal symptoms, that is. fek sick
because you stopped or cut down on drinking. 1 ..2..3..4..5..6..7..8..9.. 10
11. Noticed a change in your personality. 1 ..2..3..4..5..6..7..8..9.. 10
12. Felt that you had a problem with school. 1 ..2..3..4..5..6..7..8..9..10
13. Missed a day (or part of a day)
o f school or work. 1 ..2..3..4..S..6..7..8..9.. 10
1 4. Tried to cut down on drinking. 1.2. J..4..5..6..7..8..9.. 10
13. Suddenly found yourself in a place that you
could not remember gening to. 1..2..3..4..5..6..7..8..9..10
16. Passed out or fainted suddenly. 1 ..2..3..4..5..6..7..8..9.. 10
Reproduced with permission ofthe copyright owner. Further reproduction prohibited without permission.
17. Had a fight, argument, or bad feelings with
a friend.
18. Had a fight, argument, or bad feelings with
a family member.
19. Kept drinking when you promised
yourself not to.
20. Felt you were going crazy.
21. Had a bad time.
22. Felt physically or psychologically
dependent on alcohol.
23. Was told by a friend or neighbor to stop
or cut down drinking.
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
1..2..3..4..5..6..7..8..9.. 10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
Age:
Sex:
Year in college:
Did one of your parents have a drinking problem?
Was one of your parents ever treated for alcoholism?
To the following questions answer True or False as they apply to you:
I have never envied anyone's success ..................T/F
I read all the editorials in the newspaper ...............T/F
I like everyone I know ............................... T/F
I never feel down .................................................... T/F
have never been to a shopping mall .....................T/F
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Applying the theory of reasoned action to condom use: The effect of immediate consequences on intention-behavior consistency

PDF

An evaluation of a brief HIV /AIDS prevention intervention using normative feedback to promote risk reduction among sexually active college students

PDF

Alcohol expectancies and consumption: Age and sex differences

PDF

Articulated thoughts about intentions to commit anti-gay hate crimes

PDF

Content analysis of articulated thoughts of chronic worriers

PDF

Hedonic aspects of conditioned taste aversion in rats and humans

PDF

Affection and conflict in family relationships

PDF

Hyperactive symptoms, cognitive functioning, and drinking habits

PDF

History of depression, antidepressant treatment, and other psychiatric illness as risk factors for Alzheimer's disease in a twin sample

PDF

Associations and mechanisms among attention deficit hyperactivity disorder symptoms, cognitive functioning, and drinking habits

PDF

Depressed children and the social and behavioral attributes of their best friend

PDF

Five quantile test: A comparison of independent groups

PDF

Cognitive functioning and dementia following cancer: A Swedish twin study

PDF

Articulated thoughts regarding cognitions toward older adults

PDF

Aging and the use of context and frequency information in ambiguity resolution

PDF

Is there more to discrete prepulses than meets the eye?

PDF

Depression in offspring following severe prenatal stress

PDF

Head injury and dementia: A co-twin control study of Swedish twins

PDF

Alcohol's impact on triggered displaced aggression

PDF

Insights into the nature of phonological and surface dyslexia: Evidence from a novel word learning task

Asset Metadata

Creator Levy, Boaz (author)

Core Title Expectancies for alternative behaviors predict drinking problems: A test of a cognitive reformulation of the matching law

School Graduate School

Degree Master of Arts

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag health sciences, public health,OAI-PMH Harvest,psychology, behavioral,Psychology, clinical

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Earleywine, Mitchell (committee chair), [illegible] (committee member), Davison, Gerald C. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-290945

Unique identifier UC11341491

Identifier 1409641.pdf (filename),usctheses-c16-290945 (legacy record id)

Legacy Identifier 1409641-0.pdf

Dmrecord 290945

Document Type Thesis

Rights Levy, Boaz

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA