The Role of Society in Promoting or Deterring Alcohol Use and Addiction

Journal of Consulting and Clinical
1999, Vol. 67, No. 3, 394-404

Copyright 1999 by the American Psychological Association, Inc.
0022-006X/99/S3.00

Personality and Substance Use Disorders:
II. Alcoholism Versus Drug Use Disorders

Matt McGue
University of Minnesota

Wendy Slutske
University of Missouri

William G. lacono
University of Minnesota

The relationship between personality and substance use disorders was investigated in a community-based
sample of 638 individuals who were alcoholic and/or had a drug use disorder, and 1,530 individuals who
did not have a substance use disorder. Personality was assessed by the Multidimensional Personality
Questionnaire; substance use diagnoses were based on standard criteria as assessed by interview. Data
were analyzed using a 3-factor (Gender X Alcoholism X Drug Use Disorder) multivariate analysis of
variance. The significant alcoholism main effect was associated primarily with negative emotionality,
whereas the significant drag use disorder main effect was associated primarily with constraint. No
significant interactions with gender were observed. These findings suggest that the elevated levels of
behavioral disinhibition observed with alcoholic individuals may be attributable to a subset of alcoholic
individuals who also abuse drugs other than alcohol.

Several recent developments in psychology, as well as in the
substance abuse research field, have helped spark a renewed in-
terest in the role of personality in the etiology of alcoholism and
other substance use disorders. First, behavioral genetic studies
have consistently indicated that genetic factors contribute to risk of
substance abuse (McGue, 1995; Pickens et al., 1991) and have
thereby implicated the existence of inherited, individual-level risk
factors for substance use disorders. Personality characteristics con-
stitute an important instance of an individual-level risk factor that
is both associated with substance abuse risk (e.g., Sher & Trull,
1994) and substantially heritable (e.g., Tellegen et al., 1988).
Second, theoretical developments in the personality field have led
to the advancement of hierarchical models of personality based on
either a three- (Cloninger, 1987b; Eysenck, 1967; Tellegen, 1991)
or a five-factor (Costa & McCrae, 1992) model. These models
have facilitated research on personality not only by providing a
theoretically coherent structure for the vast array of hypothesized
personality traits but also by suggesting systematic approaches to
the assessment of the major dimensions of personality. Finally,
several prominent substance use researchers have proposed theo-
retical models that accord personality factors a central role in the

Matt McGue and William G. lacono, Department of , Uni-
versity of Minnesota; Wendy Slutske, Department of , Univer-
sity of Missouri.

This research was supported in part by U.S. Public Health Service
Grants AA00175, AA09367, and DA05147.

Correspondence concerning this article should be addressed to Matt
McGue, Department of , University of Minnesota, 75 East
River Road, Minneapolis, Minnesota 55455. Electronic mail may be sent to
[email protected].

development of substance use disorders (Cloninger, 1987a; Tarter,
Alterman, & Edwards, 1985).

The vast majority of research linking personality to substance
use disorders has focused on alcoholism. This research (reviewed
by Sher & Trull, 1994) is generally consistent in implicating two
major dimensions of personality in the etiology of alcoholism:
negative emotionality, or the tendency to experience negative
mood states and psychological distress, and behavioral disinhibi-
tion, or the inability or unwillingness to inhibit behavioral im-
pulses. Compared with nonalcoholic individuals, alcoholic indi-
viduals, on average, score higher on these two dimensions of
personality, and there is some evidence, albeit limited, suggesting
that these differences predate alcoholism onset (e.g., Caspi, Mof-
fitt, Newman, & Silva, 1996; Cloninger, Sigvardsson, Reich, &
Bohman, 1988; Kammeier, Hoffman, & Loper, 1973). Moreover,
the association of these dimensions with alcoholism risk is robust,
holding, for example, for both men and women (McGue, Slutske,
Taylor, & lacono, 1997). Nonetheless, mean differences between
representative samples of alcoholic and nonalcoholic individuals
on indicators of negative emotionality and behavioral disinhibition
appear only to be moderate in magnitude (McGue et al., 1997). If
personality influences alcoholism risk it does so as one of a myriad
of risk factors in a multifactorial system rather than as the effect of
a unique configuration of personality characteristics, the posses-
sion of which leads inevitably to an addiction (i.e., an “addictive
personality”; Nathan, 1988).

The personality correlates of drug use disorders have been less
widely researched than the personality correlates of alcoholism.
Moreover, much of the research that does exist in this area suffers
from serious methodological shortcomings, including the use of
nonrepresentative, treatment-ascertained samples; poorly defined
control groups; and psychometrically weak personality measures

394

PERSONALITY AND SUBSTANCE USE DISORDERS 395

(Sutker & Allain, 1988). Sensation seeking is the most widely
studied personality correlate of drug use disorders, with the pre-
dictable finding that indicators of this personality characteristic
have been consistently associated with drug use disorder status
(Kosten, Ball, & Rounsaville, 1994; Luthar, Anton, Merikangas, &
Rounsaville, 1992; Sutker, Archer, & Allain, 1978; Vukov, Baba-
Milkic, Lecic, Mijalkovic, & Marinkovic, 1995). The recent study
by Krueger, Caspi, Moffitt, Silva, and McGee (1996) is the only

published investigation of the personality correlates of substance
use disorders that used both a comprehensive assessment of per-
sonality (the Multidimensional Personality Questionnaire [MPQ]),

and a community-based ascertainment scheme (a representative
birth cohort from New Zealand). These investigators reported that
a composite diagnosis of substance dependence was associated
with elevated levels of negative emotionality and depressed levels

of constraint relative to non-substance-dependent controls. Unfor-
tunately, Krueger et al. did not break down their findings by either
gender or type of substance dependence, making it difficult to
evaluate the generality of the correlations they reported. Although
there have been few studies of the personality correlates of sub-
stance use disorders, the personality correlates of drug use have
been extensively investigated, especially in adolescent popula-
tions. These studies have consistently reported that adolescent drug
users score higher than adolescent non-drug users on indicators of
behavioral disinhibition (e.g., Labouvie & McGee, 1986; Wills,
Vaccaro, & McNamara, 1994).

The present study is the second in a series exploring the role of
personality in the etiology of substance use disorders. In the first
study (McGue et al., 1997), we investigated the association of
alcoholism with personality and determined whether that associa-
tion was moderated by gender or alcoholism subtype. In this
second study, we contrasted the personality correlates of alcohol-
ism with the personality correlates of drug use disorders. Although
the use of a cross-sectional design did not allow us to determine
whether the personality correlates we observed were causes or
consequences of a substance use disorder, the present study did
address several of the major limitations of research relating per-
sonality and addictions (Sutker & Allain, 1988). In particular, the
present study was based on a large sample of individuals with (n =
638) and without (n — 1,530) a substance use disorder. All
individuals were ascertained using the same community-based
procedure, and all individuals completed a comprehensive self-
report personality inventory that is based on a theoretically mean-
ingful three-factor model.

Method

Sample

The sample was composed of parent participants in the Minnesota Twin
Family Study (MTFS). The MTFS is a prospective study of the develop-
ment of substance use and related disorders in a large sample of adolescent
twins and their parents (lacono, Lykken, & McGue, 1996; McGue, Lykken,
& lacono, 1996). Twin families were ascertained from Minnesota birth
records and located using various public databases, including telephone
directories and driver’s license registrations (Lykken, Bouchard, McGue,
& Tellegen, 1990). For any given birth year, we were able to locate more
than 90% of twin births. Families were excluded from participation if they
lived farther than a day’s drive from our laboratories in Minneapolis,
Minnesota, or if either of the twins in the family had a physical or cognitive

handicap that would preclude their completing our day-long, in-person
intake assessment. Among eligible families, 17% refused participation,
with 78% of the nonrefusing families ultimately completing the intake
assessment. Through telephone interviews, mail surveys, and access to
information recorded on the birth records, we were able to determine the
socioeconomic status (SES) and complete brief mental health assessments
on 69% of the nonparticipating families. Analysis of these data indicated
that participating families differed minimally from nonparticipating fami-
lies, with the largest differences occurring on measures of SES in which
parents in participating families averaged 0.6 years more education than
parents in nonparticipating families. Consistent with Minnesota demo-
graphics, over 98% of participating parents were Caucasian.

Substance use diagnoses were made from the clinical interviews of 2,698
parents (1,384 mothers, 1,314 fathers) and were based on Diagnostic and
Statistical Manual of Mental Disorders (3rd ed., rev.; DSM-III-R; Amer-
ican Psychiatric Association, 1987) lifetime criteria at two levels of cer-
tainty: definite (met all criteria for dependence or abuse) and probable (all
but one Criterion A dependence symptom were satisfied). Diagnoses were
used to form four groups in a fully crossed 2 X 2 design (Alcohol Use
Disorder x Drug Use Disorder). The first grouping factor was based on
DSM-III-R diagnoses of alcohol dependence and distinguished individuals
with a definite diagnosis of alcohol dependence (designated “alcohol yes”)
from those who did not meet criteria for alcohol dependence at either the
definite or probable levels of certainty, (designated “alcohol no”). The
second factor was based on DSM-III-R diagnoses of psychoactive sub-
stance abuse or dependence for the following eight substances: amphet-
amines, cannabis, cocaine, hallucinogens, inhalants, opiates, PCP-related
substances, and sedatives. Individuals with a definite diagnosis of sub-
stance abuse or dependence for at least one of the eight substances
composed one group (designated “drug yes”), whereas those not diagnosed
with a substance use disorder at either the definite (for dependence or
abuse) or probable (for dependence) level for any of the substances formed
the second group (designated “drug no”). Note that nicotine was not used
as one of eight substances to determine group status and that individuals
with a probable diagnosis of alcohol dependence, as well as individuals
with no more than a probable dependence diagnosis for the eight sub-
stances, were not classified into one of the four groups (n = 237, or 9%).

Measures

Demographics. Age, educational attainment in years, current marital
status, and occupation were all assessed by self-report. On the basis of the
occupation reports, we rated occupational status using the Hollingshead
system (Hollingshead & Redlich, 1958), which ranges from 1 (profes-
sional) to 7 (unskilled labor). Because full-time homemakers are not scaled
in the Hollingshead system, we considered homemakers as “missing” in the
analysis on occupational status presented here. Occupational status codes
were not available for 123 participants (11%) in the female sample.

Clinical assessment. Participants were assessed in person by trained
lay interviewers. Substance use disorders were assessed using an expanded
version of the Substance Abuse Module (Robins, Babor, & Cottier, 1987),
which was developed as a supplement to the World Health Organization’s
Composite International Diagnostic Interview (Robins et al., 1988). Major
depression was assessed using the Structured Clinical Interview for DSM-
III-R (Spitzer, Williams, & Gibbon, 1987), whereas conduct disorder and
antisocial personality disorder (ASPD) were assessed using a clinical
interview developed by MTFS staff. Prior to the assignment of individual
diagnoses, a clinical case conference was completed in which the evidence
for every symptom was reviewed (including, if necessary, the replaying of
audiotapes from the interviews) by at least two advanced graduate students
in clinical psychology. After consensus was reached, life-time diagnoses
using DSM-III-R criteria were assigned by computer algorithm.

Because the DSM-III-R criteria are meant to be used to diagnose acute
disorders, the lifetime substance use diagnoses were made at two levels of
certainty to reduce the likelihood that affected individuals were classified

396 McGUE, SLUTSKE, AND IACONO

as “unaffected.” A definite diagnosis was made if all necessary DSM-III-K
criteria were satisfied; a probable diagnosis was made if all but one of the
criteria were met. Because DSM-IIl-R substance abuse is defined by the
presence of a single symptom, this approach was used for dependence only.
As described previously, individuals with probable diagnoses of substance
dependence were excluded from the sample. Findings for all of the relevant
disorders presented here are based on lifetime definite diagnoses only.
Because our diagnoses are lifetime rather than current, we provide descrip-
tive data on two categorical indicators of the recency of substance use
problems. (Our intake assessment procedures did not allow for the diag-
nosis of current alcohol dependence and substance abuse and dependence.)
Alcohol recency distinguished individuals who had at least one of the
symptoms used to diagnose alcohol dependence in the past year from those
who did not; drug recency distinguished individuals who had used any of
the eight illicit substances in the past year from those who had not.
Participants were also asked whether they had ever been treated for
alcoholism.

Personality. Personality was assessed by self-report using the 198-
item version of the MPQ (Tellegen, 1982; Tellegen & Waller, in press).
The MPQ was developed using factor analysis and consists of 11 psycho-
metrically sound primary scales and three higher order factor scales, the
latter being linear-weighted composites of the primary scales. The MPQ is
organized around three broad dimensions of personality: positive emotion-
ality, or the tendency to be actively and pleasurably engaged with one’s
work and social environments; negative emotionality, or the tendency to
experience psychological distress and negative mood states; and constraint,
or the tendency to be cautious, to inhibit behavioral impulses, and to
endorse conventional moral values (i.e., the complement of behavioral
disinhibition). Four of the MPQ primary scales (Well-Being, Social Po-
tency, Achievement, and Social Closeness) load principally on positive
emotionality, three (Stress Reaction, Alienation, and Aggression) load
principally on negative emotionality, and three (Control, Harm Avoidance,
and Traditionalism) load principally on constraint. The eleventh primary
scale, Absorption, does not load principally on any of the three higher order
factors.

Prior to their scheduled in-person assessment, parents were mailed the
MPQ, along with several other self-report instruments, which they were
asked to complete and either return by mail or bring with them to the
assessment. If we failed to obtain a completed MPQ from a parent by the
time of the in-person assessment, he or she was asked to complete the form
at home and return it to us by mail. One additional telephone reminder was
made if we still did not receive a completed MPQ. Of the 2,461 parents
classified into one of the four substance use diagnosis groups, 236 (10%)
did not return a completed MPQ and 57 (2%) returned a form but had either
skipped too many items or did not pass the MPQ validity scale screens. We
were more likely to receive a completed MPQ from women than from men
(91% vs. 85%) and from those not having an alcohol or drug use diagnosis
than from those having these diagnoses (90% vs. 83%).

After deleting those with missing or nonvalid MPQs, we found that the
sample sizes for the four female groups were 1,000 (84% of the sample) for
the alcohol no/drug no group, 66 (6%) for the alcohol yes/drug no
group, 85 (7%) for the alcohol no/drug yes group, and 40 (3%) for the
alcohol yes/drug yes group. The comparable figures for the four male
groups were 530 (54%), 246 (25%), 68 (7%), and 133 (14%), respectively.
The proportionate representation of gender in the four groups reflects the
well-documented gender difference in prevalence for substance use disor-
ders (e.g., Anthony & Helzer, 1991; Helzer, Bucholz, & Robins, 1992;
Warner, Kessler, Hughes, Anthony, & Nelson, 1995), whereas the non-
independence of the two classification variables reflects the substantial
comorbidity that exists between alcoholism and other substance use dis-
orders (Helzer & Pryzbeck, 1988).

Statistical Procedures

The major study questions were addressed with a three-factor multivar-
iate analysis of variance (MANOVA). The three factors were gender (male
vs. female), alcohol use disorder (no vs. yes), and drug use disorder (no vs.
yes). Significant MANOVA effects were followed up with univariate
analyses of variance (ANOVAs). Two separate MANOVAs were com-
puted. The first had as dependent variables the 11 primary scales, whereas
the second, because the higher order scales are linear composites of the
primary scales, had as dependent variables the three higher order scales. In
all analyses, we evaluated effects using a regression approach in order to
control for the correlation between the two independent variables. That is,
the alcohol effect was evaluated net of the drug effect, and, conversely, the
drug effect was evaluated net of the alcohol effect.

Results

Demographic and Clinical Comparisons

Demographic and clinical comparisons are given in Table 1 for
the four female groups and in Table 2 for the four male groups. We
assessed group differences separately for men and women on the
demographic and clinical indicators using either a one-factor
(group membership) ANOVA for quantitative measures or a log-
linear analysis for categorical variables. Because the drug use
disorder classification was based on the substance use diagnoses,
we compared rates of substance use disorders only for the alcohol
no/drug yes and the alcohol yes/drug yes groups.

Among women, those with a substance use disorder were on
average somewhat younger and less likely to be married than those
not having a substance use diagnosis, and alcoholic women had on
average about 1 year less of education than did nonalcoholic
women. Among men, those with a substance use disorder were less
likely to be married than those not having a substance use disorder,
and alcoholic men had on average 1 year less of education and
lower occupational status than did nonalcoholic men. As has been
reported in other community-ascertained epidemiological samples
(Helzer & Pryzbeck, 1988), the majority of alcoholic participants
in both the male and female samples had not been treated for
alcoholism. In the year previous to their assessment, a little less
than 50% of the alcoholic men and women were positive for at
least one alcohol-dependence symptom and approximately 20% to
25% of men and women with a drug diagnosis had used a sub-
stance illicitly.

As expected, rates of conduct disorder, ASPD, and nicotine
dependence were consistently higher among those having a sub-
stance use disorder than among those not having a substance use
disorder. Among men with a drug use disorder, alcoholism was
significantly associated with higher rates of multiple-substance
abuse or dependence. Nonalcoholic men with a drug use disorder
met criteria for either abuse or dependence for an average of 1.3
(SD = 0.7) of the eight substances, with only 24% of these
individuals being multiple-substance abusers (i.e., having a posi-
tive diagnosis for more than one of the eight substances). In
contrast, alcoholic men with a drug use disorder met criteria for
abuse or dependence for an average of 2.1 (SD = 1.4) of the eight
substances, and 54% of these individuals had a positive diagnosis
for more than one of the substances. Among women with a drug
use disorder, however, alcoholism was not associated with
multiple-substance abuse; nonalcoholic women with a drug use
disorder met criteria for abuse or dependence for an average of 1.4

PERSONALITY AND SUBSTANCE USE DISORDERS 397

Table 1
Demographic and Clinical Characteristics of the Female Substance Use Disorder Samples

Substance use disorder group

Alcohol no/
Characteristic drug no

Sample size
Total sample size
Personality sample sizeb

%
Demographics

Age (years)
M
SD

Education (years)
M
SD

Occupation
M
SD

Married (%)
Substance use diagnoses, treatment, and recency

Amphetamines (%)
Cannabis (%)
Cocaine (%)
Hallucinogens (%)
Inhalants (%)
Opiates (%)
PCP-related substances (%)
Sedatives (%)
No. of drug diagnoses

M
SD

Alcohol treatment (%)
Alcohol recency (%)
Substance recency (%)

Other clinical diagnoses (%)
Conduct disorder
ASPD
Major depression
Nicotine dependence

1,092
1,000

92

42.1
5.1

13.9
1.9

3.7
1.6

90

0
0
0
0
0
0
0
0

0.0
0.0
0
8
1

1
0

20
24

Alcohol yes/
drug no

76
66
87

40.3
5.3

13.0
1.7

4.0
1.6

76

0
0
0
0
0
0
0
0

0.0
0.0

17
36
3

9
3

33
73

Alcohol no/
drug yes

94
85
90

39.5
4.6

14.1
2.1

3.3
1.3

80

34
73
11
11
2
4
1
5

1.4
0.9
2

14
17

5
0

35
48

Alcohol yes/
drug yes p’

50
40
80

37.8
4.1

13.2
1.8

3.9
1.7

68

33
83
20
13
0
8
3

13

1.7
1.1

20
55
20

23
3

45
70

.03

<.001 .003 ns <.001 ns ns ns ns ns ns ns ns ns •C.001 <.001 <.001 <.001 <.001 <.001 <.001 Note. AH diagnoses are based on lifetime Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; DSM-III-R; American Psychiatric Association, 1987) criteria at a definite level of certainty. An individual was considered positive for each of the eight DSM-III—R substance use disorders if that individual met DSM—II1-R criteria for either abuse or dependence for that substance. An individual was considered to have an alcohol use disorder if he or she met DSM-III-R criteria for alcohol dependence (i.e., "alcohol yes"), and an individual was considered to have a drug use disorder (i.e., "drug yes") if he or she was positive for at least one of the eight substance use disorders. ASPD = antisocial personality disorder. 'p values are for the appropriate statistic for testing equality of group means or percentages. For the eight substance use diagnoses used in forming the drag use disorder grouping, the p value corresponds to the two-group comparison; in all other cases, it corresponds to the four-group comparison, ns = not significant at p < .05. b The personality sample size gives the size of the sample participating in this study. The remaining descriptive data in this table are based on this sample. (SD = 0.9) of the substances, whereas the corresponding mean for alcoholic women with a drug use disorder was 1.7 (SD = 1.1). The Association of Personality With Alcohol and Drug Use Disorders In the Gender X Alcohol X Drug MANOVA of the 11 MPQ primary scales, the gender main effect was statistically significant, F(ll,2150) = 52.1, p< .001, although none of the interaction terms involving gender were significant atp < .05. Both the alcohol, F(ll, 2150) = 4.60, p < .001, and drag, F(ll, 2150) = 9.01, p < .001, main effects were statistically significant, as was the Alcohol X Drug interaction, F(l 1,2150) = 2.16,p < .05. The MANOVA for the three higher order MPQ scales paralleled the results obtained with the primary scales: The gender effect was significant, F(3, 2158) = 17.6, p < .001, but none of the interaction terms involving gender were significant. Both the alcohol, F(3, 2158) = 8.52, p < .001, and drug, F(3, 2158) = 28.7, p < .001, main effects, as well as the Alcohol X Drug interaction, F(3, 1903) = 3.80, p < .05, were statistically significant. Follow-up ANOVAs for the significant MANOVA ef- fects are summarized in Table 3. 398 McGUE, SLUTSKE, AND IACONO Table 2 Demographic and Clinical Characteristics of the Male Substance Use Disorder Samples Substance use Alcohol no/ Characteristic drug no Sample size Total sample size Personality sample sizeb % Demographics Age (years) M SD Education (years) M SD Occupation M SD Married (%) Substance use diagnoses, treatment, and recency Amphetamines (%) Cannabis (%) Cocaine (%) Hallucinogens (%) Inhalants (%) Opiates (%) PCP-related substances (%) Sedatives (%) No. of drug diagnoses M S£>

Alcohol treatment (%)
Alcohol recency (%)
Substance recency (%)

Other clinical diagnoses (%)
Conduct disorder
ASPD
Major depression
Nicotine dependence

604
530

88

44.9
5.7

14.6
2.4

3.4
1.8

92

0
0
0
0
0
0
0
0

0.0
0.0
1

11
1

8
0
9

25

Alcohol yes/
drug no

287
246
86

44.6
5.9

13.7
2.2

4.1
1.7

84

0
0
0
0
0
0
0
0

0.0
0.0

18
46

2

15
3

15
55

disorder group

Alcohol no/
drug yes

90
68
76

41.1
3.5

14.6
2.3

3.1
1.7

84

16
94

2
15
0
2
2
3

1.3
0.7
2

28
24

19
7

19
49

Alcohol yes/
drug yes p’

168
133
79

40.7
4.7

13.6
1.9

4.1
1.8

79

49
91
20
20
2

10
2

11

2.1
1.4

38
50
26

17
23
17
67

.002

<.001 <.001 <.001 <.001 <.001 ns <.001 ns ns .03 ns .04 <.001 <.001 <.001 <-001 .005 <.001 .005 <.001 Note. All diagnoses are based on lifetime Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; DSM—II1-R; American Psychiatric Association, 1987) criteria at a definite level of certainty. An individual was considered positive for each of the eight DSM-IIl-R substance use disorders if that individual met DSM-HI-R criteria for either abuse or dependence for that substance. An individual was considered to have an alcohol use disorder if he or she met DSM—IIl—R criteria for alcohol dependence (i.e., "alcohol yes"), and an individual was considered to have a drug use disorder (i.e., "drug yes") if he or she was positive for at least one of the eight substance use disorders. ASPD = antisocial personality disorder. a p values are for the appropriate statistic for testing equality of group means or percentages. For the eight substance use diagnoses used in forming the drug use disorder grouping, the p value corresponds to the two-group comparison; in all other cases, it corresponds to the four-group comparison, ns = not significant at p < .05. b The personality sample size gives the size of the sample participating in this study. The remaining descriptive data in this table are based on this sample. MPQ scale means and standard deviations are given in Table 4 for the four female groups and in Table 5 for the four male groups. Because of the large size of our sample, modest effect sizes (ESs) could be statistically significant in the present study. Therefore, to facilitate interpretation and estimation of ES, we transformed scores separately for men and women so that each scale had a mean of 50.0 and a standard deviation of 10.0 in the alcohol no/drug no group (i.e., a T-score metric). Although this transfor- mation has the effect of removing the gender main effect, it allows us to focus on the ESs that are of principalinterest here (i.e., those associated with substance use disorders). Moreover, the transfor- mation serves to highlight the replicability of findings across the male and female samples. The pattern of significant ANOVA results summarized in Ta- ble 3 tends to be organized around the hierarchical structure of the MPQ. Alcoholism is primarily associated with elevated levels of negative emotionality. A significant alcohol effect was observed for the Negative Emotionality higher order scale, as well as for each of the three primary scales that load principally on this factor (i.e., Stress Reaction, Alienation, and Aggression). Despite this consistency of statistical results, the ESs associated with the alco- hol effect were only moderate in magnitude. For example, alco- PERSONALITY AND SUBSTANCE USE DISORDERS 399 Table 3 Results of Follow-Up ANOVA Tests of Significant MANOVA Effects MPQ scale Primary scale Weil-Being Social Potency Achievement Social Closeness Stress Reaction Alienation Aggression Control Harm Avoidance Traditionalism Absorption" Higher order scale PE NE CN Principal factor loading PE PE PE PE NE NE NE CN CN CN — — ANOVA effect p value Gender ns M > F***
M > F *
p > M***
F > M***

ns
M > F***

ns
F > M***

ns
p>M***

ns
ns

F > M***

Alcohol (A)

ns
ns
ns
ns

Yes > No***
Yes > No***
Yes > No***
No > Yes***

ns
ns
ns

ns
Yes > No***

ns

Drug (D)

ns
Yes > No**
No > Yes**

ns
ns

Yes > No*
Yes > No*
No > Yes***
No > Yes***
No > Yes***

ns

ns
ns

No > Yes***

A X D

<.01 ns ns <.01 <.05 ns ns ns ns …