Addictive Behaviors, Vol. 5, pp. 107-112, Printedin the USA. All rightsreserved.
03064603/80/020107-06$02.00/O Copyright 0 1980 PergamonPressLtd
1980
AN INVESTIGATION OF RELATIONSHIPS BETWEEN LEVEL OF ALCOHOL USE ‘IMPAIRMENT AND PERSONALITY CHARACTERISTICS RICHARD L. WILLIAMS, KENNETH U. GUTSCH, RICHARD KAZELSKIS, J. PENELOPE VERSTEGEN University
and JOAN SCANLON
of Southern
Mississippi
Abstract-This study sought to determine the relationships between levels of alcohol use impairment and personality characteristics. Subjects were 166 persons selected because of known or suspected alcohol use patterns. Subjects were administered the +tlcohol Use Inventory and Perscnality Research Form. Six alcohol use groups were determined by the raw scores of the Akohol Use Inventory. Multiple discriminant analysis was used to evaluate personality differences among the groups. It appears that at least 2 personality variables distinguish different drinking patterns. The variables of Aggression and Play (a composite variable of Affiliation, Order, and Play) appear to be indicative of alcohol use patterns and may provide a measurement of proclivity to them.
Use of the terms alcoholic and alcoholism has resulted in more obfuscation of the diagnostic issue than clarification. Jacobson (1976) stated that (this) may be an oversimplified representation of a complex multidimensional problem and the acceptance of the concept may lead to faulty understanding of etiology and treatment. He further stated that rather than continue the chimerical pursuit of a typical alcoholic or a unitary alcoholism, it would seem more reasonable and prudent to entertain the idea that there may be several alcoholisms (Jacobson, 1976). Several assessment models have been developed that differentiate specific types and dimensions of alcohol use. These models are usually constructed such that they measure level of alcohol impairment or progression of the syndrome. The model that appears to have the most reliability and validity is the Alcohol Use Inventory (AUI) developed by Wanberg et al. (1977). This inventory allows for discrimination of 16 specific patterns of alcohol use and 5 broader dimensions of alcohol use patterns with a final summary scale of General Alcoholism that indicates the degree of alcohol use impairment. Previous research has not been able to identify personality characteristics that are specific to alcohol use (National Institute of Alcohol Abuse and Alcoholism (NIAAA), 1971). NIAAA (1971) maintains that current methods of assessment (objective and projective tests) have failed to identify a common personality structure of the alcoholic patient which would be predictive of alcoholism. One possible explanation is researchers have assumed (all) alcoholics share common attributes regarded as alcoholic traits. An alternative hypothesis is that different personality types may be related to different patterns of alcohol use. Typically, research has reflected an inherent assumption that there is a single personality type for a person who is alcoholic, thereby leading to research bias (Wanberg et al., 1977). Varela (1970) offered that in a study of drinking habits, whether in alcoholics or nonalcoholics, it is possible to focus on 2 different aspects: first, behavioral facts related to the intake of the alcoholic beverages; and second, the tendencies inducing the individual to drink alcohol. NIAAA (1971) findings indicated that specific alcoholic personality traits are common to alcohol addicted individuals. However, NIAAA (1971) showed that the studies thus far have been restricted to populations already addicted and then commonly referred to as alcoholics with no further behavioral descriptions. The idea that personality is integral to predicting alcohol addiction has been noted by several researchers (Horn et al., 1974; Jones, 1968; Kissin et al., 1968). Personality tests have been considered a useful prognostic instrument for future alcohol use (Graham, 1976; MacAndrew, 1979). 107
108
RICHARD
L. WILLIAMS et al.
The most consistent problem with most research on alcoholism and personality has been a lack of a clear definition of what constitutes alcoholism. Jacobson (1976) stated that there may be several alcoholisms. The Rand Report (1976) said “there is no such thing as an alcoholic”. Further, Wanberg et al. (1977) said that there is no cross-validated set of symptoms that define alcoholism. Since it is now possible to tentatively assess different drinking patterns through the use of such instruments as the AIcohol Use Inventory, this study examined the personality of a selected group of subjects determined to have different levels of alcohol use and explored what personality differences could be elucidated as a result of different alcohol use impairment levels. METHOD
Subjects
In order to include subjects manifesting a wide range of alcoholism impairment levels, several different populations provided the subjects for this study. The original population was 190 selected subjects. They were undergraduate and graduate students at the University of Southern Mississippi, in-patients and out-patients of the Gulf Coast Mental Health Center, and the professional staff of the Gulf Coast Mental Health Center. The final sample was composed of 166 subjects. Fifteen subjects had infrequency scores greater than 1 on the Personality Research Form (PRF) and were eliminated from the population. A score greater than 1 indicated an implausible or pseudo-random response pattern. Five subjects were found to be non-drinkers; 4 did not complete the questionnaires. Eighty-five males and 81 females were distributed equally throughout the alcohol use impairment levels. Procedure. All subjects were administered the Alcohol Use Inuentory and the Personality Research Form. The 166 subjects were classified into 6 groups representing alcohol impairment levels using the raw scores of the Alcohol Use Inventory. These 6 groups were classified as group l-Social, group 2-Light Moderate, group 3-Moderate, group &Light Heavy, group S-Heavy, and group 6-Alcoholic. A multiple discriminant analysis separating the 6 groups was based on the 15 PRF scores. All raw data for all 15 PRF scores were derived, including 10 demographic and 2 alcohol variables, for all 166 cases. A discriminant analysis and univariate F-ratio were computed for each of the 15 PRF scales. Group means and standard deviations for each AU1 group were calculated. Correlation of personality variables with the discriminant functions was established to determine contribution among variables. Additionally, discriminant scores functions were computed. RESULTS
The strategy used in the analysis comes from Bargmann (1969). Bargmann emphasized interpretation of just the first discriminant function in that “the second discriminant function would qualify as best after elimination of the effect due to the first one”. Putting emphasis on the manipulation of observable variables rather than artificial mathematical variables appears to be more productive. He suggested that it might be “preferable to make a separate discriminant analysis (first eigenvector) on the observable variates which Table Discriminant function
1. Results Canonical correlation
F significance
0.542 0.443 0.358 0.300 0.156
1
2 3 4 5 *Degrees
of discriminant
of freedom
are approximate.
1.756 1.326 1.008 0.752 0.335
analysis
of all PRF variables Degrees of freedom* 751704 561574 391438 241295 1 l/148
Probability 0.005 NS NS NS NS
(P)
Alcohol Table 2. Correlation
of personality
Variable first run
All variables
AC Af
0.113 -0.171 0.681 0.073 0.028 0.067 0.080 - 0.037 0.082 -0.193 0.312 - 0.284 0.153 -0.123 - 0.059 0.542
Ag Au Do En Ex Ha Im Nu Or Pl Sr Un In Rc
109
personalities variables
with the discriminant
functions
First function second run
Second function second run
Third function* second run
0.037 0.336
0.518
0.862
0.096 0.142 0.331 0.268 -0.202 0.104 0.208 - 0.388 - 0.678 0.038 0.205 - 0.007 0.458
0.066 -0.175 -0.164 -0.152 0.442 - 0.440
0.464 - 0.482
-0.184 0.381
0.334
*Function indicates that on each successive computer run the variables with plus or minus 0.20 correlations were omitted to determine those variables contributing to significance.
.9 .6
3 0 -3 -6 -9
Fig. 1. Plot of discriminant
scores contrast,
first run, Aggression.
were eliminated because of their low correlation versus the total discriminant function” (Bargmann, 1969). The results of the direct discriminant functions analysis for all PRF variables are presented in Table 1. Only the first of the 5 possible discriminant functions was significant [F(75,704) = 1.756, P < O.OOl]. The canonical correlation for that function was 0.542, indicating a strong relationship between the group and personalify vectors. The correlations of each personality variable with the discriminant function is shown in Table 2. The personality variable of Aggression dominates the first function with a correlation of 0.681. The closest correlation to this is that of Order (0.312) which is less than half that of Aggression. The group effects (Discriminant Score Contrast) are presented in Fig. 1. Aggression begins low in group 1, increases to groups 3 and 4, and then declines. Table 3 presents a summary of significance of first discriminant functions for each successive analysis for the remaining 14 PRF variables after removal of Aggression, Table 2 depicts correlation of personality variables with these discriminant functions. Table
3. Summary
of tests of significance of first discriminant analysis after removal of Aggression
Run
Number of variables
1 2 3
14 8 3
*Degrees
of freedom
Canonical correlation 0.458 0.381 0.334
are approximate.
F significance 1.341 1.550 2.037
function
Degrees of freedom* 701704 401670 151437
for successive
Probability (P) 0.05 0.05 0.02
110
RICHARD L. WILLIAMS et al. Table 4. Univariate F significance*
Variable AC Af
Probability
0.494 3.385 6.787 1.389 0.486 2.101 1.414 1.031 0.69 1 1.206 1.906 4.050 0.675 1.079 0.265
Ag Au Do En Ex Ha Im NU Or Pl Sr Un In *Degrees
F tests
of freedom
(P)
NS 0.00 1 0.00 1 NS NS NS NS NS NS NS NS 0.001 NS NS NS
= 5/160 for all F’s
Variables correlating less than 0.20 with the discriminant functions were dropped at each successive step. The resulting function received substantial correlation from Affiliation (0.852) Play (-0.482) and Order (0.464). The canonical correlation for this function was 0.433, indicating a prominent relationship between the group and personality vectors. The univariate group comparisons are presented in Table 4. In this analysis, 3 varisignificant : Affiliation [F(5,160) = 3.385, P < O.OOl], Aggression ables were [F(5,160) = 6.787, P < O.OOl], and Play [F(5,160) = 4.050, P < O.OOl]. The group effects (Discriminant Score Contrast) are presented graphically in Fig. 2 and indicate as level of alcohol use impairment increases, Affiliation and Order decrease. Although not shown, the variable of Play is inversely related to Affiliation and Order and increases from group 1 through 6. Thus, Play increases as Affiliation and Order decrease. DISCUSSION
The present study revealed that the personality types of the 6 different alcohol use impairment groups were significantly different (P < 0.01 and P < 0.05) on the variables of Aggression and Play. While the analysis of these variables revealed sig‘nificant differences, several other factors merit further consideration. The personality variables of Affiliation and Order approached significance and as such were strong contributors to the variable Play, even though they were inversely related to it. The interplay of Aggression and Play in the six AU1 groups is important in terms of understanding the personality dynamics as the groups increased in level of alcohol use impairment. The personality trait adjectives for Aggression are: aggressive, hostile, quarrelsome, irritable, argumentative, threatening, attacking, and retaliative (Jackson, 1967).
6
4
.-•
2
\
0
.l.
-2
\.
-4
\
.
-6 I
2
Fig. 2. Plot of discriminant
3 scores contrast,
4 second
5 run, Affiliation,
6 Order.
Alcohol
111
personalities o- Affihtion l - Aggression +-Order
16 I5 14 13 12 IO 9 8
6 5 4 3
I Fig. 3. Summary
2
of group
3
means for Aggression,
4
5
Affiliation,
6
Order
and Play for all groups.
o- Ploy l
12
II IO 9
- Agqesslon
cl
0
8 7 6 5 Malt%
Fig. 4. Summary
of group
kmoles
Totol
means for the variables Play and Aggression total subject population.
for males, females and the
At the other end of the personality spectrum is the trait of Play. Play is described as: playful, jovial, jolly, pleasure-seeking, merry, fun-loving, carefree, and blithe (Jackson, 1967). Figure 3 depicts the relationship of these variables as alcohol use impairment increases from group 1 through group 6. Both Aggression and Play increase initially as an individual displays more problems associated with alcohol consumption. At the level of Light Moderate drinkers, the Play variable begins to decline slightly as Aggression rises sharply to Moderate drinkers. In other words, the initial drinking patterns seem to be dominated by the personality trait of Play. Aggression is very low initially, even though it increases rapidly to group 3. Play remains more prominent than Aggression in all groups. Aggression and the composites of Play, Affiliation and Order, decline towards the Light-Heavy drinking group. Figure 4 indicates that there is almost no difference between males and females on the Play variable when compared on sex or to the total population. Aggression appears more pronounced in males than in females and the total sample. MacAndrew (1979) reported on his scale to determine alcoholism that “whatever the scale is tapping, it appears not only to be relatively stable, but quite independent of the immediate sequelae and at least somewhat independent of the longer term consequences of substance abuse”. It therefore seems that the personality differences detected may be independent of alcohol use per se and reflect the type of person who might easily fit into a specific consumptive pattern. It is conceivable that different personality types drink in specific patterns. If so, it is reasonable to postulate that there are several alcoholic personalities. These alcoholic personalities appear to be directly related to consumptive patterns and, as such, may be predictive of them.
112
RICHARD L. WILLIAMS et al. REFERENCES
Bargmann, R. E. Exploratory techniques involving artificial variables. In P. R. Krishnaiah (Ed.), Multioariate Analysis--II. New York: Academic Press, 1969. Graham, F. M. An analysis of the alcoholic personality as evidenced by the 26 personality factor questionnaire; an artempt to predict proclioity to alcoholism. Unpublished doctoral dissertation, University of Southern Mississippi, 1976. Horn, J. L., Wanberg, D. W. & Adams, G. H. Diagnosis of alcoholism: factors of drinking, background and current conditions in alcoholics. Quarterly Journal ofStudieson Akohol, 1974, 35, 147-175. Jackson, D. N. Personality Research Form Manual. Goshen, NY: Research Psychological Press, 1967. Jacobson, G. R. The Alcoholisms; Detection, Assessment, and Diagnosis. New York: Human Services Press, 1976. Jones, M. C. Personality correlates and antecedents of drinking patterns in adult males. Journal of Consulting and Clinical Psychology, 1968, 32, 2-12. Kissin, B., Rosenblott, S. M. & Machover, S. Prognostic factors in alcoholism. Psychiatric Research Reports, 1968, 24, 2243. MacAndrew, C. On the possibility of the psychometric detection of persons who are prone to the abuse of alcohol and other substances. Addictive Behaviors, 1979, 4, 1 I-20. National Institute of Alcohol Abuse and Alcoholism. Alcohol and Health, First special report to the US Congress, DHEW, Rockville, MD. 1971. Rand Report. Alcoholism and Treatment. Prepared by D. J. Armor, J. M. Polich & H. B. Stambul. Santa Monica, CA: Rand Corp., 1976. Varela, A. The role of personality in alcohol abuse. In R. R. Popham (Ed.), Alcohol and Alcoholism. Toronto: Addiction Research Foundation, 1970. Wanberg, K. W., Horn, J. L. 8~ Foster, M. A differential assessment model for alcoholism: the scales of the alcohol use inventory. Quarterly Journal of Studieson Alcohol, 1977, 3, 512-543.