ALCOHOL SOCIOLOGY HOME PAGE
Drinking and Drinking Problems:
Some Notes on the Ascription
of Problems to Drinking
When Keller (1960) first outlined a definition of alcoholism that was specially tailored to the needs of survey studies, he elaborated a set of dimensions that prefigured the operational measures of drinking problems in subsequent studies. In Keller's view, however, the things to be measured were not just disjointed bits of information about the prevalences of various alcohol-related problems, but rather they were to provide a variety of indicators of alcoholism that, taken together, would allow the analyst to decide whether or not a particular drinker was an alcoholic. In short, Keller, like Jellinek (1952) before him, wanted to distinguish between instances of alcohol problems that were disease-based and instances which were merely misbehaviors involving alcohol. Keller's orientation, in other words, was to use the definition he provided as the basis for operational measures of a putative addiction to alcohol, and the idea of addiction provided the conceptual link between drinking problems and the drinking itself.
This focus on the need to discriminate addiction- or disease-based drinking problems from nonaddictive or nondisease problems did not always travel well into epidemiological research. Knupfer (1967), for example, referred to respondents who positively responded to drinking problem measures as "problem drinkers," and she pointed out that her use of the term "problem drinker" rather than "alcoholic" was not accidental: "We wish to avoid getting into the question, 'what is a real alcoholic?,' or 'does the person have the disease called alcoholism?' We take the point of view that a problem -- any problem connected fairly closely with drinking constitutes a drinking problem."
There were, perhaps, a variety of reasons for this departure. First of all, no consensual definition of alcoholism was available to provide for the discrimination of alcoholics (Bailey, 1966). Hence, survey researchers tended to use an ecletic array of problems indicators drawn from a variety of perspectives. Eclecticism is itself a methodological tradition in survey research, in part because such projects are often too expensive to be committed to one or two definitions of the "dependent variable" in situations where many more are available. Also, although the social control and management of particular individuals might be much influenced by the claim that the behavior was disease caused, the problem of diagnosis was not central to survey studies. No respondents' fates hung on the discriminations. Moreover, survey researchers were disturbed by the number of cases that reported serious drinking problems but did not conform very well to any description of the alcoholism syndrome. Mulford (1968), for example, found that the strict application of Jellinek's definition of gamma alcoholism virtually eliminated all cases from a general population sample. Finally, survey research raised doubts about the definitional procedure which analytically disregarded the vast majority of the reported drinking problems in their samples. If "alcoholism" was, after all, to provide the master concept for a theoretical account of problematic behaviors associated with drinking, was it not unfortunate that this theoretical approach would deal only with a small "span" (Stinchcombe, 1974) of the phenomena to be accounted for (Clark and Cahalan, 1973)? In a sense, then, the drinking problems approach grew out of efforts to operationalize various dimensions of the alcoholism concept without assuming that alcohol addiction was the source.of those problems. It was an argument for greater dependence on empirical findings but also included the tendency to play down the moral and political arguments for viewing alcoholism as largely a disease process. Thus, methodological and situational factors as well as empirical findings from epidemiological studies led some researchers to look for other theoretical frameworks which would explain drinking problems and yet not rely so heavily upon addictive alcoholism as a part of that explanation.
But a fundamental problem was inherent in this departure: once drinking problems were conceptually separated from a presumed underlying addiction, in what sense should they be considered drinking problems? From a research standpoint, calling something a "drinking problem" seemed to imply that drinking was somehow an independent causal force in the problem or that the hunt for an adequate theoretical account of drinking problems was best organized around the study of drinking (cf. Gusfield, 1974). Of course, many problematic behaviors which were labeled "drinking" problems in alcohol surveys could be classified under many different headings by different observers: so, for example, a respondent who "spends too much money on drinking" might be seen as manifesting a "budgeting" problem, an "income" problem, a "guilt" problem, or a problem finding other things to spend his money on. Any one of these alternatives could form the focus for a study of the same behavior.
It was recognized early on that the "drinking" in "drinking problems" sometimes referred to quite different features of alcohol use. Some problems such as arrests were "drunkenness problems" while others such as "health problems" were often due to years of heavy drinking (Knupfer, 1960). Analyses of the interrelationships among different problems also tended to show that problems drawn from different spheres of life did not always highly intercorrelate with one another (Cahalan and Room, 1974). Finally, subtracting the notion that an addiction to alcohol provided the connecting link between drinking and drinking problems begged a number of additional questions: Why did the respondent do the drinking that was ostensibly causing a problem for him? Were these "problems" in fact considered problematic by the respondent? Did these "problems" outweigh the pleasures of drinking or the pain of altering a desired drinking pattern? In short, once it was recognized that addiction was not the only sort of"glue" that could hold a person in a pattern of behaviors that produced drinking problems, the explanation of problems was opened up to an indefinite number of potential causal scenarios. And not all of these scenarios focused on the drinking dimensions of drinking problems.
One result of the above consideration has been a tendency to regard the term "drinking problems" as referring to "a problem which some observer has ascribed, wholly or in part, to some aspect of drinking." In some areas the appropriateness of this causal ascription is reasonably clear. For example, one can estimate the contribution of drinking to mortality, and do such studies without much regard to the opinions of drinkers concerning the danger associated with various drinking patterns. In other areas, however, the process of ascription is problematic. Some drinking problems are ascribed by the analyst, as when a particular pattern of consumption is called, say, heavy drinking. Some seem to presume the universality of certain preferences, as when "feeling guilty about drinking" is called a problem on the assumption that people would rather not feel guilty about their activities. Still other drinking problems may involve multiple layers of ascription. When someone ascribes the loss of a job to drinking, the researcher may call this a drinking problem, making the assumption that the respondent would have preferred to keep the job. The respondent himself may have made an ascription about the reasons for his boss's action -- which might be thought of as an ascription of an ascription.
These ascriptions are infused with the respondent's "account" (Scott and Lyman, 1968, p. 46) of the events he experienced in the sense that they provide a linguistic device which can be employed when an action is subjected to valuative inquiry. Ascriptions may also be thought of as "quasi-theories" (Hewitt and Hall, 1973) or ad hoc explanations with which the respondent (or the analyst) brings order to the experiences he has reported. Such accounts are thoroughly entwined in the cultural definition of alcohol and the acceptability of explanations based on drinking in the respondent's cultural setting. Thus, for example, an account based on drunkenness among men of the Urban African Township of Rhodesia will not exculpate the actor or reintegrate the situation because drunken behaviors are there considered intentional (May, 1973).
The influence of variations in ascriptive behaviors is one of the little attended. to subjects in the epidemiology of drinking problems. This inattention may be due in part to the lingering influence of the concept alcoholism. In the classical alcoholism perpsective, variations in ascriptive practices such as those described above must be viewed as a kind of error associated with the reporting of problems; error which obscures the dividing line between "real" alcoholics and other nonalcoholics.
If, on the other hand, one considers the ascriptive component of drinking problems to be a worthwhile issue in its own right, the processes and variations in ascriptions are not seen merely as a source of error, but rather as a call to investigate the relationship between drinking and problems ascribed to drinking. As a tentative first step in this direction, I would like to turn now to a brief analysis of the relationship between consumption and one drinking problem area, "spouse problems" associated with drinking.
The selection of only one drinking problem and the choice to use "spouse problems" requires a word of explanation. As mentioned earlier, different drinking problems often have different correlates. Consequently, adding problems together and treating the sums as a measure of the degree of "problem drinking" in a given respondent may obscure the patterns of association between a single problem and a set of other variables. The use of spouse problems, rather than one of the dozen or so remaining problem scales in our surveys, is based on the fact that spouse problems are the most prevalent single problems reported in general population surveys. And in addition to providing an adequate number of cases, they are the sorts of problems that proffer at least a couple of levels of ascription.
Drinking and Spouse Problems
Perhaps the simplest way to examine the relationship between drinking and drinking problems is to take a look at the bivariate relationships between spouse problems (in this case) and a number of measures of consumption practices. Nonabstaining married men in our panel sample of men in the U.S. population were asked a variety of questions about the responses of their wives to their drinking. These responses are here scaled into a simple three-level index: "No problem," a "mild-level problem" (the spouse "showed concern?' over the respondent's drinking or "indicated that he should cut down"), and a "higher-level problem" (the wife got angry about the respondent's drinking, had threatened to leave or kick him out because of his drinking, a separation actually occurred, or the respondent reported that his drinking was having a harmful effect on his marriage).
Respondents were also questioned in detail about their consumption practices. These measures have been converted into a variety of consumption scales, of which five are used in this analysis: (1) overall volume of drinking, (2) highest frequency of drinking, (3) mean quantity of drinking at a sitting, (4) frequency of "getting high or tight," and (5) a "current intake scale" which combines several of these dimensions. The operational definitions of these scales are described in Figure 1.
The data for this analysis are drawn from a national panel sample of adult men (aged 21-59 at the time of the first interview) interviewed first in 1969 and again in 1973. A description of the Time I sample can be found in Cahalan and Room (1974), and a description of the follow-up sample in Cahalan and Roizen (1974) and Friedman (1974). For the purposes of this analysis it is necessary to mention only a couple of points about the data: first, we will be looking at the set of men who at Time I were current nonabstainers (i.e., reported that they drank at least as often as once a year in the past year), currently married, and who later responded to the second wave of the survey. This group (N=513) amounts to 71 per cent of the full panel sample and 52 per cent of the full Time I sample (which included respondents who were not reinterviewed or died by the time of the second wave). Secondly, the consumption measures and spouse problems used in this analysis as "current" measures refer to relatively recent rather than long past events and behaviors. The definition of the current frame, however, is somewhat different for the consumption and the spouse problems scales. Questions about drinking practices are usually phrased in the present tense and refer to behaviors that stretch back over no more than a year's time--detailed information about more distant drinking practices is likely to be more unreliable. The spouse problems questions, on the other hand, refer to events within the "past three years." The use of a 3-year time frame for spouse problems is based on the inherent infrequency of such events as separations or divorces, so that a somewhat greater time period is allotted in order that these events can be caught. The discrepancy in time frames is, of course, a potential source of some correlative independence between problems and consumption.
The bivariate relationships between spouse problems and the five consumption scales are shown in Table 1. As I mentioned earlier, unmarried and abstaining men have been removed from the tables so that all of the remaining cases are "at risk" of a "spouse problem."
Among these respondents, 75 per cent reported no current spouse problem, 9 percent a mild-level problem, and 16 percent a higher-level problem. The bivariate sub-tables show correlation coefficients ranging from a low of .231 (spouse problems by frequency of drinking) to a high of .407 (by "current intake"). None of the consumption scales, alone, provides an efficient "sort" of the spouse problems scores such that a "true positive"/"true negative" split on spouse problems can be drawn by cutting the consumption distributions at one place or another. Instead the patterns show a scattering of cases throughout the cells which is sufficiently spread for some "high consumption" respondents to nevertheless report no spouse problems and some "high spouse problems" respondents to nevertheless turn up in the lowest consumption category. Where the conditional probability of reporting a high spouse problem, based on scoring high on a consumption measure, is highest ("Score 3" on the "current intake" scale), the stringency of this category is such that only 10 percent of the high spouse problems group is thereby captured.
Consumption measures that tapped the quantity of drinking per sitting ("current intake," "frequency high and tight," and "mean quantity per sitting") show stronger associations with spouse problems than do "frequency of drinking" or "overall volume." The relatively weak contribution to problems by frequency and overall volume of drinking is shown in Table 2, a stepwise regression of spouse problems on the five consumption variables. Frequency and volume entered last and next to last, contributed almost nothing to the multiple correlation coefficient, and showed negligible standardized regression coefficients. Although the skewedness of the distributions on which the correlations were calculated argues against overinterpreting the regression, most of the association between problems and consumption is provided by "current intake" and "frequency of getting high or tight," which suggests that periodic drunkenness is commonly associated with spouse problems. The multiple correlation coefficient climbs only a little through the addition of new consumption variables, suggesting, of course, that the influences of each of these variables is only very slightly additive.
The cross-sectional associations and the multiple regression, in short, show a moderate association between drinking and problems while leaving a considerable portion of the variance for nondrinking factors. The stock account for these patterns is, of course, the degree of normative variation in conforming drinking in the U.S., variations which are fully broad enough to permit both some unsanctioned heavy drinking and some relatively light drinking which nevertheless elicits problems. In a sense, then, these cross-sectional tabulations can be seen as broadly sketching in the conditional probabilities of spouse problems with no controls for variations in norms and contexts. We infer that the correlative independence of consumption and problems is in part a sign of contextual variations, but the tables themselves, of course, do not provide that conclusion.
In order to more closely assess the influence of drinking on drinking problems it is desirable to have a means for varying drinking while controlling for the influence of contextual and other nonconsumption factors. Fortunately, the panel data provide one way to approximate such a control.
Changes in Drinking and Changes in Spouse Problems
Table 3 presents the bivariate relationships between changes in consumption scales and changes in spouse problems. The sample size in these tables has been reduced by the number of respondents who were divorced, separated, or widowed by the time of the second wave of interviewing.
On first appearances, these tables seem remarkable for the lack of apparent relationships between changes in drinking and changes in spouse problems. In fact this outcome is part "finding" and part artifact. At least three artifactual influences should be pointed out: First, at least half, and often a considerably greater portion of the sample, reports no change on each of the variables in Table 3. Consequently, these tables tend to show a high concentration of cases along the "cross" defined by the "no change" column and the "no change" row. Such patterning is of course a strong damper on the product moment coefficient, although the patterning is noteworthy in that it shows that the modal group of changers in each category of spouse problems change is usually made up of respondents who have not changed their drinking practices so much that these consumption measures would register the change. The second artifactual influence derives from the disparities in the marginals between most of the consumption scales and the spouse problems scale. In all but one case, rather more respondents show changes on consumption than show changes on spouse problems. Hence, many changers on consumption are "forced" into the 'no change' row of the spouse problems scale. This, too, can be regarded as "the way the data came out" save for the fact that marginals themselves are premised on arbitrary cutting points. Therefore, the tendency for correlation coefficients to be lowered by a large discrepancy in the number of changers on either axis of the table is more a byproduct of cutting points than levels of association. Finally, as in most cross-tabulations of change scores, there are "ceiling and floor" effects to be reckoned with. Some of the respondents who reported a high level of consumption and no spouse problem at Time 1, for example, may have dropped to a low level of consumption, but could not drop lower on the spouse problems scale. The status of these effects is equivocal but the relatively frequent occurrence of respondents who are high on consumption without reporting problems implies that they have a significant influence on the correlation statistics.
A stepwise regression of change in spouse problems on change in consumption generated a nonsignificant multiple correlation coefficient that claimed to explain only about 4 per cent of the variance. A second regression which removed all ceiling and floor cases from the sample improved the explained variance to about 9 percent. The altering of cutting points might add a bit more, but the pattern in the tables is not such that it would generate a much higher level of multiple correlation.
Even discounting for the influences of artifacts, there seems to be a good deal of correlative independence between "drinking," as these scales have measured it, and this "drinking problem." What should we make of this?
Calling something a "drinking problem," of course, has a number of important implications and functions for research. By defining the phenomena that will be focussed on, it has a large share in determining the sorts of data that will be collected, the literatures used, the outlets for publication, the audiences that may be watching, the agencies that may support it, and the shape of the intellectual and policy outputs that are sought. In many instances, research associated with social problems takes its "problem definition" from the prevalent cultural definition of the "problem" and draws,its support from the agencies whose boundaries have been marked out by that same definition. From a research standpoint, however, the utility of a particular conceptualization is borne by the utility (or potential utility) of the theoretical propositions of which that conceptualization is a part. And that utility is ultimately tied up with the amount of order and comprehensibility that is brought to a domain which was puzzling or chaotic beforehand.
The "drinking problems" concept is itself an evolutionary product of the disease conceptualization of alcoholism and the methods and findings of epidemiological studies. This analysis has suggested that the concept sometimes has a strong ascriptive component, and future studies of the epidemiology of drinking problems might benefit from careful attention to the cultural and interpersonal mechanics at work in the ascription of problems to drinking.
Bailey, Margaret B. (1966), "Some Issues in Epidemiologic Surveys of Alcoholism," Presented at the Annual Meeting, American Public Health Associationi, Epidemiology Section, San Francisco, California, November 22 1966.
Cahalan, Don (1970), Problem Drinkers, San Francisco: Jossey-Bass.
Cahalan, Don and Ronald Roizen (1974), "Changes in Drinking Problems in a National Sample of Men," Presented at the North American Congress on Alcohol and Drug Problems, San Francisco.
Cahalan., Don and Robin Room (1974), Problem Drinking Among American Men, New Brunswick, N.J.: Rutgers Center of Alcohol Studies.
Clark, Walter and Don Cahalan (1973), "Changes in Problem Drinking over a Four-Year Span," Presented at the annual meeting of the American Public Health Association, San Francisco, November.
Edwards, Griffith (1973), "Epidemiology Applied to Alcoholism," Quart. J. Stud. Alc. 34:28-56.
Freedman, Gail (1974), "A Technical.Report on the National III -National IV Longitudinal Study," Social Research Group Working Paper F33.
Gusfield Joseph (forthcoming), "The Prevention of Drinking Problems," in W.J. Filstead, J.J. Rossi, and M. Keller (eds.), Alcohol: New Thinking and New Directions, Chicago: Ballinger, in press.
Hewitt, John P. and Peter M. Hall (1973), "Social Problem, Problatic Situations., and Quasi-Theories," American Sociological Review, 38:367-374.
Jellinek., E.M. (1952), "Phases of Alcohol Addiction," Quart. J. Stud. Alc. 13:673-684.
Keller Mark (1960), "Definitions of Alcoholism," Quart. J. Stud. Alc. 21:125-134.
Knupfer, Genevieve (1960), "Use of Alcoholic Beverages by Society and its Cultural Implications," California's Health 18:17-21.
Knupfer, Genevieve (1967), "The Epidemiology of Problem Drinking," Amer. J. Public Health, 57: 973-986.
May, John (1973), Drinking in a Rhodesian African Township, Occasional Paper No. 8, Dept. of Sociology, University of Rhodesia, Salisbury.
Mulford, Harold (1968), "'Alcoholics', 'Alcoholism', and 'Problem Drinkers': Social Objects-in-the-Making: The 'Experts' Contribution," Presented at the 28th International Congress on Alcohol and Alcoholism, Washington, D.C.
Scott, Marvin B. and Stanford M. Lyman. (1968), "Accounts," American Sociological Review, 33:46-62.
Stinchcombe, Arthur L. (1974), "Theoretical Domains and Measurement, Part II," Acta Sociologica, 16:76-97.