Further Reading

Alcohol & Alcoholism, 32, 51-64, 1997.

Utilizing Culture and Behaviour in Epidemiological Models of Alcohol Consumption and Consequences for Western Nations

Stanton Peele
Fellow, The Lindesmith Center, New York



Cultural differences in alcohol consumption are inescapable, but have been difficult to establish as predictor variables in epidemiological models. With respect to dependent variables, the behavioural outcomes of alcohol use have not been operationalized as successfully as the health outcomes. This study examined cultural differences in drinking by employing Levine's distinction between Temperance and non-Temperance cultures, along with other cultural, consumption, and policy predictor variables, among 21 Western countries. Dependent variables included the prevalence of Alcoholics Anonymous (AA) groups (as a measure of behavioural and social problems) and a range of alcohol consumption and health measures. Level of consumption was an important determinant of the health consequences of drinking among Western nations, but not so important in determining behavioural outcomes. Culture, on the other hand, is largely determinative of behavioral outcomes and also quite critical for some health outcomes. An inverse relationship between alcohol consumption and AA membership strongly indicated that consumption is modified by cultural styles in producing drinking behaviours. Temperance cultures, which are largely Protestant, have far more AA groups and higher rates of coronary heart disease mortality, but lower cirrhosis mortality. Overall mortality does not vary according to national alcohol consumption or cultural distinctions. The percentage of alcohol consumed as wine is a strong inverse predictor of mortality in the 55-64 age group, but the change in absolute national wine consumption is directly associated with overall all-age mortality. In conclusion, religious and cultural distinctions among Western nations strongly predict behavioural drinking problems and also enhance the prediction of death rates from diseases related to alcohol consumption. Social engineering techniques which attempt to modify well-established cultural drinking practices can have counterproductive results.


This study represents an attempt to operationalize cultural factors in the epidemiology of alcohol use and to measure the impact of such cultural variations on an expanded range of behavioral as well as health outcomes. The need for an epidemiological model that (a) gives proper weight to culture as a predictor variable, and (b) considers behavioural as well as health consequences, arises from both the successes and the limitations of recent research which relates alcohol consumption to disease incidence and mortality.

Both case comparison and cohort epidemiologic research, as well as cross-cultural analysis, have now firmly established that alcohol reduces coronary heart disease (CHD) incidence and mortality (Criqui and Ringel 1994; Gaziano et al. 1993; Klatsky et al. 1992; Rimm et al. 1991; Stampfer et al. 1988; Suh et al. 1992). Prospective epidemiological studies also find that overall mortality is reduced by moderate alcohol consumption (Boffetta and Garfinkel 1990; Doll et al. 1994; Fuchs et al. 1995; Grnbæk 1994; Klatsky 1992). These benefits occur primarily for middle-aged men and women, for whom heart disease is the primary cause of death. However, they also apply to all adults at risk for heart disease, a substantial majority of both female and male adults (Fuchs et al. 1995).

Impressive as these findings are, to have real policy implications they must be considered in the light of an older body of research on cross-cultural differences in drinking styles. Some cultures, notably Mediterranean and other wine-drinking societies, unquestionably socialize the use of alcohol more effectively and display fewer behavioural drinking problems than other cultures (Blum and Blum 1969; Lolli et al. 1958; Maloff et al. 1982). These differences are reflected in epidemiological and community studies through lower rates of drinking problems and alcoholism (Cahalan and Room 1974; Glassner and Berg 1980; Greeley et al. 1980). Vaillant (1983), for example, found that Irish-Americans in an urban Boston setting were seven times as likely to become alcohol dependent over their lifetimes as were Mediterranean-Americans (Italian, Greek, and Jewish). Yet, quantitative cross-cultural research has rarely found systematic differences in drinking behavior (see Whitehead and Harvey 1974).

The question that remains is how culture, as a predictor variable, fits into an overall epidemiological model. To define the relevant cultural differences solely by the consumption patterns (i.e., amount and style) that different cultures have developed historically is tautological. It begs the question of whether the health and behavioural consequences of alcohol use in a given culture are attributable primarily to such consumption patterns or to some larger cultural gestalt. This is a crucial determination to make at the policy level, since it tells us whether it is possible to replace one culture's consumption patterns with patterns that have been more successful in other cultures.

Predictor Variables

The current study employs the cross-cultural epidemiologic model which Criqui and Ringel (1994) used to analyze the impact of alcohol consumption on heart disease and overall mortality independent of diet. The primary predictors examined are consumption and culture. Two variables are utilized to capture the overall character of a culture. One is religion. The other is Levine's (1992) concept of Temperance versus non-Temperance cultures. In regression analyses, 1990 consumption and Temperance were analyzed first; other consumption data and/or policies such as taxation were then considered to see if they provided a better fit.

Temperance cultures are strongly concerned with alcohol abuse and maintain activist approaches to combating drinking problems, because ostentatious behavioural drinking problems are more apparent in these societies than in non-Temperance cultures (Levine 1992; Peele 1993). Temperance countries consume less alcohol, and a smaller proportion of their beverage alcohol as wine, than non-Temperance cultures (Levine 1992). Temperance cultures also have significantly higher rates of CHD (Peele 1993), due to a strong inverse association between societal alcohol consumption and CHD (LaPorte et al. 1980). However, the increased incidence of accidents, cirrhosis, and cancer resulting from higher levels of alcohol consumption counterbalance the cardiovascular benefits of drinking in terms of overall mortality (Boffetta and Garfinkel 1990; Klatsky et al. 1992), which may also neutralize the benefits of higher alcohol consumption cross-nationally (Criqui and Ringel 1994).

The Temperance/non-Temperance distinction represents, of course, just one way of operationalizing the impact of culture. Other conceptualizations have been attempted, as exemplified by Heath's (1984, 1995) efforts to identify cultural factors in drinking behaviour. As an anthropologist, however, Heath has not systematized these variables, or categorized cultures or societies, so as to incorporate cultural analysis within a comprehensive epidemiological framework. By contrast, Levine's classification lends itself to epidemiological analysis. The classification of particular cultures in his model may be questioned and objections based on historical or geographical complexities may be raised. However, fine-tuning (see the shifting of Ireland from the non-Temperance to the Temperance category as outlined in the Methods section below) do not change the major findings obtained by using this model (cf. Peele 1993).

Dependent Variables

The current study also expands the scope of the dependent measures in Criqui and Ringel's model by examining behavioural consequences (accidental deaths, suicides, murder, and Alcoholics Anonymous membership) along with health outcomes (e.g., deaths from stroke, cirrhosis, cancer, and hypertension as well as heart disease and total mortality). The study focusses primarily on AA membership and health measures previously related to alcohol, namely CHD and cirrhosis.

Until now, the behavioural and social problems resulting from alcohol use have not been operationalized as successfully as the health outcomes. The use of AA groups per million population in the current study for this purpose is a response to this methodological challenge. AA group membership has previously been analyzed qualitatively and quantitatively by Levine (1992) and Mäkelä (1990). Whereas estimates of the number of individual AA members at a given place and time fluctuate and may be inflated, AA groups are a more stable measure. At a minimum, AA groups express a concern about drinking behaviour. But the huge cross-cultural variations in AA groups per capita that emerge from this study appear to capture real behavioural phenomena as represented by the amount of public drunkenness commonly observed in these societies.

The current study examines the following clusters of potential relationships between culture and drinking, health, and alcohol policy, in some cases offering tests of specific hypotheses:

  1. Do non-Temperance cultures have less affinity for AA? Does the Temperance distinction help to explain the negative association between alcohol consumption in a society and AA groups in that society?
  2. Do drinking patterns in non-Temperance cultures enhance the health benefits of alcohol and buffer the negative consequences of drinking? Do cultural variables enhance the predictive value of alcohol consumption on health outcomes in these societies? Which sources of mortality are higher in Temperance cultures and are positively associated with alcohol consumption? Do non-Temperance and higher alcohol-consuming cultures have lower overall mortality? Are the same general relationships between alcohol and various non-CHD disease states discovered in within-nation epidemiologic studies evident in cross-cultural analyses?
  3. Are alcohol policies aimed at lowering overall consumption or at changing the balance of consumption from one type of beverage alcohol to another associated with greater or lower mortality rates for various diseases?

The current study does not examine the relationship between the institution of alcohol control policies and changes in health outcomes. Rather, these policies are examined for their relationship at one point in time with health outcomes.


The Relationship of Religion and Alcohol Consumption

The 21 primary countries in this analysis were classified as either predominantly Protestant or Catholic (there were four countries in which neither religion was dominant) (see Table 1). Catholic countries consume significantly more alcohol than Protestant (or neither religion dominant) countries (F=6.76, 20 df, p=<.01). The correlation between percentage Catholic and total consumption is also highly significant (r=.64, p=<.005). Catholic nations consume twice the percentage of their total beverage alcohol in the form of wine as do Protestant countries (F=5.78, 20 df, p=<.05). This difference narrowed marginally, and the percentage wine consumed declined in Catholic countries and increased in Protestant countries from 1980 to 1990 (F=5.04, 19 df, p=<.05; missing data for Iceland). Differences between Temperance and non-Temperance cultures are more significant than those based solely on religion, in terms both of total alcohol consumption (p=<.001) and percentage consumed as wine (p=<.001) or spirits (p=<.005), but not beer.

Table 1. Temperance distinctions, alcohol consumption, and AA membership
Temperance Countries Non-Temperance Countries
Country Consumptiona % Protestantb AA Groupsc Country Consumptiona % Protestantb AA Groupsc
Iceland 3.9 95 (P) 784 Netherlands 8.4 24 (N) 12
Norway 4.0 94 (P) 28 Italy 8.6 2 (C) 6
Sweden 5.5 95 (P) 33 Denmark 9.8 83 (P) 22
Canada 7.1 32 (N) 177 Belgium 9.9 3 (C) 53
Ireland 7.2 4 (C) 201 Portugal 9.9 2 (C) 1
U.S. 7.2 56 (P) 164 Spain 10.4 1 (C) 8
U.K. 7.6 49 (P) 51 Switzerland 10.8 48 (N) 22
Finland 7.8 84 (P) 110 Austria 11.5 6 (C) 92
New Zealand 7.8 47 (P) 102 Germany 12.6 44 (N) 26
Australia 8.3 51 (P) 56 France 13.2 0 (C) 7
Luxembourg 13.6 3 (C) --
6.6 61 170 10.8 20 25
+/- 1.6 +/- 31 +/- 224 +/- 1.7 +/- 27 +/- 28
a  1990 liters beverage alcohol consumed per capita annually
b  dominant religious group in parentheses (P=Protestant, C=Catholic, N=Neither)
c  1991 AA groups per million population

AA Membership and Cultural Differences

Levine (1992) described a drinking pattern in non-Temperance cultures in which more alcohol is consumed and alcohol is more often consumed in the form of wine at meals and family gatherings. In Temperance cultures, less alcohol is consumed and it is more often consumed in the form of beer/spirits and in less socially controlled environments, usually with only men present, as in bars or at sporting events. As a result, drinking leads to more acting out and overt behavioural problems of the type recognized by AA.

Consistent with this description, AA groups are more common in Temperance (170 AA groups/million population) than in non-Temperance (25 groups/million) cultures (but the wide variation in Temperance cultures makes significance only borderline: t=2.04, 9 df3, p=<.10; see Table 1). National differences in AA membership are often remarkable. The country with the lowest per capita alcohol consumption in Europe (Iceland) has by far the highest density of AA groups (784 per million population). The wettest country in Europe (Luxembourg) was listed by AA in 1991 as providing no data on AA groups, suggesting there were no, or very few, AA groups in that country. The next wettest country, France, has 7 groups per million population. This cultural pattern leads to a surprising negative correlation between alcohol consumption and AA groups (Table 3). There are also highly significant correlations between number of AA groups (1991) and types of alcohol consumed: percentage spirits consumed is most positively correlated and percentage wine consumed (1987) is most negatively correlated (the negative correlation between AA groups and proportion wine consumed appeared, but was nonsignificant, with 1990 consumption data).

Table 2. Temperance countries and alcohol consumption
Alcohol Consumption
total consumption 6.6 +/- 1.6 10.8 +/- 1.7 =< .001a
percent wine 17.7 +/- 6.6 43.7 +/- 18.7 =< .001b
percent beer 53.1 +/- 10.3 40.4 +/- 17.1 n.s.
percent spirits 29.2 +/- 11.0 15.9 +/- 6.0 =< .005c
a  t=5.69, 19 df (df=n-2 for t-test)
b  t=4.31, 13 df (adjusted for unequal variances)
c  t=3.38, 14 df (adjusted for unequal variances)
Table 3. Correlations with 1991 AA groups/million population
AA Correlate Correlation R2 Probability
percent Catholic -.28 .08 n.s.
percent Protestant .35 .12 n.s.
total consumption 1990 -.52 .28 =< .05
percent wine 1987 -.64 .41 =< .005a
percent spirits 1990 -.77 .60 =< .00001
a  Not significant for 1990 percent wine; significant (p =< .01) for 1990 consumption for 27-nation analysis.

A regression model which accounts for 82 percent of the variance in AA membership has as its most highly positive predictors the percentage of alcohol consumed as beer and a national advertising policy encouraging beer/wine vs. spirits consumption (Table 4). The Temperance categorization of a country is also highly significant. In multiple regression including the Temperance variable, percentage Protestant and AA membership are negatively related, although they are positively correlated in a simple regression. This finding suggests that in Temperance societies, the lower the percentage of majority Protestants, the more AA groups appear, as an assertion of anti-alcohol attitudes in the face of a present challenge from a pro-alcohol cultural perspective. This might help to explain the small percentage of AA groups in overwhelmingly Protestant Norway and Sweden and the high number in multicultural U.S. With such small sample sizes, however, such a proposed interaction can only be speculative.

Table 4. Regression on 1991 AA groups/million populationa
Variable t Probability
percent beer 1987 4.67 =< .0005
percent of giving beer/wine preference in advertising 4.26 =< .001
temperance classification 3.77 =< .005
percent Protestant -3.57 =< .005
a  Adjusted R2 = .82, F = 21.08, 18 df, p =< .00001 (df=n-1 for ANOVA; Iceland, Luxembourg missing)


The consistency of these results with one another and with epidemiological studies supports their validity. Cross-national studies have the added value of tapping differences in national/ cultural attitudes and policies which anthropologists typically assess in individual or comparative ethnographic studies (Heath 1995). The current study broadens the base of previous cross-cultural studies of drinking by incorporating cultural variables as primary predictors in statistical analyses, along with, secondarily, policy measures. In addition, quantitative differences in behavioral consequences of drinking are estimated by AA membership. Results are consistent with other cross-cultural studies of drinking and health outcomes (Criqui and Ringel 1994), with individual cultural studies of drinking (Lolli et al. 1958), and with epidemiologic studies of drinking problems among ethnic groups within the U.S. (Greeley et al. 1980; Vaillant 1983).

The current study finds culture to be a substantial determinant of alcohol consumption among Western nations. One measure of culture, religion, is associated with significant differences both in amount of alcohol consumed and in patterns of consumption (although these differences may be narrowing). For example, total alcohol consumption and the proportion of alcohol consumed as wine are strongly correlated with Catholicism. Moreover, defining nations in terms of their historical Temperance orientation enhances the association between national identity and style and amount of alcohol consumed. Temperance nations consume far less alcohol and a far smaller proportion of that alcohol as wine than non-Temperance nations.

These results support the notion that drinking is more "socialized" in some cultures than in others. That is, in Western nations where alcohol is consumed as wine, probably more often in integrated social settings (such as meals and religious ceremonies, and also cafes where family members of both genders and different ages participate) rather than in settings devoted exclusively to drinking (such as male-dominated bars), alcohol use is behaviourally benign. There is less acting out of the type labelled as loss-of-control drinking that provides the grounds for large-scale alcoholism treatment and AA membership. Beer consumption and a Temperance tradition are, indeed, the best predictors in a powerfully predictive model of AA membership.

Alcohol consumption has been shown to be negatively correlated with CHD cross-nationally (LaPorte et al. 1980), while moderate alcohol consumption has been found to prevent CHD incidence and mortality in prospective cohort studies (Peele 1993). The current analysis finds that both greater total consumption and (even more strongly) consuming alcohol as wine prevent heart disease mortality on a national scale. However, the most powerful predictors in a regression model on CHD mortality are degree of Protestantism in a country and percentage of total revenues generated by taxation of alcohol. Protestantism may be a proxy for a confabulation of variables, including lifestyle and dietary factors. In the same vein, Temperance cultures have far higher CHD mortality rates.

Taxation policies toward alcohol, examined at one point in time, may serve primarily as a measure of traditional drinking practices and attitudes in a culture -- i.e., a cultural orientation toward restricting drinking and/or reducing proportionally high percentages of alcohol consumed as spirits. The policy measures included as predictors here are part of a further analysis currently being conducted with this data base. The current, limited finding suggests that taxing alcohol at a high rate and depending on such taxation as a source of revenue could have the counterproductive impact of elevating mortality due to heart disease, which is reduced by drinking.

Other sources of mortality may be directly related to alcohol consumption as well. For example, the fact that cerebrovascular mortality is significantly correlated with alcohol consumption in this study suggests that the U.N. statistics used refer to hemorrhagic stroke. The more common form of stroke, ischemic or occlusive, may be included in the "other ischaemic" disease category. This type of stroke behaves like CHD in response to (that is, is reduced by) drinking (Gaziano et al. 1993; Klatsky et al. 1992; Stampfer et al. 1988). In this study, wine consumption is associated with stroke while spirit consumption was inversely correlated. One finding with policy implications is that the increase in beer consumption in a country between 1985 and 1990 most strongly predicts mortality from stroke (presumably hemorrhagic).

Cirrhosis deaths have long been related to alcohol consumption. This analysis finds cirrhosis mortality strongly positively correlated with total consumption and, independently, with the percentage of alcohol consumed as wine. It is not clear why consuming alcohol in the form of wine causes cirrhosis while consuming alcohol in the form of beer is strongly inversely correlated with cirrhosis mortality. Thus, non-Temperance countries have far higher cirrhosis rates, and the Temperance classification is the single best predictor for cirrhosis. Nonetheless, changes in absolute beer and in wine consumption (as well as the percentage of alcohol stemming from importation of wine) are directly and inversely, respectively, associated with cirrhosis death. Again, the effect of changing consumption patterns is not what would be expected from projecting the static relationship between types of alcohol consumed in a society and disease prevalence.

No other source of death is so clearly related to overall alcohol consumption as CHD or cirrhosis. There are relationships to proportion of alcohol consumed in various forms: national beer consumption is inversely associated with hypertension deaths; spirits consumption is inversely related to diabetes and auto accident mortality; wine consumption correlates directly with stomach cancer, diabetes, and hypertension mortality. Violent and accidental deaths are not associated with the Temperance distinction, nor is any other type of mortality other than CHD and cirrhosis among 55-64 year olds.

As in Criqui and Ringel (1994), overall mortality in the current study is not related to total alcohol consumption, but it is related to the type of beverage consumed. In particular, the most powerful regression models for total mortality have as their strongest predictors either beer (in a direct relationship) or wine (in an inverse relationship). Wine has been shown sometimes (Tunstall-Pedoe et al. 1995) but not consistently to reduce CHD more than other forms of beverage alcohol. Woodward and Tunstall-Pedoe (1995) found that wine drinkers had a lower risk of heart disease than beer drinkers in the Scottish Heart Health Study. Ironically, Tunstall-Pedoe et al. (1994) had questioned the low coronary disease mortality rate claimed for France, which is often attributed to wine drinking (Renaud and de Lorgeril 1992). Grønbæk et al. (1995) found that wine, but not beer or spirits, reduced risk of overall mortality by one half for those who drank from three to five glasses a day. In a complex additional relationship in the current study, a policy encouraging consumption of alcohol other than spirits also inversely predicts total 55-64 mortality.


This study demonstrates that culture is a primary force shaping both the way alcohol is used and the consequences of its use in a given society. The present effort represents a first step in deconstructing and operationalizing the pervasive influence of culture on drinking. Nonetheless, this application of Levine's cultural taxonomy in an epidemiologic model has generated stronger correlations with behavioural consequences, and equally strong correlations with some health consequences, to those achieved with the primary predictors of consumption level and beverage type.

Thus, religious and cultural variables (specifically the Temperance/non-Temperance distinction) have strong independent predictive power for drinking outcomes such as AA membership, and also for mortality from several diseases. Overall these differences in mortality appear to balance out, yielding about equal total mortality rates for the two major cultural categories. The Temperance distinction bespeaks broader differences than drinking behaviors, including dietary and other lifestyle differences, which are independently related to CHD cross-culturally (Criqui and Ringel 1994).

The substantial reduction in CHD mortality and elevation in cirrhosis deaths found cross-culturally due to drinking replicate findings based on cohort and case-matching epidemiologic studies. However, this study also finds mortality from diabetes and stroke to be related to alcohol consumption, findings inconsistent with epidemiological research (cf. Perry et al. 1995; Rimm et al. 1995). This same discrepancy between cross-cultural and epidemiological research is found as well in the absence of a relationship between drinking and overall mortality. The advantages of moderate drinking for overall mortality have now been firmly established by over a dozen prospective studies (Poikolainen 1995), including the American Cancer Society (Boffetta and Garfinkel 1990), Kaiser (Klatsky et al 1992), Nurses (Fuchs et al. 1995), and several European (Doll et al. 1994; Grønbæk et al. 1994) studies. The discrepancy between prospective epidemiologic and cross-cultural studies may occur because elevation in societal alcohol consumption does not necessarily lead to a proportional increase in moderate drinkers.

Another casualty of the shift to a cross-cultural research design is the "mind-body" effect found in epidemiological research in the United States by Harburg and the present author. This research demonstrates that drinking in a setting and manner resembling the socialized drinking of non-Temperance nations can produce better physiological outcomes, such as less severe hangover (Harburg et al. 1993) and lower blood pressure (Harburg et al. 1994). The Temperance outlook may reflect deep-seated perspectives and feelings about alcohol that influence how alcohol consumption and intoxication are experienced and even how alcohol is processed by the body. However, not all individual differences measurable within cultures can be established cross-culturally with society-wide data.

Most if not all countries, but particularly Temperance nations, seek to improve behavioural and health outcomes by modifying alcoholic beverage consumption patterns. However, such policies can also have iatrogenic effects. One popular policy is to raise taxes so as to reduce overall consumption, an application of the "Ledermann hypothesis" which holds that the impact of reduced overall consumption in a society will be felt mainly at the high end of the consumption curve as an alleviation of the most severe alcohol problems experienced by that society. However, this policy runs the risk that it will increase cardiovascular mortality, since a reduction in alcohol consumption is experienced broadly throughout the society. In the current study, the level of revenue-generating taxes on alcohol in a nation strongly predicts CHD mortality.

Some societies, taking a different approach to encouraging healthier drinking patterns, enact a policy preference for wine and/or beer over spirits. Yet alcohol consumed as beer (as opposed to spirits) is associated with greater premature adult (age 55-64) mortality. Even where one beverage has positive health correlates- such as beer's inverse relationship to cirrhosis mortality-a shift away from wine consumption toward beer does not prevent, but actually increases death from cirrhosis. And, while overall mortality is inversely predicted by percentage wine consumption, absolute increases in wine consumption from 1980-1990 are significantly correlated with all age morality.

Policy changes must overcome a heavy weight of cultural inertia to have their intended impact. Cultural patterns of alcohol consumption have such integrity and equilibrium that trying to transform what appear to be less healthful to more healthful levels and forms of consumption, without sensitive attention to the cultural context, often is futile or worse. Specifically, the following iatrogenic results may occur:

  1. there are benefits to alcohol consumption (e.g., in preventing CHD) that lowering consumption may reduce;
  2. at the same time, raising consumption does not necessarily produce health benefits that outweigh negative outcomes for a society;
  3. policies geared towards encouraging one type of beverage alcohol over another may lead to disease mortality specifically associated with that type of alcohol (wine and cirrhosis) or to greater overall mortality (beer and total mortality in the 55-64 age group);
  4. however, even when the proportion consumed as a type of beverage alcohol is associated with lower or higher levels of mortality (beer and wine, respectively, with cirrhosis deaths; wine and beer with overall mortality), absolute increases or decreases in consumption of the beverages may not enhance these benefits, or may even reverse them.

This study indicates dangers in social engineering directed abstractly at modifying alcohol consumption, at least when it does not take into account existing societal drinking styles. Nonetheless, a future paper based on these data will explore potential health and mortality benefits associated with a variety of beverage alcohol policies.


The Wine Institute underwrote some of the costs of this study. The Brewers Association of Canada donated a copy of the document Alcoholic Beverage Taxation and Control Policies. Deborah McComber conducted the statistical analyses reported in this paper. Archie Brodsky assisted in integrating reviewers' comments from an earlier draft. I have benefitted from discussions about Temperance nations with Harry Levine, Robin Room, and George Bretherton. An earlier version of this paper was presented at the International Conference on Social and Health Effects of Different Drinking Patterns, Toronto, November 13-17, 1995.


Babor, T.F., Hesselbrock, M., Radouco-Thomas, S., Feguer, L., Ferrant, J.P. and Choquette, K. (1986) Concepts of alcoholism among American, French-Canadian, and French alcoholics. In Alcohol and Culture, Babor, T.F., ed., pp. 98-109. New York: Annals of the New York Academy of Sciences.

Blum, R.H. and Blum, E.M. (1969) A cultural case study. In Drugs: Vol. I. Society and Drugs, Blum, R.H. et al., eds., pp. 188-227. San Francisco: Jossey-Bass.

Boffetta, P. and Garfinkel, L. (1990) Alcohol drinking and mortality among men enrolled in an American Cancer Society prospective study. Epidemiology 1: 342-348.

Bretherton G. (1991) Against the flowing tide: whiskey and temperance in the making of modern Ireland. In Drinking: Behavior and Belief in Modern History, Barrows, S. and Room, R., eds., pp. 147-164. Berkeley: University of California.

Brewers Association of Canada (1993) Alcohol Beverage Taxation and Control Policies (8th edition). Ottawa, Ontario.

Cahalan, D. and Room, R. (1974) Problem Drinking Among American Men. New Brunswick, NJ: Rutgers Center of Alcohol Studies.

Criqui, M.H. and Ringel, B.L. (1994) Does diet or alcohol explain the French paradox? Lancet 344: 1719-1723.

Doll, R., Peto, R., Hall, E., Wheatley, K. and Gray, R. (1994) Mortality in relation to consumption of alcohol: 13 years' observations on male British doctors. BMJ 309: 911-918.

Fuchs, C.S., Stampfer, M.J., Colditz, G.A. et al. (1995) Alcohol consumption and mortality among women. New England Journal of Medicine 332: 1245-1250.

Gaziano, J.M., Buring, J.E., Breslow, J.L. et al. (1993) Moderate alcohol intake, increased levels of high-density lipoprotein and its subfractions, and decreased risk of myocardial infarction. New England Journal of Medicine 329: 1829-1834.

Glassner, B. and Berg, B. (1980) How Jews avoid drinking problems. American Sociological Review 45: 647-664.

Greeley, A.M., McCready, W.C. and Theisen, G. (1980) Ethnic Drinking Subcultures. New York: Praeger.

Grønbæk, M., Deis, A., Sorensen, T.I.A., Becker, U., Borch-Johnsen, K., Müller, C., Schnohr, P. and Jensen, G. (1994) Influence of sex, age, body mass index, and smoking on alcohol intake and mortality. BMJ 308: 302-306.

Grønbæk, M., Deis, A., Sorensen, T.I.A., Becker, U., Schnohr, P. and Jensen, G. (1995) Mortality associated with moderate intake of wine, beer, or spirits. BMJ 310: 1165-1169.

Harburg, E., Gleiberman, L., DiFranceisco, W. and Peele, S. (1994) Towards a concept of sensible drinking and an illustration of measure. Alcohol & Alcoholism 29: 439-450.

Harburg, E., Gunn, R., Gleiberman, L. and Schork, A. (1993) Psychosocial factors, alcohol use, and hangover signs among social drinkers. Journal Clinical Epidemiol. 46: 413-422.

Heath, D.B. (1984) Cross-cultural studies of alcohol use. In Recent Developments in Alcoholism (vol. 2), Galanter, M., ed., pp. 405-415. New York: Plenum.

Heath, D.B. (1995) An anthropological view of alcohol and culture in international perspective. In International Handbook on Alcohol and Culture, Heath, D.B., ed., pp. 328-347. London: Greenwood.

Klatsky, A.L., Armstrong M.A. and Friedman, G.D. (1992) Alcohol and mortality. Annals of Internal Medicine 117: 646-654.

LaPorte, R.E., Cresanta, J.L. and Kuller, L.H. (1980) The relationship of alcohol consumption to atherosclerotic heart disease. Preventive Medicine 9: 22-40.

Levine, H.G. (1992) Temperance cultures: Alcohol as a problem in Nordic and English-speaking cultures. In The Nature of Alcohol and Drug-Related Problems, Lader, M., Edwards, G. and Drummond, C., eds., pp. 16-36. New York: Oxford University Press.

Lolli, G., Serianni, E., Golder, G.M., and Luzzatto-Fegiz, P. (1958) Alcohol in Italian Culture. Glencoe, IL: Free Press.

Mäkelä, K. (1990) Social and cultural preconditions of A.A. and factors associated with the strength of A.A. Paper presented at the 16th Annual Alcohol Epidemiology Symposium of the Kettil Bruun Society for Social and Epidemiological Research on Alcohol, Budapest, June 3-8.

Maloff, D., Becker, H.S., Fonaroff, A. and Rodin, J. (1982) Informal social controls and their influence on substance use. In Control Over Intoxicant Use, Zinberg, N.E. and Harding, W.M., eds., pp. 53-76. New York: Human Sciences Press.

Peele, S. (1993) The conflict between public health goals and the Temperance mentality. American J Public Health 83: 805-810.

Perry, I.J., Wannamethee, S.G., Walker, M.K., Thomson, A.G., Whincup, P.H. and Shaper, A.G. (1995) Prospective study of risk factors for development of non-insulin dependent diabetes in middle aged British men. BMJ 310: 560-564.

Poikolainen, K. (1995) Alcohol and mortality. Journal of Clinical Epidemiology 48: 455-465.

Renaud, S. and de Lorgeril, M. (1992) Wine, alcohol, platelets and the French Paradox for coronary heart disease. Lancet 339: 1523-1526.

Rimm, E.B., Chan, J., Stampfer, M.J., Colditz, G.A. and Willett, W.C. (1995) Prospective study of cigarette smoking, alcohol use, and the risk of diabetes in men. BMJ 310: 555-559.

Rimm, E.B., Giovannucci, E.L., Willett, W.C., Colditz, G.A., Ascherio, A., Rosner, B. and Stampfer, M.J. (1991) Prospective study of alcohol consumption and risk of coronary disease in men. Lancet 338: 464-468.

Stampfer, M.J., Colditz, G.A., Willett, W.C., Speizer, F.E. and Hennekens, C.H. (1988) A prospective study of moderate alcohol consumption and the risk of coronary heart disease and stroke in women. New England Journal of Medicine 319: 267-273.

Suh, I., Shaten, B.J., Cutler, J.A. and Kuller, L.H. (1992) Alcohol use and mortality from coronary heart disease: the role of high density lipoprotein. Annals of Internal Medicine 116: 881-887.

Tunstall-Pedoe, H., Kuulasmaa, K., Amouyel, P., Arveiler, D., Rajakangas, A.M. and Pajak, A. (1994) Myocardial infarction and coronary deaths in the World Health Organization MONICA project. Circulation 90: 583-612.

Vaillant, G.E. (1983) The Natural History of Alcoholism. Cambridge, MA: Harvard University Press.

Whitehead, P.C. and Harvey, C. (1974) Explaining alcoholism: an empirical test and reformulation. Journal of Health Social Behavior 15: 57-65.

Woodward, M. and Tunstall-Pedoe, H. (1995) Alcohol consumption, diet, coronary risk factors, and prevalent coronary heart disease in men and women in the Scottish heart health study. Journal of Epidemiology and Community Health 49: 354-362.

Zar, J.H. (1984) Biostatistical Analysis (2nd edition). Englewood Cliffs, NJ: Prentice Hall.