Further Reading


Drug and Alcohol Dependence, 20:173-201, 1987

Why Do Controlled-Drinking Outcomes Vary by Investigator, by Country and by Era?

Cultural Conceptions of Relapse and Remission in Alcoholism

Stanton Peele
Morristown, New Jersey



Variations in the reported rates of controlled drinking by former alcoholics are notable, at times startling. Reports of such outcomes (which in some cases involved a large percentage of subjects) were common for a brief period ending in the mid- to late 1970s. By the early 1980s, a consensus had emerged in the United States that severely alcoholic subjects and patients could not resume moderate drinking. Yet—at a point in the mid-1980s when the rejection of the possibility of a return to controlled drinking appeared to be unanimous—a new burst of studies reported resumption of controlled drinking was quite plausible and did not depend on the initial severity of alcoholics' drinking problems. Variations in controlled-drinking outcomes—and in views about the possibility of such outcomes—involve changes in the scientific climate and differences in individual and cultural outlooks. These cultural factors have clinical implications as well as contributing to the power of scientific models of recovery from alcoholism.

Key words: Expectations—Beliefs and alcoholism—Controlled drinking—Behavior therapy—Therapy efficacy—Natural remission

Introduction and Historical Overview

Twenty-five years after Davies' [1] report that 7 out of a group of 93 treated British alcoholics had returned to moderate drinking, Edwards [2] and Roizen [3] analyzed reactions to Davies' article. Nearly all of the 18 comments on the article published in the Quarterly Journal of Studies on Alcohol were negative, most extremely so. Respondents, who were all physicians, based their objections to Davies' findings on their clinical experience with alcoholic patients. The respondents furthermore expressed a consensus against controlled-drinking in America that, according to Edwards, expressed 'an ideology with nineteenth century roots, but [which] in the 1960s.... had been given new strength and definition under the conjoint influence of Alcoholics Anonymous (AA), the American National Council on Alcoholism and the Yale School' [2, p.25]. At the time it appeared, the Davies' article and its critiques created relatively little stir [3], probably because the article posed no real challenge to accepted medical [4] and folk wisdom that abstinence was an absolute necessity for recovery from alcoholism.

Two responses to Davies' article, however, endorsed and even extended Davies' findings. Myerson [5] and Selzer [6] claimed that the hostile atmosphere surrounding such results stifled genuine scientific debate and stemmed in part from the involvement of many recovering alcoholics in the field who tended to 'preach rather than practice' [5, p. 325]. Selzer recounted similar hostile reactions to his own 1957 report [7] of treated alcoholics who achieved moderation (the percentage of moderation outcomes in this study was twice as high—13 of 83 subjects—as that reported by Davies). Giesbrecht and Pernanen [8] discovered that outcome or follow-up research (like Selzer's and Davies') increased in the 1960s, at the same time as clinical studies more often relied on changes or improvements in drinking patterns as outcome criteria.

Through the 1960s and 70s, a number of studies revealed substantial rates of non-abstinent remission for alcoholism [9]. These included controlled-drinking outcomes for 23% (compared with 25% abstainers) of treated alcoholics interviewed 1 year after leaving the hospital by Pokorny et al. [10], 24% (compared with 29% abstainers) of women alcoholics treated at a psychiatric hospital at a 2-year follow-up conducted by Schuckit and Winokur [11], and 44% (compared with 38% abstainers) of alcoholics studied 1 year after undergoing inpatient group therapy by Anderson and Ray [12]. Among a group of alcoholics that was largely untreated, Goodwin et al. [13] noted at a follow-up period of 8 years that 18% were moderate drinkers (compared with only 8% abstainers) and that a large additional group (14%) drank to excess on occasion but were still judged to be in remission.

The debate about resuming controlled drinking became far more heated when the first Rand report appeared in 1976 [14]. This study of NIAAA-funded treatment centers found 22% of alcoholics to be drinking moderately (compared with 24% abstainers) at 18 months after treatment, leading immediately to a highly publicized rebuttal campaign organized by the National Council on Alcoholism (NCA). A 4-year follow-up of this study population by the Rand investigators continued to find substantial nonproblem drinking [15]. These well-publicized findings did not change prevailing attitudes in the treatment field—the directors of the NIAAA at the time of the two Rand reports each declared that abstinence remained 'the appropriate goal in the treatment of alcoholism' [16, p. 1341].

At around the same time the Rand results were being compiled in the early and mid-1970s, several groups of behavior therapists published reports that many alcoholics had benefited from controlled-drinking (CD) therapy [17,18]. The most controversial of these behavioral-training investigations was conducted by Sobell and Sobell [19,20], who found that moderation training for gamma (i.e. loss of control [21]) alcoholics led to better outcomes 1 and 2 years after treatment than did standard hospital abstinence treatment. This and similar findings by behavioral researchers remained for the most part esoteric exercises, and like the Rand reports, had little or no impact on standard treatment for alcoholics.

Nevertheless, CD treatment and research continued throughout the 1970s. In 1983, Miller [22] indicated 21 of 22 studies had demonstrated substantial benefits from CD therapy at follow-ups of from 1—2 years (see Miller and Hester [23, Table 2.1] and Heather and Robertson [24, Tables 6.3 and 6.4] for detailed outlines of these studies). This research found greater benefits for problem drinkers who were less severely dependent on alcohol, although no comparative study had shown moderation training to be less effective than abstinence as a treatment for any group of alcoholics. Despite the absence of a single case of strong evidence to contraindicate CD therapy for alcoholics, beginning in the mid-1970s behavioral researchers became increasingly conservative in recommending this therapy for severe cases of alcoholism [16]. By the early 1980s, the leading practitioners of CD therapy in the United States claimed it was not suitable for physically dependent alcoholics (i.e. those who displayed withdrawal symptoms following abstinence [25,26]).

At the same time, several outcome studies disputed the Rand reports' contention that CD remission was no more unstable than was that due to abstinence. Paredes et al. [27] reported that abstinence led to more stable remission than controlled drinking. Another research group that had previously reported substantial CD outcomes [28] also found, in 1981, that abstinence remission was more stable than moderate-drinking outcomes between 6 months and 2 years [29]. However, in a study of hospital-based treatment conducted by Gottheil et al. [30], alcoholics who moderated their drinking did not relapse more frequently than abstainers between 6 months and 2 years. Gottheil and his colleagues furthermore compared their results with those from the Rand studies and Paredes et al., noting that despite differences in treatment goals (the Gottheil study did not require abstinence) and follow-up criteria, 'similarities seemed to far outweigh differences in the findings' (p. 563).

In the 1980s, a number of studies strongly disputed both the possibility of moderate drinking by alcoholics and specific earlier reports of CD outcomes. The most publicized of these studies was a follow-up of the Sobells' research [19,20] conducted over 9 years by Pendery et al. [31] and published in Science. The study found that only one of the Sobells' group of 20 alcoholics who was taught to control his consumption actually became a moderate drinker, and the authors claimed this man was not a gamma alcoholic originally. Edwards [32], reporting a later follow-up of CD outcome subjects in the Davies study [1], found only two (one of whom had a low level of alcohol dependence) had engaged in trouble-free drinking continuously after treatment.

Vaillant [33], in a long-term longitudinal study, reported frequent controlled drinking by subjects but noted that these outcomes were unstable over the long term. Vaillant was especially dubious about more severely dependent drinkers achieving moderation: 'There appeared to be a point of no return beyond which efforts to return to social drinking became analogous to driving a car without a spare tire. Disaster was simply a matter of time' [p. 225]. Edwards et al. [34] found that drinkers who could sustain controlled drinking over a lengthy (12-year) follow-up period came entirely from among those less severely dependent on alcohol. Finally, Helzer et al. [35] reported in the New England Journal of Medicine that only 1.6% of hospitalized alcoholics had resumed stable moderate drinking at from 5 to 7 years after treatment.

By the mid-1980s, many prominent sources had concluded that controlled drinking was not a viable alternative in alcoholism treatment. In a review article on this question, the principal authors of the New England Journal study questioned whether controlled drinking 'is a realistic treatment goal when so few seem able to sustain it for long periods of time.... One fairly consistent finding,' these authors further noted, 'is that alcoholics who are able to return to social drinking tend to be milder cases' [36, p. 120]. A leading behavioral researcher declared: 'responsible clinicians had concluded that the available data do not justify continued use of CD treatment with alcoholics' [37, p. 434]. A psychologist active in alcohol-dependence syndrome research in Britain failed to find a 'convincing case of a prolonged return to controlled drinking following a significant period of alcohol dependence' [38, p. 456].

This broad-based and firm rejection of the possibility of controlled drinking came after a decade (beginning with the first Rand report) of intense reevaluation of this issue. It was quite surprising, therefore, when a number of studies—also appearing in the mid-1980s—questioned this emergent consensus. In each case, the research found that severely dependent alcoholics could resume moderate drinking and/or that level of severity of alcoholism was unrelated to moderation outcome. McCabe [39], for example, reported a 16-year follow-up of 57 individuals diagnosed and treated for alcohol dependence in Scotland. He found that 14.5% of the subjects were abstinent and 20% were controlled drinkers.

In Sweden, Nordström and Berglund [40] conducted another long-term (21 + 4 years) follow-up of patients admitted for inpatient alcoholism treatment in Sweden. Of 84 patients found to have met the criteria for alcohol dependence, 15 were abstaining and 22 were social drinkers. Among a 'Good Social Adjustment Group' that was the primary focus of the study, social drinkers (38%) were almost twice as frequent as abstainers (20%). Abstainers had more instances of relapse in this study, and severity of alcohol dependence was not related to outcome. In a 5—6-year follow-up of chronic alcoholics receiving either abstinence-oriented or CD treatment, Rychtarik et al. [41] found 20.4% were abstinent and 18.4% drinking moderately; no measure of alcohol dependence distinguished between the two groups.

Two British studies evaluated interactions among patient beliefs and past experiences, the type of treatment they received (CD vs. abstinence), and outcome at 1 year. Both studies found substantial CD outcomes. Orford and Keddie [42] found there was 'no relationship between level of dependence/severity and the type of drinking outcome (abstinence or CD)' (p. 495). Elal-Lawrence et al., reporting results on 45 successful abstainers and 50 controlled drinkers after 1 year: 'Of the variables measuring the severity of the problem — duration, daily intake, reported number of alcohol-related symptoms...—none of them discriminated between the outcome groups' [43, p. 45]. Lastly, another British team of investigators, Heather et al. [44], found that subjects 'reporting signs of late dependence' (p. 32) benefited more from moderation instructions than did other problem drinkers.

Given that controlled drinking for alcoholics had apparently been conclusively rejected, at least in America, the appearance of a number of studies disputing this conclusion indicated just how unlikely it is that the controlled-drinking issue will ever entirely disappear. The concurrent appearance of these positive CD findings also highlighted a more basic question: what accounts for historical changes in the receptiveness of the climate for controlled drinking and in the reportings of the frequency of such outcomes, as well as for the major differences in the views and results of different groups of investigators? This article explores some factors related to the investigators, the era (or point in time) in which the research was conducted, and the national, professional, or popular culture that may help to explain such divergent research results and conclusions.

The Causes and Consequences of Recent Shifts in Controlled-Drinking Outcomes

Reactions to the Rand reports

The reaction to the first Rand report was the strongest and most critical that had yet appeared to any piece of alcoholism research (and may have been unique for research in any scientific field in the twentieth century) [16]. As a result, the significance of this research did not come so much from its actual results, which—as its authors pointed out—were unexceptional in relation to prior data on alcoholism outcomes [14]. Instead, the climate engendered in the aftermath of the reports was to have important implications for views of alcoholism and methods for assessing outcomes.

Criticisms of the first report concerned the (1) length of the follow-up period (18 months), (2) completion rate of interviews (62%), (3) exclusive reliance on subject self-reports, (4) initial classification of subjects and their degree of alcoholism, (5) limiting assessment of drinking to a 30-day period, and (6) overgenerous criteria for normal or controlled drinking. The second report [15], released in 1980, (1) extended the study to a 4-year follow-up period, (2) completed outcome data for 85% of target sample, (3) employed unannounced breathalyzer tests as well as questioning collaterals in one-third of cases, (4) segmented the study population into three groups based on symptoms of alcohol dependence, (4) lengthened the assessment period of drinking problems to 6 months, and (5) tightened the definition of controlled drinking (which was called 'normal' drinking in the first report and 'nonproblem' drinking in the second).

The non-problem drinking category included both high consumption (up to 5 oz ethanol on a given day, with an average consumption on drinking days of no more than 3 oz daily) and low consumption (no more than 3 oz on 1 day and average less than 2 oz) drinkers. The second report emphasized consequences of drinking and symptoms of alcohol dependence over consumption measures in categorizing non-problem drinking. Whereas the first report permitted a 'normal' drinker to manifest two serious drinking symptoms in the previous month, the second eliminated from the non-problem category anyone who had a single health, legal, or family drinking problem in the previous 6 months or who had shown any signs of alcohol dependence (e.g. tremors, morning drinking, missed meals, blackout) 30 days prior to their last drink.

The percentage of non-problem drinkers was reduced in the second Rand report from 22 to 18% (10% with high and 8% with low consumption, together comprising 39% of all those in remission). This reduction was due largely to the changed criteria rather than to attrition of moderation outcomes. Comparison of clients in remission at 18 months and 4 years showed CD outcomes were not more unstable than was abstinence. For those experiencing fewer than 11 symptoms of dependence, controlled drinking was the more frequent outcome. At the highest level of dependence, abstinence outcomes predominated. Nonetheless, more than a quarter of those having more than 11 symptoms of dependence on admission who achieved remission did so through non-problematic drinking. The second Rand report results therefore found significant numbers of severely alcohol-dependent subjects who engaged in non-problematic drinking. (Overall, the Rand study population was severely alcoholic: nearly all subjects reported symptoms of alcohol dependence at admission to treatment, and median alcohol consumption was 17 drinks/day).

The second Rand report elicited a large number of positive reviews by social scientists [45,46]. Writing several years after the second report appeared, Nathan and Niaura [37] declared that 'in terms of subject numbers, design scope, and follow-up intervals as well as sampling methods and procedures, the four-year Rand study continues at the state-of-the-art of survey research' [p. 416]. Nonetheless, these authors asserted, 'abstinence ought to be the goal of treatment for alcoholism' (p. 418). As Nathan and Niaura's statement demonstrates, the Rand results did not change attitudes in the field toward CD treatment. When NIAAA administrators claimed the second report had reversed the earlier Rand finding that alcoholics could control their drinking, the Rand investigators publicly and vigorously rejected this contention [47]. Nonetheless, the impression remains to this day in the alcoholism field that the idea that alcoholics can drink again was 'a sad conclusion the Rand Corporation in 1975 came to, but has since repudiated' (pers. commun., Patrick O'Keefe, September 16, 1986).

Changing criteria for controlled drinking

The Rand reports revealed a degree of opposition to controlled drinking in the United States that social scientific investigators and clinicians could not ignore. As Room [48, p. 63n] reported: 'The present author knows of two cases where public funding for studies was cut off over the issue of 'controlled drinking' in about 1976' in connection with a California State Alcoholism Board resolution 'during the Rand controversy' that public funds not be expended 'to support research or treatment programs that advocate so-called 'controlled drinking' practices'. At the same time, researchers became more cautious in labeling CD outcomes and relating them to initial classification of severity of alcohol dependence and alcoholism in treatment clients. Prior to the Rand reports, for example, investigators had tended to classify as alcoholic anyone who ended up in alcoholism treatment [10,11,12].

The Rand investigators themselves pioneered this change, and their second report is now often cited by alcohol-dependence investigators as a seminal study in indicating treatment outcomes shift in relation to initial severity of drinking problem, or degree of alcohol dependence [49]. The Rand investigators also led the way toward stricter labeling of CD outcomes by eliminating from that category drinkers who showed any subsequent signs of alcohol dependence in their second study, whether or not subjects reduced either their level of drinking and/or the number of dependence symptoms. In addition, the Rand reports focused attention on the length of outcome follow-up period (which was the primary point in conducting the second study). Overall, the Rand reports presaged longer follow-up periods, the examination of continuous drinking behavior over this period, and greater care generally in identifying CD findings.

Pendery et al. [31] applied such stricter standards to the work of the Sobells. The Pendery group, for example, questioned the accuracy of diagnoses of gamma alcoholism in the Sobells' subjects who showed the greatest improvement due to CD therapy. They also tracked subjects for almost a decade, while chronicling all recorded instances of hospitalizations and emphasizing uncontrolled binges during the 2-year follow-up period for which the Sobells reported their data [19,20] and an additional third-year follow-up by Caddy et al. [50]. Many of these individual incidents diverged sharply from an image of successful controlled drinking. Cook [51] analyzed how very different images were carried out of the same data by the different research teams.

In this light, the standards for successful outcomes had shifted from the early 1970s when the Sobells conducted their research to the 1980s when the Pendery et al. study appeared. The Sobells' and Caddy et al.'s analyses indicated that CD subjects had fewer days of drunkenness than did subjects given standard abstinence treatment. In today's atmosphere, however, there is less tolerance for the idea that subjects continue to get drunk within the context of an overall improvement in functioning and moderation of drinking problems. Identifying in treated subjects periodic (or even occasional) instances of intoxication seemingly vitiates the idea that treatment has been helpful or that subjects have recovered from alcoholism. That only three of the Sobells' CD-treatment subjects had no drunk days during the second year, and many had had several severe drinking episodes, provided substantial fuel for the Pendery et al. critique.

Edwards [32] likewise extended the follow-up period in Davies' [1] research, challenged initial diagnoses of alcoholism, and pointed out drinking problems that Davies missed or neglected, apparently because subjects often did drink normally and had improved their conditions overall. Other research from the 1960s and 70s would seem to be open to similar challenges. These earlier clinical investigations often were more concerned about global measures and impressions of psychological adjustment than they were about moment-by-moment measures of drinking or drunken misbehavior. Fitzgerald et al. [52], for example, reported that 32% of patients treated for alcoholism showed 'good adjustment with drinking' (compared with 34% showing 'good adjustment without drinking'), without detailing actual drinking behavior. Gerard and Saenger [53] neglected patients' alcohol consumption and drinking patterns in favor of assessing patients' psychological functioning in the CD outcomes they reported.

Outcome research today is far more likely to scrutinize whether subjects have actually improved in the face of continued drinking. As controlled drinking itself became the focus of outcome results in Davies' study and the Rand reports, investigators became concerned to measure exactly the extent of controlled drinking, often employing extremely stringent criteria. Investigations such as Vaillant's [33] and Helzer et al.'s [35], for instance, had as primary foci the exact nature and extent of non-problematic drinking. The behavioral investigation of alcoholism has also had this effect, because this research turned to precise measures of consumption to replace vaguer psychological diagnoses [54]. Thus, Elal-Lawrence's CD research reported successful CD outcomes based exclusively on consumption measures. Paradoxically, the Sobells' research was a part of this process, because it used as its primary measure 'days functioning well'—which simply meant the combined number of days in which subjects either abstained or drank less than the equivalent of 6 oz of 86-proof alcohol.

Potential drawbacks of revised standards for controlled drinking

If rigorous current methodologies reveal earlier CD research to be seriously flawed, then it may be best to discard this research. Helzer et al. discounted 'the existing literature on controlled drinking because of small or unrepresentative samples, failure to define moderate drinking, acceptance of brief periods of moderate drinking as a stable outcome, failure to verify subjects' claims, and.... [inadequacy] of duration or subject-relocation rates' [35, p. 1678]. Another perspective, however, is offered by sociologists Giesbrecht and Pernanen, when they commented about changes they measured between 1940 and 1972 (including utilization of CD, abstinence and other remission criteria in research): 'that they are caused less by accumulating scientific knowledge than by changes in conceptions and structurings of research and knowledge' [8, p. 193].

Are there complementary costs to discounting much pre-1980s research on controlled drinking, along with the assessment methods the research relied on? In focusing solely on whether subjects can achieve moderation, or else discarding this goal in favor of abstinence, the alcoholism field has drastically de-emphasized issues of patient adjustment that do not correlate exactly with drinking behavior. Is it completely safe to assume that absence of drunkenness is the sine qua non of successful treatment, or can sober alcoholics manifest significant problems, problems that may even appear after the elimination of alcoholism? Pattison [55] has been the most consistent advocate of basing treatment evaluations on psychosocial health rather than on patterns of drinking, but for the time being this remains a distinctly minority position.

A related possibility is that patients may improve—in terms of their drinking and/or overall functioning—without achieving abstinence or strictly defined controlled drinking. This question is particularly relevant because of the low rates of successful outcomes (and especially of abstinence) reported by several important studies of conventional alcoholism treatment. For instance, the Rand reports found only 7% of clients at NIAAA treatment centers abstained throughout the 4-year follow-up period. Gottheil et al. [56], noting 10% was a typical abstinence rate among treated populations, pointed out that between 33 and 59% of their own VA patients 'engaged in some degree of moderate drinking' following treatment:

If the definition of successful remission is restricted to abstinence, these treatment centers cannot be considered especially effective and would be difficult to justify from cost-benefit analyses. If the remission criteria are relaxed to include moderate levels of drinking, success rates increase to a more respectable range.... [Moreover] when the moderate drinking groups were included in the remission category, remitters did significantly and consistently better than nonremitters at subsequent follow-up assessments. (p. 564)

What is more, the research and researchers that have been most prominent in disputing CD outcomes have themselves demonstrated severe limitations in conventional hospital treatment geared toward abstinence. For example, the Pendery et al. critique of the Sobells' work failed to report any data on the hospital abstinence group with which the Sobells compared their CD treatment group. Yet such relapse was common in the hospital group; as Pendery et al. noted, 'all agree [the abstinence group] fared badly' (p. 173). Relapse was likewise very evident among 100 patients Vaillant [33] treated in a hospital setting with an abstinence goal: 'only 5 patients in the Clinic sample never relapsed to alcoholic drinking' (p. 284). Vaillant indicated that treatment at the hospital clinic produced outcomes after 2 and 8 years that 'were no better than the natural history of the disorder' (pp. 284—285). Edwards et al. [57] randomly assigned alcoholic patients to a single informational counseling session or to intensive inpatient treatment with outpatient follow-up. Outcomes for the two groups did not differ after 2 years. It is impossible to evaluate CD treatments or patients' ability to sustain moderation without considering these limitations in standard treatments and outcomes.

The intense concentration on CD outcomes does not seem to be matched with comparable caution in evaluating abstinence outcomes and treatment. For example, Vaillant [33] also reported (in addition to his clinical results) 40 year longitudinal data on drinking problems in an inner-city group of men. Vaillant found that 20% of those who had abused alcohol were controlled drinkers at their last assessment, while 34% were abstaining (this represents 102 surviving subjects who had abused alcohol; 71 of 110 of the initial subjects were classified as alcohol dependent). Vaillant was not very sanguine about CD outcomes, however, particularly for more severely alcoholic subjects, because he found that their efforts to moderate their drinking were unstable and frequently led to relapse.

Vaillant defined men as abstinent who in the previous year were 'using alcohol less often than once a month' and 'had engaged in not more than one episode of intoxication and that of less than a week in duration' (p. 184). This is a permissive definition of abstinence, and does not correspond with either most people's commonsensical notions or the Alcoholics Anonymous (AA) view of what comprises abstinence. Yet controlled drinkers in this study were not allowed to show a single sign of dependence (like binge or morning drinking) in the previous year (p. 233). Making the definitions of relapse more equivalent would seemingly increase relapse for those called abstainers and decrease relapse among controlled drinkers (that is, increase the prevalence and durability of moderation outcomes).

The non-comparability of definitions may be even more severe in the case of Helzer et al. [35] in comparison with the Rand studies. In discussing outcomes for alcoholic hospital patients in a 5—8-year period (the abstract referred to a 5—7-year period) following hospital treatment, the Helzer group classified 1.6% as moderate drinkers. In addition, the investigators created a separate category of 4.6% alcoholic patients who had no drinking problems and drank moderately, but who drank during less than 30 of the previous 36 months. Lastly, these investigators identified as a separate group heavy drinkers (12% of the sample) who had had at least 7 drinks on 4 or more days within a single month in the previous 3 years. These drinkers had given no indication of having any alcohol-related problems, nor did the investigators find any records of such problems.

Although Helzer et al. concluded almost no alcoholic patients became moderate drinkers, these data could be interpreted to show that 18% of alcoholic patients continued to drink without showing any drinking problems or signs of dependence (compared with the 15% in this study who abstained). For such a hospitalized subject population, in which three-quarters of the women and two-thirds of the men were unemployed, this level of non-problem drinking would actually be quite a remarkable finding. In fact, the second Rand study [15] reported almost identical results: 8% of subjects were drinking small quantities of alcohol while 10% sometimes drank heavily but did not manifest adverse consequences or symptoms of dependence. The Rand investigators labeled this entire group non-problem drinkers, causing those who endorsed conventional treatment precepts of abstinence to attack the study as unreliable and ill-advised. By applying wholly different perspectives on the essential element in remission (dependence symptoms vs. consumption), the Rand investigators and Helzer et al. ended up in diametrically opposed positions on the matter of controlled drinking.

The Helzer group (like the Rand investigators) attempted to verify reports by drinkers that they had not experienced alcohol-related problems. Thus this research team conducted collateral interviews to confirm subject self-reports, but only in the case where subjects had indicated that they were controlled drinkers. Even where no problems were found through collateral measures, these researchers simply regarded as denial that those who had drunk at all heavily during one period over 3 years did not report drinking problems; this despite their finding that patients' self-reports of whether they had achieved the study's definition of moderate drinking (regular drinking rarely or never leading to intoxication) corresponded very closely to the researchers' assessments.

Seemingly, Helzer et al. and Vaillant were more concerned to validate CD than abstinence outcomes, a caution very typical in the field. It is certainly possible that patients drinking with problems might report moderate drinking to disguise their problems. Yet, in an abstinence treatment setting, it is also plausible that patients who claim to be abstaining may also be covering up drinking problems. There is an additional potential self-report error in a situation where patients have received abstinence treatment: they may disguise instances of moderate drinking while claiming to be abstinent. Data indicate that all such self-report errors occur, and furthermore are not uncommon (see comments by Fuller, Workshop on the Validity of Self-Report in Alcoholism Treatment Research, Clinical and Treatment Research Subcommittee of the Alcohol Psychosocial Research Review Committee, Washington, DC, 1986).

The Helzer et al. study results indicate little benefit from hospital treatment of alcoholism, at least for severely alcoholic populations. Actually, only one of four groups of subjects in the study received inpatient alcoholism treatment at the hospital. This group had the lowest remission rate— among survivors, one-half that for medical/surgical patients. Of those treated in the alcoholism unit, 'only 7 percent survived and recovered from their alcoholism' (p. 1680). Thus Helzer et al. rejected decisively the value of CD treatment in a study which did not actually administer such treatment, and in which the recovery rate of below 10% for standard treatment was significantly worse than the typical untreated remission rates found among the community populations with which Vaillant compared his treated hospital group [33, p. 286].

The emerging focus on expectations in CD research

The six studies cited in the introduction to this paper [3944] have, as a group, responded to criticisms typically leveled at earlier work reporting controlled-drinking outcomes. Each took care to establish the initial presence or degree of alcoholism, using Jellinek's [21] classification system or measures of alcohol dependence (defined either as a specific syndrome marked by withdrawal symptoms, or else gradated in terms of numbers of symptoms in alcohol dependence) [15,58,59]. The studies in addition have been careful to define moderate or non-problem drinking and have relied on combinations of measures to corroborate moderate drinking including collateral interviews, biological tests, and hospital and other records.

Five of the six studies—as well as establishing that alcoholic or alcohol-dependent subjects did achieve controlled drinking—found no relationship between severity of alcohol dependence and CD outcomes. In the sixth study, McCabe [39] classified subjects in terms of gamma, delta (inability to abstain), and epsilon (binge drinking) alcoholism [21], but did not relate controlled drinking to initial diagnoses. All subjects, however, qualified for one of the three alcoholism categories, and 17 of 19 subjects in remission had been classified gamma or delta alcoholics while 11 of those in remission were controlled drinkers.

The studies also addressed other criticisms against previous CD research, such as the endurance of controlled-drinking outcomes. McCabe [39] and Nordström and Berglund [40] reported on follow-up data extending from 16 years to over two decades. In both cases, the number of long-term controlled-drinking subjects exceeded abstainers. All of Nordström and Berglund's cases were defined as alcohol dependent, and even subjects who had experienced delirium tremens in the past were more likely to be controlled drinkers than to abstain. In the United States, Rychtarik et al.'s [41] assessment of chronic alcoholics receiving treatment with either an abstinence or CD goal found that at 5—6 years following treatment, 20% became abstinent and 18% controlled drinkers.

Two of these CD studies, by Elal-Lawrence et al. [43] and Orford and Keddie [42], furthermore applied sophisticated research designs to comparisons of CD and abstinence treatment and outcomes. Both studies contrasted the effects of patients' beliefs and expectations with objective measures of alcohol dependence and found the former to be more important for outcomes than the latter. The emphasis on expectations and alcoholic behavior has been a major focus of psychological research on alcoholism and would seem to comprise an important component in alcoholism theory and treatment. A large body of research, for example, has examined the exaggerated expectations for emotional relief and other benefits alcoholics and heavy drinkers anticipate from drinking [60,61].

In addition, research on expectancies has focused on their effects on craving and relapse. Marlatt et al. [62], in a classic study, found gamma alcoholics drank more when they believed they were consuming alcohol (but were not) than when they actually drank alcohol (but believed they were not). Research of this kind has clearly indicated that 'what alcoholics think the effects of alcohol are on their behavior influences that behavior as much or more than the pharmacologic effects of the drug.... Expectancies are relevant to craving and loss of control because many alcoholics do in fact subscribe to the view that craving and loss of control are universal among alcohol dependent individuals' [54]. Although the authors of this quote defended abstinence as the appropriate goal in treatment, the ideas they expressed would seem to support the notion that convincing people they can or cannot be controlled drinkers (or patients' prior convictions in this regard) would significantly affect controlled-drinking outcomes.

Based on exactly this assumption, Heather et al. [63] found that those believing in the 'one drink, then drunk' axiom were less likely than other alcoholics to drink moderately following treatment. Heather and his coworkers [64] also reported that subjects' beliefs about alcoholism and about their particular drinking problems significantly affected which patients relapsed and which maintained harm-free drinking, while patients' severity of alcohol dependence did not. Elal-Lawrence et al. [43] likewise found that 'alcoholism treatment outcome is most closely associated with patients' own cognitive and attitudinal orientation, past behavioral expectations, the experience of abstinence and the freedom of having his or her own goal choice' (p. 46), while Orford and Keddie [42] found support for the idea that abstinence or controlled-drinking outcomes are relatively likely 'the more a person is persuaded that one goal is possible' (p. 496).

The studies discussed in this section overall represent a movement into a new era of research sophistication. This is far from saying they are immune from criticism. Definitions of alcohol dependence and alcoholism vary from one study to the next and, in addition, in the longitudinal research [39,40] were constructed post hoc. Use of different criteria to identify alcoholics is typical in the field, however, and may not be a bad thing as different dimensions of severity of alcoholism yield different insights and benefits. The controlled studies of CD and abstinence therapy [4143], on the other hand, suffer from the very complexity of the conclusions they uncover; they do not offer simple criteria for predicting controlled drinking. All things considered, nonetheless, the results of these studies cannot in good faith be dismissed as research aberrations traceable to sloppy or inadequate research designs.

The Cultural Analysis of Research, Treatment and Remission in Alcoholism

Perhaps the shifting empirical support for controlled-drinking represents a model of science in which evidence is gathered and interpreted until one hypothesis gains sufficient support to become the dominant theory. In this view, opinions may see-saw back and forth for a time, but during this process the entire body of evidence proceeds toward an emergent scientific consensus that transcends each component hypothesis. Working against this notion of accumulated scientific progress in alcoholism remission is that each side in the debate simultaneously claims the mantle of emergent scientific reality—i.e. that controlled-drinking findings represent the overthrow of a now outmoded disease paradigm [65], and that discarding unsubstantiated controlled-drinking findings leaves a purified scientific data base that points clearly in the opposite direction [31,32,36].

From this perspective, it is doubtful this debate will be resolved along decisive evidentiary lines. An alternate model of this debate, therefore, is that each side represents a different cultural view, where culture may be defined in terms of traditional ethnic and national terms, but also in terms of professional and scientific cultures.

Scientific frameworks for interpreting remission—explanatory cultures

Scientists with different views and working in different eras may not be evaluating the same questions in terms of comparable measures. The evolution to the Helzer et al. [35 study from the Rand reports [14,15] suggests a complete shift in the conception of what being a controlled drinker means between research conducted in the 1970s and the 1980s. A single period of heavy drinking (involving as few as 4 days) in the previous 3 years was sufficient to disqualify subjects in the Helzer et al. study from the moderate-drinking category. At the same time, drinking anything less than an average of 10 months a year during these years also disqualified subjects as moderate drinkers. Both these cut-off points for controlled drinking differed drastically from those imposed in the Rand reports.

Perhaps an even starker contrast with Helzer et al.'s and other current definitions and conceptions of controlled drinking and remission is provided in Goodwin et al.'s [13] report on 93 alcoholic felons eight years after their release from prison. Goodwin et al. found that 'frequency and quantity of drinking could be omitted without affecting the diagnosis [of alcoholism]' (p. 137). Instead, their measures focused on binge drinking, loss of control, and legal consequences and social problems associated with drinking. This study classified 38 of the prisoners to be in remission: 7 were abstinent and 17 were classified as moderate drinkers (drinking regularly while 'rarely getting intoxicated'). Also classified as being in remission were eight men who got drunk regularly on weekends, and another six who had switched from spirits to beer and still 'drank almost daily and sometimes excessively'. None of these men, however, had experienced alcohol-related social, job or legal problems in the prior 2 years.

The Goodwin et al. analysis might be said to be incompatible with any contemporary views of alcoholism. The alcoholism concept has become more rigidly defined as a self-perpetuating entity, so that no clinical model accepts the idea that the alcoholic in remission can reduce alcoholic symptoms while drinking regularly or heavily. For example, the one outcome study in the post-Rand period cited by Taylor et al. [36] that provided support for controlled drinking, by Gottheil et al. [30], defined controlled drinking as drinking on no more than 15 of the last 30 days with no intoxication. Goodwin et al. instead interpreted their data with an existential view of their subjects' lives. That is, subjects substantially improved their lives in terms of very central and concrete measures: this highly antisocial group no longer got arrested or got in other kinds of trouble when drunk in a way that had previously marred their lives. (Nordström and Berglund [66] present a related discussion of 'atypical' alcohol abuse in improved 'Type II' alcoholics.)

Helzer, Robins et al.'s [35] definition of and findings about remission in alcoholism also contrasts with the same two chief investigators' (Robins, Helzer et al. [67]) notable research with narcotics addicts. In their study of American soldiers who had been addicted to narcotics in Vietnam, these investigators asked the question 'Does recovery from addiction require abstinence?' Their findings: 'Half of the men who had been addicted in Vietnam used heroin on their return, but only one-eighth became readdicted to heroin. Even when heroin was used frequently, that is, more than once a week for a considerable period of time, only one-half of those who used it frequently became readdicted' (pp. 222—223). Abstinence, they found, was not necessary—rather, it was unusual—for recovered addicts.

The controlled use of heroin by former addicts (indeed, controlled heroin use by anyone) might be considered a more radical outcome than the resumption of controlled drinking by alcoholics. The image of heroin addiction is of a persistently high need for and intake of the drug. Thus, although veterans might use the drug to become intoxicated more than once a week, Robins et al. could classify them as non-addicted when these users regularly abstained without difficulty. This is quite a different model of remission from that Helzer et al. applied to alcoholism. It seems that different explanatory cultures prevail for narcotic addiction and alcoholism, although there has always been an abundance of evidence from naturalistic research that heroin addicts—like alcoholics—often voluntarily enter and withdraw from periods of heavy narcotics usage [61]. Interestingly, one of the important thrusts in alcoholism theory and research has been the development of a model of alcohol dependence based on intense periods of heavy drinking and the appearance of withdrawal symptoms upon cessation of drinking [49] — a replica of the narcotics addiction or drug dependence model.

Treatment cultures

One of the remarkable aspects of the Rand studies was that so much controlled drinking appeared in a patient population treated in centers where abstinence almost certainly was emphasized as the only acceptable goal. The first Rand report contrasted those who had minimal contact with treatment centers and those who received substantial treatment. Among the group with minimal contact who also did not attend AA, 31% were normal drinkers at 18 months and 16% were abstinent, while among those who had minimal contact and attended AA, there were no normal drinkers. Several other studies have found less contact with treatment agencies or AA is associated with greater frequency of CD outcomes [12,29,68]. Similarly, none of Vaillant's clinical population became controlled drinkers; among those in his community population who did so, none relied on a therapy program.

Pokorny et al. [10], on the other hand, noted with surprise that they found so much controlled drinking among patients treated in a ward that conveyed the view that life-long abstinence was absolutely necessary. In the Pokorny et al. study, abstinence was the typical form of remission immediately after discharge, while controlled drinking became more evident the more time that had elapsed since treatment. This pattern suggests more controlled drinking will appear the longer patients are separated from abstinence settings and cultures. In an unusually long (15 year) follow-up reported in the 1970s, Hyman [69] found as many treated alcoholics were drinking daily without problems as were abstaining (in each case 25% of surviving ambulatory subjects). This and other findings from recent long-term follow-up studies [39,40] directly contradict the notion that controlled drinking becomes less likely over the life span.

Similar increases in controlled drinking over time have also been noted in patients treated with behavior therapy aimed at controlled drinking [41]. The learning theory interpretation of these data is that patients improve with practice their use of the techniques they have been taught in therapy. One interpretation, however, can account for long-term increases in controlled drinking after both kinds of therapy: the longer people are out of therapy of any kind, the more likely they are to develop new identities other than those of alcoholic or patient and thereby to achieve a normal drinking pattern. This pattern will not appear, of course, when patients continue to be involved (or subsequently become involved) in standard abstinence programs. For example, nearly all patients in the Sobells' study later entered abstinence programs, as a result of which many patients actively rejected controlled-drinking and the therapists who taught it to them when questioned later [70].

Nordström and Berglund found abstainers reported less internal control of behavior and less social stability. In this long-term follow-up study of a treated population, abstinence outcomes prevailed initially and those who became controlled drinkers showed little improvement after treatment, despite advantages (such as social stability) that ordinarily predict favorable treatment outcomes. However the majority of the subjects who achieved remission gradually shifted from alcohol abuse to controlled drinking, in most cases 10 and more years following treatment. Since average age of onset of problem drinking was nearly 30, with treatment following on the average 5 years later, CD remissions apparently occurred most often when subjects were 50 and 60 years old. Indeed, this corresponds with the age period when a large number of untreated drinkers show remission for their drinking problems [71]. In a sense, Nordström and Berglund's subjects seem to have relied on their social stability and internal behavioral orientation to reject treatment inputs and to persevere in their drinking until it attenuated with age.

The analyses by Elal-Lawrence et al. [42] and by Orford and Keddie [43] suggest different possibilities for the reduction of controlled-drinking through participation in abstinence programs. Elal-Lawrence emphasized the goodness of the match between treatment goal and patients' beliefs and experiences: when these were aligned, patients succeeded better at either abstinence or controlled-drinking; when they were opposed, relapse was most likely. In this case, forcing a person who does not accept abstinence into a treatment framework that accepts only abstinence can eliminate controlled drinking but will have little impact on the numbers who successfully abstain. Orford and Keddie, on the other hand, emphasized primarily the persuasion of patients that they can attain one goal or the other. In this model, the more intense and consistent the persuasion effort toward one type of outcome, the greater will be the prevalence of that outcome.

Helzer et al. [35] presented as one possibility in their research that 'For any alcoholics who are capable of drinking moderately but are incapable of abstinence, treatment efforts directed only at the latter goal will be doomed to failure' (p. 1678). These investigators offered little support for this idea on the grounds that so few patients achieved the study's definition of moderate drinking, although none was encouraged to do so. In other words, their research did not directly test this idea as a hypothesis. However, their absolute remission rate for those in alcoholism treatment of 7% might be considered evidence that conventional treatment discourages non-abstinence outcomes without producing an increase in abstention.

Sanchez-Craig and Lei [72] compared the success of abstinence and CD treatment for problem drinkers with lighter and heavier consumption. They found lighter problem drinkers did not differ in successful outcomes between the two treatments, but that heavier drinkers did better in CD treatment. Abstinence treatment did not succeed generally in encouraging abstinence for any group, while it did reduce the likelihood of heavier drinkers becoming moderate drinkers. Unlike the other recent studies reported here that have found controlled drinking among alcohol-dependent patients, this study was limited to 'early-stage problem drinkers' and classified subjects according to self-reported drinking levels. Nonetheless, a later reanalysis of the data (Sanchez-Craig, private communication, November 24, 1986) found that the same results held for level of alcohol dependence, including some drinkers with high levels of dependence.

Miller [73] has presented a theoretical review of motivational issues in treatment. Conventional alcoholism treatment dictates goals and rejects self-assessments by clients—such as that they can moderate their drinking— that contradict prevailing treatment philosophy. A body of experimental and clinical evidence indicates that such an approach attacks clients' self-efficacy [74,75], and that commitment to action is enhanced instead when therapy accepts and reinforces clients' perceptions and personal goals. The large majority of patients refuse or prove unable to cooperate with the insistence in conventional treatment programs that they abstain. The therapy then defines this as failure and, paradoxically, attributes the failure to the absence of patient motivation.

Non-treatment cultures and denial

Other data support the idea that less involvement in therapy is a positive prognosticator of controlled use patterns. Robins et al. [67] found that the large majority of formerly narcotic-addicted subjects became controlled or occasional heroin users, while Helzer et al. [35] found controlled drinking was almost non-existent among alcohol patients. Helzer et al.'s subjects were all hospitalized, while subjects in Robins et al. seldom underwent treatment. Indeed, Robins et al. concluded their paper with the following paragraph:

Certainly our results are different from what we expected in a number of ways. It is uncomfortable presenting results that differ so much from clinical experience with addicts in treatment. But one should not too readily assume that differences are entirely due to our special sample. After all, when veterans used heroin in the United States two to three years after Vietnam, only one in six came to treatment. (p. 230)

Waldorf [76] found the principal difference between heroin addicts who achieved remission on their own or through treatment was that the latter considered abstinence essential, while the former often tried narcotics again.

Goodwin et al. [13], in finding a non-abstinent remission rate of 33% among untreated alcoholics (a rate dwarfing non-problem drinking rates in such treated populations as Davies' [1] and the Rand reports [14,15]), were also aware that their results violated treatment precepts and wisdom. The investigators sought another explanation 'rather than conclude that treatment had adverse effects on alcoholics', while noting 'symptomatically the untreated alcoholism may be just as severe' as that which drives some to treatment (p. 144) (subjects in this study were all categorized as 'unequivocal alcoholics'). Goodwin et al. did not, however, report how their untreated alcoholics differed from treated alcoholics in ways that influenced outcomes. The group of felons that Goodwin et al. studied seemed especially unlikely to accept therapy and conventional treatment goals. The possibility is that this therapeutic recalcitrance contributed to their unusually high CD rates.

Cynical wisdom is that those who refuse to seek treatment are practicing denial and have no chance at remission. Roizen et al. [77] examined the remission of drinking problems and alcoholism symptoms in a general population of men at two points 4 years apart. There were both substantial drinking problems and substantial remission of drinking problems across the board for this subject population. Nonetheless, when the investigators eliminated treated alcoholics, of 521 untreated drinkers only one who displayed any drinking problems at point 1 was abstaining 4 years later. Room [78] analyzed this and other puzzling discrepancies between the alcoholism found in clinical populations and problem drinking described by survey research. Once treated drinkers are removed from such surveys, almost no cases appear of the classic alcoholism syndrome, defined as the inevitable concurrence of a group of symptoms including loss of control. The non-appearance of this syndrome is not due to respondents' denial of drinking problems in general, since they readily confess a host of drinking problems and other socially disapproved behaviors.

Room [78] discussed how such findings seemingly indicate that all of those with fully developed alcoholism have entered treatment. Mulford [79] examined comparable data gathered for both clinical alcoholics and general population problem drinkers. Whereas 67% of the clinical population reported the three most common clinical symptoms of alcoholism from the Iowa Alcoholic Stages Index, 2% of the problem drinkers did so (which translates into a general population rate of less than 1%). About three-quarters of the clinical population reported loss of control, while the general population prevalence rate was less than 1%. Mulford summarized: 'The findings of this study indicate that the prevalence of persons in the general population having the symptoms of alcoholism like clinic alcoholics is probably around 1%, as Room [78] has speculated'. Furthermore, Mulford maintained, 'If 1.7 million Americans are already being treated for alcoholism, there would appear to be little unmet need for more alcoholism treatment' (p. 492).

A more radical explanation for these data, of course, is that problem drinkers may only report the full alcoholism syndrome after, and as a result of, having been in treatment. In his anthropological study of Alcoholics Anonymous, Rudy [80] noted the typical explanation for the more severe and consistent symptomatology reported by AA members relative to non-AA problem drinkers is that 'AA affiliates have more complications or that they have fewer rationalizations and better memories. However, there is another possible explanation for these differences: members of AA may learn the alcoholic role of AA ideology perceives it' (p. 87). Rudy observed "AA alcoholics are different from other alcoholics, not because there are more 'gamma alcoholics' or 'alcohol addicts' in AA, but because they come to see themselves and to reconstruct their lives by utilizing the views and ideology of AA" (p. xiv). Rudy cited the confusion new AA members often showed about whether they had undergone alcoholic blackout—a sine qua non for the AA definition of alcoholism. Recruits were quickly instructed that even the failure to recall blackout was evidence for this phenomenon, and those who became actively engaged in the group uniformly reported the symptom.

Data presented by natural remission studies suggest that untreated drinkers, even those reporting severe addiction and alcoholism problems, frequently achieve remission—perhaps as frequently as do treated addicts and alcoholics. These drinkers may best be characterized by a preference for dealing with addictive problems in their own ways, rather than by the classical concept of denial. A study by Miller et al. [81] bears on this question of patient self-identification and outcome. This study (like others discussed in this article) examined the relationship between CD outcomes and severity of alcohol dependence and the possibility of controlled drinking by heavily dependent drinkers. Miller et al. reported follow-up of from 3 to 8 years for problem drinkers treated with CD therapy. Twenty-eight percent of the problem drinkers were abstinent compared with only 15% who became 'asymptomatic drinkers'.

This level of controlled drinking is far below that Miller and Hester [23] previously reported from CD therapy. On the other hand, although subjects were solicited on the basis that they were not severely alcohol dependent, 76% of this sample was judged alcohol dependent according to appearance of withdrawal signs and 100% according to appearance of tolerance, two-thirds were classified either gamma or delta alcoholics, and three-quarters had reached the chronic or crucial stages of Jellinek's [82] developmental model of alcoholism. As a result, 11 of 14 of asymptomatic drinkers 'were clearly diagnosable as manifesting Alcohol Dependence, and nine were classifiable at intake as either gamma (3) or delta (6) alcoholics'. Thus, although the CD rate from this therapy was unusually low, the population in which this outcome appeared was strongly alcoholic, unlike the typical CD clients Miller and Hester had described.

Miller et al.'s work differed from other recent studies cited in this article in finding that level of alcohol dependence was strongly related to outcome. However, in keeping with several of these studies, the strongest single predictor was 'intake self-label', or clients' self-assessment. Indeed, despite the high level of alcohol dependence in asymptomatic drinkers, 8 of 14 described themselves as not having a drinking problem! What appears to have occurred in this study is that the denial of often quite severe alcohol problems in a group who acknowledged a need to change their drinking habits was a positive predictor of achieving a very strict definition of controlled-drinking (no signs of alcohol abuse or dependence for 12 months). Other psychological research suggests that those who see their problems as having remediable causes are more likely to overcome problems in general [83].

We see in both natural groups and treated patients who deny they are alcoholic that people regularly refuse to turn over either their labeling or their therapeutic goals to others. This refusal is tied in very basic ways to both the person's outlook and prognosis. Furthermore, to identify this attitude as anti-therapeutic (as by labeling it denial) is not justified according to the lack of success of treatment that runs counter to patients' personal beliefs or goals or according to people's demonstrated ability to change their behavior in line with their own agendas. One study of respondents in a typical community offering almost no CD service found a number of people who reported having eliminated a drinking problem without entering treatment [84]. Most of these self-cures had reduced their drinking. A majority of these subjects, not surprisingly, claimed controlled drinking was possible for alcoholics. A large majority of those from the same community who had never had a drinking problem thought such moderation was impossible, the view held by an even larger majority who had been in treatment for alcoholism.

National cultures

National differences exist in views of controlled drinking, or at least in the acceptance of discussions of controlled drinking as a possible outcome for alcoholism. Miller [85] emphasized that European audiences he spoke to— particularly in Scandinavia and Britain—were a world apart from those in the United States in their belief that CD therapy could be valid for even severely alcohol-dependent drinkers. He noted a similar readiness to utilize CD therapy in non-European countries such as Australia and Japan. Miller found that only in Germany among the European nations he visited, where alcoholism treatment was hospital-based and largely medically supervised, did the commitment to abstinence as the sole goal of alcoholism treatment approach the climate in America.

Miller may have sampled in Britain and Scandinavia non-medical specialists (including psychologists, social workers and others) who gave a skewed picture of attitudes toward controlled drinking in their countries. For example, medical approaches in Britain may not differ substantially from those in America. An editorial in the leading British medical publication, Lancet, concluded in 1986 (relying heavily on Helzer et al.'s findings [35]) that the idea 'that abstinence is the only generally viable alternative to continued alcoholism has received convincing support' [86, p. 720]. Some British psychologists who favor the alcohol dependence concept have also claimed severe alcohol dependence rules out the possibility of controlled drinking [38].

Nonetheless, national differences in this regard seem to be real. Although not based on a systematic survey, Nathan—a behaviorist—reported 'there is no alcoholism center in the United States using the technique [CD therapy] as official policy' [16, p. 1341]. This would contrast dramatically with a survey of British treatment facilities [87] showing that 93% accepted the value of CD treatment in principle, while 70% actually offered it (the survey included Councils on Alcoholism which, in the United States, are the greatest seat of opposition to controlled drinking). A survey of treatment facilities in Ontario, Canada—a nation influenced as it were from both directions— revealed an intermediate level (37%) of acceptance of controlled drinking by alcoholism programs [88].

Orford [89] detected an overall movement in Britain toward the ' abandonment of 'alcoholism' as a disease analogy, and the legitimizing of reduced or more sensible drinking as a possible goal' (p. 250), a trend not at all visible in the United States. Orford furthermore analyzed some national differences in this respect:

In Britain,....only a tiny minority of men abstain totally from alcohol....in other parts of the world abstinence is more acceptable even for younger men—Ireland, the USA with its relatively recent history of prohibition and the stronger influence of Puritanism than in Britain, and of course the Islamic world. (p. 252)

Perhaps as a result of such national differences, most of the notable refutations of CD outcomes in the 1980s have been American-based (the major exception being the work of Edwards, a psychiatrist, and his colleagues [32,34]), while recent findings of substantial controlled drinking among treated alcoholics have been almost exclusively European in origin (with one exception [41]).

How exactly these differences in national climates influence the outlooks of individual practitioners and researchers is captured in a report Miller sent from Europe [90] as he analyzed the culture shock he experienced:

Addressing audiences of alcoholism professionals [in Britain] on the subject of controlled drinking, I was astounded to find that my ideas which are seen as so radical in America were regarded as quite non-controversial, if not a bit old-fashioned....Here in Norway, where A.A. has never really obtained a strong foothold, I likewise find an openness and excitement about new models and approaches....It is difficult to appreciate the immensity of the effects of our current zeitgeist upon theory, research and practice until one steps outside of this pervasive milieu.... What I had not appreciated was the extent to which my own perspectives had been influenced by America's nearly total dedication to the Alcoholics Anonymous view of drinking problems....(pp. 11—12)

Investigator variables

Ethnic and national views very strongly affect attitudes towards alcohol and drinking practices both cross-culturally [91] and within individual countries with diverse populations, such as the United States [33]. There are national and ethnic variations in acceptance of the disease view of alcoholism: for example, Jewish Americans seem especially resistant to the idea that alcoholism is an uncontrollable disease [92]. Although analyzing research results in terms of investigators' ethnic origins runs counter to both scientific custom and democratic traditions in America, it would seem the ethnic, regional, and national differences that apply to drinkers themselves could also affect scientists and clinicians in America and elsewhere.

Another investigator variable that may affect CD findings is professional training and background. Although there are some exceptions in the United States [6,7] (and perhaps more in Europe [40]), anti-CD findings and perspectives have most often been announced by physicians. Among psychologists, although behaviorists have been those most visible in conducting research from a non-disease framework, the behavioral identification of differential goals based on client characteristics has focused increasingly on severity of drinking problems [49,93]. Other, more psychodynamically oriented therapists may be more open to social, cognitive and personality determinants in controlled drinking, and perhaps to be more accepting of controlled drinking overall. For example, in a survey of alcoholism services in a Western city, Vance et al. [84] found that although treatment agencies almost never did so, 7 out of 8 private psychologists questioned offered controlled drinking as a regular option in treatment.

Patient variables: Expectations and cultural background

The single most important prognosticator of CD behavioral training indicated by Miller and Hester [93] was severity of drinking problems or alcohol dependence, an assessment in keeping with current clinical wisdom in the field. However, these authors gave little attention to the expectations and outlooks—including self-assessment and beliefs about alcoholism—that Miller et al. [81], Heather et al. [63,64], Orford and Keddie [42], and Elal-Lawrence et al. [43] found most important to outcomes. Subjective variables such as expectations may underlie or mediate other client traits and outcomes in alcoholism. For example, Brown [94] found that changed expectations about the effects of alcohol predicted the degree of both abstinence and controlled drinking following treatment; Miller et al. [81] reported similar data. When patients no longer looked to alcohol to provide necessary or welcome emotional benefits, they were more successful both at abstaining and reducing their drinking. Similarly, the work of several researchers discussed in this article has shown clients' expectations about the possibility of achieving controlled drinking or abstinence affects the prevalence of these outcomes.

Considered as an objective indicator, past success at moderate drinking could indicate a less severe variety of alcoholism. Orford and Keddie and Elal-Lawrence et al., however, viewed these factors as operating through their influence on patients' expectation of achieving success through one style of remission over the other. In this case, objective and subjective versions of the same variable point in the same direction. In other cases, predictions from considering the same factor either objectively or subjectively may be opposed. Such a case is provided by family history of alcoholism. Miller and Hester [93] indicated family history of alcoholism should probably be considered as predicting greater success at abstinence. However, two research teams—Elal-Lawrence et al. and Sanchez-Craig et al. [95]—have reported finding that such positive family histories led to greater success at controlled drinking.

Miller and Hester considered family history to be indicative of an inherited strain of alcoholism and to favor abstinence (certainly a strong trend of thought in the United States today), while the results of these other non-American studies suggested instead that having examples of alcohol abuse alerted people to the need to respond to a drinking problem at an early stage. Vaillant [33] did not find that number of alcoholic relatives predicted whether alcohol abusers achieved abstinence or controlled drinking. He did find ethnic background (Irish vs. Italian) affected these outcomes which he analyzed as the result of global differences in views of drinking between these cultures. Such cultural differences affect basic outlooks and responses to treatment. Babor et al. [96] found French clinical populations did not accept the disease viewpoint that American alcoholics in treatment endorsed (French-Canadians were intermediate to the two groups). Within the United States different ethnic and religious groups display different symptomatology and severity of problems in alcoholism treatment as well as different prognoses and aftercare conduct [97].

Social, ethnic and cultural differences are rarely considered in matching clients with treatment or tailoring treatment to clients however. Nor are other differences in patient outlook like those discussed in this section usually taken into account. Clients who have a choice will probably gravitate toward treatment and counselors whose views are compatible with their own. Most often however, those with alcohol problems do not have any choices in treatment options [98]. At the same time real differences in acceptance of efforts at controlled drinking may exist below the surface of apparent unanimity. Gerard and Saenger [53] reported highly variable rates of controlled drinking depending on the specific treatment site studied (from no such drinkers to twice as many controlled drinkers as abstainers). Yet the rate was not influenced by the type of treatment the center supposedly practiced.

The United States is a pluralistic society and significant ethnic and individual differences in attitudes toward drinking and toward dealing with alcohol problems will never disappear entirely no matter what standard wisdom dictates. For the most part these differences are sources of conflict and impediments both to scientific understanding and to agreement on and success at achieving treatment goals. The analysis in this article is a plea for bringing such cultural differences to the surface, where they may increase the power of scientific analysis and the efficacy of treatment.


It is impossible to explain the major variations in alcoholism treatment and outcomes and particularly controlled-drinking outcomes—variations over time, cross-culturally, according to investigator and treatment environment—without reference to the explanatory framework that prevailed in a particular research setting. These frameworks—or explanatory cultures— are the result of different ethnic and national attitudes toward alcohol, of various professional outlooks and of changing attitudes about appropriate research methods standards and results that characterize different scientific eras. By their nature these explanatory cultures are not open to scrutiny by their members. Rather such Zeitgeists simply pervade the assumptions and thinking of culture members sometimes to such a degree that they become received opinion that only those in another cultural setting are able to recognize, let alone to question.

Analysis of the various cultures that play a role in determining treatment outcomes could enable us to remove explanatory cultures as an impediment to understanding and instead incorporate them in our scientific models, as well as making them useful ingredients in treatment. A number of cultural factors that affect controlled-drinking research findings and outcomes have been analyzed, and are summarized in the accompanying table (see Table 1).

At the same time that this analysis offers an optimistic view of the possibility of utilizing a cultural dimension in explaining alcoholism remission, it also indicates the difficulty in overcoming cultural inertia and beliefs about drinking and treatment. In this sense, positive behavioral, psychological, and sociological findings about controlled-drinking outcomes and treatment are cultural aberrations that have never really had a chance to have a major impact on American thinking. There is no reason to expect this to change, and certainly research findings by themselves will not be sufficient to bring about such change.


Table 1. Cultural Factors in Controlled-Drinking Outcomes
Cultural Dimensions More + toward CD (a) More - toward CD
National Culture Most European and developed nations (e.g. Australian, Japanese) [85] British [87,89] Canadian [88] German [85] American [16]
Ethnicity and other subcultural groups in America Italian and other Mediterranean and low-alcoholism groups [33,92] Irish, conservative Protestant, dry regions, low SES [14,71,89]
Professional culture Sociological [77-79] Psychodynamic [12,52,55,94] Behavioral [54,59,93] Medical [33,86]
Era (b) 1970 - 1976, post-1986? 1960 - 1970 1976 - 1980 pre-1960     1980-1986

(a) The labels 'more' or 'less' positive toward controlled drinking are, obviously, relativistic statements and do not mean that controlled drinking was the dominant approach in any category or time span.

(b) Of all the variables, 'era' is the hardest to pin down, since research is conducted over years and reporting of completed research can take additional years; nonetheless, this paper argues different attitudes toward controlled drinking are palpable at different times and are real influences on scientific findings and reports.


Archie Brodsky and Haley Peele assisted me in the preparation of an earlier draft of this article, and Nick Heather, Reid Hester, Alan Marlatt, Barbara McCrady, William Miller, Peter Nathan, Goran Nordström, Ron Roizen, Robin Room, Martha Sanchez-Craig, and Mark and Linda Sobell provided me with helpful information and comments.


  1. D.L. Davies, Q.J. Stud. Alcohol, 23 (1962) 94.
  2. G. Edwards, Drug Alcohol Depend., 15 (1985) 19.
  3. R. Roizen, The great-controlled drinking controversy, in: M. Galanter, (Ed.), Recent Developments in Alcoholism (Vol. 5), Plenum, New York, 1987, pp. 245­279.
  4. I. Zwerling and M. Rosenbaum, Alcoholic addiction and personality (nonpsychotic conditions), in: S. Arieti (Ed.), American Handbook of Psychiatry (Vol. 1), Basic Books, New York 1959, pp. 623­644.
  5. D.J. Myerson, Q.J. Stud. Alcohol, 24 (1963) 325.
  6. M.L. Selzer, Q.J. Stud. Alcohol, 24 (1963) 113.
  7. M.L. Selzer and W.H. Holloway, Q.J. Stud. Alcohol, 18 (1957) 98
  8. N. Giesbrecht and K. Pernanen, Sociological perspectives on the alcoholism treatment literature since 1940, in: M. Galanter (Ed.), Recent Developments in Alcoholism (Vol. 5), Plenum, New York, 1987, pp. 175­202.
  9. E.M. Pattison, Non-abstinent drinking goals in the treatment of alcoholics, in: R.J. Gibbons et al. (Eds.), Research Advances in Alcohol and Drug Problems (Vol. 3), Wiley, New York 1976, pp. 401-455.
  10. A.D. Pokorny, B.A. Miller and S.E. Cleveland, Q.J. Stud. Alcohol, 29 (1968) 364.
  11. M.A. Schuckit and G.A. Winokur, Dis. Nerv. Syst., 33 (1972) 672.
  12. W. Anderson and O. Ray, Abstainers, non-destructive drinkers and relapsers: One year after a four-week in-patient group-oriented alcoholism treatment program, in: F. Seixas (Ed.), Currents in Alcoholism (Vol. 2), Grune and Stratton, New York, 1977.
  13. D.W. Goodwin, J.B. Crane and S.B. Guze, Q.J. Stud. Alcohol, 32 (1971) 136.
  14. D.J. Armor, J.M. Polich and H.B. Stambul, Alcoholism and Treatment, Wiley, New York, 1978.
  15. J.M. Polich, D.J. Armor and H.B. Braiker, The Course of Alcoholism: Four Years After Treatment, Wiley, New York, 1981.
  16. S. Peele, Am. Psychol., 39 (1984) 1337.
  17. G.R. Caddy and S.H. Lovibund, Behav. Ther., 7 (1976) 223.
  18. H.H. Schaefer, Psychol. Rep., 29 (1971) 587.
  19. M.B. Sobell and L.C. Sobell, Behav. Res. Ther., 11 (1973) 599.
  20. M.B. Sobell and L.C. Sobell, Behav. Res. Ther., 14 (1976) 195.
  21. E.M. Jellinek, The Disease Concept of Alcoholism, Millhouse, New Haven, 1960.
  22. W.R. Miller, J. Stud. Alcohol, 44 (1983) 68.
  23. W.R. Miller and R.K. Hester, Treating the problem drinker: Modern approaches, in: W.R. Miller (Ed.), The Addictive Behaviors: Treatment of Alcoholism, Drug Abuse, Smoking, and Obesity, Pergamon Press, Oxford, 1980, pp. 11­141.
  24. N. Heather and I. Robertson, Controlled Drinking, Methuen, New York,.1981.
  25. A.R. Lang and G.A. Marlatt, Problem drinking: A social learning perspective, in: R.J. Gatchel (Ed.), Handbook of Psychology and Health, Erlbaum, Hillsdale, NJ, 1982, pp.121 - 169.
  26. W.R. Miller and R.E. Muñoz, How to Control Your Drinking (Second Edition), University of New Mexico Press, Albuquerque, 1982.
  27. A. Paredes, D. Gregory, O.H. Rundell and H.L. Williams, Alcoholism Clin. Exp. Res., 3 (1979) 3.
  28. E.J. Bromet and R. Moos, Br. J. Addict., 74 (1979) 183.
  29. J.W. Finney and R.H. Moos, J. Stud. Alcohol, 42 (1981) 94.
  30. E. Gottheil, C.C. Thornton, T.E. Skoloda et al., Follow-up study of alcoholics at 6, 12 and 24 months, in: M. Galanter (Ed.), Currents in Alcoholism (Vol. 6), Treatment, Rehabilitation and Epidemiology, Grune & Stratton, New York, 1979, pp. 91­109.
  31. M.L. Pendery, I.M. Maltzman and L.J. West, Science, 217 (1982) 169.
  32. G. Edwards, J. Stud. Alcohol, 46 (1985) 181.
  33. G.E. Vaillant, The Natural History of Alcoholism, Harvard University Press, Cambridge, MA, 1983.
  34. G. Edwards, A. Duckitt, E. Oppenheimer et al., Lancet, 2 (1983) 269.
  35. J.E. Helzer, L.N. Robins, J.R. Taylor et al., N. Engl. J. Med., 312 (1985) 1678.
  36. J.R. Taylor, J.E. Helzer and L.N. Robins, J. Stud. Alcohol, 47 (1986) 115.
  37. P. Nathan and R.S. Niaura, Behavioral assessment and treatment of alcoholism, in: J.H. Mendelson and N.K. Mello (Eds.), The Diagnosis and Treatment of Alcoholism (Second Edition), McGraw-Hill, New York, 1985, pp. 391­455.
  38. T. Stockwell, Br. J. Addict., 81 (1986) 455.
  39. R.J.R. McCabe, Alcohol Alcoholism, 21 (1986) 85.
  40. B. Nordström and M. Berglund, J. Stud. Alcohol, 48 (1987) 95.
  41. R. G. Rychtarik, D.W. Foy, T. Scott et al., J. Consult. Clin. Psychol., 55 (1987) 106.
  42. J. Orford and A. Keddie, Br. J. Addict., 81 (1986) 495.
  43. G. Elal-Lawrence, P.D. Slade and M.E. Dewey, J. Stud. Alcohol, 47 (1986) 41.
  44. N. Heather, B. Whitton and I. Robertson, Br. J. Clin. Psychol., 25 (1986) 19.
  45. D.E. Beauchamp et al., J. Stud. Alcohol, 41 (1980) 760.
  46. R.J. Hodgson et al., Br. J. Addict., 75 (1980) 343.
  47. J.E. Brody, N.Y. Times, Jan. 30,1980, p. 20.
  48. R. Room, Sociological aspects of the disease theory of alcoholism, in: R.G. Smart, F.B. Glaser, Y. Israel et al. (Eds.), Research Advances in Alcohol and Drug Problems, Vol. 7, Plenum, New York, 1983, pp. 47­91.
  49. R. Hodgson and T. Stockwell, The theoretical and empirical basis of the alcohol dependence model: A social learning perspective, in: N. Heather, I. Robertson and P. Davis (Eds.), The Misuse of Alcohol, New York University, New York, 1985, pp. 17­34.
  50. G.R. Caddy, H.J. Addington, Jr. and D. Perkins, Behav. Res. Ther., 16 (1978) 345.
  51. D.R. Cook, J. Stud. Alcohol, 46 (1985) 433.
  52. B.J. Fitzgerald, R.A. Pasewark and R. Clark, Q.J. Stud. Alcohol, 32 (1971) 636.
  53. D.L. Gerard and G. Saenger, Out-patient Treatment of Alcoholism: A Study of Outcome and Its Determinants, University of Toronto Press, Toronto, 1966.
  54. P.E. Nathan and B.S. McCrady, Drugs and Society, 1 (1987) 109.
  55. E.M. Pattison, Addict. Behav., 1 (1976) 177.
  56. E. Gottheil, C.C. Thornton, T.E. Skoloda and A.L. Alterman, Am. J. Psychiatry, 139 (1982) 560.
  57. G. Edwards, J. Orford, S. Egert et al., J. Stud. Alcohol, 38 (1977) 1004.
  58. R. Caetano, Drug Alcohol Depend., 15 (1985) 81.
  59. T. Stockwell, D. Murphy and R. Hodgson, Br. J. Addict., 78 (1983) 145.
  60. M.S. Goldman, S.A. Brown and B.A. Christiansen, Expectancy theory: Thinking about drinking, in: H.T. Blane and K.E. Leonard (Eds.), Psychological Theories of Drinking and Alcoholism, Guilford, New York, 1987, pp. 181­226.
  61. S. Peele, The Meaning of Addiction: Compulsive Experience and Its Interpretation, Lexington Books, Lexington, MA, 1985.
  62. G.A. Marlatt, B. Demming and J.B. Reid, J. Abnorm. Psychol., 81 (1973) 233.
  63. N. Heather, M. Winton and S. Rollnick, Psychol. Rep., 50 (1982) 379.
  64. N. Heather, S. Rollnick and M. Winton, Br. J. Clin. Psychol., 22 (1983) 11.
  65. M.B. Sobell and L.C. Sobell, Behav. Res. Ther., 22 (1984) 413.
  66. G. Nordström and M. Berglund, Br. J. Addict., in press.
  67. L.N. Robins, J.E. Helzer, M. Hesselbrock and E. Wish, Vietnam veterans three years after Vietnam: How our study changed our view of heroin, in: L. Brill and C. Winick (Eds.), Yearbook of Substance Use and Abuse (Vol. 2), Human Sciences Press, New York, 1980, pp. 213 - 230.
  68. J. Orford, E. Oppenheimer and G. Edwards, Behav. Res. Ther., 14 (1976) 409.
  69. H.H. Hyman, Ann. N.Y. Acad. Sci., 273 (1976) 613.
  70. S. Peele, Psychol. Today, April (1983) 38.
  71. D. Cahalan, I.H. Cisin and H.M. Crossley, American Drinking Practices, Rutgers Center of Alcohol Studies, New Brunswick, NJ, 1969.
  72. M. Sanchez-Craig and H. Lei, Br. J. Addict., 81 (1986) 505.
  73. W.R. Miller, Psychol. Bull., 98 (1985) 84.
  74. H.M. Annis and C.S. Davis, Self-efficacy and the prevention of alcoholic relapse, in: T. Baker and D. Cannon (Eds.), Addictive Disorders, Praeger Publishing Co., New York, in press.
  75. S.G. Curry and G.A. Marlatt, Building self-confidence, self-efficacy, and self-control, in: W.M. Cox (Ed.), Treatment and Prevention of Alcohol Problems, Academic Press, New York, pp. 117­137.
  76. D. Waldorf, J. Drug Issues, 13 (1983) 237.
  77. R. Roizen, D. Cahalan and P. Shanks, Spontaneous remission among untreated problem drinkers, in: D. Kandel (Ed.), Longitudinal Research on Drug Use: Empirical Findings and Methodological Issues, Hemisphere Publishing, Washington, DC, 1978, pp. 197­221.
  78. R. Room, Treatment seeking populations and larger realities, in: G. Edwards and M. Grant (Eds.), Alcoholism Treatment in Transition, Croom Helm, London, 1980, pp. 205­224.
  79. H.A. Mulford, Symptoms of alcoholism: Clinic alcoholics vs. problem drinkers at large, 34th International Congress on Alcoholism and Drug Dependence, Calgary, 1985.
  80. D.R. Rudy, Becoming Alcoholic, Southern Illinois University Press, Carbondale, 1986.
  81. W.R. Miller, A.L. Leckman. M. Tinkcom et al., Long-term follow-up of controlled drinking therapies, paper presented at Annual Meeting of American Psychological Association, Washington, DC, 1986.
  82. E.M. Jellinek, Q.J. Stud. Alcohol, 13 (1952) 673.
  83. S. Nolen-Hoeksema, J.S. Girgus and M.E.P. Seligman, J. Pers. Soc. Psychol., 51 (1986) 435.
  84. B.K. Vance, S.L. Carroll, P. Steinsiek and B. Helm, Alcoholism, abstinence, and self-control: A social psychological exploration of alcohol problems, poster presentation at Convention of the Oklahoma Psychological Association, Tulsa, Oklahoma, 1985.
  85. W.R. Miller, Haunted by the Zeitgeist: Reflections on contrasting treatment goals and concepts of alcoholism in Europe and the United States, in: T.F. Babor (Ed.), Alcohol and Culture: Comparative Perspectives from Europe and America, Annals of the New York Academy of Sciences (Vol. 472), New York, 1986, pp. 110­129.
  86. Lancet, March 29 (1986) 719.
  87. I.H. Robertson and N. Heather, Br. J. Alcohol Alcoholism, 17 (1982) 102.
  88. B.R. Rush and A.C. Ogborne, J. Stud. Alcohol, 47 (1986) 146.
  89. J. Orford, Br. J. Addict., 82 (1987) 250.
  90. W.R. Miller, Bull. Soc. Psychol. Addict. Behav., 2 (1983) 11.
  91. D.B. Heath, Cross-cultural studies of alcohol use, in: M. Galanter (Ed.), Recent Developments in Alcoholism (Vol. 2), Plenum, New York, 1984, pp. 405­415.
  92. B. Glassner and B. Berg, J. Stud. Alcohol, 45 (1984) 16.
  93. W.R. Miller and R.K. Hester, Matching problem drinkers with optimal treatments, in: W.R. Miller and N. Heather (Eds.), Treating Addictive Behaviors: Processes of Change, Plenum Press, New York, 1986, pp. 175­203.
  94. S. Brown, J. Stud. Alcohol, 46 (1985) 304.
  95. M. Sanchez-Craig, D. Wilkinson and K. Walker, Theory and methods for secondary prevention of alcohol problems: A cognitively based approach, in: W.M. Cox (Ed.), Treatment and Prevention of Alcohol Problems, Academic Press, New York, 1987, pp. 287­331.
  96. T.F. Babor, M. Hesselbrock, S. Radouco-Thomas et al., Concepts of alcoholism among American, French-Canadian, and French alcoholics, in: T. F. Babor (Ed.), Alcohol and Culture, Annals of the New York Academy of Science, New York, 1986, pp. 98­109.
  97. T.F. Babor and J.H. Mendelson, Ethnic/religious differences in the manifestation and treatment of alcoholism, in: T.F. Babor (Ed.), Alcohol and Culture, Annals of the New York Academy of Science, New York, 1986, pp. 46­59.
  98. M. Sanchez-Craig, Br. J. Addict., 81 (1986) 597.