Counting the Saved
Statistical Claims in Official Recovery Literature Versus Independent Follow-up Studies
The Document and Its Purpose
The founding text of AA is not a research report. First published in 1939, it is a testimonial and instructional volume written by members for prospective members. Its numbers function as rhetorical instruments intended to persuade a reader that the program works. Anyone weighing its statistics should begin with that fact about genre: the text is evidence of what early members believed and wished to convey, not evidence of measured outcomes.
What the Volume Claims
Two claims carry the statistical weight. From the chapter “How It Works,” the text states that “rarely have we seen a person fail who has thoroughly followed our path.” In the Foreword to the Second Edition, written in 1955 about the fellowship’s early decades, it reports that of those who came and really tried, about half recovered at once, a quarter after relapses, and most of the rest improved while remaining involved.
Both claims share a structure. They supply no denominator, no stated follow-up period, no operational definition of recovery, and no method for counting.
What the Volume Admits
The text is more candid than its popular reputation sometimes suggests. It concedes that some people are constitutionally unable to be honest with themselves and that others suffer from grave emotional and mental disorders. It does not promise universal success.
This candor performs analytic work. Failure is attributed to the individual — to those who “cannot or will not” give themselves to the program. The claim is thereby protected from disconfirmation: anyone who does not recover is understood not to have followed the path thoroughly. A success rate defined over “those who really tried” is circular, because the evidence of trying is judged by the outcome.
What the Volume Omits
A source-critical reading asks what a competent reader would need and does not receive.
The denominator is missing. “Those who really tried” excludes everyone who attended once, twice, or many times and left. The reference population consists of survivors of attrition.
The timeframe is missing. Recovery is reported at no fixed interval, so a person recovered for six weeks and one recovered for six years count the same.
The sample is undescribed. Early membership was small, self-selected, and heavily white, male, and middle-class. The text does not say how its early figures were gathered.
A comparison group is absent. Nothing shows what would have happened to the same people without the program.
Attrition data are silent. The text does not report how many left.
What AA’s Own Later Data Show
The fellowship’s triennial membership surveys paint a different picture. They sample current attendees, so they measure retention rather than success among all who ever came. They show wide variation in length of continuous recovery, with a substantial share of members under a year. These sources remain retention studies. Surveying people who remain tells you about people who remain.
What Independent Studies Found
Independent research has been mixed rather than uniformly negative.
Early controlled trials from the 1960s and 1970s, including Ditman’s work with court-referred offenders and the Brandsma trial, found groups assigned to AA-style conditions doing no better, and sometimes worse, than alternatives. Both studies have well-known limitations: coerced participation, small samples, and imperfect fit between the intervention delivered and the voluntary nature of ordinary meetings.
Longitudinal work by Vaillant followed men over decades and found that recovery rates among those who attended AA were not clearly superior to natural remission. He treated AA as one helpful route among several rather than a unique cure. Attendance was associated with better outcomes, but the design could not separate the effect of attendance from the characteristics of those who continued to attend.
Get Fringe Logbook’s stories in your inbox
Join Medium for free to get updates from this writer.
Project MATCH, a large multi-site trial in the 1990s, compared twelve-step facilitation with cognitive-behavioral and motivational therapies. Outcomes were broadly comparable across conditions, with some advantage for twelve-step facilitation on continuous abstinence in certain arms.
The 2020 Cochrane review by Kelly and colleagues examined dozens of studies of AA and professionally delivered twelve-step facilitation. It concluded that structured, manualized twelve-step facilitation performs as well as or better than other active treatments for continuous abstinence, often at lower cost. The authors note the limits of the evidence: the strongest findings concern clinician-delivered facilitation more than ordinary voluntary meeting attendance, and trial participants differ from the general population of people who seek help.
Large population surveys show that many people recover, a substantial share of them without formal treatment. This undercuts any claim that a single program uniquely accounts for recovery.
The Core Methodological Problem
All outcome claims about AA face selection bias. People who attend regularly are more motivated, on average, than those who do not, so correlations between attendance and recovery cannot be read as causation. Randomized trials address this problem but struggle with a voluntary, anonymous fellowship that cannot be easily assigned or monitored. The authors of the founding text cannot be faulted for failing to solve a problem that researchers still face. They can be noted for presenting uncontrolled impressions as rates.
A Fair Reading
The statistics in the founding text are best read as testimony — early, sincere, produced without research training, and shaped by the movement’s need to persuade. Independent studies suggest that AA and its clinical derivatives help many people, perform comparably to other approaches, and are less miraculous than the founding claims imply. The honest summary is neither that the program works for seventy-five percent nor that it does not work. The official claim is unverifiable as stated, while the independent evidence supports a more modest and more useful conclusion.
Reading Checklist for Recovery Claims
What is the denominator, and who was excluded?
How long was follow-up, and how was recovery verified?
Is there a comparison group, and how were participants assigned?
Does the source measure outcomes or retention?
Who produced the figure, and what did they need it to show?
References
Alcoholics Anonymous. Alcoholics Anonymous. 4th ed. New York: Alcoholics Anonymous World Services, 2001. Foreword to the Second Edition and Chapter 5, “How It Works.” https://www.aa.org/the-big-book
Ditman, K. S., et al. “A Controlled Experiment on the Use of Court Probation for Drunk Arrests.” American Journal of Psychiatry 124, no. 2 (1967): 160 — 163. https://doi.org/10.1176/ajp.124.2.160
Brandsma, J. M., M. C. Maultsby, and R. J. Welsh. Outpatient Treatment of Alcoholism: A Review and Comparative Study. Baltimore: University Park Press, 1980.
Vaillant, G. E. The Natural History of Alcoholism Revisited. Cambridge, MA: Harvard University Press, 1995. See also Vaillant, G. E., et al. “Prospective Study of Alcoholism Treatment: Eight-Year Follow-up.” American Journal of Medicine 75, no. 3 (1983): 455 — 463. https://doi.org/10.1016/0002-9343(83)90349-2
Project MATCH Research Group. “Matching Alcoholism Treatments to Client Heterogeneity: Project MATCH Posttreatment Drinking Outcomes.” Journal of Studies on Alcohol 58, no. 1 (1997): 7 — 29.
Kelly, J. F., K. Humphreys, and M. Ferri. “Alcoholics Anonymous and Other 12-Step Programs for Alcohol Use Disorder.” Cochrane Database of Systematic Reviews 2020, Issue 3. Art. No.: CD012880. https://doi.org/10.1002/14651858.CD012880.pub2
Alcoholics Anonymous. Triennial Membership Surveys (various years). New York: Alcoholics Anonymous World Services. Summaries available via https://www.aa.org/membership-surveys
