In a 1993 controlled comparison of therapist styles with problem drinkers, Miller, Benefield and Tonigan found that the more the therapist confronted, the more the client was drinking a year later. Confrontation did not merely fail; it predicted worse drinking at twelve months. A family member confronting at eleven at night starts further behind that baseline. The protocol below survives that finding: timing criteria, an opening script, answers to the five predictable deflections, and the single-request close.
SACENDU surveillance data show that in Gauteng in the first half of 2018, 59% of admissions to substance-use treatment centres were referred by themselves, family or friends, against only 2% by a healthcare professional. The family is the country's main admissions channel. The referral is almost always driven by one steady relative, not a committee.
Timing criteria
Three conditions, all required: sober, private, not mid-conflict. Sober, because an intoxicated person retains little of what is said and will use any display of anger to end the conversation. Private, because an audience converts any concession into a public defeat; no children in the room. One subject only: the drinking, no other grievances. The test is whether the person could repeat tomorrow what was said tonight; if not, wait.
For two weeks before the conversation, keep a written record of dates, incidents, amounts and money. "You said you had two beers, but the bank statement shows two bottle-store purchases" is specific and verifiable; "you drink too much" is deniable.
The opening script
Facts, not labels: "I want to talk about your drinking. In the last month you missed two Mondays at work, you slept on the couch three times, and you drove after drinking on Saturday. I am not willing to treat this as normal." A label, even an accurate one, invites an argument about the word instead of the pattern. "Always" and "never" are omitted because they hand the person a single exception to dispute.
The five deflections
The replies are predictable. Families report the same five, with little variation. Each gets one answer, then a return to the record.
"You drink too." Answer: "We can talk about my drinking another day. Today is about yours."
"You are exaggerating." Read one dated entry from the record and stop. The dated record answers the objection; no further argument is required.
"You are the reason I drink." Answer: "That may be worth raising with a counsellor. It does not change what happened on these dates."
"I can stop whenever I want." Answer: "Good. Then the assessment this week will confirm it." The claim becomes the request.
"This is not the time." Answer: "Name a time this week and I will hold it." A postponement without a date is a refusal.
Answer each deflection once, then return to the list; the conversation is lost the moment the subject changes.
The single-request close
Close on one request verifiable within a week: a GP assessment, a phone call to a treatment centre, alcohol out of the house. "If you ever drink again I am leaving" fails in a first conversation because most families will not execute it on a first refusal, and an unexecuted threat teaches the drinker that the family's statements carry no consequence.
Consequences belong in the second conversation, after the single request has been refused. A credible consequence meets three criteria: the family can execute it alone, will execute it, and can sustain it indefinitely.
A worked example. A first conversation ends in refusal. The second, a week later, attaches: "I will no longer phone your employer when you cannot work. The next absence, you explain yourself." That passes all three criteria. "I am leaving" passes none unless it is true.
The staged intervention underperforms
The lounge full of relatives reading letters is a film convention, not a method. In 1999 Miller, Meyers and Tonigan ran a randomized trial comparing three ways families can engage a treatment-refusing drinker, each with the same 12 hours of family contact. CRAFT (Community Reinforcement and Family Training), which coaches the family in the skills above, engaged 64% of drinkers into treatment; the staged Johnson-style confrontation engaged 30%; Al-Anon facilitation 13%. Training the family engaged more than twice as many drinkers as staging the confrontation.
If the person agrees to stop
A dependent daily drinker must not quit alone at home. Withdrawal seizures typically occur 8 to 48 hours after the last drink, often before anything looks serious. Delirium tremens begins around 48 to 72 hours after the last drink in roughly 3 to 5% of hospitalised withdrawal cases; mortality is around 1 to 5% treated and at least 15% untreated (StatPearls, 2023). The first stop after a successful conversation is a medical assessment, not a locked liquor cabinet. Where the drinking history calls for it, Harmony arranges medical detox off-site before the residential programme begins.
The endpoint
The first conversation is not measured by surrender. Its endpoint: dated facts stated without shouting, one verifiable request made, the second conversation left open. That is the stage the referral statistics depend on.
Sources and further reading
Miller, Meyers & Tonigan, randomized trial of three family strategies for engaging treatment-refusing drinkers (CRAFT vs Johnson vs Al-Anon), Journal of Consulting and Clinical Psychology, 1999
Miller, Benefield & Tonigan, controlled comparison of therapist styles with problem drinkers, Journal of Consulting and Clinical Psychology, 1993
SACENDU / South African Medical Research Council, treatment admissions surveillance data, 2018
StatPearls (NCBI Bookshelf), Alcohol Withdrawal Syndrome, 2023
Gad Avnon
Director and Lead Addiction Counsellor, Harmony Retreat
This article is general information, not medical advice. Alcohol and some drug withdrawals can be medically dangerous. If someone is acutely unwell or at risk, seek medical help immediately.