Westbrook Recovery

Relapse Warning Signs in 2026: The Signs and the Measured Map Behind Them

Last updated

Relapse is a process with stages, and the warning signs that matter belong to the earliest one, where nothing looks like using. The recognised early signs are about self-care collapsing: bottling feelings up, isolating, going to meetings without speaking, sleeping and eating badly, putting everyone else first. Behind the signs sits a measured map: across the original 1985 sample of 137 people and its later replications, the situations preceding relapse cluster the same way every time: negative emotional states first at about 37%, social pressure second at about 24%, interpersonal conflict third at about 15%. Read the signs as a trend, not a moment, and the map tells you what to do about each.

The short version

  1. The first stage contains no thoughts of using — emotional relapse is self-care collapsing, which is why it is missed and why it is the cheapest place to intervene
  2. Three situation types precede most relapses — negative emotional states ~37%, social pressure ~24%, interpersonal conflict ~15% in the original 1985 count of 137 people
  3. The pattern has survived 41 years of replications — a 2026 study of 428 smokers found the same top three in the same order
  4. About 70% of relapses happen in the first three months — the window when the early signs matter most is exactly when they are easiest to miss
  5. Returning to an old dose can be fatal — tolerance falls during abstinence; a relapse is not like a flare-up of asthma, however tidy the comparison

Step by step

  1. Learn your own early signs by name

    Not the generic list: yours. Which meeting you skip first, what your sleep does, who you stop answering. Write them down while things are calm; the list is for the version of you who will not feel like reading it.

  2. Name the situation, not just the mood

    The measured map exists to be used: is this a negative emotional state, a pressure situation, or a conflict? Each has a different counter-move, and naming which one you are in halves the work.

  3. Tell one person the same day

    Sponsor, counselor, a friend from a meeting, the SAMHSA helpline if nobody else. The stage where telling is easy is the stage where it works best; silence is itself one of the signs.

  4. Change the setting, not just the intention

    If the situation is social pressure, leave the setting. If it is a conflict, schedule the conversation instead of holding it. If it is exhaustion, the fix is sleep before it is willpower.

Three cards for the stages of relapse: emotional weeks before with self-care collapsing, mental days to weeks of internal argument, physical the moment of use with the least leverage
The leverage is early. The warning signs live in the stage where nobody is thinking about using yet.
Six cards of stage-one signs: bottling feelings up, isolating, meetings without speaking, poor sleep and food, everyone else’s problems first, restless irritable discontent
None of the six mentions the substance. That absence is exactly why this stage is missed.
The HALT check as four letters: hungry, angry, lonely, tired, the four states that lower the guard
Three of four are body states with mechanical fixes. The check converts a vague slide into a specific one.
Two cards for stage two: the using side with cravings and bargaining, the sober side with the memories and reasons
The side that wins is the side that gets fed. Contact feeds the sober side; silence feeds the other.
Two cards on timing: about 70 percent of relapses happen in the first three months of the 2026 replication; the early signs matter most when they are easiest to miss
The risk window is front-loaded. That is when the boring signs deserve the most attention.

The relapse map, measured

Our analysis

Every warning-signs page lists the signs. We went to the documents that counted what actually precedes relapse, and the same three situations top the table across three substances and forty-one years of replications.

37%
of episodes preceded by negative emotional states
the top row of the original 1985 table (N=137), and still the top row in the 2026 replication at 38.65%
24%
preceded by social pressure
18% for alcohol, 25% for smokers, 34% for heroin: it scales with the social life of the substance
15%
preceded by interpersonal conflict
the third row everywhere; arguments with partners, family, employers
59% vs 42%
intrapersonal versus interpersonal
most relapses start inside the person, not between people
~70%
of relapses in the first three months
from the 2026 replication: the risk window is front-loaded
41 years
the pattern has held
1985 original to 2026 replication, same top three in the same order

How we got this: Read in full on 15 October 2026: the Marlatt and Gordon 1985 table via the Penn State training document (verified verbatim), the Cambridge 2026 replication, the University of Amsterdam determinant analysis, and the 2015 Yale Journal of Biology and Medicine stages model. The table above is the original count, cell by cell.

What these numbers cannot tell you: The 1985 sample is small (N=137) and old; replications cover smokers and alcohol, not every substance. Percentages describe what preceded episodes, not any individual’s risk. Your own warning-signs list, written with your counselor, outranks any published table.

The measured map of relapse situations

Before the signs, the terrain. In 1985 Marlatt and Gordon published the original analysis of high-risk situations for relapse, built from interviews with 137 people across three substances: 70 recovering from alcohol, 35 from smoking, 32 from heroin. Each relapse episode was assigned the situation that preceded it. The table below is that original count, read from the primary document on 15 October 2026, and it has an unusual property for this field: it keeps replicating. A 2026 study of 428 smokers in treatment found the same top three situations in the same order, and roughly 70% of the relapses inside the first three months.

  • Negative emotional states

    Alcohol (n=70)
    38%
    Smokers (n=35)
    43%
    Heroin (n=32)
    28%
    All (N=137)
    37%
  • Social pressure

    Alcohol (n=70)
    18%
    Smokers (n=35)
    25%
    Heroin (n=32)
    34%
    All (N=137)
    24%
  • Interpersonal conflict

    Alcohol (n=70)
    18%
    Smokers (n=35)
    12%
    Heroin (n=32)
    13%
    All (N=137)
    15%
  • Urges and temptations

    Alcohol (n=70)
    11%
    Smokers (n=35)
    6%
    Heroin (n=32)
    none recorded
    All (N=137)
    8%
  • Positive emotional states

    Alcohol (n=70)
    none recorded
    Smokers (n=35)
    14%
    Heroin (n=32)
    16%
    All (N=137)
    9%
  • Testing personal control

    Alcohol (n=70)
    9%
    Smokers (n=35)
    none recorded
    Heroin (n=32)
    none recorded
    All (N=137)
    4%
  • Negative physical states

    Alcohol (n=70)
    3%
    Smokers (n=35)
    none recorded
    Heroin (n=32)
    9%
    All (N=137)
    4%
The original taxonomy, by substance: share of relapse episodes by preceding situation (Marlatt & Gordon 1985, N=137; dashes mean none recorded)

What the numbers say in words

Most relapses start inside the person, not between people: intrapersonal situations account for 59% of the original sample against 42% interpersonal. The single largest row everywhere is negative emotional states: anger, anxiety, frustration, boredom, loneliness. That is the measured answer to why the warning signs below look like a bad month rather than like craving.

Where the substances differ

The one cell that moves is social pressure: 18% for alcohol, 25% for smokers, 34% for heroin, the highest number in the heroin column. Substances with a scene, a circle, a route back through people carry the interpersonal risk; solitary ones carry the emotional kind. If your substance has a social life attached to it, social pressure is your top row.

Why the 2026 replication matters

Forty-one years later, a different country, a different substance, a different treatment era, and the ranking holds: negative emotional states 38.65%, then social pressure 34.46%, with the rest of the taxonomy trailing. When a pattern survives that long it stops being a finding and becomes furniture: you can build on it. The build here is simple: watch your emotional state like a gauge, because the gauge is what the numbers say comes first.

The three stages, and why the order matters

The clinical model the whole field uses, set out in a 2015 Yale Journal of Biology and Medicine paper on relapse prevention, divides the process into three: emotional relapse, mental relapse, physical relapse. In the first, the person is not thinking about using at all. In the second, part of them is. The third is the use itself. The order is the point: each stage is easier to leave than the one after it, and the warning signs live almost entirely in the first two.

Stage one: emotional relapse

Nothing here looks like addiction. The person remembers the last relapse and does not want another; meanwhile their self-care is quietly collapsing. Because there are no thoughts of using, denial is built into the stage, which is why it is so often the people around who notice first. The six recognised signs: bottling up emotions, isolating, not going to meetings, going but not speaking, focusing on other people’s problems, and poor eating and sleeping habits.

Stage two: mental relapse

The internal argument starts. Cravings, remembering the good parts and not the endings, bargaining about controlled use, planning around people who would notice. The paper is blunt about the mechanics: this stage is a war, and the side that wins is the side that gets fed. Telling someone about the argument is the feed for the sober side, which is why the advice at this stage is always contact, not willpower.

Stage three: physical relapse

The use. The paper adds the sentence worth memorising: when people do not understand relapse prevention, they think it involves saying no just before they are about to use, but that is the final and most difficult stage to stop, which is why people relapse. Almost all popular advice aims at this moment. Almost all the leverage sits earlier.

The early signs, read as a trend

The lists online give you the signs; the map above gives you the reading. A single skipped meeting is a scheduling fact. A month of skipped meetings, unanswered texts, takeout and bad sleep is a trend, and the trend is the sign. Read yourself the way the taxonomy says relapse happens: through accumulating negative emotional states, not through sudden temptation.

Isolation is the load-bearing sign

Across the handouts and the clinical lists, withdrawal from contact comes first and predicts most: skipped meetings, unanswered messages, eating alone, dropping the people who would notice. It is also the sign with the clearest counter-move, because contact is the treatment for it and it costs nothing but the sentence out loud.

Irritability, restlessness, discontent

The emotional surface of stage one. From the inside it feels like everyone else became annoying; from the outside it reads as a short fuse. In the taxonomy this is the negative-emotional reservoir filling: the 37% row assembling itself one irritation at a time.

Self-care collapsing

Sleep first, then food, then the small disciplines. The published model treats self-care not as wellness decoration but as the thing emotional relapse actually is. Poor sleep in particular is both a sign and a cause, which is why the boring fixes are the real ones here.

The HALT check

The oldest self-check in recovery survives because it operationalises the largest row of the table: hungry, angry, lonely, tired. Three of the four are body states, not character flaws, and all four push the same direction as the negative-emotional row. The practice is a question, asked honestly, before decisions: which of the four am I right now, and what fixes that one?

Why four body states outrank willpower

Each HALT state is a negative emotional state with a mechanical fix. Hungry is fixed by eating; tired by sleep; lonely by contact; angry by naming it. The check works because it converts a vague slide into a specific one, and specific states have specific exits.

Using it as a daily habit, not a crisis tool

The handbooks that teach HALT teach it as routine: a morning question, not a 2 a.m. one. The value compounds because the same four states are also the earliest visible signs to the people around you, which makes HALT a shared vocabulary between you and them.

The 37-signs list, mapped onto the map

A huge share of the searching for this topic is for one specific document: the 37 warning signs, Terry Gorski’s checklist, circulated in worksheets for decades. The list is real and useful, and it is also misread when read flat. The 37 items are not 37 equal things: they are the three stages and the taxonomy, written out long. Isolating, skipping meetings, poor sleep sit in stage one. Bargaining, romanticising, craving sit in stage two. The value of the long list is granularity; the value of the map is knowing which stage a sign belongs to, because the response differs by stage.

How to use a long checklist without drowning in it

Do not score all 37 weekly. Pick your five: the ones that have preceded your own bad stretches before. The research version of this is the personal warning-signs list that Miller and Harris built for exactly this purpose, turning the long checklist into a short personal one. A short list you actually read beats a long one you frame.

What the list cannot tell you

A checklist cannot weight itself. The measured map does the weighting: if you are tracking limited energy, track the emotional and contact rows first, because those are the rows the numbers say precede most episodes. The list is the vocabulary; the taxonomy is the grammar.

By substance: what shifts and what does not

The signs are the same across substances; the weights are not. The 1985 table’s moving cell is social pressure, and it moves with the social life of the substance: lowest for alcohol at 18%, higher for smokers at 25%, highest for heroin at 34%. The emotional rows barely move at all. Reading your own risk through the right column is more useful than reading the average one.

Alcohol: the conflict and boredom substance

In the original sample the alcohol column leads on interpersonal conflict (18%) and on testing personal control (9%, the only column where it appears). The profile matches the drug: legal, domestic, argued about at home, and endlessly renegotiated by the person who wants to believe one is fine.

Opioids: the people problem

The heroin column’s top cell is social pressure at 34%, and its physical-state row (9%) is the only one that shows up there either. The practical reading: the circle is the risk, and early signs include drifting back toward it. Contacts and places are not nostalgia here; they are the mechanism.

For family and friends reading the signs

You are more likely to see stage one than the person in it, because denial is built into that stage. The handoff matters: said as surveillance, it feeds the hiding; said as care, it opens the talking. If you are reading these signs for someone else, the family playbook is how to get someone into rehab. Name the observable, not the fear: not “I think you are going to relapse” but “I noticed you have skipped three meetings and you are not sleeping.” The first invites a fight; the second invites a sentence back.

What to do with what you notice

Offer the counter-move, not the verdict: food, sleep, a ride to a meeting, your own presence. If the trend continues, the next level is a counselor or a helpline, and making that call yourself for guidance is legitimate and common. The SAMHSA National Helpline is free, confidential, 24/7, and used by families exactly like this.

Protecting yourself inside the watching

Reading someone else’s warning signs is exhausting work, and the family version of stage one exists too: resentment, sleeplessness, obsession with monitoring. Al-Anon and similar groups exist for precisely this. You cannot watch someone sober; you can only keep the door open and yourself intact.

If it happens: the next 24 hours

A lapse is the initial use; a relapse is the return to uncontrolled use. What decides which one you are in happens in the next day, and the deciding factor is whether the episode stops. If home is the risky place right now, the structured option is sober living. Stop it fast: get to a safe place and a safe person, call the sponsor or counselor or helpline, and refuse the self-sentencing. The binge after the lapse is what the shame is for, and declining the shame is the whole move.

The one difference that can kill

Unlike a flare-up of asthma, a relapse after abstinence carries a specific lethal risk: tolerance has fallen, and the old dose is now an overdose. The measured course of the opioid withdrawal itself is on the opioid withdrawal timeline page. This is why the first call after a lapse involving opioids is not optional, and why the naloxone conversation belongs in every relapse plan for opioid users.

Reading the slide afterwards

When it is stable, debrief like the taxonomy: which stage-one signs were missed, which situation type preceded it, what the counter-move should have been. This is not paperwork; it is how the next slide gets interrupted earlier. A relapse plan rewritten after a lapse is stronger than the one written before it, because it now contains your own data.

What we read, and how

On 15 October 2026 we read the primary documents behind this topic rather than the listicles: the full 1985 Marlatt and Gordon table as republished in the Penn State training text (verified cell by cell), the 2026 Spanish Journal of Psychology replication on 428 smokers, the University of Amsterdam analysis of 42 self-reported determinants in alcohol dependence, and the 2015 Yale Journal of Biology and Medicine paper that defines the three stages. The table, the substance-by-substance reading, and the framing of the signs against the measured map are ours.

The limits of this reading

The original sample is from 1985 and small by modern standards; the replications are on smokers and on alcohol, not every substance. The percentages describe the situations preceding episodes, not any individual’s probability. For your own risk, the personal five-signs list written with your counselor outweighs any table, including ours.

Grouped bar chart of the 1985 Marlatt and Gordon table (N=137): negative emotional states lead at 38/43/28 percent for alcohol, smokers and heroin; social pressure 18/25/34; interpersonal conflict 18/12/13
The same top three in the same order across three substances. The pattern has since replicated in 2026 on 428 smokers.
Two cards splitting the 1985 sample: intrapersonal situations 59 percent versus interpersonal 42 percent
Most relapses start inside the person, not between people. Feelings and urges outrank places.
Two cards on who notices first: the person in stage one is not thinking of using, while the people around notice the isolation, irritability, skipped meetings and bad sleep
Denial is part of stage one, which is why the first reader of the signs is usually someone else.
Three cards matching the response to the stage: restore self-care at stage one, tell someone and change the setting at stage two, stop the slide at stage three
The response that works matches the stage you are actually in.

Where this sits in the system of care

Addiction treatment is organised into levels of care, and clinicians choose between them using the ASAM criteria (six dimensions covering withdrawal risk, medical and psychiatric conditions, readiness, relapse risk and, decisively, the environment you return to. Facilities are licensed by the state and many hold additional accreditation from bodies such as the Joint Commission or CARF; that status is worth confirming for any specific programme. Percentages below are our count of how many facilities in the federal directory report offering each level of care.

  1. Detox19.5% of facilitiesmedical withdrawal management, 3–7 daysDetox vs rehab
  2. Residential21.4% of facilitieslive on site, 24-hour staffingInpatient vs outpatient
  3. PHP20–30 structured hours a week, sleep at homePHP vs IOP
  4. IOP9–19 hours a week, usually compatible with workPHP vs IOP
  5. Standard outpatient85.4% of facilities1–8 hours a week, the maintenance levelMutual aid options
“Remaining in treatment for an adequate period of time is critical.”

Programmes for specific groups

Facilities also report dedicated programming, and the coverage is wider than most people expect, worth asking about before assuming a programme will not fit. From our count: 42.7% run programmes for adolescents, 39.2% for veterans, 38% for pregnant and postpartum women, and 25% provide services in Spanish. Medication for opioid use disorder is available at 59.2% (buprenorphine) and 13.7% (methadone).

Cost and coverage vary by level: residential care is the tier insurers scrutinise most and almost always requires prior authorisation before admission. See what your plan covers and how to pay without insurance. Family involvement is a standard part of most programmes, and aftercare planning ( step-down to a lower level, sober living, relapse prevention) should be discussed before admission, not at discharge. If a mental health condition is also present, ask specifically whether the programme is staffed for co-occurring disorders: 69.5% of facilities in the directory report treating them, but that is a self-reported field worth confirming.

Two cards on relapse as a chronic-condition event: rates in the range of other chronic illnesses per NIDA, but unlike asthma a return to an old dose after abstinence can be fatal
Not failure and not fate. And with opioids, not safe to improvise.
Four steps for the 24 hours after a lapse: stop the episode with a safe person, call the sponsor counselor or helpline, decline the self-sentencing, read which stage-one signs were missed
The binge after the lapse is what the shame feeds. Declining it is the whole move.
Two cards for family: name what you see as care, not surveillance; an open door feeds the talking, interrogation feeds the hiding
I notice you skipping meetings invites a sentence back. I think you are relapsing invites a fight.
Two cards for after a relapse: review the map for missed signs and the triggering situation, then adjust the plan with meetings, boundaries or a higher level of care
A plan rewritten after a lapse is stronger than the one written before it: it contains your own data.
Two cards on where help is: the SAMHSA National Helpline free and confidential around the clock, and ongoing contacts like a counselor, sponsor or meeting
The contacts that interrupt the process only work if they exist before the process starts.
Diagram of the step-down pathway through addiction treatment: detox, residential, partial hospitalisation, intensive outpatient, then outpatient and aftercare
Nobody needs every step. The ladder exists so that support can be reduced as stability improves, not so that everyone starts at the top.
Bar chart of how many US treatment facilities offer each level of care: outpatient 85.4%, residential 21.4%, medical detox 19.5%, hospital inpatient 5.9%
Outpatient care is nearly universal; residential and detox are the scarce tiers. That scarcity is part of why they cost what they do and why waits exist.
Bar chart of medication for opioid use disorder availability: buprenorphine at 59.2% of facilities, methadone at 13.7%, certified opioid treatment programmes at 12.8%
The four-to-one gap between buprenorphine and methadone availability is a regulatory fact, not a clinical judgement about which works better.
Checklist of the six ASAM assessment dimensions: withdrawal risk, medical conditions, emotional and behavioural conditions, readiness to change, relapse potential, and recovery environment
This is the framework US programmes and insurers actually use. The sixth dimension — what home is like — is the one families most often underestimate.
Three figures: 69.5% of facilities treat co-occurring disorders, 59.8% offer trauma-informed care, 19.5% run medical detox
Co-occurring capability is self-reported. Ask what it means concretely, which clinicians are on staff, and whether psychiatric medication can be prescribed and reviewed on site.
Checklist of six questions to ask a treatment facility about network status, prior authorisation, all-in cost, medication availability, aftercare and accreditation
How plainly a programme answers these is itself information about how it operates.
Bar chart of payment types accepted by US addiction treatment facilities: cash 91.6%, private insurance 80.7%, Medicaid 83.4%, Medicare 55.2%, state-financed 54.5%, military 49.6%, sliding scale 40.0%, payment assistance 22.5%
Accepting your insurer is not the same as your plan covering the level of care you need. The two questions have to be asked separately.
Table showing when to call 911, 988 or the SAMHSA National Helpline for overdose, immediate danger, suicidal thoughts, severe distress and treatment questions
All three are free and staffed around the clock. 988 is the Suicide & Crisis Lifeline; the SAMHSA National Helpline is 1-800-662-4357.

Common questions

What are the early warning signs of relapse?

The earliest stage has no thoughts of using in it, which is why it is missed. The recognised signs: bottling up feelings, isolating from people and meetings, attending but not speaking, poor sleep and eating, focusing on everyone else’s problems, growing irritability and restlessness. Read as a trend over weeks rather than a single day, these precede most episodes.

What are the three stages of relapse?

Emotional, mental, physical, in that order. Emotional relapse is self-care collapsing while the person is not thinking about using. Mental relapse is the internal argument: cravings, bargaining, planning. Physical relapse is the use. Each stage is easier to leave than the next, which is the entire argument for learning the early signs.

What are the 37 warning signs of relapse?

Terry Gorski’s long checklist, circulated in worksheets for decades. The 37 items are the three stages written out long: isolation and poor self-care in stage one, bargaining and romanticising in stage two. The useful version is not the 37 but your five: the signs that have preceded your own bad stretches, written down while you are calm.

What does HALT stand for?

Hungry, angry, lonely, tired: four states that lower the guard. Each one is a fixable body or contact state, and each feeds the negative-emotional row that tops the measured relapse map. The check works as a daily habit, asked before decisions, not during crises.

Is a lapse the same as a relapse?

No, and the next 24 hours decide which it becomes. A lapse is the initial use; a relapse is the return to uncontrolled use. The counter-moves: stop the episode, reach a safe person the same day, decline the shame spiral that turns one into the other. After opioids in particular, remember that tolerance has fallen and the old dose can kill.

What are the biggest relapse triggers?

Measured rather than guessed: negative emotional states precede about 37% of episodes, social pressure about 24%, and interpersonal conflict about 15%, in the original 137-person count. Forty-one years of replications keep the same top three in the same order. Triggers are less about places and more about states.

What should I do if I notice these signs in someone I love?

Name what you observe, not what you fear: skipped meetings, bad sleep, unanswered texts, not “you are going to relapse.” Offer the counter-move (food, sleep, a ride to a meeting), and if the trend continues call a counselor or the SAMHSA National Helpline for guidance; families use it for exactly this. Protect your own sleep and sanity while you watch.

How do I build a relapse prevention plan?

Short and personal beats long and generic: your five early warning signs by name, your top situations from the measured map (emotional states, pressure, conflict), one person to call the same day, and the counter-moves that fit your life. Rewrite it after any lapse with what you learned; that version is stronger because it contains your own data.

Sources

Compare levels of care

Check what your plan covers

How this page was made

Who wrote it:
the Westbrook Recovery editorial team : researchers and writers, not clinicians. No doctor has reviewed this page, and we do not claim otherwise. What that means in practice: we report what public clinical sources say and show you where to read them, rather than offering a clinical opinion of our own.
Checked against:
7 primary sources listed above: federal agency material and published clinical criteria. Where a figure varies by state, plan or facility, we say so instead of publishing a number that looks authoritative and is not.
Original analysis:
the facility proportions on this page are our own count of SAMHSA’s public directory. The full method is documented , including what we count, what we deliberately do not publish, and an error the approach caught. You can repeat the pull and check the numbers.
Last reviewed:
. This date changes when the content is actually re-checked, not automatically on every visit.
What this page is not:
a medical recommendation, a diagnosis, or a referral. Decisions about levels of care and medication belong with a clinician who knows your history. We do not operate facilities, take placement fees, or receive payment for sending anyone anywhere.

Related on this site

More guides

Weighing levels of care? See PHP vs IOP and what your insurance covers.

This page is information, not medical or legal advice. Laws on commitment and treatment differ by state. If you are in crisis, call or text 988 in the US, or reach the SAMHSA National Helpline at 1-800-662-4357 — free, confidential, 24/7.