There continues to be an ongoing gap between how addiction gets treated and how it actually works.
Knowledge and behaviour are handled by different systems. Insight lives largely in the prefrontal cortex — the part of the brain that plans, weighs consequences, and holds “I know this is bad for me.” But addiction increasingly reshapes subcortical circuits involved in craving, habit, and stress reactivity. Under cue exposure or stress, those older circuits can override the newer, knowledge-holding ones. This is why someone can recite their relapse triggers perfectly in a therapist’s office or recount pages of 12-step literature from memory and still act against that knowledge the moment they’re actually standing in the triggering situation.
Wiers and colleagues’ work on “dual process” models of addiction — explicit cognition versus automatic, implicit approach biases — captures this well. Cognitive bias modification research grew directly out of the finding that explicit knowledge doesn’t touch these automatic processes.
The transition from goal-directed use to habit. Everitt and Robbins’ work describes how repeated substance use shifts control from goal-directed brain circuits (which respond to consequences and knowledge) to habitual, and eventually compulsive, circuits centred in the dorsal striatum. Compulsive behaviour is, by definition, less responsive to knowing better. Decades of use can entrench this shift very deeply.
For example:
Early: goal-directed. Bad day → “I want to relax” → decide to drink. The action is chosen and tied to an expected outcome. If consequences got bad enough, the person could still stop — this runs through prefrontal cortex and ventral striatum.
Middle: habit. After years of stress → drink → relief, the sequence no longer needs a decision. Walking in the door at 6pm triggers pouring a drink before any conscious “do I want this?” moment happens. The cue now drives the action directly. Control has shifted to dorsolateral striatum — same circuitry as driving a familiar route on autopilot.
Late: Compulsive. In Everitt and Robbins’ animal studies, drug-seeking gets paired with a mild shock. Most animals stop once punished. But a subset — roughly a third, matching the rate of severe addiction in humans — keeps seeking the drug even though it now reliably delivers a shock. That’s the model for compulsion: behaviour persisting despite known, immediate harm.
The human version: Someone with DUI’s, lost relationships, health scares — every piece of proof the behaviour is destructive — still relapses. Not from lack of knowledge; they’ve lived the consequences repeatedly. The circuitry driving the behaviour at this stage doesn’t consult the consequence-tracking system the way it did early on.
This is also why pure education or insight-oriented therapy often under-performs at this stage — it’s arguing with a system that isn’t listening. Interventions that interrupt the cue-response link itself (contingency management, environmental restructuring, craving-blunting medication, removing triggers) tend to have more traction, because they target the habit/compulsion circuit rather than trying to out-inform it.
Koob and Volkow’s allostasis model. Their research reframes chronic addiction as a shift from positive reinforcement (seeking a high) to negative reinforcement (avoiding a increasingly severe dysphoric, anxious, irritable state without the substance). Long-term users aren’t chasing a reward anymore — they’re fleeing a state. That flight response is fast, automatic, and largely indifferent to intellectual understanding.
Unresolved comorbidity. A large fraction of people with treatment-resistant, decades-long relapse patterns have under-treated trauma, complex PTSD, personality disorders, or conditions like ADHD sitting underneath the substance use. If the underlying driver isn’t directly addressed, coping skills training treats the symptom while the engine keeps running. This is a common finding in dual-diagnosis literature — outcomes are much worse when co-occurring conditions go unaddressed even amid extensive substance-focused treatment.
Recovery capital, not just recovery knowledge. William White’s work on recovery capital argues sustained recovery depends on social, financial, physical, and community resources — stable housing, relationships, purpose, employment — not just internal coping skills. Someone can have excellent DBT (Dialectical Behaviour Therapy) skills and still relapse repeatedly if their environment, relationships, or life structure haven’t meaningfully changed.
Relapse as expected, not exceptional. Prochaska and DiClemente’s stages-of-change model treats relapse as a normal part of the cycle, not a failure of knowledge. But when relapse repeats over decades despite intensive treatment, researchers increasingly look at severity markers — poly-substance use, genetic loading, age of onset, impulsivity and risk-taking — that predict a harder course regardless of how much psychoeducation someone has absorbed.
The consistent variable across all of this: Insight is necessary but nowhere near sufficient. Addiction at that level of chronicity behaves more like a deeply conditioned neurobiological and habitual process than an information deficit. That’s also part of why chronic relapse is increasingly framed like a chronic illness — closer to diabetes or hypertension in its relapsing-remitting course — rather than something insight alone should be expected to resolve.
Here’s what the literature and clinical guidelines point to for people who’ve cycled through standard treatment repeatedly without sustained success. I’ll organise by approach type.
Pharmacological — often underused
- For opioid use disorder, methadone and buprenorphine substantially outperform abstinence-only approaches in treatment-resistant populations — this is some of the strongest evidence in the entire addiction field, yet many long-term relapsers have never actually been offered maintenance medication, often due to program philosophy (abstinence-only 12-step programs) rather than clinical reasoning.
- For alcohol, naltrexone (blocks the reward), acamprosate (reduces post-acute withdrawal dysregulation), and disulfiram (aversive) are under-prescribed. Naltrexone specifically blunts the reward signal that drives compulsive use — targeting the biology rather than relying on willpower.
- Emerging: psychedelic-assisted therapy (psilocybin, ketamine) is showing promise in trials for alcohol and other use disorders, though this is newer evidence and not yet standard of care everywhere.
Interrupting the cue-response cycle directly
- Contingency management — tangible rewards for verified abstinence — has some of the best effect sizes in the literature, particularly for stimulant use disorders where no strong pharmacological option exists. It works by directly competing with the habit circuit rather than arguing with it.
- Environmental restructuring — changing routes, routines, social circles, physically removing cues — matters more than intellectual insight at this stage, because the behaviour has become cue-triggered rather than decision-driven.
Treating what’s underneath
- Systematic assessment for under-treated trauma, complex PTSD, ADHD, and personality disorders. Trauma-focused therapies (EMDR, prolonged exposure, Seeking Safety) address the driver rather than just the symptom, and are often skipped in standard programs focused on the substance alone.
- Integrated dual-diagnosis treatment (treating psychiatric conditions and addiction concurrently, by the same team) outperforms sequential or parallel treatment where the person bounces between separate providers.
Building recovery capital, not just coping skills
- Housing First and supported employment models — stabilising the environment before or alongside treatment — show better outcomes than clinical treatment alone in chronic, treatment-resistant cases. William White’s recovery capital framework: outcomes track with social support, stable housing, purpose, and community as much as with therapy itself.
- Recovery community centres and mutual-aid alternatives beyond 12-step (SMART Recovery, Refuge Recovery) matter for people whose repeated 12-step involvement hasn’t translated to sustained sobriety — fit between the person and the model isn’t universal.
Reframing the model of care itself
- Treating addiction like a chronic relapsing-remitting illness (the diabetes/hypertension model) rather than an acute, curable event changes what “treatment resistant” even means. Long-term, low-intensity, indefinite maintenance care (medication, periodic check-ins, ongoing peer support) shows better outcomes than repeated cycles of intensive short-term treatment followed by discharge.
- Harm reduction as a bridge or destination — not necessarily full abstinence — for people who haven’t achieved abstinence-based sobriety after years of trying. This reduces mortality and morbidity even when abstinence isn’t currently achievable, and can sometimes become a pathway toward it later.
- Multidimensional Family Therapy / systemic approaches — treating the family/relational system rather than the individual alone — show better outcomes for people whose relapse is tangled up in unchanged relational dynamics.