Contingency Management: Applied Behavior Analysis's Evidence-Based Contribution to Substance Use Treatment

Why Contingency Management Is Behavior Analysis, Not a Borrowed Tool

Contingency management (rewards for verified behavior change) runs all through substance use research. It is often framed as a tool from some other field. It is not. CM is operant conditioning (behavior shaped by its consequences) put to work on a serious social problem. A target behavior gets checked in an objective way. If the target is met, a reinforcer is delivered. If it is not met, the reinforcer is withheld. Behavior analysts have been doing this for decades.

That framing was the heart of this talk. CM is one of the clearest cases of ABA producing a treatment. It has a strong evidence base in substance use care. The field did not borrow it. The field built it, tested it, and then scaled it.

One warning for readers new to this literature. Much of the CM writing leans hard on brain and reward pathway talk. Foster noted that this language is common in the field and let it be. The behavior analytic read is simpler. Reinforcement that is immediate and certain makes a behavior more likely.

Contingency management is not a borrowed technique that just happens to resemble ABA.

From the talk — Hal Foster

Mapping a CM Session Onto the Three-Term Contingency

A CM session maps onto the three-term contingency (antecedent, behavior, and consequence). The antecedent is a scheduled drug screen. The rule is stated to the client in advance: a negative result earns a reinforcer. The behavior is giving a sample that tests negative for the target drug. The consequence is a reinforcer delivered right away. That may be a voucher, a draw for a prize, or a gift card.

Foster raised a wrinkle that behavior analysts will spot fast. Not using a drug is not a response you can see. So when abstinence (not using the target drug) is the goal, the plan works differently. It looks much like differential reinforcement of other behavior, or DRO (reward for not doing something). The reinforcer follows the absence of a response across a set interval. A lab test verifies it. Foster was careful to label this as an analytic point, not a finding from the CM research.

Plenty of programs reinforce behavior you can see directly. Attendance, medication adherence, and other health promoting behaviors all work as targets. Foster named adherence as a personal favorite target. When attendance is the target, you reinforce a real response instead of an absence. That difference should show up in how you write the plan and the data sheet.

This is definitely an analytic point for discussion and not a claim derived from the contingency management literature.

From the talk — Hal Foster

Four Design Decisions in Every Contingency Management Plan

Four choices define a CM plan, and behavior analysts make these calls already. If you have built a token economy (points traded for rewards later), you already know this work.

First, verify the target in an objective way. A test result or a check-in counts. Self report does not. Some phone apps add a location check-in at a set time and place. Second, deliver the reinforcer as soon as the result is verified. Third, pick a magnitude and a schedule. Roll and Shoptaw reported a pilot in 2006 with people who were dependent on methamphetamine. Schedules that raised the reinforcer with each straight negative result worked better. A positive test reset the amount. The authors called those findings preliminary. Fourth, hold the contingency. A positive or missed test means no reinforcer that day.

Verification will never be airtight, and Foster was blunt about that limit.

For a clinician-facing guide to building a program, Foster pointed to Nancy Petry's 2012 book. It walks through how to select target behaviors and plan a program. Some of its numbers are dated now. The design logic still holds up.

there's always ways to trick these things, but we do the best we can.

From the talk — Hal Foster

The Evidence Base for Stimulant Use Disorder

Stimulant use disorder (problem use of cocaine or meth) covers cocaine and methamphetamine use disorders. For opioid use disorder there are approved medicines, such as methadone and buprenorphine (a medicine for opioid use disorder). Those work well. For stimulant use disorder there is no approved medicine at all. A 2021 paper noted that no medicine for cocaine use disorder is approved in the United States or Europe. The American Society of Addiction Medicine and the American Academy of Addiction Psychiatry published a joint guideline in 2024. It notes that some medicines, including psychostimulants, may be used off-label (not approved for that use). So care rests mostly on behavioral and psychosocial work.

The largest comparison of cocaine treatments ran in JAMA Network Open in 2020. It pooled 157 clinical trials, 402 treatment groups, and 15,842 adults over 22 years. The outcome was objective: a negative cocaine test. Only CM programs were linked to negative tests at a significant level. The odds ratio was 2.13 (how much more likely something is). The 95% confidence interval ran from 1.62 to 2.80 (the likely range of a result). That link held in every sensitivity analysis. Psychotherapy as a category did not show it. Two limits matter. The review covered cocaine only, and the outcome was a test result, not recovery over time.

Two reviews in PLOS One covered cocaine and amphetamine type stimulants together. One was a review of other reviews plus a meta-analysis (pooling results from many studies) of randomized trials. It found enough evidence to support CM for stimulant use disorder and called for better medication research. The second was a network meta-analysis (compares many treatments at once) of psychosocial care. Against treatment as usual, CM plus the community reinforcement approach (builds rewarding drug-free activities) stood out. It was the only option that raised abstinence by the end of treatment. The odds ratio was 2.84, with an interval of 1.24 to 6.51. Dropouts were lower too. The authors flagged study quality and lack of blinding. Foster added one caveat that spans all of this research. Abstinence was the goal in every study. CM does not require that. Reduced use or safer use can be the target instead.

Stimulant use is also common among people taking medicine for opioid use disorder, and it can undercut their care. A 2021 review in JAMA Psychiatry pooled 74 randomized trials and 10,444 adults in that group. CM was linked to abstinence across stimulants, multiple substances, illicit opioids, and cigarettes. It was also linked to better therapy attendance and medication adherence. So one plan can move several socially significant behaviors in the same person. The 2024 guideline goes further. Recommendation five says CM should be a primary component of the treatment plan. A key takeaway in that guideline states CM has shown the best effectiveness for stimulant use disorders.

I do hope that the statement regarding lack of on-label approved medications for stimulant use disorder will render my CEU out of date as soon as possible.

From the talk — Hal Foster

From 1994 Trials to a National Rollout

CM did not appear overnight. In 1994, Stephen Higgins and colleagues published a study titled Incentives Improve Outcome in Outpatient Behavioral Treatment of Cocaine Dependence. In 2005, Nancy Petry and colleagues ran a multi-site study through the National Drug Abuse Treatment Clinical Trials Network. It tested prize-based incentives for people who use stimulants in community outpatient programs.

In 2011, the Department of Veterans Affairs began a national rollout of CM, mostly for stimulant use. A 2018 evaluation covered 94 sites and 2,060 patients. Patients attended more than half of their scheduled sessions. About 91% of the urine samples turned in were negative for the target substance. The authors concluded that these outcomes looked like the ones from controlled trials.

That matters for how you pitch CM to a team. The common worry is that tight research controls carry the effect. Here a large health system got similar numbers in regular clinics. The design features do the work, not the lab.

And that was a large health system delivering a behavioral analytic intervention at scale.

From the talk — Hal Foster

Digital and Telehealth Delivery: What Holds Up and What Breaks

The same features that make CM work make it hard to run in a clinic. One paper noted that CM usually calls for in-person visits two to three times a week. Each visit verifies abstinence and hands over the incentive. That is a heavy load on clients and staff. Clinics also have to stock instant screens and track who earned what and when. For someone with a job or no car, a narrow weekday window is a wall. That is high response effort (how hard a behavior is). Digital platforms try to keep the contingency intact. They move the test and the payout onto a phone or a video call.

Two studies showed gains. Campbell and colleagues ran a multi-site randomized trial in 2014. It appeared in the American Journal of Psychiatry and included 507 adults in 10 outpatient addiction programs. Half got usual care. Half got usual care plus an internet delivered program. It had 65 interactive modules and prize-based incentives tied to abstinence and treatment adherence. The program replaced about two hours of standard care each week. Dropout was lower, with a hazard ratio of 0.72 (relative risk over time). Abstinence was higher, with an odds ratio of 1.62. The effect was larger for people who tested positive at entry, at 2.18. Psychiatric News reported in 2017 that this product became a prescription digital therapeutic (an app a doctor prescribes). It was the first one the FDA cleared. The current version may differ from the tested one.

DeFulio and colleagues in 2021 looked at 121 patients on buprenorphine who also had stimulant use disorder. Half received CM through a smartphone and smart card platform. Half were matched controls. The CM group had higher rates of abstinence and clinic attendance. The gaps reached significance in the last two of four 30-day periods. This was a retrospective comparison, not a randomized trial.

Two other studies show where digital delivery strains. A single arm pilot in Washington State ran a fully remote phone-based program for patients who used methamphetamine. Of 28 patients who consented, 15 started the 12-week program, an uptake rate of 54%. Engagement varied widely and verified abstinence was generally low. No one rated the app negatively, but the authors named onboarding and engagement as real problems. A 2024 telehealth case series with veterans who had cocaine use disorder was smaller still. Most participants could join sessions, run their own oral fluid tests, and show results on camera. Reading results near the test cutoff was sometimes unreliable. Both failures should feel familiar.

You can't reinforce behavior that you cannot measure reliably.

From the talk — Hal Foster

Language That Changes How Clinicians Respond

Foster opened the hour with words, not procedures, and the reason is data. Kelly and Westerhoff gave highly trained clinicians a vignette in 2010. The text was identical except for one phrase. When the person was called a substance abuser, clinicians judged more harshly. They were more likely to see that person as deserving blame and punishment. The same person described as having a substance use disorder (a DSM diagnosis for problem use) drew a different response.

Ashford, Brown, and Curtis tested 1,022 participants. They used an implicit association task (a test of hidden bias) and a social distance measure. Terms like addict, alcoholic, opioid addict, and substance abuser were strongly tied to negative bias. Health professionals held a greater negative association with substance abuser. Recurrence of use carried a stronger positive association than relapse. A small pilot of 54 people in recovery looked at self-labeling. The label a person chose was not significantly tied to outcomes such as recovery capital, self-esteem, or internalized stigma. The authors still urged professionals to drop stigmatizing labels. Foster's read: the evidence is clearest about us, the professionals.

Two rules follow for CM work. CM is built around drug screening, so result language comes up constantly. Call a result positive or negative. Not pass or fail. Not clean or dirty.

Keep diagnosis separate from behavior as well. Using a drug and having a substance use disorder are not the same thing. The DSM-5 text revision (the manual used to diagnose mental health conditions) sets the rule. It requires a pattern of use that causes clinically significant impairment or distress. It takes at least two of 11 criteria in a 12-month period, with severity set by the count. Many people who use drugs do not meet those criteria. Person who uses drugs describes behavior, which should feel natural to a behavior analyst. Diagnosing sits outside our scope, so do not apply the label to anyone who has not been diagnosed. When a client uses a different word for themselves, follow their lead in conversation.

a test result is a measurement and not a description of a person or their value.

From the talk — Hal Foster

Policy Moves, Scope Limits, and Where BCBAs Fit

Policy is catching up. California's public insurance program began covering CM in early 2023 through a structured outpatient program. It runs 24 weeks for people with stimulant use disorder. Participants earn gift cards for screens that are negative for stimulants. Six or more months of recovery support services follow. SAMHSA (a federal behavioral health agency) also raised its grant limit on CM incentives tenfold per patient per year. Incentives must be vouchers, goods, or gift cards, and cash is not allowed. Magnitude drives reinforcer effectiveness, so Foster called that change a very big deal.

The ethics code lines up well with this work. We select interventions that fit behavioral principles and have scientific support. We prioritize positive reinforcement procedures. We practice in our scope, and we enter new areas only after documented study, training, supervised experience, consultation, or co-treatment. We deliver non-behavioral services only with the right education, training, and credentials. Diagnosing or treating substance use disorder falls outside most BCBA (board certified behavior analyst) scopes. Collaboration is in the client's best interest.

Work next to licensed clinicians who have real experience with substance use disorders. Read your state license law too, since scope of practice rules vary by state.

One attendee asked how any plan can compete with a drug as powerful as a stimulant. Foster reframed the goal. CM is for people who want to cut down or try recovery. It does not drag anyone in, and it does not have to be abstinence based. Many clients are contacting very little reinforcement at that point in life. Small earnings for check-ins, meetings, and counseling visits start to add up. Over a month, trust from family or a group can return. Natural reinforcement does not out-muscle a stimulant. It can make the drug less relevant.

the BCBA's contribution to a CM program is most likely going to be in designing the reinforcement schedule, measurement, treatment integrity, and staff training

From the talk — Hal Foster