Reinforcer Value Predicts Little of Resistance to Extinction, Study of 90 Adults Finds
A Vanderbilt team tested whether a progressive ratio reinforcer assessment predicts resistance to extinction. It did, a little, but not enough for clinical use.
A study published on November 8, 2026 in the Journal of Applied Behavior Analysis tested whether measuring how much someone values a reinforcer can predict their resistance to extinction, meaning how long a behavior persists after the reinforcer stops. Joseph Lambert, Allyson Hayward, Caitlin Price and Jessica Logan of Vanderbilt University studied 90 adults with disabilities working for snacks or songs, not challenging behavior. The measure predicted persistence only weakly, and the authors concluded it isn't ready for the BCBAs who plan extinction-based treatment.
Key Dates for the Reinforcer Value and Extinction Study
| Date | What happened | Source |
|---|---|---|
| May 26, 2025 | The manuscript is received by the Journal of Applied Behavior Analysis. | Lambert et al., article history |
| October 21, 2026 | The journal accepts the paper. | Lambert et al., article history |
| November 8, 2026 | The paper is published online, open access, as e70097. | Lambert et al., JABA |
| Winter 2027 | The paper appears in volume 60, issue 1 of the journal. | Lambert et al., JABA |
What the Study Found About Resistance to Extinction
The team measured each participant's reinforcer value three ways, then put responding on extinction and tracked how it persisted. The best of the three measures was the breakpoint from a progressive ratio assessment: the last schedule value at which a person still earned the reinforcer before quitting. In an exploratory median regression, it was a statistically significant predictor of three of the four persistence measures: peak response rate, how long responding took to stop, and how many expected reinforcers a participant "missed" before giving up.
On peak rate, adding the breakpoint to a model that already held the known moderators more than doubled the variation explained, from 13.1% to 28.7%. That was the high point. The authors write that their models "never accounted for more than 30% of the variance," a level of control they say no earlier human operant study of persistence had reached, and still "a far cry from the level needed to be clinically useful."
Who Took Part in the Study
The researchers enrolled 111 adults with disabilities over two years through organizations in two states. Of those, 90 completed every phase used in this analysis, an attrition rate of 18.92%. Sixteen left over scheduling conflicts or loss of interest, two left for unrelated health reasons and one moved. Two more were excluded after researcher fidelity errors during extinction.
Participants' mean age was 40.3 years, ranging from 18 to 78. Seventy-five had an intellectual disability and 16 were autistic, with smaller numbers reported as having other health impairments, traumatic brain injuries, or orthopedic, speech or visual impairments. Sessions ran at a day program, workshop or university setting for 64 participants and at home for 26. Because the work was translational, not treatment, the team chose consenting adults it could pay, at a minimum-wage rate of $7.25 an hour in gift cards.
How the Study Measured Reinforcer Value With a Progressive Ratio Schedule
Each participant picked one reinforcer to work for: a preferred snack or, for people with dietary restrictions, 30 seconds of a favorite song. The target response was chosen to be arbitrary, to reduce the chance it had any history with those reinforcers. That's why 88 participants rolled a die and two squeezed a clothespin. The study calls the value measures relative reinforcer efficacy (RRE), and it compared three of them.
| Measure of reinforcer value | What the participant did | What was recorded |
|---|---|---|
| Demand intensity (NCR) | Ate or listened freely, with no response required | How many reinforcers they consumed |
| Progressive ratio breakpoint | Rolled the die as the price rose by three after every second reinforcer (FR 1, FR 4, FR 7 and so on) | The last schedule value that still earned a reinforcer |
| Fixed ratio 1 consumption | Rolled once per reinforcer for a whole session | How many reinforcers they earned |
A progressive ratio schedule finds the point where a reinforcer stops being worth the rising effort. Each participant ran three progressive ratio sessions, each paired with a control session where responding produced nothing. A separate progressive fixed-ratio analysis held the price constant within a session and raised it across sessions, which identified each person's Pmax, the price that produced the most total responding. In these assessments, sessions ended when a participant stopped for 60 seconds, said they were done or touched a stop card, and none ran into the one-hour limit.
The Five Conditions Before Extinction
The study reused the design of a 2024 paper by Lambert and colleagues, which had already published extinction data for 69 of these participants. People were matched in groups of four by breakpoint and randomly assigned to one of four conditions. Inelastic conditions set the baseline price below Pmax, a rich schedule. Elastic conditions set it above Pmax, a lean one. That was crossed with deprivation: EO participants had consumed little just before extinction, and AO participants had consumed close to their fill. A fifth condition added 22 more participants at Pmax, without randomization.
| Condition | Baseline schedule | Just before extinction |
|---|---|---|
| Inelastic EO | Rich (below Pmax) | Had consumed little |
| Inelastic AO | Rich (below Pmax) | Had consumed close to their fill |
| Elastic EO | Lean (above Pmax) | Had consumed little |
| Elastic AO | Lean (above Pmax) | Had consumed close to their fill |
| Pmax EO (not randomized) | Set at Pmax | Had consumed little |
Extinction ran as one continuous session of up to an hour and ended when responding fell to 10% or less of the last three baseline sessions for 10 minutes, or when the participant dissented. In the 2024 paper, deprived participants persisted longer than those near their fill, especially on lean schedules, yet no condition fully prevented extinction bursts. The authors suspected reinforcer value was one missing piece.
Four Ways to Measure Persistence in Extinction
Persistence has no single definition, so the team tested four: peak response rate during extinction, latency to extinction, peak rate as a proportion of baseline, and reinforcers missed (total extinction responses divided by the baseline schedule value). The table shows how much of the variation in each the models explained, in percent, first with only the known moderators (price and prior consumption), then with each value measure added on its own. The base model holds only the known moderators, "% of base" is peak rate as a proportion of baseline, and "Missed" is reinforcers missed.
| Model | Rate | Latency | % of base | Missed |
|---|---|---|---|---|
| Base model | 13.1 | 16.1 | 12.0 | 13.0 |
| + Breakpoint | 28.7 | 24.7 | 17.9 | 25.1 |
| + FR 1 | 20.9 | 23.5 | 15.6 | 17.6 |
| + NCR | 16.3 | 24.5 | 16.6 | 17.3 |
The breakpoint came out ahead on every outcome, though only just on latency (24.7% against 24.5% for free consumption). Mean interobserver agreement ran from 98.96% to 99.35% across the three phases, scored on 30.85% to 42.12% of sessions, and mean procedural fidelity ran from 99.49% to 99.76%.
Statistically Significant, Not Clinically Useful
The preregistered models assumed normally distributed data, and the team found multiple outliers, so it added an exploratory median regression, which holds up better with skewed data. In it, the breakpoint significantly predicted peak rate (estimate 0.54, p < .01), latency (7.74, p < .01) and reinforcers missed (0.46, p < .01). FR 1 consumption predicted rate and reinforcers missed but not latency. Free consumption predicted only latency. None of the three predicted peak rate as a proportion of baseline.
The raw correlations are modest. Before other variables were controlled for, the breakpoint correlated 0.30 with peak extinction rate, 0.23 with latency, 0.24 with reinforcers missed and -0.05 with proportion of baseline. By our arithmetic, a correlation of 0.30 means the two measures share about 9% of their variance. The authors also report that the relations were "less apparent via visual inspection" of their scatterplots than in the regressions.
Their conclusion is that the findings "do not justify embedding assessments of RRE (e.g., PRA) into standard-of-care assessments (e.g., functional analysis)." Citing other studies, they add two reasons for caution: basic research has found an inverse relation between progressive ratio results and persistence, and exposure to a progressive ratio test may itself change later persistence. The last line of their discussion says they do not recommend that researchers combine progressive ratio and functional analysis methods to predict the persistence of actual challenging behavior.
Why Researchers Wanted to Predict Persistence
Extinction is effective but has side effects, and those side effects have led some researchers to argue for treatments that avoid it. The paper pushes back on avoiding it on principle, noting that "the efficacy of reinforcement-based interventions often depends on contact with extinction." The authors argue that a test showing in advance whose behavior is likely to persist could help clinicians weigh extinction before choosing it. The idea had precedent outside ABA: the paper cites substance-use studies in which people who consumed more alcohol or nicotine when it was freely available, or worked harder to keep consuming as costs rose, were more likely to relapse. Our guide to extinction in ABA covers how often bursts occur and what reduces them.
The authors proposed building that test into functional analysis, the method at the center of a functional behavior assessment. But they point out that functional analyses deliver reinforcers on FR 1 schedules, and response rates at FR 1 are often poor predictors of a reinforcer's value once the effort required goes up. A progressive ratio version could fix that in principle. This study was a first step toward the idea, run on arbitrary responses, and the authors say its results don't justify taking the next one yet.
Limitations the Authors Name
The authors list three. The 22 people in the Pmax EO condition weren't randomized, which weakens comparisons across groups. The target responses were arbitrary and had relatively short histories with the reinforcers, unlike real challenging behavior. And the study enrolled members of a protected population for a question not tied to their own clinical needs, which the team addressed with pay, third-party recruitment and assent before every session.
Two other details matter. The analysis was planned at the project's start, but the specific methods were formally preregistered only after data collection ended, though before these analyses were run. And all 90 participants were adults, 88.89% of them White, recruited in two states, so the study says nothing directly about children.
What This Means If You're Considering ABA
This paper is a worked example of the gap between statistically significant and clinically useful. The progressive ratio breakpoint passed the significance test in an exploratory analysis and still left most of what happened in extinction unexplained. The authors' own scatterplots showed the relation less clearly than the regressions did.
For practice, the finding is narrow: it doesn't support adding a progressive ratio assessment to functional analysis to predict how long behavior will persist. How bursts are planned for, and what research says reduces them, is covered in our extinction guide. The concepts the paper uses, from motivating operations and schedules of reinforcement to extinction as an operation and a process, all appear on the BCBA Test Content Outline, so the paper is useful reading for anyone preparing for that exam.
The study came out of Vanderbilt's Department of Special Education with funding from the National Institute of Child Health and Human Development. If research like this is where you see yourself, our guide to the PhD in ABA explains how doctoral training in behavior analysis works.
More ABA News
- BCBA 2027 Changes: The December 31, 2026 Deadline for the 2022 Rules BACB & Certification · November 6, 2026
- Only 51% of First-Time BCBA Candidates Passed in 2025 Exams & Pass Rates · November 6, 2026
- New York Cuts Its Fee-for-Service ABA Technician Rate to $9.63 Policy & Funding · November 6, 2026
Common questions
Questions About This Story
01What is resistance to extinction in ABA?+
02What is a progressive ratio schedule?+
03Can a progressive ratio assessment predict persistence in extinction?+
04Who conducted the 2026 reinforcer value and extinction study?+
05Does this study change how extinction is used in ABA therapy?+
Sources
Primary sources are the board, agency or organization that made the decision or published the data. Each shows the date we read it.
- [1] Lambert, Hayward, Price and Logan, Relative reinforcer efficacy and persistence in extinction are significantly correlated but not clinically actionable, Journal of Applied Behavior Analysis 60(1):e70097 (published online November 8, 2026; open access on PMC) Primary Read November 14, 2026
- [2] Same article at the publisher (DOI 10.1002/jaba.70097) Primary Read November 14, 2026