Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Mar 7.
Published in final edited form as: Behav Processes. 2025 Sep 26;232:105278. doi: 10.1016/j.beproc.2025.105278

A Computational Perspective on Rapid Reacquisition: Phenotype-Specific Effects of Reinforcement History

John Michael Falligant 1, Shane H Phillips 1, Michael Kranak 2,3, Bryan Klapes 4
PMCID: PMC12964443  NIHMSID: NIHMS2145166  PMID: 41016658

Abstract

Rapid reacquisition refers to the recurrence of a previously eliminated behavior following the response-dependent reintroduction of the reinforcer that originally maintained it. Woods and Bouton (2007) demonstrated that rapid reacquisition was attenuated when behavior was decelerated using intermittent reinforcement rather than extinction—a finding attributed to the reduced discriminability of reinforcer reintroduction under intermittent schedules. The present study used artificial organisms (AOs) animated by the Evolutionary Theory of Behavior Dynamics (ETBD) to evaluate the extent to which rapid reacquisition and its mitigation can be captured within a computational framework. AOs with phenotypes characterized by a high sensitivity to environmental contingencies exhibited reduced reacquisition of the target response following intermittent reinforcement compared to extinction, replicating prior findings. In contrast, AOs with diminished sensitivity to environmental changes showed little difference in reacquisition across response elimination conditions; instead, their relapse was primarily driven by the density of reinforcement in the reacquisition challenge condition. These findings revealed that susceptibility to disruption varied systematically with mutation rate, offering a computational perspective on how reinforcement sensitivity modulates behavioral persistence. These findings suggest that the impact of reinforcement history on relapse is phenotype-dependent and potentially shaped by operant variability. This study is the first to demonstrate rapid reacquisition using the ETBD and provides a foundation for future theoretical and translational investigations of relapse and resistance to change.

Keywords: behavioral phenotypes, complexity theory, computational modeling, intermittent reinforcement, relapse

The Evolutionary Theory of Behavior Dynamics (ETBD; McDowell, 2004) is a computational model of operant behavior grounded in three core Darwinian principles: (1) selection, (2) reproduction, and (3) mutation. These low-level rules are applied iteratively to simulate a range of potential actions made by an artificial organism (AO), akin to the behavioral repertoire of a subject in an operant laboratory experiment. These low-level rules constitute a genetic algorithm. Repeated operation of the algorithm (i.e., application of the low-level rules) can lead to the emergence of complex phenomena. That is, outcomes are not explicitly programmed or predictable from the theory alone but arise as emergent properties of the system (McDowell, 2013). As such, the ETBD is a complexity theory—a system comprised of many different interacting components—aligned with Skinner’s (1984) concept of selection by consequences. The ETBD has been extensively described in the literature, with comprehensive reviews and accessible summaries (McDowell, 2019, 2023). Its capacity to produce steady-state and dynamic behavioral patterns that bear a striking resemblance to those observed in live organisms has been demonstrated across multiple studies (e.g., McDowell & Klapes, 2020; Kulubekova & McDowell, 2013).

The ETBD is well-suited to model many transdiagnostic behavioral processes relevant to the Research Domain Criteria (RDoC; Insel et al., 2010) framework, particularly within the positive valence systems domain. Processes such as reward learning, behavioral flexibility, and impulsivity are central to its application (e.g., Higginbotham et al., 2025) and have been used to model various forms of psychopathology, including attention-deficit/hyperactivity disorder (Popa, 2013) and stereotypic movement disorder with self-injury (Morris et al., 2021). Beyond modeling psychopathology, the ETBD has successfully modeled the effects of numerous clinical procedures, including functional analyses of automatically maintained behavior, noncontingent reinforcement, competing stimulus assessments, and punishment (Hagopian & Falligant, 2023; Morris & Lucia, 2023). More recently, the ETBD has been applied to model relapse phenomena, which broadly refer to the recurrence of a previously reinforced but currently eliminated behavior resulting from some treatment challenge. For example, it has replicated findings that resurgence increases exponentially as a function of the relative downshift in alternative reinforcement (Falligant, Klapes et al., 2022), captured the effects of resurgence and repetition (Falligant, Hagopian et al., 2022), and demonstrated that the resurgence patterns of its AOs conform to predictions of quantitative models of resurgence (Klapes et al., 2023). A critical next step in evaluating the translational and functional utility of the ETBD is to explore its ability to model other relapse-related behavioral phenomena beyond resurgence. Examining its generality to other forms of relapse is essential for assessing its broader relevance to behavioral science and applied practice. One such phenomenon is reinstatement—a widely studied relapse effect, particularly in behavioral pharmacology and basic behavioral science domains, with significant theoretical and applied implications (e.g., Bossert et al., 2013; Krishnan et al., 2023; Podlesnik & Shahan, 2009; Shaham et al., 2003; Zuzina & Vinarskaya, 2022).

Operant reinstatement refers to the recurrence of previously reinforced but currently eliminated or reduced response following the re-presentation of the reinforcer that previously maintained the behavior. Reinstatement is typically evaluated within a three-phase arrangement. In Phase 1, a target response (e.g., lever pressing) is established and maintained via a reinforcement contingency. In Phase 2, the reinforcement contingency is suspended for the target response (i.e., extinction), leading to a marked reduction in the frequency of the target response. In Phase 3, the specific reinforcer that maintained the target response in Phase 1 is reintroduced while the reinforcement contingency remains suspended. If target responding recurs in Phase 3, reinstatement is said to have occurred.

Reinstatement has primarily been studied in basic and translational research preparations, though it has clear treatment-related implications in applied settings (see Kranak & Falligant, 2023). Consider the following example. At home, a child with a neurodevelopmental disability characterized by communication deficits and challenging behavior engages in operant vomiting maintained by caregiver attention, specifically in the form of statements of concern (e.g., “Are you okay?”; Roth et al., 2025). When the child vomits, the caregiver’s expression of concern—intended to provide comfort—unintentionally reinforces and maintains the behavior (Phase 1). To address this, a therapist designs a function-based intervention in which such attention is withheld contingent on vomiting (i.e., extinction; Phase 2). However, during a bout of actual illness, the child vomits and the caregiver naturally delivers statements of concern, or even in the absence of vomiting, another event (e.g., the child coughing) prompts the caregiver to say “Are you okay?” (Phase 3). This reintroduces the specific reinforcer that previously maintained the target response, potentially triggering reinstatement and a sudden reemergence of operant vomiting—sometimes at the same or even greater intensity or frequency than before extinction was implemented.

Within the relapse literature, reinstatement is sometimes categorized as either response-independent or response-dependent1 (e.g., Wathen & Podlesnik, 2018). Response-independent reinstatement, the more commonly studied paradigm, involves the noncontingent delivery of reinforcers that previously maintained the target behavior. In contrast, response-dependent reinstatement—also referred to as rapid reacquisition2—occurs when the reinforcer is briefly reintroduced contingent on the target response, even though the reinforcement contingency is otherwise suspended (as in the above example). For clarity and consistency, we use the term rapid reacquisition throughout this paper to refer to this flavor of response-dependent reinstatement.

Rapid reacquisition paradigms closely mirror treatment integrity errors—specifically, the inadvertent reinforcement of problem behavior—commonly associated with treatment failure in applied settings (Kranak & Falligant, 2023). Preliminary research suggests that rapid reacquisition in Phase 3 can be reduced by using progressively leaner reinforcement schedules during Phase 2 instead of relying solely on complete extinction (e.g., Kranak et al., 2022). In an innovative series of studies, Woods and Bouton (2007) compared the effects of intermittent reinforcement and extinction in Phase 2 on rapid reacquisition in Phase 3 across two groups of rodents. Subjects exposed to progressively leaner intermittent reinforcement schedules during Phase 2 showed a slower decline in target responding during the response elimination phase, as expected, but also exhibited lower levels of reacquisition in Phase 3. Conversely, subjects whose target responses were eliminated through extinction in Phase 2 showed a more rapid reduction in behavior, but also exhibited greater reacquisition in Phase 3. In other words, intermittently reinforcing target responding during response elimination may slow the rate of behavior reduction, but it also substantially diminishes the magnitude of reacquisition that follows.

The purpose of the present study was to systematically replicate Woods and Bouton’s (2007) study using AOs animated by the ETBD. Specifically, this study aimed to (1) determine if the ETBD can model rapid reacquisition and (2) assess if intermittent reinforcement attenuates relapse relative to extinction, as observed by Woods and Bouton. This question is particularly important because the ETBD, in its most recent published form, lacks explicitly defined mechanisms for establishing stimulus control (although see Berg, 2011, Chi, 2024, and Lanovaz & Falligant, 2024 for unpublished proposed implementations), a key process thought to underlie rapid reacquisition. Despite this missing component, the ETBD has successfully replicated other relapse phenomena, such as the effects of repeated exposure on resurgence, which are also thought to also involve discriminative repertoires. Testing if AOs emergently demonstrate rapid reacquisition would therefore provide a critical validation of the ETBD’s generality and its utility in modeling complex behavioral phenomena.

General Method

Genetic Algorithm

The ETBD is a genetic algorithm that operates by iterating through a series of computational cycles, each termed a “generation.” During each generation, the AO emits one behavior from a range of potential behaviors (i.e., a “behavioral repertoire”), referred to as a “population” represented by integers distributed between 0 and 1023. Artificial organisms can exhibit 1024 individual behaviors, with the integer value of each behavior known as its phenotype. Within this range, behaviors are typically organized into “response classes” constituting contiguous sets of integers that represent a collection of behaviors that constitute the same response and are functionally linked together by their production of the same consequences during computational experiments. For example, a rat using its right and left paws to manipulate the same lever may be represented as phenotypes 471 and 472, respectively, as the behaviors are topographically different but still belong to the same response class from the experimenter’s perspective. ETBD studies employing a concurrent schedules typically place the two response classes for the alternatives between 471 to 511 and 512 to 552, while ETBD studies that simulate experiments with a single defined operandum typically place a “target” response class between 471 to 511 and a “background” response class between 512 to 470, which aims to capture all other possible phenotypes (c.f., McDowell & Klapes, 2020). These integers are also expressed in binary form as 10-bit strings, referred to as genotypes (e.g., the integer 500 corresponds to the binary string 0111110100). Thus, every behavior in the population is characterized by both a genotype (a binary bit-string) and a phenotype (its corresponding integer value).

Reinforcement contingencies are applied to specific target, alternative, or background response classes within the AO’s repertoire (see Figure 1). When a response meets the reinforcement criteria, behaviors from the population are selected to serve as “parent” behaviors. These parents contribute segments of their genotypes to generate “child” behaviors. A fitness density function (FDF) is employed to determine which parent behaviors are selected, based on their proximity to the phenotype of the reinforced behavior (see McDowell, 2004), and can be programmed separately for different reinforcing contingencies. Behaviors whose phenotypes are closer to the reinforced behavior’s phenotype are more likely to be chosen as parents. For instance, if the reinforced phenotype is 510, a behavior with a phenotype of 520 would have a higher probability of selection than one with a phenotype of 550. Thus, FDFs with smaller means exert stronger selective pressure on the population, analogous to a larger magnitude reinforcer, while FDFs with larger means result in weaker selective pressure, analogous to smaller magnitude reinforcers. In the absence of reinforcement during a generation, parent behaviors are chosen randomly from the population to allow for another iteration of the algorithm without any specific selection mechanism.

Figure 1. Schematic of Genetic Algorithm.

Figure 1

Open in a new tab

Note. Schematic adapted from Morris and Lucia (2023).

The primary method of reproduction in ETBD is known as “bitwise recombination”, where each bit of the first parent’s genotype has a 50% chance of being passed on to the child behavior. If a bit is not inherited from the first parent, the corresponding bit from the second parent is used instead. This process of combining parent genotypes continues until a new population, typically consisting of the same number of behaviors (usually 100) as the original population, is formed. Following the formation of the new population, a mutation process is applied to the genotypes of the child behaviors. The “bit-flip-by-individual” method is most commonly used for mutation within the ETBD framework. A certain percentage of behaviors, determined by the “mutation rate,” are selected at random, and one bit within the genotype of each selected behavior is randomly flipped—either from 0 to 1 or from 1 to 0. Depending on which bit is altered, the mutation can result in either minor or substantial changes to the behavior’s phenotype (for example, flipping the first bit of the genotype 0000000000 to 1000000000 changes the phenotype from 0 to 1023, whereas flipping the last bit to 0000000001 changes it only to 1). After mutation, a behavior is randomly chosen from the new population, and the cycle is repeated for the desired number of generations specified by the experimenter.

General Method

Background

Woods and Bouton (2007) had 32 rats lever press in three different phases that included shaping and conditioning (Phase 1), response elimination (Phase 2), and a reacquisition test (Phase 3). Once the rats’ lever pressing was established, they responded on a VI 30-s schedule of reinforcement for five, 30-min sessions for conditioning. Rats were then assigned to four different groups, which varied based on whether they experienced extinction (EXT) or partial reinforcement (PRF) during Phase 2 and a VI 2- (EXT2 and PRF2) or 8-min (EXT8 and PRF8) schedule during Phase 3. The response elimination phase consisted of eight 60-min sessions, with the intermittent reinforcement procedure consisting of schedules that got progressively leaner from VI 4-min to VI 32-min. Reacquisition of responses was then evaluated by having rats respond on two 60-min sessions in which a VI 2-min or VI 8-min schedule was operating, based on the group. Responses were measured as a rate (average number of lever presses per minute) and analyzed as a percentage of baseline responding.

Procedure

To replicate the approach from Woods and Bouton, we created 360 AOs divided into three mutation rate conditions (2%, 10%, and 20%). Within each mutation rate, AOs were assigned to one of four groups defined by the Phase 2 response elimination method (intermittent reinforcement vs. extinction) and the Phase 3 reacquisition schedule (RI 120 vs. RI 480), with 30 AOs per group. The target response, composed of 41 phenotypes (ranging from 471 to 511), was reinforced using a linear FDF with a mean of 40. A background response class, composed of 100 phenotypes, was reinforced on a RI 60 reinforcement schedule, operating on an RI 60 schedule and associated with a linear FDF of 60.

A mutation rate of either 2%, 10%, or 20% was applied to each new generation of behavior, consistent with values successfully used in prior ETBD studies of both applied (e.g., Morris & Lucia, 2023; Morris & McDowell, 2023; Hagopian & Falligant, 2023) and basic (e.g., Klapes et al., 2023; McDowell, 2004) behavioral processes. In the ETBD, mutation rate shapes an organism’s sensitivity to reinforcement. Lower values bias persistence in recently reinforced topographies, but can increase adaptive behavioral variability when reinforcement sources vary. Higher mutation rate values diminish sensitivity to reinforcement history, producing greater variability driven by mutation dynamics—that is, behaviors mutating out of the reinforced response class—rather than by variability shaped by alternative reinforcement sources. At extreme levels, higher mutation rates lead to disorganized responding (Popa et al., 2016) consistent with impulsive behavior in live organisms (McDowell, 2023; Popa, 2013).

In Phase 1, each group of AOs completed five conditioning sessions on a RI-30 schedule. In Phase 2, AOs in the EXT groups received eight response elimination sessions (n.b., because extinction schedules do not deliver any reinforcers, these schedules result in random parental selection for all iterations; Figure 1), whereas AOs in the PRF groups were exposed to a series of progressively leaner blocks of schedules similar to that used by Woods and Bouton (2007; viz., RI 240, RI 480, RI 720, RI 960, RI 1200, RI 1440, RI 1680, and RI 1920). In Phase 3, the target response was reinforced on either a RI 120 (Groups EXT2 and PRF2) or a RI 480 (Groups EXT8 and PRF8) schedule. All sessions were comprised of 1000 generations. We examined behavior during the initial transition to the response elimination phase in the first session of Phase 2, as well as during the first and second reacquisition tests, corresponding to sessions one and two of Phase 3. Behavior in Phase 3 was also compared to behavior observed in the final session of Phase 2 to evaluate the magnitude of reacquisition relative to response levels prior to the reintroduction of the reinforcer. Data from Phase 2 (response elimination) and Phase 3 (reacquisition test) sessions are expressed as a proportion of baseline, with responses per 1,000 generations multiplied by 100 to convert target responding into a percentage.

Results

Figure 2 shows individual and mean AO response patterns across the response elimination and reacquisition phases for each mutation rate. A two-way ANOVA revealed significant main effects of both the Phase 2 response elimination strategy, F(1, 354) = 22.2, p < .001, and mutation rate, F(1, 354) = 13.7, p < .001 (Figure 3). Post hoc comparisons indicated that extinction produced significantly greater reductions in target responding than intermittent reinforcement at the 2% (q = 3.98, p = .002) and 10% (q = 2.77, p = .018) mutation rates, but not at the 20% mutation rate (q = 1.41, p = .40). These findings suggest that as sensitivity to reinforcement decreases, the differential efficacy of these response elimination procedures diminishes in this computational environment.

Figure 2. Individual and Mean AO Response Patterns Across Response Elimination and Reacquisition Phases by Mutation Rate.

Figure 2.

Open in a new tab

Note. Each thin line represents an individual AO’s target responding during each session of Phase 2 and Phase 3. Groups are defined by mutation rate (2%, 10%, 20%), response elimination condition, and reacquisition schedule. Thicker lines represent the grand mean.

Figure 3. Mean (SEM) Target Responding in Phase 2.

Figure 3.

Open in a new tab

Note. * = p < .05, ** = p < .01, *** = p < .001

To more formally assess resistance to change, we used Nevin’s (1992) behavioral momentum framework and examined the log proportion of baseline responding during the response elimination phase (Figure 4, left panel). This analysis revealed clear differences across mutation rates and disruptor conditions. A two-way ANOVA revealed significant main effects of both mutation rate, F(2, 349) = 35.47, p < .001, and disruptor, F(1, 349) = 20.27, p < .001, as well as a significant interaction, F(2, 349) = 3.17, p = .043. Extinction produced significantly lower log-proportions of baseline responding than intermittent reinforcement, consistent with greater behavioral disruption. This effect was especially pronounced at lower mutation rates. For AOs with 2% (q = 4.65, p = .002) and 10% (q = 2.77 p = .018) mutation rates extinction resulted in significantly greater disruption relative to AOs with 20% mutation rates. In contrast, the relative disruptive effects of intermittent reinforcement did not significantly vary across mutation rates. To examine overall sensitivity to disruption irrespective of disruptor type (i.e., extinction vs. intermittent reinforcement), we calculated the geometric mean of the proportion of baseline responding across extinction and intermittent reinforcement conditions for each mutation rate group, and then log-transformed these values.

Figure 4. Resistance to Change by Disruptor and Mutation Rate.

Figure 4.

Open in a new tab

Note. Error bars correspond to 95% confidence intervals.

A one-way ANOVA revealed a significant effect of mutation rate on resistance to change, F(2, 170) = 6.91, p = .001, suggesting that overall resistance to disruption varies systematically with sensitivity to environmental changes (Figure 4, right panel). Tukey’s post hoc comparisons indicated that AOs with a 2% mutation rate exhibited significantly greater sensitivity to disruption (i.e., lower log-transformed geometric mean of proportion of baseline) than AOs with a 20% mutation rate (q = 5.22, p < .001), with a marginal difference observed between 2% and 10% mutation rates (q = 3.22, p = .06). No difference was observed between 10% and 20% mutation rates. Together, these findings suggest that behavioral phenotypes characterized by low mutation rates (i.e., high sensitivity to reinforcement) are more susceptible to disruption by both extinction and intermittent reinforcement. In contrast, phenotypes with higher mutation rates—reflecting diminished sensitivity to reinforcement—demonstrated more uniform resistance across disruptor types. This pattern implies that resistance to change under extinction, but not intermittent reinforcement, is moderated by underlying behavioral phenotype.

Next, we examined the effects of the response elimination tactic used in Phase 2 (extinction vs. intermittent reinforcement) and the schedule of reinforcement during the reacquisition test in Phase 3 on the change in target responding between the final session of Phase 2 and the first session of Phase 3. For AOs with a 2% mutation rate, there was a significant difference in the change in target responding between the final session of Phase 2 and first session of Phase 3 by group F(1, 116) = 5.8, p = .02, such that the increase in target responding in Phase 3 was greater for AOs exposed to extinction in Phase 2 relative to intermittent reinforcement (Figure 5, top panel). There was no effect of the schedule of reinforcement (RI 120 vs RI 480) in the reacquisition challenge F(1, 116) = .16, p = .69. In contrast, for AOs with a 10% mutation rate, there was no significant effect of the disruptor (extinction vs. intermittent reinforcement) on the change in target responding between the final session of Phase 2 and the first session of Phase 3, F(1, 116) = 0.04, p = .84 (Figure 5, middle panel). A similar pattern was observed for the 20% mutation rate group, with no effect of the Phase 2 response disruption tactic on reacquisition in Phase 3, F(1, 116) = 0.34, p = .56 (Figure 5, bottom panel). Interestingly, for both the 10% and 20% groups, the Phase 3 schedule had a significant effect on reacquisition in Phase 3 (10%: F(1, 116) = 17.40, p < .001; 20%: F(1, 116) = 14.02, p < .001). The denser schedule produced greater reacquisition than the leaner schedule following both extinction (10%: q = 2.85, p = .01; 20%: q = 3.06, p = .006) and intermittent reinforcement (10%: q = 3.05, p = .006; 20%: q = 2.24, p = .053). Thus, at 10% and 20% mutation rates, reacquisition magnitude was determined by the Phase 3 schedule rather than the Phase 2 disruptor; at 2%, no such effect of the Phase 3 schedule was observed.

Figure 5.

Figure 5.

Open in a new tab

Reacquisition in Phase 3 (Mean + SEM)

Discussion

The pattern of results suggests that the influence of response elimination history on response reacquisition is modulated by the AO’s sensitivity to environmental changes. Specifically, AOs with a 2% mutation rate exhibited reduced reacquisition when target behavior had been previously exposed to intermittent reinforcement rather than extinction. In contrast, AOs with 10% and 20% mutation rates— characterized by diminished sensitivity to reinforcement and disorganized, impulsive-like responding (McDowell, 2023; Popa & McDowell, 2016)—did not show differential reacquisition based on the Phase 2 condition. Instead, relapse in these groups was solely influenced by the density of reinforcement in Phase 3, with denser schedules producing higher levels of reacquisition regardless of whether behavior had previously contacted extinction or intermittent reinforcement. These findings suggest a shift in the controlling variables of behavior based on the underlying phenotype: AOs with low mutation rates may be more influenced by temporally extended contingencies and reinforcement history, whereas those with high mutation rates appear more influenced by immediate, local contingencies. In other words, diminished sensitivity to environmental changes may attenuate the long-range effects of response history and render behavior more susceptible to current reinforcement conditions. This distinction provides a computational analog—and a first-order prediction—of how different behavioral phenotypes may differentially respond to relapse mitigation strategies in applied contexts.

These findings further support emerging findings that progressively thinning reinforcement during Phase 2 attenuates rapid reacquisition (Kranak et al., 2022; Woods & Bouton, 2007). Reacquisition occurred across conditions, but its magnitude was reliably lower following intermittent reinforcement than extinction. Thus, relapse mitigation is best viewed as reducing the severity of relapse rather than preventing it outright (see Kimball et al., 2023, for related discussion). Progressively thinning reinforcement schedules during response elimination may attenuate rapid reacquisition by engaging several distinct behavioral mechanisms—at least in real, living organisms. Increasing the unit price of reinforcers in Phase 2—that is, the number of responses required per reinforcer—reduces response strength by decreasing the overall efficacy of the reinforcer. As reinforcement schedules become leaner, the increased response requirement may evoke ratio strain, thereby reducing the likelihood of responding. In this context, the higher unit price functions as an abolishing operation, decreasing the value of the reinforcer and the probability of the operant behavior occurring. This suppression of responding is consistent with behavioral economic principles and may contribute to reduced relapse when reinforcement is reintroduced. The use of intermittent reinforcement during Phase 2, rather than complete extinction, also reduces the discriminability of transitions between Phase 2 and Phase 3. This diminished salience makes it less likely that the organism will detect changes in reinforcement contingencies, increasing the likelihood that the response elimination achieved in Phase 2 will generalize to Phase 3. From this perspective, the extent to which relapse-mitigation strategies successfully reduce or prevent relapse in live organisms may depend, in part, on their resemblance to the relapse-inducing conditions they aim to mimic. This interpretation implies that reinstatement is, at least in part, governed by stimulus control processes, with reinforcers delivered in Phase 3 functioning as contextual cues for previously reinforced behavior. Notably, the current implementation of the ETBD does not include an explicit mechanism for simulating stimulus control. Thus, the observed inoculative effect of progressively leaner reinforcement schedules in our study occurred without any expressively programmed discriminative stimuli—marking a key departure from prior studies in which schedule thinning was paired with explicit contextual cues (e.g., Kranak et al., 2022). These findings suggest that a history of intermittent reinforcement may reduce relapse even in the absence of salient contextual signals. Future research should directly compare reinstatement outcomes following progressively leaner reinforcement schedules with and without the inclusion of programmed discriminative stimuli to clarify the role of contextual control in relapse mitigation.

This study is the first to demonstrate rapid reacquisition using the ETBD, further showcasing the generality of this computational algorithm for modeling relapse phenomena. This is a promising development, as the ETBD could theoretically be used to test novel hypotheses regarding the controlling variables of relapse, uncover additional mechanisms, and evaluate innovative treatment approaches. Moreover, examining how the computational dynamics of the ETBD give rise to behavioral patterns like rapid reacquisition may offer unique insights into the emergent properties of relapse and the conditions under which mitigation strategies succeed or fail. However, realizing this potential will require replication and thoughtful investigation across the basic-to-applied continuum to ensure the robustness and translational relevance of these findings.

Author Note

Manuscript preparation was partially supported by Grants 1R21HD113881-01, 1R21HD112724-01A1, and 1R21HD113794-01A1 from the National Institute of Child Health and Human Development (NICHD), as well as a NARSAD Young Investigator Grant from the Brain and Behavior Research Foundation (BBRF). This content is solely the responsibility of the authors and does not necessarily reflect the official views of the NICHD or BBRF. Portions of this document were prepared with the assistance of ChatGPT to help clarify concepts and wording. The content has been reviewed and approved for accuracy and clarity by the first author.

Footnotes

1

Although there is no direct evidence suggesting these reinstatement paradigms (response-independent vs. response-dependent) involve distinct behavioral processes, some studies indicate that reinstatement may be greater under response-dependent conditions due to enhanced response-reinforcer contiguity (Baum, 2012; Podlesnik & Shahan, 2009).

2

It is not clear whether response-dependent reinstatement and rapid reacquisition are one and the same, nor if this distinction is even meaningful (e.g., Kranak & Falligant, 2023). While both involve relapse caused by the contingent delivery of reinforcers in Phase 3, their distinctions remain an area for further exploration and discussion (Bouton, 2024).

References

  1. Baum WM (2012). Rethinking reinforcement: Allocation, induction, and contingency. Journal of the Experimental Analysis of Behavior, 97(1), 101–124. 10.1901/jeab.2012.97-101 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Berg J (2011). Implementation of stimulus control in a computational model [Doctoral dissertation], Emory University. https://etd.library.emory.edu/concern/etds/8910jv394?locale=en [Google Scholar]
  3. Bossert JM, Marchant NJ, Calu DJ, & Shaham Y (2013). The reinstatement model of drug relapse: Recent neurobiological findings, emerging research topics, and translational research. Psychopharmacology, 229(3), 453–476. 10.1007/s00213-013-3120-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bouton ME (2019). Extinction of instrumental (operant) learning: interference, varieties of context, and mechanisms of contextual control. Psychopharmacology, 236(1), 7–19. 10.1007/s00213-018-5076-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bouton ME (2024). Habit and persistence. Journal of the Experimental Analysis of Behavior, 1211(1), 88–96. 10.1002/jeab.894 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bouton ME, Winterbauer NE, & Todd TP (2012). Relapse processes after the extinction of instrumental learning: Renewal, resurgence, and reacquisition. Behavioural Processes, 90(1), 130–141. 10.1016/j.beproc.2012.03.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chi C (2024). Implementation of discriminative stimuli in the Evolutionary Theory of Behavior Dynamics [Doctoral dissertation], Emory University. https://etd.library.emory.edu/concern/etds/05741s996?locale=en [Google Scholar]
  8. Falligant JM, Hagopian L, Laureano B, & Klapes B (2022). Examining resurgence and repetition with the evolutionary theory of behavior dynamics. Behavioural Processes, 203, 104776. 10.1016/j.beproc.2022.104776 [DOI] [PubMed] [Google Scholar]
  9. Falligant JM, Klapes B, & Hagopian LP (2022). Modeling resurgence with an evolutionary theory of behavior dynamics. Behavioural Processes, 197, 104623. 10.1016/j.beproc.2022.104623 [DOI] [PubMed] [Google Scholar]
  10. Fisher WW, Greer BD, Mitteer DR, & Fuhrman AM (2022). Translating quantitative theories of behavior into improved clinical treatment for problem behavior. Behavioural Processes, 198, 104639. 10.1016/j.beproc.2022.104639 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Hagopian L (2020). The consecutive controlled case series: Design, data-analytics, and reporting methods supporting the study of generality. Journal of Applied Behavior Analysis, 53(2), 596–619. 10.1002/jaba.691 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Higginbotham R, Dallery J, & McDowell JJ (2025). The Evolutionary Theory of Behavior Dynamics predicts delay discounting. Perspectives on Behavior Science. Advance online publication. 10.1007/s40614-025-00443-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Insel T, Cuthbert B, Garvey M, Heinssen R, Pine DS, Quinn K, Sanislow C, & Wang P (2010). Research domain criteria (RDoC): Toward a new classification framework for research on mental disorders. The American Journal of Psychiatry, 167(7), 748–751. 10.1176/appi.ajp.2010.09091379 [DOI] [PubMed] [Google Scholar]
  14. Kimball RT, Greer BD, Fuhrman AM, & Lambert JM (2023). Relapse and its mitigation: Toward behavioral inoculation. Journal of Applied Behavior Analysis, 56(2), 282–301. 10.002/jaba.971 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Klapes B, Falligant JM, & Hagopian LP (2023). Modeling and quantifying resurgence in the Evolutionary Theory of Behavior Dynamics. Behavioural Processes, 104860. 10.1016/j.beproc.2023.104860 [DOI] [PubMed] [Google Scholar]
  16. Kranak MP, & Falligant JM (2023). Treatment integrity, commission errors, and reinstatement: A primer. Education and Treatment of Children, 46(3), 249–262. 10.1007/s43494-023-00104-x [DOI] [Google Scholar]
  17. Kranak MP, Rooker GW, Arevalo AR, & Robinson TP (2022). Mitigating reinstatement using progressive ratio schedules: A proof of concept. Behavioral Development, 27(1-2), 21–35. 2. 10.1037/bdb0000108 [DOI] [Google Scholar]
  18. Krishnan S, Bevins RA, & de Wit H (2023). Place conditioning in humans: Opportunities for translational research. Psychopharmacology. 10.1007/s00213-023-06316-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kulubekova S, & McDowell JJ (2013). Computational model of selection by consequences: Patterns of preference change on concurrent schedules. Journal of the Experimental Analysis of Behavior, 100(2), 147–164. 10.1002/jeab.40 [DOI] [PubMed] [Google Scholar]
  20. Lanovaz MJ, & Falligant JM (2024). Establishing discriminative repertoires in the Evolutionary Theory of Behavior Dynamics: Preliminary outcomes and future directions. PsyArXiv. 10.31234/osf.io/euvzd [DOI] [Google Scholar]
  21. McDowell JJ (2004). A computational model of selection by consequences. Journal of the Experimental Analysis of Behavior, 81(3), 297–317. 10.1901/jeab.2004.81-297 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. McDowell JJ (2013). A quantitative evolutionary theory of adaptive behavior dynamics. Psychological Review, 120(4), 731–750. 10.1037/a0034244 [DOI] [PubMed] [Google Scholar]
  23. McDowell JJ (2019). On the current status of the evolutionary theory of behavior dynamics. Journal of the Experimental Analysis of Behavior, 111(1), 130–145. 10.1002/jeab.495 [DOI] [PubMed] [Google Scholar]
  24. McDowell JJ (2023). Creating and studying the behavior of artificial organisms animated by an evolutionary theory of behavior dynamics. Perspectives on Behavior Science, 1–18. 10.1007/s40614-023-00366-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. McDowell JJ, & Klapes B (2020). All Behavior is choice: Revisiting an evolutionary theory’s account of behavior on single schedules. Journal of the Experimental Analysis of Behavior, 114(3), 430–446. 10.1002/jeab.630 [DOI] [PubMed] [Google Scholar]
  26. Morris SL, & Lucia SM (2023). Applying the evolutionary theory of behavior dynamics to model the subtypes of automatically reinforced self-injurious behavior. Journal of Applied Behavior Analysis, 56(2), 442–457. 10.1002/jaba.982 [DOI] [PubMed] [Google Scholar]
  27. Morris SL, & McDowell JJ (2021). Modeling subtypes of automatically reinforced self-injurious behavior with the evolutionary theory of behavior dynamics. Perspectives on Behavior Science, 44, 581–603. 10.1007/s40614-021-00297-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Nevin JA (1992). An integrative model for the study of behavioral momentum. Journal of the Experimental Analysis of Behavior, 57(3), 301–316. 10.1901/jeab.1992.57-301 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Podlesnik CA, & Shahan TA (2009). Behavioral momentum and relapse of extinguished operant responding. Learning & Behavior, 37(4), 357–364. 10.3758/LB.37.4.357 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Popa A (2013). The Evolutionary Theory of Behavior Dynamics: Complexity, Darwinism, and the emergence of high-level phenotypes [Doctoral dissertation], Emory University. http://pid.emory.edu/ark:/25593/f8hw4 [Google Scholar]
  31. Popa A, & McDowell JJ (2016). Behavioral variability in an evolutionary theory of behavior dynamics. Journal of the Experimental Analysis of Behavior, 105(2), 270–290. 10.1002/jeab.199 [DOI] [PubMed] [Google Scholar]
  32. Roth H, Chesbrough E, & Falligant JM (2025). Response-class hierarchies in self-induced vomiting: Functional assessment and treatment. Manuscript in revision. Behavioural Processes. [Google Scholar]
  33. Shaham Y, Shalev U, Lu L, De Wit H, & Stewart J (2003). The reinstatement model of drug relapse: History, methodology and major findings. Psychopharmacology, 168(1), 3–20. 10.1007/s00213-002-1224-x [DOI] [PubMed] [Google Scholar]
  34. Skinner BF (1984). Selection by consequences. Behavioral and Brain Sciences, 7(4), 477–481. 10.1017/S0140525X0002673X [DOI] [Google Scholar]
  35. Stephens M, Wessel M, Falligant JM, & Kranak MP (2025). Reacquisition and reinstatement of problem behavior: Knowns and unknowns. In Kranak MP (Ed.), Behavioral treatments for individuals with autism: Promoting longevity and durability. Springer. [Google Scholar]
  36. Wathen SN, & Podlesnik CA (2018). Laboratory models of treatment relapse and mitigation techniques. Behavior Analysis: Research and Practice, 18(4), 362–387. 10.1037/bar0000119 [DOI] [Google Scholar]
  37. Woods AM, & Bouton ME (2007). Occasional reinforced responses during extinction can slow the rate of reacquisition of an operant response. Learning and Motivation, 38(1), 56–74. 10.1016/j.lmot.2006.07.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Zuzina AB, & Vinarskaya AK (2022). Increased his-tone acetylation levels or a serotonin precursor reinstate the context memory impaired by the serotonin receptor blocker methiothepin. Neuroscience & Behavioral Physiology, 1–13. 10.1007/s11055-022-01303-636590598 [DOI] [Google Scholar]

RESOURCES