Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Aug 1.
Published in final edited form as: Behav Processes. 2024 Jul 26;220:105082. doi: 10.1016/j.beproc.2024.105082

Resurgence Mitigation Across Extended Extinction Following Four and Eight Cycles of On/Off Alternative Reinforcement

Timothy A Shahan 1, Gabrielle M Sutton 1, Matias Avellaneda 1
PMCID: PMC11317034  NIHMSID: NIHMS2015213  PMID: 39069280

Abstract

Resurgence is an increase in an extinguished operant response resulting from a worsening of conditions (e.g., extinction) for a more recently reinforced alternative behavior. Previous research has shown that exposure to cycles of alternative reinforcement available versus unavailable (i.e., on/off alternative reinforcement) across sessions can reduce subsequent resurgence. Most previous assessments of the procedure have examined target operant responding during only single-session resurgence tests, and it remains unclear if exposure to relatively few cycles of on/off alternative reinforcement can maintain low rates of target behavior across extended exposure to extinction. This experiment with rats examined the effects of 4 or 8 cycles of on/off alternative reinforcement on subsequent resurgence during a 10-session extinction test. The results show that exposure to 4 cycles of on/off alternative reinforcement is as effective as 8 cycles in producing low rates of target behavior during treatment and across extended extinction. This result is consistent with extant theories of resurgence and suggests that on/off alternative reinforcement could have translational utility following relatively few cycles of exposure.

Keywords: resurgence, contingency discrimination training, relapse, operant behavior

Resurgence is an increase in an extinguished operant response with a worsening of conditions (e.g., extinction) for a more recently reinforced alternative behavior (Lattal and Wacker, 2015; Shahan and Craig, 2017). Resurgence has received considerable attention because of its similarity to relapse of problem behavior (e.g., Briggs et al., 2018; Lieving et al., 2004; Volkert et al., 2009; Wacker et al., 2011) following treatments employing differential reinforcement of alternative behavior (i.e., DRA; Petscher et al., 2009).

One method that reduces resurgence is exposure to alternating sessions with alternative reinforcement available versus unavailable during extinction of the target response (Schepers and Bouton, 2015; Shahan et al., 2020; Trask et al., 2018). This effect has played an important role in current theories of resurgence by highlighting the role of context change in the phenomenon. Bouton’s context theory suggests resurgence results from a change in the extinction context resulting from removal of the stimulus effects of alternative reinforcement (Winterbauer and Bouton, 2010). The theory suggests experiencing extinction of target behavior with and without alternative reinforcement reduces resurgence by teaching organisms to discriminate that reinforcement for target behavior remains unavailable even when alternative reinforcement is absent (Schepers and Bouton, 2015; Trask et al., 2018). Resurgence as Choice in Context theory (i.e., RaC2; Shahan et al., 2020) suggests resurgence is governed by the changing relative values of target and alternative responses over time, but also includes a quantitative formalization of the discrimination from context theory to account for the effects of on/off alternative reinforcement. Because the discrimination underlying the effects of on/off alternative reinforcement on resurgence in both theories is based on the stimulus effects of the reinforcement contingencies themselves, we refer to the procedure as contingency discrimination training (i.e., CDT).

Little is known about factors influencing the efficacy of CDT. Schepers and Bouton (2015) showed the effects of CDT in an experiment where rats initially pressed a target lever for food. Next, target pressing was extinguished and reinforcement provided by an alternative lever for 4 cycles of CDT (i.e., 4 “on” sessions; on, off, on, off, on, off, on) or an equivalent duration (i.e., 7 sessions) of “All On” alternative reinforcement. CDT reduced responding compared to All On alternative reinforcement in a single-session resurgence test with no alternative reinforcement available. Trask, et al. (2018) found that while 13 cycles of CDT reduced resurgence in a single test session, three cycles did not. Shahan et al. (2020) compared one group of rats exposed to 16 CDT cycles and groups exposed to All On alternative reinforcement for the equivalent of 2, 4, 8, 12, and 16 cycles of CDT. This allowed comparison of CDT and All On alternative reinforcement in single “off” sessions across the five durations of exposure to CDT. CDT reduced resurgence in all single-session tests except following only 2 cycles. In an extended resurgence test conducted after the 16th cycle, the rate of target behavior during the resurgence test for the CDT group remained low across 10 days of testing. In contrast, in an experiment with crowdsourced human participants, Smith and Greer (2023) saw evidence that 6 cycles of CDT reduced initial resurgence during a 6-session resurgence test, but a potential increasing trend in target responding is apparent in their data across further test sessions. This potential trend suggests fewer cycles of CDT than the 16 used by Shahan et al. could generate only a short-lived reduction of target responding during extinction. However, there are many differences between Shahan et al. and Smith and Greer including but not limited to species, reinforcer, response type, reinforcement schedules, and session durations.

Based on the extant theories of resurgence described above, there is no reason to expect smaller numbers of CDT cycles to generate increasing rates of target behavior across extended extinction. Although context theory makes no explicit predictions about longer testing durations, there seems little reason to expect it should reduce the relevant discrimination. RaC2, which formalizes the impact of the relevant discrimination, predicts decreases in target responding across extended extinction training (see Shahan et al., 2020). Thus, if target responding does increase across extinction following fewer cycles of CDT, it could challenge these theories.

Finally, because CDT has reduced resurgence in animal experiments, an ongoing clinical trial is examining its efficacy in reducing resurgence of severe problem behavior following DRA (ClinicalTrials.gov, NCT05537610). The trial employs 16 cycles of CDT because only that number of cycles has been shown to maintain low levels of target responding across extinction. If low levels of target responding might be maintained across extinction following fewer CDT cycles, future clinical implementation might be simplified. However, the increasing trend in target responding across extinction in Smith and Greer could suggest fewer CDT cycles might be insufficient to produce lasting benefits.

Thus, using methods comparable to Shahan et al. (2020) we examined the effects of CDT on resurgence of rats during a 10-session resurgence test following either 4 or 8 cycles of CDT or equivalent All On alternative reinforcement. Theses durations were chosen because 4 cycles of CDT is the smallest number of cycles shown to reduce resurgence in single-session tests, 8 cycles is intermediate to the 16 cycles known to maintain low levels of target responding across extended extinction, and these values bracket the 6 cycles examined in Smith and Greer.

Method

Subjects

Thirty-six naïve male Long-Evans rats (Charles River, Portage, MI) 70–90 days old were used. They were housed individually in a humidity- and temperature-controlled colony on a 12:12 hr light/dark cycle with unlimited access to water and maintained 80% of free-feeding weights. All procedures were in accordance with the University’s Institutional Animal Care and Use Committee.

Apparatus

Ten Med Associates chambers (30 cm × 24 cm × 21 cm) were used. A house light on the left wall provided illumination. Two retractable levers with lights above were located at both sides of a pellet dispenser (45-mg grain-based; Bio Serv, Flemington, NJ) on the right wall.

Procedure

Training.

Sessions lasted 30 minutes (excluding food deliveries) and were conducted 7 days/week at the same time. The first four sessions were magazine training, during which all chamber lights were off, levers retracted, and pellets were delivered on a variable-time 60-s schedule (accompanied by 3-s receptacle illuminations). In the following session, one lever (i.e., target lever, counterbalanced left/right) was extended with its stimulus light and the house light illuminated except during food deliveries. In this session only, the first press produced a food pellet, and then further presses were reinforced on a VI 10-s schedule.

Phase 1: Baseline.

Rats experienced 32 baseline sessions. Stimulus conditions were identical to training, but target-lever pressing was reinforced on a VI 30-s schedule.

Phase 2: Treatment.

Following Phase 1, rats were assigned to one of four groups (n=9) exposed to either 4 or 8 cycles of CDT or a comparable number of sessions of All On alternative reinforcement. Target responding was placed on extinction and the alternative lever was extended with its stimulus light on. For the CDT groups, the alternative lever cycled between entire sessions of VI 10-s reinforcement or extinction, where the number of cycles refers to the number of “on” sessions prior to extended “off” resurgence testing (e.g., 4 cycle CDT = on, off, on, off, on, off, on). For the All On groups, the alternative lever produced food on a VI 10-s schedule for a comparable number of Phase 2 sessions as the corresponding CDT group (e.g., 4 cycle All On group = 7 on sessions). A changeover delay prevented alternative responses from producing reinforcement for 3-s following a target response.

Phase 3: Resurgence Test.

Both target and alternative levers were placed on extinction for 10 sessions for all groups.

Data Analysis.

R v4.2.1 (rstatix package) was used with α = .05, two-tailed tests, and Greenhouse-Geiser corrections for sphericity violations.

Results

Target responses/min averaged across the final baseline 3 sessions for CDT 4, CDT 8, All On 4, and All On 8 groups were 28.93, 29.33, 29.55, and 28.83, respectively, and did not differ significantly F(3,32) = 0.005, p = .999, η2G = .005. Figure 1 (top panels) shows target rates across sessions of Phase 2. Separate type x session mixed ANOVAs for different Phase 2 durations revealed target rates were higher for CDT groups than for All On groups (4 cycles: F[1,16] = 31.68, p < .001, η2G = .314, 8 cycles: F[1,16] = 16.74, p < .001, η2G = .22), decreased across sessions (4 cycles: F[2.09, 33.48] = 59.07, p < .001, η2G = .74, 8 cycles: F[1.73, 27.71] = 32.58, p < .001, η2G = .60), and decreased more for CDT groups (4 cycles: F[2.09, 33.48] = 23.50, p < .001, η2G = .53, 8 cycles: F[1.73, 27.71] = 13.62, p < .001, η2G = .38). Separate session type × session ANOVAs revealed target response rates for CDT groups were higher for “on” than for “off” sessions (4 cycles: F[1,8] = 31.58, p < .001, η2G = .55, 8 cycles: F[1,8] = 10.48, p = .001, η2G = .28) and decreased across sessions (4 cycles: F[1.13,9.02] = 86.63, p < .001, η2G = .71, 8 cycles: F[1.23,9.88] = 51.11, p < .001, η2G = .60), more for “on” than “off” sessions, but marginally for 8 cycles (4 cycles: F[1.18,9.44] = 10.87, p = .007, η2G = .27, 8 cycles: F[1.52,12.15] = 12.15, p = .052, η2G = .094). Figure 1 (bottom panels) shows alternative rates across Phase 2 sessions. Alternative rates were higher for All On than CDT groups (4 cycles: F[1,16] = 11.75, p = .003, η2G = .35, 8 cycles: F[1,16] = 10.23, p = .006, η2G = .32), increased across sessions (4 cycles: F[3.69, 59.04] = 27.71, p < .001, η2G = .32, 8 cycles: F[2.76, 44.13] = 33.40, p < .001, η2G = .34), more so for All On than CDT groups (4 cycles: F[3.69, 59.04] = 15.47, p < .001,η2G = .21, 8 cycles: F[2.76, 44.13] = 21.81, p < .001, η2G = .26). Separate ANOVAs revealed alternative response rates for CDT groups were higher for “on” than “off” sessions (4 cycles: F[1,8] = 43.21, p < .001, η2G = .62, 8 cycles: F[1, 8] = 64.97, p < .001, η2G = .73), increased across sessions (4 cycles: F[2,16] = 4.01, p =.039, η2G = .06, 8 cycles: F[2.08, 16.68] = 13.73, p < .001, η2G = .12), with the difference between “on” and “off” sessions increasing significantly across sessions only for the 8 cycle group, F(1.90, 15.17) = 25.01, p < .001, η2G = .18.

Figure 1.

Figure 1.

Open in a new tab

Target (top panels) and alternative (bottom panels) mean response rates across sessions for CDT and All On groups exposed to 4 (left panels) or 8 (right panels) cycles of CDT or the equivalent duration of All On alternative reinforcement. Error bars represent ± SEM.

Figure 2 (top panels) shows target response rates in the final session of Phase 2 and all sessions of Phase 3 (resurgence test). A two-way ANOVA revealed target rates in the final session of Phase 2 did not differ for CDT versus All On, F(1,32) = 1.57, p = .220, η2G = .047, or for 4 versus 8 cycles, F(1,32) = .111, p = 0.742, η2G = .003, and there was no significant interaction, F(1,32) = 0.241, p = .627, η2G = .007. Thus, target rates were low and did not differ for CDT versus All On alternative reinforcement prior to resurgence testing. A mixed 2 × 2 × 2 (phase × type × number) ANOVA comparing target rates during the final session of Phase 2 and the first session of Phase 3 revealed significant main effects of phase, F(1,32) = 115.56, p < .001, η2G = .599, and type F(1,32) = 7.28, p = .001, η2G = .118, and a significant phase × type interaction F(1,32) = 16.22, p < .001, η2G = .173. No other effects were significant. Thus, resurgence occurred and was larger for All On than CDT, but did not differ for 4 versus 8 cycles. Considering responding across all 10 sessions of Phase 3, a 2 × 2 × 10 (type × number × session) mixed ANOVA revealed a significant main effect of session, F(3.22,103.2) = 24.13, p <.001, η2G = .29, and a significant type × session interaction, F(3.22,103.2) = 6.29, p < .001, η2G = .09. No other effects were significant. Thus, across resurgence testing, target response rates were initially higher following All On than CDT, and as a result, responding decreased more across continued sessions for All On. The fact that the impact of CDT versus All On did not depend on the number of cycles suggests CDT remained as effective across extinction following 4 cycles as following 8 cycles. Separate one-way ANOVAs revealed target rates decreased across sessions for both CDT, F(4.57, 77.71) = 7.906, p < .001, η2G = .164, and All On, F(2.79, 47.36) = 17.372, p < .001, η2G = .372, but more so for All On (i.e., the significant 2 way interaction above). Thus, there was no evidence of an increase in target responding across extinction following CDT, instead there was a significant decrease.

Figure 2.

Figure 2.

Open in a new tab

Target (top panels) and alternative (bottom panels) mean response rates in the final session of Phase 2 and across all Phase 3 resurgence test sessions for CDT and All On groups exposed to 4 (left panels) or 8 (right panels) cycles of CDT or the equivalent duration of All On alternative reinforcement. Error bars represent ± SEM.

Figure 2 (bottom panels) shows alternative rates in the last session of Phase 2 and across all sessions of Phase 3. A two-way ANOVA revealed alternative response rates in the final session of Phase 2 were marginally higher for All On than CDT, F(1,32) = 4.136, p = .050, η2G = .114, but did not differ for 4 versus 8 cycles, F(1,32) = 2.46, p = 0.127, η2G = .073, and there was no significant interaction, F(1,32) = 0.016, p = .899, η2G < .001). Finally, there was a significant 3-way interaction between treatment type, number, and session, F(1.5, 47.87) =4.42, p = .003, η2G = .093, resulting from higher response rates for All On that decreased more slowly across sessions than for CDT.

Discussion

This experiment demonstrated that exposure to as few as four cycles of CDT is sufficient and as effective as eight cycles in producing low rates of target behavior during treatment and during extended extinction. Importantly, the low rates of target behavior following both four and eight cycles of CDT showed no evidence of increasing across the ten sessions of extinction of alternative behavior, and in fact target responding decreased across extinction. This outcome is consistent with extant theories suggesting CDT teaches organisms to discriminate that reinforcement for target behavior remains unavailable even when alternative reinforcement is not available. Although the source of the potential increasing trend in target responding across extinction in Smith and Greer (2023) with crowdsourced humans is unclear, it deserves further experimental attention as it is the only published examination of CDT with humans. Further research should examine the effects of duration of CDT exposure in humans in other non-crowdsourced preparations. In addition, other differences between Smith & Greer and research with rats should be examined, including session durations, the number of sessions per day, and schedules of reinforcement for target and alternative behaviors. Based on current theories, such factors would be expected to alter the efficacy of CDT to the extent that they alter the relevant hypothesized discrimination.

Finally, although the resurgence in target response rates following All On alternative reinforcement was temporary (e.g., Craig & Shahan, 2016), the ability to mitigate such increases is important for preventing errors of commission where problem behavior is reinforced (e.g., St Peter Pipkin et al., 2010). Thus, the present results suggest CDT might have translational utility even following relatively small numbers of cycles of on/off alternative reinforcement.

Highlights.

  1. Treatment with 4 and 8 on/off alternative reinforcement cycles similarly reduced target behavior

  2. 4 and 8 cycles of on/off alternative reinforcement both reduced resurgence in extended extinction

  3. On/off alternative reinforcement may have translational utility following relatively few cycles

Acknowledgments

This work was funded by grant R01HD093734 (TAS) from the Eunice K. Shriver National Institute of Child Health and Human Development. The authors thank Santiago Rojas Otero for help conducting the experiment.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Briggs AM, Fisher WW, Greer BD, Kimball RT, 2018. Prevalence of resurgence of destructive behavior when thinning reinforcement schedules during functional communication training. J. Appl. Behav. Anal 51, 620–633. 10.1002/jaba.472 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Craig AR, Shahan TA, 2016. Behavioral momentum theory fails to account for the effects of reinforcement rate on resurgence. J. Exp. Anal. Behav 105, 375–392. 10.1002/jeab.207 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Lattal KA, Wacker D, 2015. Some dimensions of recurrent operant behavior. Rev. Mex. Análisis Conducta 41, 1–13. 10.5514/rmac.v41.i2.63716 [DOI] [Google Scholar]
  4. Lieving GA, Hagopian LP, Long ES, O’Connor J, 2004. Response-class hierarchies and resurgence of severe problem behavior. Psychol. Rec 54, 621–634. 10.1007/BF03395495 [DOI] [Google Scholar]
  5. Petscher ES, Rey C, Bailey JS, 2009. A review of empirical support for differential reinforcement of alternative behavior. Res. Dev. Disabil 30, 409–425. 10.1016/j.ridd.2008.08.008 [DOI] [PubMed] [Google Scholar]
  6. Schepers ST, Bouton ME, 2015. Effects of reinforcer distribution during response elimination on resurgence of an instrumental behavior. J. Exp. Psychol. Anim. Learn. Cogn 41, 179–192. 10.1037/xan0000061 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Shahan TA, Browning KO, Nall RW, 2020. Resurgence as Choice in Context: Treatment duration and on/off alternative reinforcement. J. Exp. Anal. Behav 113, 57–76. 10.1002/jeab.563 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Shahan TA, Craig AR, 2017. Resurgence as Choice. Behav. Processes 141, 100–127. 10.1016/j.beproc.2016.10.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Smith SW, Greer BD, 2023. Translational evaluation of on/off alternative reinforcement cycling. J. Exp. Anal. Behav 120, 429–439. 10.1002/jeab.879 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. St Peter Pipkin C, Vollmer TR, Sloman KN, 2010. Effects of treatment integrity failures during differential reinforcement of alternative behavior: a translational model. J. Appl. Behav. Anal 43, 47–70. 10.1901/jaba.2010.43-47 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Trask S, Keim CL, Bouton ME, 2018. Factors that encourage generalization from extinction to test reduce resurgence of an extinguished operant response. J. Exp. Anal. Behav 110, 11–23. 10.1002/jeab.446 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Volkert VM, Lerman DC, Call NA, Trosclair-Lasserre N, 2009. An evaluation of resurgence during treatment with functional communication training. J. Appl. Behav. Anal 42, 145–160. 10.1901/jaba.2009.42-145 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Wacker DP, Harding JW, Berg WK, Lee JF, Schieltz KM, Padilla YC, Nevin JA, Shahan TA, 2011. An evaluation of persistence of treatment effects during long-term treatment of destructive behavior. J. Exp. Anal. Behav 96, 261–282. 10.1901/jeab.2011.96-261 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Winterbauer NE, Bouton ME, 2010. Mechanisms of resurgence of an extinguished instrumental behavior. J. Exp. Psychol. Anim. Behav. Process 36, 343–353. 10.1037/a0017365 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES