Relapse: An Introduction

Quitting smoking is easy…I’ve done it many times. This is an old joke, often attributed to Mark Twain, although there seems to be no evidence that he actually ever said it. Nevertheless, like other effective behavior, the joke worked in the past, and there is a good chance it will work well into the future. The reason the joke works is that even non-scientists can relate to what it conveys about the difficulty of maintaining long-term behavior change--previously eliminated behavior tends to return. Scientifically, it has become quite clear that behavior suppressed by a variety of means (e.g., extinction, punishment, omission contingencies, alternative reinforcement, various combinations of these) tends to recur following a wide range of changes in circumstances. Such recurrence goes by different names depending on the specific changes in circumstance that provoke it (e.g., resurgence, renewal, reinstatement, spontaneous recovery), all of which comprise an area of research that is often referred to more generically by the moniker “relapse.”

Numerous histories/reviews of the remarkably long study of different relapse-like phenomena across different eras and various research domains can be found elsewhere (e.g., Epstein, 1985, 2015; Bouton & Swartzentruber, 1991; Shalev, Grimm, & Shaham, 2002; Bouton, 2002; Vervliet, Baeyens, Van den Bergh, & Hermans, 2013; McConnel & Miller, 2014; Lattal & Wacker, 2015; Kestner & Peterson, 2017; Khoo Gibson, Prasad, & McNalley, 2017; Podlesnik, Kelley, Jimenez-Gomez, & Bouton; 2017). A quick perusal of the pages of JEAB and JABA in recent years makes it clear that such phenomena have become the focus of increasing interest in basic, translational, and applied research in behavior analysis. The name “relapse” for the research area has probably stuck because of the obvious clinical relevance of the return of problem behavior (and likely its fit with the priorities of funding agencies). Indeed, it is true that although interventions for a wide range of behavioral problems (e.g., addictions, destructive behavior in intellectual and developmental disabilities, anxiety disorders, eating disorders, etc.) are highly effective initially, relapse of problem behavior is disconcertingly widespread. However, the name is not entirely appropriate because in common usage the term “relapse” refers to the return of inappropriate or unwanted behavior. Obviously, the phenomena falling under into this category could just as easily result in the return of appropriate or desirable behavior (e.g., Epstein, 1985; Shahan & Chase, 2002; Novak, Erath, & DiGennaro-Reed, this issue; Williams & St. Peter, this issue). Further, in animal studies where arbitrary responses are used, it is clear that judgements about appropriate or inappropriate behavior are completely meaningless. Given these perfectly reasonable concerns, some have suggested the name “recurrence” (see Lattal & Wacker, 2015) for the general research area. Nevertheless, there is little doubt that interest within and beyond the field has been driven by the hope of making progress toward reducing the return of various types of unwanted behavior. So, setting aside concerns about purity of nomenclature, and in the hopes of maintaining and expanding the reach of such work published in JEAB, I suggest acceptance and continued use of the name “relapse” for the general area of research. Thus, I have named this special issue accordingly.

Theoretically, although the various relapse phenomena may themselves go by different names, there is a long-standing and ongoing effort to account for them all with the same set of basic processes (e.g., Bouton & Swartzentruber, 1991; Bouton, Winterbauer & Todd, 2012; McConnel & Miller, 2014; see also Lattal & Wacker, 2015; Nevin et al., 2017). In addition, relapse phenomena continue to play a pivotal role in vigorous theoretical debates about the fundamental nature and neurobiological bases of learning and memory (e.g., Bouton & Swartzentruber, 1991; Bouton, 2002; Redish, Jensen, Johnson, & Kurth-Nelson, 2007; McConnel & Miller, 2014; Dunsmoor, Niv, Daw, & Phelps, 2015; Gallistel & Papachristos, this issue). The reason relapse phenomena have played a pivotal role in these debates is that they have long raised questions about the nature of learning and extinction (and other suppression techniques) as portrayed by theories of conditioning and learning. If learning represents the build-up or accumulation of response strength, associative values, or connections, and extinction undoes this with some sort of dissipation, decrease, or weakening, how can an extinguished response suddenly reappear with some change in circumstances? As an examination of the citations above will make clear, there have been a variety of approaches to this problem. For example, perhaps extinction and relapse phenomena result from the imposition of new learning on old--possibly inhibitory learning that is contextually dependent or dissipates at a different rate. Or, alternatively, perhaps learning, extinction, and relapse phenomena suggest something entirely different involving explicit learning that or when reinforcement is or is not coming--at least for a while.

In the Skinnerian tradition there has been little formal theorizing about relapse phenomena, with a tendency to instead settle for simply describing, defining, or naming as a substitute (see Shahan, 2017, for related discussion). This lack of theoretical effort directed at relapse phenomena in the operant tradition is unfortunate because the phenomena draw attention to what might be considered some of the core scientific questions about operant learning and behavioral regulation. For example, how is a history of past experiences carried forward and integrated with changing circumstances across time? Once a relation between events has been learned, how does an organism learn when things have changed? Given that an organism does learn that things have changed, how does it decide if or when old learning might be relevant again? In my view, what makes relapse phenomena interesting from a basic-science perspective is that they are especially useful proving grounds for these and related fundamental theoretical questions. I suspect that a reluctance to engage with the type of questions raised may have prevented many who publish in JEAB from seriously entering the theoretical fray. But, the theoretical questions involved could have the potential to truly advance and fundamentally alter our understanding of behavior while simultaneously making contact with highly impactful research domains with widespread popular interest (e.g., neuroscience, reinforcement learning, artificial intelligence). So, with any luck, maybe something in this special issue will serve as a call to arms for the established or aspiring theorists among us.

Regardless, I do hope it will be apparent from the pages of this special issue that the study of relapse seems special in representing a research area that has both critical basic-theoretical importance and easily recognized (and communicated) significance for a wide range of issues of human concern. Like previous examples of such areas of research (e.g., delay discounting), relapse phenomena provide us with an opportunity to study the fundamental scientific issues of our field, contact other fields concerned with similar problems, justify our existence to society at large, and hopefully do some good along the way. To that end, the purpose of this special issue is to bring together work on relapse from a variety of areas and approaches in the hopes of encouraging inspiration, integration, and innovation.

Theoretical, Perspective, and Review Papers

The special issue opens with seven papers providing theoretical perspectives and/or reviews of aspects of relapse including basic theoretical issues in learning and memory, addictions, resurgence, anxiety disorders, and human research on renewal.

Gallistel and Papachristos provide a brief review of evidence supporting content-based theories of learning suggesting that organisms explicitly learn durations and numerosities in conditioning experiments (e.g., Gallistel & Gibbon, 2000; Balsam & Gallistel, 2009; Gallistel & Wilkes, 2016; Gallistel, Craig, & Shahan, 2019). They note that previous evidence of such learning has focused on learning about trial parameters (i.e., trial duration, intertrial interval), and thus they report novel experiments examining higher levels of hierarchically structured episodes in a Pavlovian anticipatory nose-poking protocol with mice. Specifically, they examined the effects of number of training sessions, trials per session, intersession interval, and the number of days over which training extended (i.e., span of training) on acquisition, extinction, and spontaneous recovery. Their results suggest that fewer training trials per session produce faster acquisition, but that more total training trials generate higher levels of responding. Although trials per session and total trials did not affect extinction or spontaneous recovery, greater spans of training generated more spontaneous recovery, regardless of the number of sessions within the span. They suggest these data support the notion that organisms learn about numbers, durations, and spacings of longer-term episodes in conditioning experiments (sessions, spans of days in training and extinction), and they argue that their results pose serious challenges to both associative conceptions of learning and memory and to standard accounts of the underlying neurobiology.

Grimm provides a review of experiments on a phenomenon known as incubation of food craving. The phenomenon was initially demonstrated with rats responding for cues previously associated with self-administered cocaine (Grimm, Hope, Wise, & Shaham, 2001) and is a variant of the widely used cue-induced reinstatement procedure (See, 2005, for review). The basic finding is that responding that produces a cue previously associated with delivery of a primary reinforcer increases (i.e., incubates) as a function of time (across a span of many weeks) without access to the primary reinforcer. As the name suggests, the phenomenon is typically interpreted as reflecting an increase in motivation for the primary reinforcer previously associated with the cue. Grimm reviews a body of work examining the effects of behavioral and neurobiological variables on incubation of craving for responding previously maintained by food reinforcers. The findings could have important implications for treating craving in food and drug addictions. Strangely, incubation of craving appears not to have been the focus of any experimental or theoretical work by behavior analysts. The phenomenon clearly deserves our attention as it suggests that there could be an increasing likelihood of the return of problem behavior with the passage of time if responding happens to produce stimuli that were once predictive of reinforcement for problem behavior.

Bickel and Athamneh describe how relapse to substance use might be understood from the perspective of Reinforcer Pathology theory. The latest version of this choice-based theory (version 2.0) suggests that excessive reinforcing value of addictive substances is determined in part by a limited temporal window of reinforcer integration (as measured by delay discounting). The theory suggests that abstinence may result from either an extended temporal window or changes in the contingent relations between substance use and its consequences. Thus, lapses might result from shorter-term decreases in the temporal window and relapse to more extended substance use might result from longer-term changes in the temporal window. Further, if treatments arranging contingencies between substance use and other consequences do not alter the temporal window, then relapse might be expected when those contingencies are no longer in effect. As a result, the authors suggest that multi-component treatments arranging contingencies on abstinence and focusing on reducing delay discounting may be more effective for preventing relapse.

Shahan, Browning, and Nall provide a review, quantitative assessment, and extension of Resurgence as Choice (i.e., RaC) theory. The theory is an extension of the generalized matching law that includes a means to incorporate histories of experiences like those known to generate resurgence (e.g., extinction, shifting reinforcer allocations across time). They present novel data with rats showing that increases in treatment duration with extinction plus alternative reinforcement produce only small, but reliable decreases in subsequent resurgence. In addition, they replicate previous findings showing that exposure to cycling on/off alternative reinforcement reduces ultimate resurgence. Fits of RaC to the data reveal some inadequacies of the model and they propose a revised model that incorporates a role for local signaling effects of reinforcer deliveries above and beyond the values of the target and alternative behaviors. The new model provides an excellent account of the data and represents an integration of RaC and a quantitative statement of some aspects of context theory. The new model also suggests a means by which other relapse phenomena might be incorporated into a more general choice-based quantitative theory.

Lattal and Oliver provide a review of animal and human studies on resurgence in which a control response was employed. Such control responses are often employed in the hopes of ruling out nonspecific increases in responding as an explanation for increases in a target behavior when an alternative behavior is extinguished. Their review suggests that although more common in human studies, meaningful increases in control responding are almost non-existent in animal studies. Regardless, they suggest that inclusion of such a control response is not useful because both the presence and absence of control responding are open to multiple unhelpful interpretations. They suggest alternative methods by which generalized increases in responding might be ruled out as an explanation for increases in target responding during a resurgence test.

Smith et al. provide a review of research on relapse phenomena in anxiety disorders and suggest that a better understanding of Pavlovian, operant, and relational processes might aid in the prevention of relapse. Further, they suggest that Acceptance and Commitment Therapy teaches strategies for altering the functions of aversive stimuli and for orienting behavior toward sources of reinforcement that remain in the face of changing environmental contingencies, and thus, might help to reduce relapse.

Saini and Mitteer provide a systematic review of human research on renewal and suggest that the phenomenon is robust in both laboratory and clinical settings, but they note that that relatively few studies have attempted to reduce or eliminate renewal of clinically meaningful behavior. Further, based on clinical considerations, they suggest that theories of relapse should seek to incorporate dynamic changes in context as well as combinations different relapse-inducing events (e.g., contextual renewal + resurgence induced by omission of alternative reinforcement).

Renewal

Following the review of renewal research with human subjects, the issue includes six empirical papers on renewal employing animals, humans online, avoidance behavior of humans, humans in a simulated workplace, and humans in a translational arrangement.

Rey, Thrailkill, Goldberg, and Bouton directly compare renewal of rats’ responding eliminated be either extinction or an omission contingency in a context that differed from the training context. Renewal of responding occurred similarly following both elimination procedures when there was a return to the original training context. However, the presentation of noncontigent pellets during both the elimination context and the original training context resulted in the recovery of responding following elimination by extinction but not the omission contingency. These findings suggest that as compared to behavior eliminated by extinction, behavior eliminated by omission training may be similarly susceptible to renewal, but less susceptible to reinstatement induced by reinforcer deliveries.

Craig, Sullivan, Browning, DeRosa, and Roane show that rats intermittently exposed to reinforcement for a target behavior in the previous training context (i.e., Context A) during extinction of that response in a different context (i.e., Context B) show less renewal when tested in a different context (i.e., Context C) than rats not experiencing such intermittent exposure to Context A reinforcement. This result stands in contrast to previous work by these authors (Craig, Sullivan, & Roane, 2019) showing that such intermittent exposure to reinforcement to Context A instead increases ABA renewal in a similar procedure. The authors suggest that alternating between contexts associated with reinforcement and extinction during treatment may offer a clinical strategy for reducing renewal of problem behavior in novel contexts.

In a first of its kind, Robinson and Kelley present two experiments demonstrating both renewal and resurgence in humans using Amazon’s Mechanical Turk platform as a subject pool. The fact that the platform can be used to examine these phenomena in humans is a major innovation and has the potential to both considerably accelerate the pace of relapse research with humans and to reduce its costs. The ability to ask experimental questions and have them answered very quickly and cheaply could have profound effects on both basic theoretical development and translational research aimed at quickly assessing potentially clinically relevant manipulations.

Schlund et al use an approach-avoidance task to demonstrate ABA renewal of avoidance of money loss and self-reported feelings of threat and loss expectancy in neurotypical adults. The focus on renewal of avoidance is novel, and both the procedure and findings may have important implications for research on relapse in anxiety disorders.

Novak, Blackman, Erath, and DiGennaro Reed examine renewal of appropriate behavior in a simulated workplace. In two experiments they demonstrate renewal of a check processing task and implementation of behavior-analytic teaching procedure. Their study represents an important extension of the study of relapse processes to issues of concern in organizational behavior management.

In three experiments with children in a translational arrangement, Kimball, Greer, Randall, and Briggs compare ABA renewal under the usual extinction conditions in the final two phases (i.e., BA) to renewal when extinction of the target behavior plus differential-reinforcement-of-alternative behavior (i.e., DRA) is arranged in those phases. They report that the target behavior was more disrupted in context B and that renewal was reduced with the return to context A when DRA was arranged. In addition, they demonstrate that this outcome was not due merely to the presence of a second response option in the DRA conditions. The authors suggest that further such examinations of renewal during extinction plus DRA are important because they more closely mimic conditions likely to be encountered in practice.

Resurgence

The special issue closes with five empirical articles on resurgence including basic empirical papers with animals, humans in an experimental/translational settings, and experimental evaluations of the treatment of problem behavior in the clinic or home.

In two experiments, Shvarts, Jimenez-Gomez, Bai, Thomas, Oskam, and Podlesnik examined the effects of stimuli paired uniquely with alternative reinforcer deliveries in Phase 2 on subsequent resurgence of target behavior of pigeons and children diagnosed with Autism Spectrum Disorder. Resurgence was examined during extinction of target and alternative responding both in the presence and in the absence of continued alternative-response-contingent presentations of the paired stimuli. Even though the maintenance of alternative responding did not differ between assessments, resurgence was modestly lower in both experiments when the alternative response continued to produce the paired stimulus. These findings replicate a similar modest effect of response-dependent presentations of stimuli paired with reinforcement on resurgence of rats by Craig, Browning, and Shahan (2017), and suggest that the modest size of the effect was not due to the fact that Craig et al. employed stimuli paired with both target and alternative reinforcer deliveries in Phases 1 and 2. As Shvarts et al. note, although it appears that stimuli paired with primary reinforcement might modestly reduce resurgence, the processes responsible for this effect remain unclear, and considerably more research will be required before they can be recommended for clinical use.

Williams and St. Peter found resurgence of desirable behavior in the context of college-level mathematics instruction. In two experiments, they demonstrate resurgence of a previously taught method to solve quadratic equations when novel problems were presented for which neither the first-taught nor a second-taught method could produce the correct solution. In addition, across experiments, their data suggest that the similarity of the form of the problem during the resurgence test to problems associated with either the first-taught or second-taught method might constitute a context that affects the amount of resurgence obtained.

Greer, Fisher, Retzlaff, & Fuhrman examine in a within-subject design the effects of short (i.e., at least 2 sessions) versus long (i.e., at least 6 sessions) exposures to extinction plus alternative reinforcement on subsequent resurgence of severe destructive behavior of children referred for treatment services. They find no effects of the different treatment durations on resurgence or the persistence of destructive behavior. The authors note that these findings run counter to the robust decreases in resurgence with increases in treatment duration predicted by the Behavioral Momentum Theory of resurgence (Shahan & Sweeney, 2011) and the considerably more modest decreases predicted by Resurgence as Choice (i.e., RaC; Shahan & Craig, 2017). However, they also note that some procedural complexities and the within-subject design make formal assessments of either model difficult. The findings of Shahan et al. (this issue) with rats in a group design under highly controlled circumstances suggest that the effects of longer treatment durations appear to be modest enough (consistent with RaC) that more powerful designs and a wider range of durations may be required to detect them in a clinical setting.

Fisher, Fuhrman, Greer, Mitteer, and Piazza evaluate the effects of discriminative stimuli that signal regular periods of alternative reinforcement availability (SD) versus unavailability (SΔ) on resurgence of destructive behavior of children with intellectual developmental disorder. As in previous research by this group, they find that including the SΔ during an extinction challenge reduces resurgence considerably as compared to when the SΔ was not present. Further, they isolate the effects of the stimuli by ruling out other sources of the effect in previous demonstrations (e.g., differential the length of treatment exposure, differential reinforcement rates, and the availability of time-based reinforcement during the resurgence test). These findings isolating the effects of such discriminative stimuli suggest that their use might offer an important way to mitigate resurgence in the clinic and beyond, and the effect is clearly deserving of considerable additional empirical and theoretical attention.

Finally, Suess, Schieltz, Wacker, Detrick, and Podlesnik use telehealth to deliver functional communication training (i.e., FCT) in order to reduce problem behavior of children with Autism Spectrum Disorder. Specifically, they show that initially training appropriate behavior in multiple contexts and then employing strategies to facilitate generalization from the alternative contexts to the treatment context can reduce resurgence of problem behavior more quickly across intermittent extinction challenges when compared to previous applications of FCT without such training.

Coda

The papers in this special issue make it clear that research on relapse represents a diverse and thriving area. The breadth of the topics addressed also makes it clear that the area has implications for a range of basic theoretical and applied concerns. A number of emerging issues have become clear within these pages. These include how learning about larger-scale regularities in the environment might contribute to relapse, how dynamically changing stimuli and reinforcement contingencies interact across time in order to contribute to relapse, relapse of avoidance and its role in anxiety disorders, relapse of appropriate and inappropriate behavior in educational and organizational training settings, the use of online subject pools, and more. Overall, the work presented here suggests that research on relapse is poised for additional growth and impact. Thus, like the oldie-but-goodie that it is, I suspect that the area will continue to effectively deliver to audiences well into the future.