Printer Friendly

Resurgence of previously reinforced responding: research and application.

Previously learned responses recur under a variety of conditions. Whether such recurrence is good news or bad news depends on the circumstances of the recurrence. Recurrence of a previously eliminated problem behavior is bad news, but the recurrence of a previously established strategy that facilitates constructive problem solving receives a warmer reception. One label attached to some recurrent responding is resurgence. The term is used to describe a procedure, to describe the behavioral effect of the procedure, and to describe a behavioral process (cf. Mazur, 2006). These three uses of the term often are concurrent. Laboratory studies with nonhuman and human animals have isolated some of the variables that contribute to resurgence. In this review, we consider these variables and their implications for facilitating or reducing the likelihood of past behavior recurring in applied research and practice (hereafter, application).

Resurgence is said to occur when a previously learned response recurs following a hiatus from that response, during which time some other response first is reinforced and thereafter extinguished. It is during the final condition that the previously learned response resurges. The process of resurgence thus involves three phases. In the first, or reinforcement, phase, a response, A, is reinforced. In the second, or alternative reinforcement, phase, a second response, B, is reinforced while Response A is extinguished. In the third, or resurgence, phase, Response B is extinguished while extinction remains in effect for Response A. The recurrence of Response A is labeled as resurgence.

Resurgence may be distinguished from several other circumstances wherein previously learned responses recur. Spontaneous recovery similarly occurs following extinction of such responses, and after an absence of exposure to the extinction situation. It differs from resurgence in that other responses are not (systematically) reinforced during extinction of the response that later spontaneously recovers (but see Cleland, Guerin, Foster, and Temple [2001] for a discussion of circumstances in which resurgence may be considered an instance of spontaneous recovery). Reinstatement occurs when a previously learned response recurs during a period of response-independent delivery of the reinforcer previously used to maintain the response, but after that response first is extinguished. In reinstatement, the response recurs as a result of the evocative, discriminative stimulus effects of the now response-independent presentations of the previously established reinforcer (cf. Franks & Lattal, 1976). Response induction or response generalization may be considered instances of response recurrence in which topographies develop that are similar to the reinforced one. The distinguishing features of these phenomena are important for application because they may result in the re-emergence of responding under different conditions or in different forms. Although all four types of response recurrence have implications for application, the present analysis is restricted to resurgence.

Examples of Resurgence

The first systematic experimental analysis of resurgence appears to have been that of Carey (1951; [see also Carey, 1953]). Two groups of rats were trained to lever press. With one group, a sequence of two lever-press responses that occurred within 0.25 s of one another, which were described as "doubles," first was reinforced. With the other group, only single responses were reinforced. Subsequently, the conditions were reversed for the two groups such that the "doubles" group received reinforcers following single responses and the "singles" group received reinforcers following double responses. In a final phase, lever pressing was extinguished for both groups. During extinction, as the number of instances of the last-reinforced response sequence (singles or doubles) decreased, the frequency of the other, first reinforced, pattern increased. Thus, the sequence of experimental events was: reinforce one pattern of responding, then reinforce a second such pattern while extinguishing the first pattern, then extinguish the second pattern and the first pattern resurges.

Subsequently, in a series of experiments that began with Leitenberg, Rawson, and Bath (1970), Leitenberg and his colleagues investigated the effects of the reinforcement of competing behavior on the elimination of previously established responding. In a prototypical experiment in this series, rats were trained to press a lever located to the right of the food tray. Then, reinforcement of responding on this lever was discontinued and responding on a lever located to the left of the food tray was reinforced. Responding on the right lever decreased rapidly under this condition. When, however, left-lever responding was extinguished the rats resumed responding on the right lever. Such an effect thereafter was systematically replicated by Epstein and Skinner (1980; see also Epstein, 1983, 1985) and labeled resurgence.

Resurgence can occur under a variety of conditions related to application. Resurgence of undesirable responses can occur when a behavioral treatment is implemented improperly or discontinued, as in the case of treatment integrity failures. In such cases, during behavioral treatment it is likely that the reinforcement contingencies during pretreatment conditions favored problem behavior. These pretreatment conditions may be analogous to the reinforcement phase in a typical resurgence experiment (that is, one response--problem behavior--is reinforced and an alternative response is not). A common intervention for ameliorating problem behavior is differential reinforcement of alternative behavior (DRA). During conventional (but not all) DRA procedures, problem behavior is extinguished while an alternative response is reinforced. Rates of problem behavior decrease and rates of appropriate behavior increase, making this phase similar to the alternative reinforcement phase of resurgence experiments. If, however, the DRA is discontinued or implemented improperly (e.g., when an untrained caregiver, such as a substitute teacher, babysitter, or new staff member, is responsible for carrying out the treatment program) reduced reinforcement rates or even extinction may result for both problem and appropriate behavior. This condition is similar to the resurgence phase, and problem behavior may recur. Such resurgence occurs as a function of both the client's reinforcement history (during pretreatment and treatment phases) and the current conditions of reduced reinforcement rates or extinction of the alternative response.

Desirable resurgence occurs when a previously reinforced adaptive response recurs following a period without reinforcement, and during which other responses in the repertoire are unsuccessful (cf. Epstein, 1991). Desirable resurgence may be involved in such activities as appropriate academic behavior. Consider a situation in which a student taking an examination suddenly remembers information related to a topic. This could develop as follows. Assume that a teacher praises the use of a particular strategy to solve problems (similar to the reinforcement phase), but that this strategy is unsuccessful during subsequent lessons that involve structurally different problems (similar to the alternative reinforcement phase). With neither response now reinforced (extinction; for example, during the examination, when immediate praise would not be available for engaging in either response), the previously effective strategy may resurge.

Resurgence, then, is a reliable effect that occurs across a variety of species, responses, and environmental circumstances. What follows is an analysis of some of the variables operating in each of the three phases of the resurgence process. Sources of control over resurgent responding first are identified, followed by observations about how variables in each of these phases might contribute to applications.

Variables Affecting Resurgence

Reinforcement Phase

In the laboratory, both single responses and sequences or patterns of responses have been established in the reinforcement phase (e.g., Carey, 1951; Epstein, 1983; Epstein & Skinner, 1980; Leitenberg et al., 1970; Reed & Morgan, 2006). Reed and Morgan (see also Bacha-Mendez, Reid, & Mendoza-Soylovna, 2007), in three successive conditions, reinforced three different three-response lever-pressing sequences with rats (e.g., respond on the left lever, then the right lever, then the right lever again). When all response sequences were extinguished, the rats first emitted the most recently (third) reinforced sequence. Thereafter, for 3 of the 6 rats, the first-reinforced sequence predominated and for 2 others the second-reinforced sequence was most frequent. For the 6th rat, an equal number of first- and second-reinforced response patterns occurred. During a second, identical, extinction session, all 6 rats emitted the second-reinforced sequence most frequently. Reed and Morgan noted that "it was not the case ... that behavior during extinction became immediately chaotic; rather, behavior followed orderly patterns of resurgence" (p. 313), thereby expanding on Epstein's (1983) earlier finding that resurgence was most likely to occur on an operandum where responding previously had been reinforced, rather than on one always associated with extinction. Thus, both Epstein's and Reed and Morgan's findings suggest that the conditions of reinforcement of the response in the reinforcement phase will determine the characteristics of the response that eventually resurges in the third phase.

The sequence of responses used by Reed and Morgan (2006) may be similar to response-class hierarchies in application. Response-class hierarchies are sequences of responses that are maintained by the same reinforcer. They have been demonstrated with children who engage in problem behavior (e.g., Harding et al., 2001; Richman, Wacker, Asmus, Casey, & Andelman, 1999). Lieving, Hagopian, Long, & O'Connor (2004) assessed the possibility of resurgence within the same response class hierarchy. The participants were 2 children who engaged in multiple forms of problem behavior. The experimenters first reinforced (with access to tangible items) any instances of problem behavior that occurred. Then, reinforcers were withheld for disruptive responses while being delivered for one or more other responses (aggression or aggression and cursing). Finally, reinforcers were withheld for two or more responses. During this final phase, both children engaged in higher rates of the initially reinforced problem behavior than during the previous phase, suggesting that resurgence occurred.

Given that many children undergoing treatment exhibit more than one topography of problem behavior, the possibility of the resurgence of such behavior is important. Resurgence may be particularly likely when clients have a history of reinforcement of multiple topographies of problem behavior, as in the case of response class hierarchies. Thus, the further analysis of resurgence of members of response class hierarchies may illuminate features of reinforcement-phase variables that promote or inhibit later resurgence.

The previous examples all have involved positive reinforcement in the reinforcement phase; however, resurgence also has been demonstrated after a history of negative reinforcement during that phase. Bruzek (2007) reinforced caregiving responses in a simulated environment, using a negative reinforcement procedure in which particular caregiver responses resulted in the termination of infant cries. Different caregiving responses were reinforced during the alternative reinforcement phase. When all responses were extinguished during the resurgence phase, the initially reinforced response resurged, demonstrating the generality of resurgence to instances in which negative reinforcement maintains reinforcement-phase responding. To our knowledge, this is the only explicit study of resurgence following negative reinforcement (but see the discussion of Goh & Iwata, 2004 in the next paragraph). Yet, it is not uncommon for problem behavior to be maintained by negative reinforcement (cf., Hanley, Iwata, & McCord, 2003), and increases in problem behavior during initial implementation of escape extinction have been documented (Goh & Iwata). These results suggest that resurgence of negatively reinforced responding could have important implications for applied behavior analysis. Bruzek's findings invite further investigation to determine specific variables that affect resurgence following negative reinforcement.

Many variables implicated in response maintenance have not yet been investigated in terms of their possible effects on resurgence. For example, in the early experiments conducted by Leitenberg and colleagues (e.g., Leitenberg et al., 1970), the training of the response in the first phase (hereafter, the first response) occurred for only a few sessions, typically five, which is difficult to compare with the far more extensive histories often seen before problem behavior is targeted for change. Goh and Iwata (1994) assessed the effects of escape extinction on SIB, but measured both SIB and aggression. Aggression was never reinforced during the experiment, but SIB was initially reinforced and then extinguished. During extinction, rates of aggression increased above those observed during the previous phase, suggesting resurgence. Although aggression was never directly reinforced during the study, presumably it had a long history of reinforcement outside of the experimental context, highlighting the possibility that extensive extra-experimental histories may contribute to resurgence of problem behavior during treatment.

Along with reinforcement history, the dimensions of the operant response (cf. Gilbert, 1958) and the conditions of response maintenance may be important in predicting or assessing resurgence. For example, the response rate (da Silva, Maxwell, & Lattal, 2008), tempo, topography, and duration of the response in the reinforcement phase all may affect subsequent resurgence. Currently, little is known about how parameters of the reinforcement schedule during the first phase, such as the reinforcer type, schedule value, delay to reinforcement, and quality of reinforcement, affect resurgence.

Alternative Reinforcement Phase

During the second, alternative reinforcement phase, two operations are introduced concurrently: reinforcement of the response maintained in the first phase is discontinued and an alternative response is reinforced. These operations raise questions about the status of both responses. Is the first response "eliminated" or simply "replaced" by reinforcement of the second response in this phase? How do the parameters of reinforcement operating on this latter response affect both the decline of the first response and the subsequent resurgence of that response?

Leitenberg et al. (1970) trained rats to respond on one lever, A, according to a multiple variable-interval (VI) 30-s VI 30-s schedule for five sessions. In the critical part of the second phase, responding to Lever A during one of the multiple schedule stimuli was extinguished. During the extinction component, half of the subjects were trained to respond on a second lever, B, where food was delivered according to a fixed-ratio (FR) 10 schedule. Responding on Lever B was never reinforced for the other half of the subjects; these animals received no reinforcers during the second phase. Responding on Lever A during this phase was significantly lower for the group receiving reinforcers for pressing Lever B than for group not receiving reinforcers. In the third, resurgence, phase, reinforcers were not available for responding on either lever for both groups. There was an increase in Lever A responding for both groups, as compared to the second phase. This increase, however, was greater for the group for which Lever B responses had been reinforced during the second phase than it was for the group that did not receive reinforcers following Lever B responses during that phase. Leitenberg et al. took this difference as evidence that Lever A responding in the second phase had not extinguished for the group wherein Lever B responses were reinforced during that phase. Thus, the resurgence of Lever A responding was suggested to reflect, at least in part, the failure of such responding to contact the extinction contingency in the second phase because of the alternative reinforcement. Over several sessions of the resurgence phase, response rates for rats that had received alternative reinforcers during the second phase decreased, eventually equaling responding of their counterparts that had not received second-phase reinforcers. Of their results, Leitenberg et al. remarked, " ... reinforcement of competing behavior suppress(es) the behavior to be extinguished and thus may prevent extinction from taking place. When ... reinforcement of competing behavior is stopped, the extinction procedure still needs to be carried out" (p. 303).

This interpretation received additional support from Rawson, Leitenberg, Mulick, and Lefebvre (1977). Following the training of rats to press a lever under a VI 30-s schedule for five sessions, Rawson et al. compared the effects of four manipulations in the second phase on subsequent resurgence. Different groups of rats were either (a) allowed to earn reinforcers by making an alternative response (pressing a second lever), (b) placed in the chamber during this phase but with no levers present, (c) retained in their home cages, or (d) exposed to both levers A and B, but without reinforcement for responding on either lever. The final group responded the least during the resurgence test, and there were no significant differences in response rates among the rats in the other three groups. Thus, only the rats with a history of pressing levers but not receiving reinforcers (and, therefore, the rats for which both responses had been extinguished) failed to exhibit resurgence.

Leitenberg et al.'s (1970) and Rawson et al.'s (1977) findings also suggest an inverse relation between degree of extinction of the first response and its subsequent resurgence. Later, Cleland, Foster, and Temple (2000) found greater resurgence by hens when the first response was extinguished for a single session before initiating the training of the second response than when the first response was extinguished for nine sessions.

The inverse relation between degree of extinction and resurgence has implications for application. Much like the outcomes of Leitenberg et al. (1970), problem behavior during DRA procedures may not contact programmed extinction (for example, if the appropriate response occurs exclusively following a training phase) or extinction may not be a programmed feature of the DRA. The latter possibility would occur whenever a DRA procedure is used for so-called automatically reinforced behavior, in which the caregiver often cannot withhold the reinforcer following the response. Thus, discontinuation of these types of differential reinforcement procedures may result in resurged problem behavior. The more extinguished the problem behavior before the commencement of reinforcement for an appropriate response, the less likely that problem behavior will resurge if reinforcement for appropriate behavior is discontinued (cf. Leitenberg et al., 1970).

Extinction of the response trained in the first phase during the alternative reinforcement phase may diminish its subsequent resurgence, but extinction does not necessarily eliminate such resurgence. Epstein (1983) first reinforced key pecking of pigeons before extinguishing it for a varied number of sessions. Then, an alternative response was reinforced for several sessions before it too was extinguished. All 6 pigeons exhibited resurgence, although the relation between number of extinction sessions and amount or degree of resurgence was not assessed (cf. to the results of Cle land et al. [2000]).

da Silva et al. (2008) also reported resurgence following extinction of the initial response. They first established key pecking of pigeons on a concurrent schedule designed to generate different rates of the two concurrent operants. Next, they eliminated responding on both operanda by implementing a DRO 20-s contingency such that reinforcers occurred following any 20-s period without a peck on either response key. Following the complete elimination of keypecking under this procedure, reinforcement was discontinued. Responding recurred on both response keys, with relatively more responding occurring on the key correlated with the higher response rates in the first phase. The DRO schedule was implemented in the second phase in a manner typical in laboratory experiments with animals: it was introduced initially at its final value of 20 s and responding was allowed to extinguish until a 20-s period of no responding occurred (this differs from some applied studies, which introduce a DRO schedule arranging frequent reinforcer delivery that is later faded; e.g., Niemeyer & Fox, 1990). As a result, key pecking was "extinguished" at least to the point that a 20-s period of no responding occurred. Thus, despite the fact that responding was at least partially extinguished in the second phase, differential resurgence still occurred. This result, along with the earlier report of Epstein (1983), questions whether "non-extinction" of the first-trained response is necessary for subsequent resurgence.

The above results suggest that, in application, a conservative approach would be to consider that problem behavior is less likely to resurge if it is extinguished before commencing with the reinforcement of an alternative response. If, however, it is useful to maintain the first response in the organism's repertoire, then perhaps the response may be better not extinguished before commencing with alternative reinforcement. This latter situation commonly occurs in educational settings when academic skills are taught sequentially such that once a child learns one skill set, the teacher decreases attention on the first set and her efforts become focused on reinforcing correct responses related to a second skill set. Previously learned responses may be important subsequently when the child encounters problems that cannot be solved by the currently reinforced skill set, but responding from the first skill set would be effective. These suggestions are speculative, however, because neither strategy has been investigated in applied research.

Variables related to the control of the second-trained response (hereafter, second response) can affect resurgence in two related ways. The reinforcement conditions of the second response can modulate the occurrence of the first response during the alternative reinforcement phase, which might affect the extent to which the first response has extinguished by the time the resurgence phase is reached. Related to this, Leitenberg, Rawson, and Mulick (1975, Exp. 4) found that the longer the alternative reinforcement phase was in effect (i.e., the longer the second response was reinforced and the first response was not), the weaker was the resurgence of the first response. This relation has implications for application because it suggests that undesirable resurgence might be reduced simply by extended exposure to treatment conditions. For example, resurgence may be less likely during extinction following a long history of appropriately implemented DRA than during extinction following only a brief exposure to the DRA treatment. This suggestion dovetails with research suggesting that initial appropriate implementation of DRA interventions results in more robust treatment effects during later degraded treatment conditions, including lean reinforcement schedules, which may function like extinction (e.g., St. Peter Pipkin, Vollmer, & Sloman, 2008). Furthermore, the duration of DRA exposure also would be mitigated by parameters of the first response (i.e., problem behavior, in many cases of application), including reinforcement history and topography, as described at the end of the preceding section.

The reinforcement conditions of the second response also will affect its resistance to extinction in the third phase, which is likely to affect at least the onset of resurgence. That is, resurgence is assumed to depend either directly on the extinction of the second response or indirectly on the absence of reinforcement of that second response serving as a discriminative stimulus to reinstate a previously learned response (i.e., the first response; cf. Franks & Lattal, 1976). Thus, resurgence may be less likely when the second response is resistant to extinction. For example, both Carey (1951) and Reed and Morgan (2006) found that resurgence is more likely as the extinction contingency in effect for the alternative response "takes hold" during the resurgence phase. In both experiments, the first-trained response patterns began to resurge only as the second-trained patterns diminished.

One way of affecting resurgence might be to fade gradually from the reinforcement schedule in the alternative reinforcement phase to extinction in the resurgence phase. For example, Hagopian, Toole, Long, Bowman, and Lieving (2004) compared a gradual with an abrupt change to a lean reinforcement schedule with FT and DRA treatment procedures. Problem behavior occurred initially at higher rates when the reinforcement rate was decreased abruptly than when it was decreased gradually. Evaluation of the effects of the gradual reduction procedure occurred immediately after a baseline condition, not a treatment condition, for all participants, so the degree to which these results are similar to the typical resurgence procedures has yet to be determined.

Features of the second response selected for reinforcement also may impact subsequent resurgence. If the two responses are of the same form (e.g., vocal responses), response generalization or induction between the two could result in different amounts of resurgence than if the two differed in form (e.g., motor response versus vocal response). This, however, has not yet been demonstrated. Resurgence has been found when the responses in the reinforcement and alternative reinforcement phases are either topographically identical (albeit in different locations: Epstein, 1983; Lieving & Lattal, 2003) or different from one another (Cleland et al., 2000; Doughty, da Silva, & Lattal, 2007; Leitenberg et al., 1970).

The results of experiments examining the effects of different reinforcement schedule s in the alternative reinforcement phase are mixed. Doughty et al. (2007) found relatively greater resurgence, and earlier onset of resurgence, when the alternative response was maintained by a DRO as opposed to a VI schedule in the alternative reinforcement phase. This finding was counter to earlier ones reported by Mulick, Leitenberg, and Rawson (1976) and Pacitti and Smith (1977), who found no systematic differences in resistance to extinction following different reinforcement schedules in the alternative reinforcement phase. Doughty et al. suggested that procedural differences among the experiments might contribute to the different outcomes. Given the importance of the reinforcement conditions as a variable affecting outcomes, however, the question of which schedules are likely to result in both the most rapid elimination of problem behavior and in the least resurgence when treatment is modified or discontinued, as with treatment integrity failures, becomes a key issue.

Resurgence Phase

The third, resurgence, phase is where the rubber meets the road. It is in this phase that the second response is extinguished and the first-reinforced response resurges, if it is going to do so. But, a first question is, what constitutes "resurgence" of that first-reinforced response? Almost all investigations of resurgence have reported the number of responses or rate of response as the primary index of resurgence. Several other measures of resurgence, however, might be used: the latency to the onset of resurgence (whether defined as the first resurged response or something else; an example of the latter to be found in Reed & Morgan, [2006]), its duration (e.g., number of sessions with at least one response), and its magnitude. The latter, among other things, could be the total number of responses across the resurgence condition or response rates within individual sessions.

Each of these measures seem most useful in relative terms, that is as an index of the relative effects of two (or more) different variables or parameters of variables in one of the phases on subsequent resurgence. This implies a comparison of conditions, which can be done, at least within individual subjects, either sequentially or concurrently. Sequentially, resurgence does occur over repeated exposures to the three-phase procedure (Cleland et al., 2000; Lieving & Lattal, 2003). A potential problem, however, is that with repeated exposure to the three-phase sequence, resurgence effects might not be constant. An alternative was investigated by da Silva et al. (2008), who used the concurrent resurgence procedure described in the preceding section. This procedure allowed assessment of relative resurgence of two different responses maintained under different contingencies in the reinforcement phase. Combining this procedure with the different measures of the resurged response described above seems a useful approach to better understanding relative resurgence as a function of different parameters of the situation.

However defined, resurgence has been most reliably found when the reinforcement conditions maintaining the second response are simply discontinued (extinguished) by removing all reinforcement of such responding. Resurgence, however, can occur, at least to a limited degree, if the reinforcement rate for the second response is reduced. This is tantamount to local extinction, for as the reinforcement rate is reduced, there are longer and longer periods of nonreinforcement within the schedule. Lieving and Lattal (2003, Exp. 4), for example, changed the reinforcement schedule on the second operandum from VI 60 s in the alternative reinforcement phase to VI 360 s in the resurgence phase and obtained some evidence of resurgence. Similar resurgence has been obtained in applied studies when reinforcement rates are systematically reduced ("leaned" or "thinned") too rapidly. Volkert (2007) conducted a procedure similar to that used by Lieving and Lattal, but with children with developmental disabilities who engaged in problem behavior. During the first phase, problem behavior was reinforced according to an FR 1 schedule. During the alternative reinforcement phase, an appropriate alternative response was reinforced on an FR1, using a functional communication training procedure. Resurgence was tested by decreasing the reinforcement schedule for appropriate behavior from FR 1 to FR 12. Rates of problem behavior increased for all three participants. Similarly, Hanley, Iwata, and Thompson (2001) observed increases in problem behavior during reinforcement reduction procedures, even though they did not explicitly target resurgence as an outcome. The increased responding may be interpreted as resurgence, with the resurgence phase consisting of reduced reinforcement rate instead of elimination of reinforcement.

Eliminating reinforcement altogether is but one form of extinction. Following Rescorla's (1967) contingency analysis of Pavlovian conditioning, Rescorla and Skucy (1969) suggested that removing the response-reinforcer dependency constitutes extinction. The removal of the response-reinforcer dependency is a key component of time-based treatment schedules (often, erroneously, called noncontingent reinforcement, cf. Lattal & Poling, 1981). This type of extinction, however, has not been shown to bring about resurgence (Lieving & Lattal, 2003, Exp. 3). Resurgence, however, is a reliable, consistent finding when conventional extinction, defined as removal of a previously established reinforcer, is used in the context of the three-phase resurgence paradigm detailed herein. This difference between the two types of extinction raises the intriguing possibility that problem behavior reduced or eliminated with response-independent events may be less likely to resurge than that subjected to conventional extinction.

At least two factors may contribute to the recurrence of responding during the resurgence phase. The first, described in the alternative reinforcement phase section above, is that the first response is incompletely extinguished during the second phase. Resurgence in the third phase may reflect the continuation of the extinction of the first response disrupted by reinforcement of the alternative response in the second phase (Leitenberg et al., 1975). Leitenberg et al. explained this as follows:
   ... decrements in extinction responding caused by reinforcement of
   alternative behavior may be temporary because such reinforcement,
   in effect, prevents the animal from making the original response
   now programmed for extinction. The response to be extinguished is
   so quickly and completely supplanted by the new reinforced
   alternative response that it never actually undergoes the
   extinction procedure. (p. 650)

The second is that the absence of reinforcement in the third phase also may serve as a discriminative stimulus for changing over to the other response, just as it was in the alternative reinforcement phase when the first response was extinguished as the second one was being reinforced.


The research discussed in this review invites several conclusions. First, resurgence is not an isolated "behavioral oddity," limited to situations where the resurgence procedure is explicitly arranged. Resurgence is integrally related to the discontinuation, or sometimes even diminishment, of reinforcement. It is part and parcel of every extinction and reinforcement-leaning operation. Even when discrete phases are arranged, the different phases of the resurgence procedure can meld together, as illustrated by the earliest study of resurgence (Carey, 1951), as well as the later work of Reed and Morgan (2006) and Bacha-Mendez et al. (2007). Each of these investigators found that earlier response patterns resurged only as later ones extinguished, with no well-defined demarcations between extinction of the later learned pattern and resurgence of the earlier learned ones.

Second, the resurged responding that develops when extinction is effected can be a wellspring of new responding, including complex repertoires such as problem solving and creativity (e.g., Epstein, 1985, 1991; see also Wilson & Hayes, 1996). It also, however, can be the bane of attempts to reduce problem behavior through extinction. That is, extinction may induce problem behavior (e.g., aggression, Azrin, Hutchinson, & Hake, 1966) that has been established in the ontogenic, or even phylogenic past, but that is not exhibited until extinction is effected.

Third, the nature of the resurged behavior, including transient increases in response variability often observed in extinction, can be traced to previously learned responses. That is, resurged responding is the reappearance of a historically reinforced response following the intervening reinforcement of some other response or class of responses. Its relation to the study of remote behavioral history is illustrated by the previously described concurrent resurgence experiment of da Silva et al. (2008; see also Barrett, 1986). In that experiment, the extent of resurgence depended on the characteristics of two different, first-reinforced, responses that were eliminated concurrently in the alternative reinforcement phase prior to the resurgence phase.

Finally, response reduction or elimination procedures other than extinction also might be sufficient to evoke resurgence of previously reinforced responding. For example, suppression of responding by response-dependent electric shock sometimes increases the rates of other, unpunished responding (e.g., Brethower & Reynolds, 1962; Crosbie, Williams, Lattal, Anderson, & Brown, 1997; Dunham, 1972). Might some of these so-called "side effects" of punishment (Newsome, Favell & Rincover, 1983) be related to resurgence?

The research reviewed herein suggests the significance of resurgence as a behavioral process operating in both laboratory and applied environments. Many of the claims with respect to human behavior, including those made here, are extrapolative from laboratory research with nonhuman animals. Of the three studies directly examining applied implications of resurgence (Bruzek, 2007; Lieving et al., 2004; Volkert, 2007), two are recent, yet unpublished, dissertations. This is promising, however, as it suggests that basic research related to resurgence is beginning to be infused into application. The implications of resurgence in terms of understanding the origins of unwanted behavior during treatment regimens, failed treatment effects, and facilitating problem solving (cf. Epstein, 1985, 1991) , bode well for its continued investigation in both research and application.


Azrin, N. H., Hutchinson, R. R. & Hake, D. F. (1966). Extinction-induced aggression. Journal of the Experimental Analysis of Behavior, 9, 191-204.

Bacha-Mendez, G., Reid, A. K., & Mendoza-Soylovna, A. (2007). Resurgence of integrated behavioral units. Journal of the Experimental Analysis of Behavior, 87, 5-24.

Barrett, J. E. (1986). Behavioral history: Residual influences on subsequent behavior and drug effects. In N. Krasnegor & D. B. Gray (Eds.), Advances in behavioral pharmacology, Vol. 5: Developmental behavioral pharmacology (pp. 99-114). Hillsdale, NJ: Erlbaum.

Brethower, D. M., & Reynolds, G. S. (1962). A facilitative effect of punishment on unpunished behavior. Journal of the Experimental Analysis of Behavior, 5, 191-199.

Bruzek, J. L. (2007). Resurgence of infant caregiving responses (Doctoral dissertation, University of Kansas, 2007). Dissertation Abstracts International, 68, 4805.

Carey, J. P. (1951). Reinstatement of previously learned responses under conditions of extinction: A study of "regression." American Psychologist, 6, 284.

Carey, J. P. (1953). Reinstatement of learned responses under conditions of extinction: A study of regression. Unpublished doctoral dissertation, Columbia University.

Cleland, B. S., Foster, M. E., & Temple, W. (2000). Resurgence: The role of extinction. Behavioural Processes, 52, 117-129.

Cleland, B. S., Guerin, B., Foster, T. M., & Temple, W. (2001). Resurgence. The Behavior Analyst, 24, 255-260.

Crosbie, J., Williams, A.M., Lattal, K.A., Anderson, M.M., & Brown, S. (1997). Schedule interactions involving punishment with pigeons and humans. Journal of the Experimental Analysis of Behavior, 68, 161-175.

da Silva, S. P., Maxwell, M. E., & Lattal, K. A. (2008). Concurrent resurgence and remote behavioral history. Journal of the Experimental Analysis of Behavior, 90, 313-331.

Doughty, A. H., da Silva, S. P., & Lattal, K. A. (2007). Differential resurgence and response elimination. Behavioural Processes, 75, 115-128.

Dunham, P. J. (1972). Some effects of punishment upon unpunished behavior. Journal of the Experimental Analysis of Behavior, 17, 443-450.

Epstein, R. (1983). Resurgence of previously reinforced behavior during extinction. Behaviour Analysis Letters, 3, 391-397.

Epstein, R. (1985). Extinction induced resurgence: Preliminary investigations and possible implications.

Psychological Record, 35, 143-153.

Epstein, R. (1991). Skinner, creativity, and the problem of spontaneous behavior. Psychological Science, 2, 362-370.

Epstein, R., & Skinner, B. F. (1980). Resurgence of responding during the cessation of response-independent reinforcement. Proceedings of the National Academy of Sciences, USA 77, 6251-6253.

Franks, G. J., & Lattal, K. A. (1976). Antecedent reinforcement schedule training and operant response reinstatement in rats. Animal Learning and Behavior, 4, 374-378.

Gilbert, T. F. (1958). Fundamental dimensional properties of the operant. Psychological Review, 65, 272-282.

Goh, H. L. & Iwata, B. A. (1994). Behavioral persistence and variability during extinction of self-injury maintained by escape. Journal of Applied Behavior Analysis, 27, 173-174.

Hagopian, L. P., Toole, L. M., Long, E. S., Bowman, L. G., & Lieving, G. A. (2004). A comparison of dense-to-lean and fixed-lean schedules of alternative reinforcement and extinction. Journal of Applied Behavior Analysis, 37, 323-338.

Hanley, G. P., Iwata, B. A., & McCord, B. E. (2003). Functional analysis of problem behavior: A review. Journal of Applied Behavior Analysis, 36, 147-185.

Hanley, G. P., Iwata, B. A., & Thompson, R. H. (2001). Reinforcement schedule thinning following treatment with functional communication training. Journal of Applied Behavior Analysis, 34, 17-38.

Harding, J. W., Wacker, D. P., Berg, W. K., Barretto, A., Winborn, L., & Gardner, A. (2001). Analysis of response class hierarchies with attention-maintained problem behaviors. Journal of Applied Behavior Analysis, 34, 61-64.

Lattal, K. A., & Poling, A. D. (l981). Describing response-event relations: Babel revisited. The Behavior Analyst, 4, 143-152.

Leitenberg, H., Rawson, R. A., & Bath, K. (1970) Reinforcement of competing behavior during extinction. Science, 169, 301-303.

Leitenberg, H., Rawson, R. A., & Mulick, J. A. (1975). Extinction and reinforcement of alternative behavior. Journal of Comparative and Physiological Psychology, 88, 640-652.

Lieving, G., & Lattal, K. A. (2003). An experimental analysis of resurgence. Journal of the Experimental Analysis of Behavior, 80, 217-233.

Lieving, G. A., Hagopian, L. P., Long, E. S., & O'Connor, J. (2004). Response-class hierarchies and resurgence of severe problem behavior. Psychological Record, 54, 621-634.

Mazur, J. E. (2006). Learning and behavior. New York: Prentice Hall. Mulick, J. A., Leitenberg, H., & Rawson, R. A. (1976). Alternative response training, differential reinforcement of other behavior, and extinction in squirrel monkeys (Saimiri sciureus). Journal of the Experimental Analysis of Behavior, 25, 311-320.

Newsome, C., Favell, J., & Rincover, A. (1983). The side-effects of punishment. In S. Axelrod & J. Apsche (Eds.), The effects of punishment on human behavior (pp. 285-316). New York: Academic.

Niemeyer, J. A., & Fox, J. (1990). Reducing aggressive behavior during car riding through parent-implemented DRO and fading procedures. Education and Treatment of Children, 13, 21-35.

Pacitti, W. A., & Smith N. F. (1977). A direct comparison of four methods for eliminating a response. Learning and Motivation, 8, 229-237.

Rawson, R. A., Leitenberg, H., Mulick, J. A., & Lefebvre, M. F. (1977). Recovery of extinction responding in rats following discontinuation of reinforcement of alternative behavior: A test of two explanations. Animal Learning and Behavior, 4, 415-420.

Reed, P., & Morgan, T. A. (2006). Resurgence of response sequences during extinction in rats shows a primacy effect. Journal of the Experimental Analysis of Behavior, 86, 307-315.

Rescorla, R. A. (1967). Pavlovian conditioning and its proper control procedures. Psychological Review, 74, 71-80.

Rescorla, R. A., & Skucy, J. C. (1969). Effect of response-independent reinforcers during extinction. Journal of Comparative and Physiological Psychology, 67, 381-389.

Richman, D. M., Wacker, D. P., Asmus, J. M., Casey, S. D., & Andelman, M. (1999). Further analysis of problem behavior in response class hierarchies. Journal of Applied Behavior Analysis, 32, 269-283.

St. Peter Pipkin, C., Vollmer, T. R., & Sloman, K. N. (2008). Effects of treatment integrity failures during differential reinforcement of alternative behavior: A translational model. Manuscript submitted for publication.

Volkert, V. M. (2007). An applied evaluation of resurgence: Functional communication training (FCT) and treatment relapse (Doctoral dissertation, Louisiana State University, 2007). Dissertation Abstracts International, 68, 4807.

Wilson, K. G., & Hayes, S. C. (1996). Resurgence of derived stimulus relations. Journal of the Experimental Analysis of Behavior, 66, 267-281.

Contact information:

Kennon A. Lattal

Department of Psychology

West Virginia University, Morgantown, WV 26506-6040, USA.

Claire St. Peter Pipkin

Department of Psychology

West Virginia University, Morgantown, WV 26506-6040, USA.
COPYRIGHT 2009 Behavior Analyst Online
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2009 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Lattal, Kennon A.; Pipkin, Claire St. Peter
Publication:The Behavior Analyst Today
Article Type:Report
Date:Mar 22, 2009
Previous Article:Behavioral momentum: translational research in intellectual and developmental disabilities.
Next Article:Laboratory evaluations of reinforcement contingencies.

Related Articles
Extinction-induced response variability in humans.
Effects of differing instructional histories on the resurgence of rule-following.
Response-class hierarchies and resurgence of severe problem behavior.
Extinction-induced response resurgence: a selective review.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters