Further evidence of renewal in automatically maintained behavior

Abstract

Renewal is a relapse phenomenon that refers to the recurrence of a previously reduced behavior following a change in stimulus conditions. Muething et al. (2022) examined the phenomenology of renewal among individuals with automatically maintained challenging behavior treated at an outpatient clinic. We replicated their findings by retrospectively examining renewal across various topographies of automatically maintained behavior treated at an inpatient hospital, and we extended their work by also examining differences across subtypes of automatically maintained self-injurious behavior. The prevalence of renewal was comparable to that observed by Muething et al., supporting the notion that automatically maintained challenging behavior is susceptible to relapse phenomena. Furthermore, renewal was twice as likely to occur for individuals with Subtype 2 versus Subtype 1 self-injurious behavior, providing additional evidence of behavioral differentiation between subtypes. Our findings suggest that even after apparent stability in treatment, practitioners should remain vigilant for the recurrence of automatically maintained behavior during generalization.

Individuals with intellectual and developmental disabilities (IDD) may engage in challenging behavior (e.g., aggression, self-injurious behavior [SIB]) to communicate their wants and needs (e.g., access to attention or preferred items, escape from instructional activities). In these cases, challenging behavior is said to be socially maintained. Sometimes, challenging behavior occurs in the absence of socially mediated consequences. In those cases, challenging behavior is said to be “automatically maintained,” as it is hypothesized that the behavior produces its own reinforcement that maintains its occurrence over time (cf. Hagopian & Frank-Crawford, 2018; Lloveras et al., 2022). One particularly dangerous and problematic topography of automatically maintained behavior is SIB. Recent research has identified subtypes of automatically maintained SIB that are hypothesized to reflect its distinct functional properties (Hagopian et al., 2015, 2017, 2023). These subtypes have been shown to be differentially affected by the presence of alternative reinforcement (Subtypes 1 and 2) or self-restraint (Subtype 3).

Behavioral interventions for challenging behavior—both socially and automatically maintained—rely heavily on reinforcement-based procedures (e.g., Hagopian et al., 2015). Although behavioral interventions for challenging behavior are very effective, they are also susceptible to a form of treatment relapse referred to as renewal (e.g., Saini & Mitteer, 2020). Renewal refers to the recurrence of a previously reinforced but currently eliminated behavior (e.g., challenging behavior) following a change in stimulus conditions (e.g., treatment setting or intervention agent) in the absence of a change in contingencies (i.e., procedural fidelity remains high; Wathen & Podlesnik, 2018). That is, a response is initially reinforced in Context A and it is subsequently reduced or eliminated in Context B. Renewal can occur following a return to a familiar (ABA renewal) or novel (ABC renewal) context (Kimball & Kranak, 2022).

Unfortunately, renewal is a relatively common clinical phenomenon (Kimball et al., 2023). Researchers have documented the occurrence and other characteristics of renewal across five consecutive controlled case series studies (CCCS; see Hagopian, 2020; e.g., Falligant, Chin, et al., 2022; Falligant et al., 2021; Haney et al., 2022; Mitteer et al., 2022; Muething et al., 2020). For example, Muething et al. (2020) found renewal in 67% (45 of 67) of cases from an outpatient clinic sample. Falligant et al. (2021) subsequently applied the analytic procedures of Muething et al. to an inpatient clinic sample and found renewal in 59% (20 of 34) of cases.

Overall, the five CCCS studies encompass 184 patients and their respective treatment applications. Of those 184 patients, only two (i.e., 1%) were found to have automatically maintained challenging behavior. Thus, 99% of the findings related to renewal in clinical practice has been gleaned from socially maintained challenging behavior. The generality of these findings to automatically maintained challenging behavior is unknown. This is problematic given both the prevalence of automatically maintained challenging behavior (Hagopian et al., 2013; Melanson & Fahmie, 2023) and the iatrogenic effects of renewal on treatment outcomes (Mitteer et al., 2018). Therefore, examining the nature of renewal among individuals with automatically maintained challenging behavior broadly and SIB specifically is essential.

Recently, Muething et al. (2022) conducted a retrospective CCCS of a sample of 30 patients from an outpatient clinic who presented with a variety of topographies of automatically maintained challenging behavior to identify prevalence estimates of renewal among this population during behavioral treatment. Muething et al. found renewal in 30% (30 of 100) of context changes, suggesting that contextual factors during treatment exert some control over automatically maintained behavior. Interestingly, patterns of renewal for automatically maintained challenging behavior tended to somewhat differ from patterns commonly observed with socially maintained challenging behavior. For example, increases in behavior following a context change were sometimes delayed and variable across sessions—this finding has implications for designing renewal-mitigation strategies for different functional classes of challenging behavior. As noted, Muething et al. evaluated renewal across a variety of topographies of automatically maintained behavior in their analysis, including SIB. Although Muething et al. reported on automatically maintained SIB, they did not evaluate the prevalence of renewal across subtypes of SIB; thus, the degree to which renewal varies across subtypes is unknown. Those potential differences could, and likely would, set the stage for additional investigations on subtype-specific treatments and recommendations for mitigating renewal of SIB. It has been hypothesized that a distinguishing feature of Subtype 1 and Subtype 2 SIB is the differential sensitivity of SIB to disruption by alternative reinforcement in the environment. This response dynamic is inferred based on the different response patterns observed during structured behavioral observations as well as the excellent response to treatment with reinforcement-based behavioral interventions for Subtype 1 but not Subtype 2 SIB (Hagopian et al., 2017). One possibility is that SIB is controlled by motivational variables that are related to the putative reinforcing biological and sensory consequences of the behavior for both Subtype 1 and Subtype 2. That is, SIB may produce reinforcing consequences that are extremely powerful for Subtype 2 SIB such that alternative sources of reinforcement cannot compete with the products of SIB. If so, one would expect to see greater renewal in Subtype 2 SIB relative to Subtype 1 SIB given basic research demonstrating that reinforcement magnitude moderates the degree of renewal following context changes (e.g., Berry et al., 2014; Kelley et al., 2015).

Replication and extension of the lone CCCS focusing on the renewal of automatically maintained challenging behavior with a different clinical sample and setting is a logical and necessary next step in this line of research. Doing so would help to identify and establish the boundaries of generality with respect to the prefatory findings of Muething et al. (2022) as well as set the stage for additional investigations on factors that may influence renewal and ways to mitigate renewal of automatically maintained challenging behavior. Accordingly, the purpose of the current investigation was to replicate and extend the analysis described by Muething et al. We conducted a retrospective CCCS study (Hagopian, 2020) and examined the prevalence and magnitude of renewal across various topographies of automatically maintained challenging behavior exhibited by individuals with IDD receiving treatment in an inpatient setting. For the subset of participants who engaged in SIB, we also conducted secondary analyses in which we evaluated the prevalence and magnitude of renewal across subtypes of automatically maintained SIB. These findings could also have implications for the generality of renewal-mitigation strategies (i.e., guide clinical practice), stimulate a continuum of reverse-translational and basic studies on automatically maintained behavior, and help us to understand better the phenomenology of automatically maintained SIB.

METHOD

Participants, setting, and general course of clinical care

Participants and their relevant demographic and clinical information (e.g., diagnoses) from this study were drawn from archived medical records of all patients admitted to an inpatient unit for the assessment and treatment of challenging behavior between 2010 and 2020. Consistent with Muething et al. (2022), to be included in the study the participants had to have (a) at least one topography of challenging behavior that was solely maintained by automatic reinforcement, (b) a completed treatment application that targeted the automatically maintained topography of behavior, and (c) experienced at least one context change at the terminal treatment phase of that evaluation. A total of 360 individuals were admitted to and discharged from the unit during this period. Of these, 72 met all the inclusion criteria (see methods below for more information on how many participants were removed at each stage of the analysis). These participants engaged in a variety of topographies of automatically maintained behavior that were included in our analysis, such as elopement, pica, disruptive behavior, and SIB. We were particularly interested in renewal among individuals with SIB. Thus, we also determined how many participants displayed at least one topography of automatically maintained SIB. In total, 40 of the 72 participants met the criteria for inclusion in the secondary analysis of SIB subtypes.

As part of the general course of care during the admission, the patient’s behavioral team, led by a doctoral-level Board Certified Behavior Analyst, conducted individualized functional analyses and subsequent treatment applications, which refer to discrete treatment evaluations conducted to decrease targeted challenging behavior. Most sessions were conducted in padded session rooms (3 × 3 m), patient bedrooms (6.1 × 6.1 m), or shared living spaces on the inpatient unit (6.7 × 6.7 m).

At least one functional analysis (Iwata et al., 1982/1994) was conducted with each patient to identify the function(s) of their challenging behavior. Some functional analyses targeted single topographies of behavior (e.g., SIB only), whereas others targeted multiple topographies simultaneously (e.g., SIB and aggression). In some cases, functional analyses were modified to further investigate the function of specific target behaviors (see Hagopian et al., 2013). Throughout admission, the behavioral team used the results from their assessments to inform the treatment evaluations to target each patient’s challenging behavior. Although the treatment protocols were highly individualized for each participant, common components included noncontingent access to competing items, response blocking, differential reinforcement of other or alternative behavior, protective equipment, and extinction.

Once reductions in challenging behavior were consistent and stable, context changes were conducted to generalize the behavior changes resulting from the interventions. These context changes included (a) person-based changes, which occurred when an individual who had not already implemented a treatment session conducted treatment for the first time (i.e., novel therapists such as new staff members, parents, school staff); (b) location-based changes, which occurred when treatment sessions were conducted in a new area where they had not been conducted previously (e.g., novel locations, such as classrooms, common rooms, novel areas in the hospital grounds or community); or (c) combined changes, which occurred when both a person change and location change were implemented simultaneously. Generalization sessions are conducted to facilitate the transfer of the final treatment package from the inpatient unit to the community. Thus, context changes were only included if they occurred during the terminal treatment phase for each treatment application (i.e., the final treatment package), which was defined as the phase of treatment immediately prior to generalization following schedule thinning.

Application of inclusion criteria

Functional analysis: Visual inspection and structured criteria

We began by visually interpreting all the functional analyses conducted for each participant to determine if any topographies of challenging behavior were automatically reinforced. This involved first using visual analysis to determine (a) whether the values frequency or duration of challenging behavior in the alone or no-interaction conditions of the functional analyses differed from those in the toy-play condition or (b) whether the challenging behavior occurred at high levels across all conditions. Both patterns of responding are generally indicative of automatically maintained behavior (Hagopian et al., 2023). Participants were excluded from the study if visual analysis did not suggest that at least one topography of behavior was likely automatically maintained. Individuals were also excluded if the visual analysis suggested that their behaviors were multiply maintained.

Next, the third and fourth authors applied structured criteria to each functional analysis to confirm that the behavior that was previously identified by visual analysis was automatically maintained (Hagopian et al., 2023). Structured criteria were further reviewed by the first or sixth author if (a) either coder disagreed with the results generated by our structured-criteria-analysis template (i.e., spreadsheet that applied the structured criteria), (b) behavior appeared to be multiply maintained, (c) there was an overall trend in the functional analysis data, or (d) the mean ranks of rate or duration of the response suggested an automatic function. Therefore, a combination of structured criteria and visual analysis was used to identify the cases that we would ultimately analyze for renewal. Of the 360 cases, we identified 157 individuals who engaged in challenging behavior maintained solely by automatic reinforcement.

Context changes

For the 157 patients with a topography of automatically maintained challenging behavior, we analyzed all treatment applications conducted to identify whether context changes occurred when the terminal treatment was in place. The third and fourth authors reviewed each case record and extracted data for context changes that occurred with the final treatment in place using procedures similar to Muething et al. (2022). However, we included combined context changes in addition to person and location changes (as defined above). Some individuals experienced more than one treatment application with a context change during admission (see Table 1).

TABLE 1.

Participant demographics and automatically maintained problem behavior(s)

Case	App(s)	Age	Sex	Level of ID	Autism	Automatically maintained behavior(s)	Subtype
1	3	16	M	Moderate	Yes	Body SIB	1
2	1	16	M	Severe	Yes	Head SIB, body SIB	2
3	1	9	M	Unspecified	Yes	SIB	1
4	1	18	M	Moderate	Yes	Skin picking	1
5	2	12	M	Severe	Yes	Head SIB, body thrashing	2
6	1	19	F	Moderate	Yes	Skin twisting	1
7	1	18	M	Severe	Yes	Body SIB	1
8	1	13	M	None	Yes	Skin SIB	1
9	2	16	M	Severe	Yes	Elopement	N/A
10	1	15	M	None	Yes	RDFS, ISB, elopement	N/A
11	2	15	M	Severe	Yes	Neck SIB	2
12	3	7	M	Moderate	Yes	LI self-biting	1
13	2	16	M	Severe	Yes	Pica	N/A
14	1	12	M	Severe	Yes	AGG	N/A
15	2	19	M	Moderate	Yes	SIB, mouth stereotypy	N/A
16	1	6	M	Unspecified	Yes	Pica	N/A
17	3	9	M	Profound	Yes	Dropping, elopement	N/A
18	2	5	M	Severe	Yes	Stereotypy	N/A
19	1	16	F	Severe	Yes	DIS	N/A
20	2	19	M	Moderate	Yes	Perseverations	N/A
21	2	9	M	Severe	Yes	SIB	1
22	4	9	M	None	Yes	Pica, DIS, disrobing	N/A
23	1	16	M	Unspecified	Yes	Disrobing	N/A
24	2	25	F	Unspecified	Yes	Scripting	N/A
25	1	10	M	Profound	Yes	HB	1
26	3	16	M	Severe	Yes	Skin picking, pica	1
27	1	13	M	Severe	Yes	Body SIB	3
28	1	7	M	Severe	Yes	Disrobing, fecal play	N/A
29	3	12	M	Severe	Yes	SIB, DIS	1
30	1	16	M	Severe	Yes	DIS	N/A
31	4	9	M	Unspecified	Yes	Head SIB	2
32	1	11	F	Unspecified	Yes	Body SIB, self-restraint	3
33	1	11	M	Severe	Yes	DIS	N/A
34	1	13	M	Moderate	Yes	Skin SIB, spitting	2
35	5	19	F	Severe	Yes	LI DIS, HI DIS, spitting	N/A
36	1	19	M	Moderate	Yes	ISB, ITO	N/A
37	1	11	M	Moderate	Yes	Pica	N/A
38	1	13	M	Severe	Yes	Spitting	N/A
39	2	15	M	Severe	Yes	Knee to head SIB	3
40	2	16	M	Severe	Yes	Rectal digging, ISB	N/A
41	1	23	M	Severe	Yes	SIB	2
42	2	14	M	Severe	Yes	SIB, HB	2
43	1	10	F	Moderate/Severe	Yes	Pica	N/A
44	1	15	M	Moderate/Severe	Yes	DIS	N/A
45	1	5	M	Severe	No	SIB, HB	1
46	1	18	M	Unspecified	Yes	Gagging, spitting, emesis	N/A
47	3	10	M	Unspecified	Yes	SIB, shirt mouthing	2
48	2	16	M	Unspecified	Yes	DIS, chin pressing	N/A
49	1	8	M	Severe	Yes	HB	1
50	3	19	M	Severe	Yes	SIB	3
51	2	11	M	Moderate/Severe	Yes	Climbing, PD	N/A
52	1	16	F	Severe	Yes	Pica	N/A
53	2	14	M	None	Yes	SIB, spit play	2, 1
54	1	10	M	Severe	Yes	Pica	N/A
55	1	19	F	None	Yes	Food DIS	N/A
56	1	10	F	Severe	Yes	SIB	3
57	2	10	M	Unspecified	Yes	Body SIB	1
58	1	8	M	Moderate	Yes	Pica, chin SIB, HB, head SIB, body SIB	2, 1, 1, 2, 1
59	2	15	M	Unspecified	Yes	SIB	1
60	1	10	M	Profound	Yes	SIB	2
61	1	12	M	Moderate	Yes	Knee to head SIB	2
62	1	15	M	None	Yes	SIB	1
63	1	13	F	Moderate	No	DIS	N/A
64	1	14	F	Moderate	Yes	DIS	N/A
65	2	16	M	Moderate	No	SIB	1
66	1	12	M	None	Yes	SIB, head pressing	1
67	2	8	M	Moderate	Yes	Head SIB, body SIB, DIS, hand mouthing	2
68	2	8	M	Moderate	Yes	Rumination, DIS	N/A
69	2	12	M	Severe	Yes	Head SIB, body SIB	2
70	1	19	M	Mild	Yes	SIB	2
71	1	19	F	Profound	Yes	Object mouthing	N/A
72	1	19	F	Moderate	Yes	Body SIB	3

Note. App = applications; M = male; F = female; ID = intellectual disability; SIB = self-injurious behavior; DIS = disruption; RDFS = rectal digging and fecal smearing; ISB = inappropriate sexual behavior; LI = low intensity; HI = high intensity; AGG = aggression; HB = head banging; ITO = inappropriate touching of others; PD = property destruction; N/A = not applicable. Subtype is listed for SIB only.

After removing patients for whom a context change did not occur when the terminal treatment was in place, the final data set for our study included 72 individuals and a total of 117 treatment applications where at least one context change occurred. In total, we identified 851 person changes, 120 location changes, and 14 combined changes (N = 985 context changes). The participant sample included 59 males (82%) and 13 females (18%) with a median age of 14 years old (range: 5–25 years). Thirty-two individuals (44%) engaged in topographies of challenging behavior other than SIB (e.g., disruptive behavior), 31 individuals (43%) engaged in SIB exclusively, and nine individuals (13%) engaged in non-SIB automatically maintained behavior in addition to SIB. Thus, 40 individuals engaged in automatically maintained SIB. Table 1 includes a summary of all participant behaviors as well as additional participant characteristics.

Subtyping of SIB

For individuals who engaged in automatically maintained SIB, behavior was characterized as Subtype 1, 2, or 3 based on the level of differentiation (LOD) between the no-interaction condition and the toy-play condition (Hagopian et al., 2023). The LOD was calculated by dividing the mean rate (or duration) of SIB in the play condition by the mean rate (or duration) in the no-interaction condition, subtracting this quotient from 1, and then multiplying that difference by 100 to convert it into a percentage. The empirically derived LOD cutoff for differentiating Subtype 1 from non-Subtype 1 is 62.5%; values at or above 62.5% indicate Subtype 1, and those below indicate non-Subtype 1 (i.e., Hagopian et al., 2023). Additionally, SIB for individuals who engaged in self-restraint was characterized as Subtype 3 if self-restraint occurred in at least 25% of 10-s intervals across 30% of the no-interaction sessions regardless of the LOD. If self-restraint met the criteria for Subtype 3, SIB was classified as such; if self-restraint did not meet the criteria for Subtype 3, SIB was classified as Subtype 1 or 2 based on the LOD as described above.

Of the 40 individuals who engaged in automatically maintained SIB, the behavior was categorized as Subtype 1 for 21 individuals (52.5%), Subtype 2 for 15 individuals (37.5%), and Subtype 3 for six individuals (15%). Two individuals engaged in multiple topographies of automatically maintained SIB, and some behaviors were identified as different subtypes. Thus, the total number of individuals who engaged in automatically maintained SIB (n = 40) is slightly lower than the sum of numbers reported for each subtype (n = 42). These two cases were excluded from our subtype analysis.

Data analysis and interrater agreement

We quantified the prevalence and magnitude of renewal following context changes. To determine the prevalence, we defined renewal as the occurrence of challenging behavior that was higher in one of the first three postchange sessions when compared with the highest occurrence of challenging behavior in any of the five prechange sessions (e.g., Briggs et al., 2018; Falligant et al., 2021; Lerman & Iwata, 1995; Muething et al., 2020, 2022). Consistent with Muething et al. (2022), when renewal was observed during any of those three postchange sessions, we also evaluated subsequent sessions to determine the magnitude of renewal and whether the effects were relatively transient. Muething et al. (2022) included up to 10 postchange sessions for these evaluations. We report on five postchange sessions rather than 10 because many of the participants’ data were limited to five postchange sessions, and this was also consistent with prior research on renewal (e.g., Falligant, Chin, et al., 2022).

There are several ways to quantify the magnitude of renewal as a function of the type of context change (i.e., person-based vs. location-based change) or the number of sessions conducted following a given context change.¹ A common approach is to express the magnitude of renewal as the quotient of the behavior rate for a given session (or as the mean rate for multiple sessions) during the postchange period by the behavior rate of the mean of the last five prechange sessions (Muething et al., 2020). However, one downside of this approach for quantifying renewal is the loss of data that occurs when the rate of behavior during the prechange sessions is zero (given that division by zero is impossible). Naturally, practitioners may be inclined to initiate generalization when challenging behavior has necessarily reached low (often zero-rate) levels. Another option is to calculate the simple difference in the level of challenging behavior during each postchange session relative to the final prechange session (e.g., Falligant, Hagopian, et al., 2022), but here there is an interpretive difficulty with across-subject heterogeneity in the range of response rates (which could be attributable to differences in the topographies of behavior or recording methods).

To minimize data attrition and normalize our response measure across a range of heterogeneous topographies of behaviors and recording methods (i.e., frequency, duration, percentage), we examined the magnitude of renewal using the log proportion rate of response suggested by Friedel et al. (2019):

$$Y=lo{g}_2\left(\frac{B_n+c}{B_{n-1}+c}\right),$$ (1)

where Y is the log proportion response rate, B is the response rate during session n (e.g., second postchange session), and c is a correction factor (here, the value of this factor is set to 1; see Friedel et al., 2019) to serve as a constant value to account for sessions in which no challenging behavior was emitted. This measure normalizes response rate on a session-by session basis and facilitates the interpretation of proportional changes in response rate across individuals. A log with a base of two means that a log proportion rate of 1 would indicate a doubling in the response rate.

A second independent rater, who was also a doctoral-level Board Certified Behavior Analyst, coded the occurrence of renewal during context changes for 32% (37 of 117) of treatment applications. Context changes coded identically were considered agreements; context changes scored nonidentically were considered disagreements. All assessed context changes were coded identically (i.e., agreement was 100%).

RESULTS AND DISCUSSION

For the primary analyses, we focused on renewal across all topographies of automatically maintained challenging behavior. Patients generally experienced multiple context changes during treatment for automatically maintained challenging behavior (M = 13.6, SD = 13.0). Individual treatment applications often included eight or more distinct programmed context changes (M = 8.4, SD = 8.8). Renewal effects (evaluated from the first three postchange sessions) were observed for approximately two thirds of participants (68.1%; 49 of 72). This means that in at least one treatment application for two thirds of participants, there was at least one instance in which automatically maintained challenging behavior increased following some type of context change. Additionally, in more than half of all treatment applications (58.1%; 68 of 117), there was at least one occasion in which automatically maintained challenging behavior increased following a context change. For many participants and applications (in which at least one instance of renewal occurred), renewal occurred in one out of every three to four context changes.

The top panel of Figure 1 depicts renewal at the level of individual context changes (i.e., person vs. location). We included data on renewal in the fourth and fifth postchange sessions as well. Across all contexts, renewal was observed in 18.8% of changes (185 of 985). Renewal was more common following location-based context changes (27.5%; 33 of 120) than following person-based context changes (17.7%; 151 of 851). This difference was small but statistically significant, χ²(1) = 6.5, p = .01. Of the few combined changes (i.e., person and location), renewal occurred infrequently (7.1%; 1 of 14).

FIGURE 1.

Renewal by context change type and number. The black portion of the bar corresponds to context changes with renewal, and the white portion of the bars corresponds to context changes without renewal (* = p < .05).

The bottom panel of Figure 1 depicts the number and proportion of context changes with renewal across applications based on the temporal order of each unique context change. That is, we quantified renewal for all context changes that represented the very first renewal challenge (e.g., novel staff member), the second renewal challenge (e.g., novel location on the hospital grounds), the third renewal challenge, and so forth, programmed across applications. We did this to examine differences in the likelihood of renewal occurring as a function of the number of successive renewal challenges across all applications. There were no pronounced changes in the prevalence of renewal up through the 20th distinct context change; renewal generally occurred in approximately 20%–30% of these changes. In subsequent context changes, the probability of renewal occurring was somewhat more variable. However, there were so few applications with this many unique changes that these results should be interpreted cautiously. Across context changes, the probability of renewal occurring did not systematically decrease as a function of successive exposure to unique renewal challenges (r = −.15, p = .31). What is apparent from these findings is that renewal effects may be persistent across successive generalization challenges (cf. Falligant et al., 2021). This is reminiscent of other findings from the relapse literature that have shown that the prevalence of resurgence (a distinct form of relapse) also does not change consistently as a function of repeated exposure to the conditions that occasion it (Falligant, Hagopian, et al., 2022; Lieving & Lattal, 2003; Redner et al., 2022).

Several studies have found that the magnitude of renewal appears to decrease across successive sessions within a given context change (e.g., Muething et al., 2020), although others have reported mixed findings (e.g., Falligant, Chin, et al., 2022; Muething et al., 2022). Figure 2 depicts the mean rate of challenging behavior across postchange sessions. A one-way analysis of variance (ANOVA) of the log proportion rates across the final prechange session and each postchange session indicated a significant difference in the magnitude of challenging behavior among context changes with renewal, F(4, 275) = 11.54, p < .001. The bottom panel of Figure 2 depicts these differences in the beeswarm plot, displaying the distribution of data points corresponding to renewal across postchange sessions. Follow-up Tukey comparisons revealed a significant difference between the first and second (p < .001), first and third (p = .001), and first and fourth (p = .003) postchange sessions. It is also worth noting the apparent increase in behavior in the fifth context change is attributable to a handful of outliers and does not represent a significant change in behavior relative to the preceding postchange sessions. Figure 3 depicts the magnitude of renewal changes across successive sessions given a specific type of context change (i.e., person-based vs. location-based). Because there were differences in the number of postchange sessions conducted across applications (e.g., some treatment applications only had a single postchange session for a given context change, whereas others had up to five postchange sessions), we used a mixed-effects restricted maximum-likelihood model instead of a two-way repeated-measures ANOVA to examine changes in renewal across successive postchange sessions and types of context changes (see Bagiella et al., 2000). Consistent with the above results, there was a significant difference in renewal across postchange sessions F(1.4, 96.7) = 7.1, p = .004, but there was not such an effect for the type of context change F(1, 269) = .30, p = .58.

FIGURE 2.

Renewal magnitude across postchange sessions. Top panel: mean rates of problem behavior by session; error bars correspond to the standard error of measurement. Bottom panel: beeswarm plot of renewal magnitude by postchange session number; each context change is represented by a data point. Different ordinates for y-axes are used in the top and bottom panels to facilitate visual inspection of the postchange data.

FIGURE 3.

Renewal magnitude across sessions by type of context change. Mean rates of problem behavior by session. Error bars correspond to the standard error of measurement.

Thus, challenging behavior was elevated across applications on the first postchange session and generally decreased across successive sessions. From a practical standpoint, this suggests that elevated levels of challenging behavior will generally decrease in one to three sessions within a given context change. However, decreases in challenging behavior across sessions within a given context change were sometimes transient or variable. This is consistent with findings from Muething et al. (2022), suggesting that increases in challenging behavior following a context change may be delayed or otherwise inconsistent across successive sessions. Interestingly, such consistencies were not observed with respect to the renewal of socially maintained challenging behavior for which contrasting results have been reported, particularly regarding the magnitude (cf. Falligant et al., 2021, and Muething et al., 2020). Therefore, practitioners should be vigilant for the recurrence of challenging behavior during generalization even after apparent treatment stability has been captured. This finding is congruent with evidence suggesting that changes in behavior during relapse tests are sometimes delayed or have a bitonic function (Craig et al., 2020; Podlesnik & Kelly, 2015).

There was also no difference in the magnitude of renewal by the number of context changes, F(47, 937) = 1.02, p = .44. Decreases in renewal may occur across successive postchange sessions following a context change. However, across distinct context changes the magnitude of renewal (at least in the initial postchange session) does not appear to change with the introduction of subsequent novel context changes. Again, this replicates findings from the relapse literature demonstrating that the magnitude of resurgence does not consistently change with repeated exposure to the conditions that occasion it (Falligant, Hagopian, et al., 2022; Gratz et al., 2019; Volkert et al., 2009; cf., Lieving & Lattal, 2003; Marsteller & St. Peter, 2014; Podlesnik et al., 2020; Redner, 2022).

Although there is clearly complexity in these effects, these results replicate the findings from Muething et al. (2022). They further establish the generality of renewal phenomena to automatically maintained challenging behavior. Figure 4 depicts the outcomes of renewal across five retrospective CCCS studies. It appears that relapse is not specific to socially maintained challenging behavior, highlighting the importance of programming for generalization and incorporating recommendations for mitigating renewal (see Kimball et al., 2023; Kimball & Kranak, 2022; Podlesnik et al., 2017) within treatments for automatically maintained behavior.

FIGURE 4.

Prevalence of renewal across large-scale published investigations. Asterisks correspond to investigations of automatically maintained problem behavior. Prevalence estimates of renewal at the case level were not reported by Muething et al. (2022). * = Study exclusively focused on automatically maintained challenging behavior.

Renewal was more prevalent across treatment applications in our findings than in those of Muething et al. (2022), although our renewal estimates at the context level were generally similar. Like Muething et al., we found inconsistent effects with respect to changes in the magnitude of renewal across successive sessions following a given context change. Although the magnitude of renewal generally decreased in one to three sessions following the initial postchange session, it did not always decline in an orderly fashion.

Renewal was more likely to occur following location-based changes than following person-based context changes, but the magnitude of renewal was equivalent across either type of context change. It is worth noting that person-based changes were much more common than location-based changes, which makes sense given that these treatment applications were conducted in a hospital-based inpatient unit. There are more opportunities to conduct generalization sessions with novel staff members and other individuals (e.g., caregivers) than in the relatively finite number of novel locations within the inpatient unit, hospital grounds, and nearby community-based settings. It is also possible that location-based changes represent a more robust renewal challenge. An additional explanation could be that because the patients in our sample were much more likely to experience a person-based change, they came to associate treatment contingencies with other people rather than other locations. Said another way, renewal may have been more likely to follow location-based changes because individuals did not have as many opportunities to pair treatment contingencies with multiple locations as they did to pair treatment contingencies with multiple people. This further highlights the necessity of replicating these procedures across other settings as well as the potential implications for programming certain renewal-mitigation tactics. For example, when programming multiple-context training, one might need to consider not only the overall context changes but also the specific type and number of context changes: person- or location-based changes (Kimball & Kranak, 2022).

In a novel extension of Muething et al. (2022), we also examined the prevalence and magnitude of renewal across subtypes of automatically maintained SIB (Figure 5). Renewal occurred more often in context changes for Subtype 2 (31.0%; 48 of 155) than in those for Subtype 1 (17.9%; 52 of 290) and Subtype 3 (25.3%; 20 of 79), respectively. This difference in relative prevalence across subtypes was significant, χ² (2) = 10.0, p = .007. That is, renewal was significantly more likely to occur following a context change for Subtype 2 than for Subtype 1, χ² (1) = 9.9, p = .001, OR = 2.1. This means that renewal was twice as likely to occur following a context change for Subtype 2 than following a context change for Subtype 1. Differences between Subtype 2 and Subtype 3 (p = .49) and Subtype 1 and Subtype 3 (p = .10) were minimal. For context changes in which renewal was observed, the results of a mixed-effects restricted maximum-likelihood model revealed a significant difference in renewal magnitude across postchange sessions, F(1.28, 45.48) = 8.22, p = .004, but a minimal difference across subtypes F(1, 142) = .09, p = .76. When examining the first postchange session, which plausibly represents a robust renewal test (cf. Craig et al., 2020), there was a significant difference in renewal magnitude across subtypes, F(2, 117) = 4.03, p = .02. That is, the magnitude of renewal was generally greater for Subtype 2 than for Subtype 1 and Subtype 3. Subsequent Tukey tests indicated that this difference in renewal magnitude was statistically significant for Subtype 2 relative to Subtype 3 (p = .04) as well as nearing significance relative to Subtype 1 (p = .06).

FIGURE 5.

Renewal prevalence and magnitude by automatically maintained SIB subtype. Mean rates of problem behavior by session. Error bars correspond to the standard error of measurement. Inset panel: data points correspond to the first session of each context change across applications with renewal (* = p < .05).

The increased prevalence of renewal among Subtype 2 is intriguing given predictions of quantitative models of behavioral persistence and findings from the laboratory and demonstrates increased relapse associated with reinforcement rate and magnitude of the target response (e.g., Berry et al., 2014; Kelley et al., 2015). In the context of automatically maintained SIB, this response dynamic is indicative of behavioral differentiation between subtypes and may support the hypothesis that SIB produces more powerful sensory consequences in Subtype 2 than in Subtype 1 SIB (Hagopian & Frank-Crawford, 2018). This finding provides another example of behavioral differentiation between subtypes and further highlights a treatment-resistant aspect of Subtype 2 SIB.

The current study included Subtypes 2 and 3 automatically maintained SIB, which have been shown to be particularly resistant to treatment with reinforcement alone (see Hagopian et al., 2015, 2018). Thus, it may be tempting to interpret the outcomes reported in this study as simply the persistence of a treatment-resistant response and not an example of renewal. However, it is important to recall that, by definition, operant renewal refers to the characteristic recurrence of behavior following a change in stimulus conditions. For renewal to be said to have occurred, SIB necessarily would have had to exceed some prechange level. If SIB occurred at high rates in treatment and continued to occur at similar rates following a context change, this would not have been classified as renewal. Also, practically speaking, it is likely that practitioners would not be inclined to transfer treatment to novel settings and people until consistently low levels of SIB have been achieved. Nonetheless, it would be interesting to explore this in future research leveraging more sophisticated and appropriate analytic techniques, such as Monte Carlo bootstrapping approaches, to assess for quantitative signatures of behavioral cyclicity and disequilibrium and further quantify the prevalence and magnitude of renewal in automatically maintained behavior (Friedel et al., 2019).

As noted, some of the analyses differed from those of prior research on renewal. For example, Muething et al. (2022) evaluated renewal for up to 10 postchange sessions, whereas we only evaluated this in up to five postchange sessions. Because these are retrospectively obtained data, we could not control how many postchange sessions were conducted or what context changes the participants experienced. This will naturally lead to variations across studies. Replication of the lone study specifically documenting the phenomenology of renewal of automatically maintained challenging behavior is a necessary contribution to the areas of renewal research specifically and relapse broadly. We replicated the findings of Muething et al. (2022), thereby further establishing the scope and characteristics, generality, and believability of renewal as they relate to automatically maintained behavior (see Hagopian, 2020). Indeed, our results, along with those from Muething et al., indicate that renewal is a phenomenon likely to be encountered in outpatient and inpatient treatment contexts across an array of functional classes of challenging behavior—including those maintained by sources of automatic reinforcement. Nevertheless, replication of these findings across additional research groups is needed. Additionally, future work should continue to examine how differences in the way we define and quantify relapse in terms of the number of prechange and postchange sessions and analytic methods (e.g., log proportion rate of response) might represent a source of heterogeneity within this literature. Future researchers may also consider analyzing renewal based on the topography of behavior (though this could prove challenging because operational definitions likely vary greatly; Kubina et al., 2022). Additionally, research exploring the conditional effectiveness of treatment approaches that inoculate against renewal (and other forms of relapse) should examine automatically maintained challenging behavior with the knowledge that there is complexity in these outcomes related to the heterogeneous nature of this functional class of behavior and apparent subtype-specific differences with respect to response to treatment and renewal characteristics.

DATA AVAILABILITY STATEMENT

Additional data beyond those presented herein are available from the corresponding author upon reasonable request.