JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2015, 48, 765–780 NUMBER 4 (WINTER)

SERIAL ALTERNATIVE RESPONSE TRAINING AS INTERVENTION

FOR TARGET RESPONSE RESURGENCE
JOSEPH M. LAMBERT
VANDERBILT UNIVERSITY

SARAH E. BLOOM AND ANDREW L. SAMAHA

UNIVERSITY OF SOUTH FLORIDA

AND

ELIZABETH DAYTON AND ANDREW M. RODEWALD

UTAH STATE UNIVERSITY

Failure to reinforce appropriate behavior could result in resurgence of previously extinguished

problem behavior and degradation of previously effective treatments such as differential
reinforcement of alternative behavior (DRA). We analyzed arbitrary responses (i.e., switch
flipping) exhibited by 3 adults with developmental disabilities to compare the effect of a traditional
DRA intervention against the effect of a serial DRA intervention on the magnitude of target
response resurgence using a 2-component multiple schedule. The target response served as an
analogue to problem behavior, and alternative responses served as analogues to socially appropriate
alternative responses. In all cases, the percentage of total responding allocated toward target
response resurgence was less in the serial DRA component than in the traditional DRA component.
Furthermore, we observed both reversion and recency for 2 of 3 subjects. Our data provide
preliminary evidence suggesting that serial DRA may produce more durable and desirable
outcomes than traditional DRA.
Key words: extinction, maintenance, resurgence, translational research

Differential reinforcement of alternative Fisher, Sullivan, Acquisto, & LeBlanc, 1998;

behavior (DRA) is a behavior-reduction strategy Kelley, Lerman, & Van Camp, 2002; Rooker,
in which caregivers provide individuals with Jessel, Kurtz, & Hagopian, 2013; Tiger, Hanley,
access to the problem behavior’s functional & Bruzek, 2008).
reinforcer contingent on socially appropriate Despite its potential to eliminate problem
alternative responses. Although DRA has shown behavior when implemented as programmed,
mixed results when used without extinction some evidence exists to demonstrate that poor
(e.g., Horner & Day, 1991; Shirley, Iwata, procedural fidelity is a threat to the efficacy of
Kahng, Mazaleski, & Lerman, 1997; Worsdell, DRA with extinction (St. Peter Pipkin, Vollmer,
Iwata, Hanley, Thompson, & Kahng, 2000), it is & Sloman, 2010). Although failing to place
most often used with extinction (e.g., Hagopian, problem behavior on extinction appears to be
more detrimental to the success of the inter-
Elizabeth Dayton is now at Melmark. Andrew Rodewald vention than failing to reinforce alternative
is now at the Johns Hopkins School of Medicine. We thank responses (e.g., St. Peter Pipkin et al., 2010),
Tim Slocum for the comments that he provided during this
project. alternative response reinforcement still plays a
Correspondence concerning this article should be role in the effectiveness of DRA with extinction.
addressed to Sarah E. Bloom, Department of Child and For example, continuing to reinforce alter-
Family Studies, University of South Florida, Tampa, Florida
33612 (e-mail: sarahbloom@usf.edu). native responses during maintenance stages of
doi: 10.1002/jaba.253 the intervention can mitigate the adverse effects

765
766 JOSEPH M. LAMBERT et al.

of extinction fidelity errors (Shirley et al., 1997; the order and magnitude of the resurgence of
Vollmer, Roane, Ringdahl, & Marcus, 1999). multiple previously extinguished responses,
Conversely, failing to reinforce alternative there is some evidence to demonstrate that
responses (or reinforcing them on too lean a behavior taught first (a primacy effect) and last
schedule) while continuing to place problem (a recency effect) in serial training programs tend
behavior on extinction can result in resurgence to resurge with greater magnitude than behavior
of problem behavior (Volkert, Lerman, Call, & taught in the middle of such programs
Trosclair-Lasserre, 2009; Wacker et al., 2013). (cf. Lieving & Lattal, 2003; Mechner & Jones,
Resurgence is a form of relapse in which a 2011; Reed & Morgan, 2006). Furthermore,
previously extinguished response recovers after Reed and Morgan (2006) cited an unpublished
an alternative response is placed on extinction study in which a serial training program yielded
(Leitenberg, Rawson, & Mulick, 1975; Rawson, “reversion,” or the resurgence of responses in the
Leitenberg, Mulick, & Lefebvre, 1977). reverse order in which they had been trained
Response resurgence is influenced by (although Reed and Morgan reported that their
various history effects. For instance, responses subjects did not demonstrate reversion).
with longer histories of reinforcement If it is possible to program serial DRA in such a
(Bruzek, Thompson, & Peters, 2009; da Silva, way that reliably produces both recency and
Maxwell, & Lattal, 2008; Doughty, Cash, reversion of previously eliminated responses,
Finch, Holloway, & Wallington, 2010) or then the technique could have a number of
responses that occur in contexts in which important implications for applied contexts in
differentially high rates of reinforcement have which socially significant human behavior (e.g.,
been delivered (Podlesnik & Shahan, 2009, appropriate communication and aggression) are
2010) appear to resurge with greater magnitude frequently targeted. For instance, because prob-
than responses with shorter histories of reinforce- lem behavior and appropriate behavior often
ment, responses emitted in contexts paired with belong to the same response class (Harding et al.,
relatively lower rates of reinforcement, or both 2009; Lalli et al., 1995; Winborn, Wacker,
(see Shahan & Sweeney, 2011, for a discussion). Richman, Asmus, & Geier, 2002; Winborn-
Conversely, prolonged (Sweeney & Shahan, Kemmerer, Ringdahl, Wacker, & Kitsukawa,
2013) or multiple (Cleland, Foster, & Temple, 2009; Winborn-Kemmerer et al., 2010), it may
2000) exposures to extinction can decrease the be possible to program an intervention involving
magnitude of target response resurgence. serial DRA to ensure that a variety of socially
Interestingly, even as responses in a response appropriate responses resurge before or in
class emerge in a hierarchical order during relatively greater magnitude than problem
behavioral escalations (Fritz, Iwata, Hammond, behavior when an initial response is ineffective.
& Bloom, 2013; Harding et al., 2001; Langdon, If so, care providers who inadvertently ignore
Carr, & Owen-DeSchryver, 2008; Najdowski, one form of appropriate communication may be
Wallace, Ellsworth, MacAleese, & Cleveland, given additional opportunities to reinforce other
2008; Smith & Churchill, 2002; see also Borrero forms of appropriate communication before
& Borrero, 2008; Lalli, Mace, Wohn, & Livezey, problem behavior resurges. This approach could
1995; Magee & Ellis, 2000; Richman, Wacker, be conceptualized as an inoculation against
Asmus, Casey, & Andelman, 1999), previously lapses in treatment fidelity.
extinguished responses also resurge during these Thus, the purpose of this study was to evaluate
same escalations (Lieving, Hagopian, Long, & whether increasing the size of a response class via
O’Connor, 2004). Although little research has a serial training program would make DRA
been conducted on the variables that influence with extinction a more durable intervention
 SERIAL ALTERNATIVE RESPONSE TRAINING 767

during lapses in procedural fidelity in which Andrea did not consistently emit the required
reinforcement was withheld. Specifically, we response for two of the original switches
evaluated whether our intervention could (a) independently (i.e., toggle light switch and
decrease the percentage of total responding personal alarm button; see failed responses in
allocated toward target response resurgence Figure 1), she responded on two switches that
when reinforcement was withheld, (b) decrease were not used with Mary or Jenny (i.e., a
the magnitude of target response resurgence different toggle light switch and a different
when reinforcement was withheld, and (c) button). We scored a response any time a subject
evaluate whether a serial training program would picked up a device from the table, performed the
produce reversion, primacy, or recency when action that would be required to produce a
reinforcement was withheld. functional consequence (e.g., turn on a light),
and then put it back on the table. We scored a
reinforcer delivery every time a therapist gave a
METHOD
subject a high-preference edible item.
Subjects
Three subjects who had been diagnosed with a Reliability
developmental disability participated in this A second trained (using training procedures
experiment. Mary was a 53-year-old woman outlined by Dempsey, Iwata, Fritz, & Rolider,
who could speak in full sentences, would 2012) observer collected reliability data through-
independently recruit attention, and attended a out the study. We calculated reliability scores by
sheltered workshop on weekdays. Jenny was a dividing session components into 10-s intervals
59-year-old woman who spoke in one- to three- for each active variable, calculating a proportion
word utterances, attended a sheltered workshop of agreement, dividing the sum of these
on weekdays, and worked part time (with proportions by the total number of intervals,
assistance) at a local store. Andrea was a and converting the result to a percentage.
32-year-old woman who occasionally spoke in Reliability for Mary averaged 93% (range,
one- to three-word utterances and who also 79% to 100%) and was collected during 47%
attended a sheltered workshop on weekdays. of her session components. Reliability for Jenny
Each subject lived in a different home of a averaged 94% (range, 80% to 100%) and was
residential service provider for adults with calculated for 42% of her session components.
developmental and intellectual disabilities. Reliability for Andrea averaged 95% (range, 82%
to 100%) and was calculated for 47% of her
Response Measurement session components. If primary and reliability
Response rates on five topographically differ- data collectors produced low reliability scores
ent devices (i.e., toggle light switch, rocker light (i.e., below 80%), both were suspended from
switch, slide light switch, cord switch, and data collection until an author reviewed all
personal alarm button) were the dependent operational definitions with them. This occurred
variables. We arbitrarily designated one switch once during the first session scored by a new
as the target response and all other switches as observer. After the review, that observer’s data
alternative responses (A1, B1, B2, or B3). We were always in at least 80% agreement with other
varied switch designation across subjects. data collectors.
That is, if we designated a specific switch as
Alternative Response A1 for one subject, we may Session Protocol
have designated the same switch as Alternative Each session consisted of two components.
Response B3 for another subject. Because One component served as the control, and one
768 JOSEPH M. LAMBERT et al.

Figure 1. Effects of reinforcement, elimination, and resurgence phases on target and alternative responding of each
subject. Responding in the control component is shown in the left panels, and responding in the test component is shown in
the right panels. Data paths labeled “failed” in the bottom right panel depict responding on two devices for which consistent
independence could not be established.

component served as the test. In the control extinction). We conducted each component in
component, we placed the target response on a separate room to enhance discrimination across
extinction and reinforced a single alternative components. Before initiating a component, we
response (analogous to traditional DRA with placed a small folding table and two chairs at the
extinction). In the test component, we placed the center of the relevant room. During each
target response on extinction and reinforced each component, the therapist sat in one chair and
of three separate alternative responses in sequen- the subject sat in the other. We counterbalanced
tial order (analogous to serial DRA with component order across sessions and days.
 SERIAL ALTERNATIVE RESPONSE TRAINING 769

For Mary and Jenny, each component lasted period in which they did not do so independ-
5 min and was separated by (approximately) a ently. We continued prompting subjects until
2-min intercomponent (transition) interval. they demonstrated independence (i.e., emitted
During target response training (early in the the reinforced response without a prompt at least
reinforcement phase), Andrea frequently stood once every 30 s) in the relevant component for a
up and left the experimental environment before single session.
a component ended. Her data showed that she In the control component, we trained subjects
stayed seated for approximately 3 min before to emit a single alternative response (A1). In the
leaving. To accommodate Andrea’s preference for test component, we trained subjects to emit three
time spent in each component, we changed her separate alternative responses (B1, B2, and B3)
component duration from 5 min to 3 min. in sequential order. That is, we trained subjects
Reinforcement phase. Before initiating a ses- to emit B1 while at the same time placing the
sion component, we placed a single device (i.e., target response on extinction. After subjects’ first
the target response) at the center of the table (the demonstration of independence with B1, we
device was the same in both components) and ceased to provide contingency reviews and
delivered a contingency review or presession prompts and reinforced B1 for an additional
forced exposure (e.g., “when you do this . . . you five sessions. Then, we placed B1 on extinction
get [high-preference edible item]). The preferred (while keeping the target response on extinction)
item had been identified via a paired-stimulus and trained subjects to emit B2. After subjects’
preference assessment (Fisher et al., 1992). first demonstration of independence with B2, we
During each component, the subject was free ceased to provide contingency reviews and
to manipulate the device at any time. However, if prompts and reinforced B2 for an additional
more than 30 s ever elapsed without a response, five sessions. Finally, we placed B2 on extinction
the therapist prompted her to make a response. (while the target response and B1 were kept on
All responses (prompted or independent) were extinction) and trained subjects to emit B3. After
reinforced on a fixed-ratio (FR) 1 schedule. subjects’ first demonstration of independence
We continued to prompt subjects to emit with B3, we ceased to provide contingency
target responses until they demonstrated reviews and prompts and reinforced B3 for an
independence (i.e., no prompted responding) additional five sessions.
across both components for a single session. Devices associated with untrained responses
After their first demonstration of independence, were never available. However, after they had
subjects stayed in this phase for a minimum of been trained, devices were never again removed.
10 additional sessions (but may have stayed For example, during the reinforcement phase,
longer depending on their tier of the multiple only the device required to emit the target
baseline). response was available in both components.
Elimination phase. During this phase, we However, by the end of the elimination phase,
placed target responding on extinction in both the devices for the target response and A1 were
control and test components and reinforced available in the control component and the
alternative responses on an FR 1 schedule. We devices for the target response, B1, B2, and B3
taught subjects to emit alternative responses in were available in the test component. We did
the same fashion as we taught them to emit the this to simulate increases in response class size
target response. That is, we provided them with a that may occur in applied settings when
forced exposure to the active contingency before alternative responses are trained during FCT
each component and then prompted them to but were not available in clients’ repertoires
emit the reinforced response following any 30-s before training.
770 JOSEPH M. LAMBERT et al.

Resurgence phase. In this phase, we placed all 33 of the control component (21 responses;
responses on extinction. Devices for the target 51.9% of average baseline responding) and was
response and A1 were available in the control greatest during Sessions 32 and 34 of the test
component, and devices for the target response component (1 response; 2.6% of average
and B1, B2, and B3 were available in the test baseline responding). During Jenny’s resurgence
component. This phase lasted for five sessions phase, target response recovery was greatest
for Mary and Jenny and 14 sessions for Andrea. during Session 41 of the control component
Andrea required additional extinction sessions (1 response; 1.8% of average baseline respond-
because we did not observe a meaningful ing) and did not occur during the test
decrease in the most recently reinforced component. Finally, during Andrea’s resurgence
responses (A1 in control and B3 in test) during phase, target response recovery was greatest
any of the first five sessions of this phase. during Session 89 of the control component
(13 responses; 117.6% of average baseline
responding) and was greatest during Sessions
RESULTS
88 and 91 of the test component (1 response;
Table 1 summarizes the total number of 7.5% of average baseline responding).
reinforcers earned by each subject in each Figure 2 shows the cumulative number of
component as well as the average rate of responses that occurred in control and test
reinforcement earned (per minute) in each components for all subjects during the resurgence
component. The table also summarizes the phase. White and black segments are stacked.
average rate of responding (per minute) during White segments represent alternative responses,
the last five sessions in which each response was and black segments represent target responses.
reinforced. For all three subjects, fewer target responses
Figure 1 displays the data from the reinforce- occurred in test segments than in control segments
ment, elimination, and resurgence phases for all during the resurgence test. Target responding also
three subjects. Data from the control components occupied a smaller percentage of total responding
are shown in left panels, and data from the test in test segments than in control segments. In
components are shown in right panels. DRA with addition, for two of three subjects, more total
extinction suppressed target responding in both responding (i.e., the sum of all responses) occurred
control and test components for all subjects. in test segments than in control segments. These
During Mary’s resurgence phase, target results suggest that larger response classes, a serial
response recovery was greatest during Session training program, or some combination of the two

Table 1
Total and Average Rate of Reinforcement Delivered to Each Subject in Each Context (Component) and Average Rate
of Responding Observed During the Final Five Sessions in Which Each Response Was Reinforced

Average rate of responding during final five sessions

Participant Total reinforcement (average) Target A1 B1 B2 B3
Mary (control) 1,078 (7.2) 8.1 8
Mary (test) 1,092 (7.3) 7.9 5.6 7.4 7.7
Jenny (control) 1,833 (9.2) 11.2 13.2
Jenny (test) 1,720 (8.6) 10 10.7 12.2 10.8
Andrea (control) 672 (2.6) 3.4 4.1
Andrea (test) 693 (2.7) 4 4.4 2.4 3.3
 SERIAL ALTERNATIVE RESPONSE TRAINING 771

(B2) occupied the largest percentage of total

relapse, demonstrating a recency effect.

DISCUSSION
In our study, teaching individuals multiple
alternative responses using a serial training
program increased the total amount of respond-
ing and decreased the percentage of total
responding allocated to the resurgence of target
behavior when reinforcement was no longer
available. In addition, it decreased the absolute
Figure 2. Bars represent the total number of responses amount as well as the relative amount of target
emitted during the resurgence phase. White and black response resurgence (when considered as a
segments are stacked. White segments represent all percentage of baseline responding). Further-
alternative responses and black segments represent target
more, when the functional reinforcer was no
responses. Numbers above bars represent the exact number
of target responses emitted and the percentage of total longer available, at least one alternative response
responding accounted for by the target response. resurged in the test condition before the target
response. The intervention was followed by
may have increased total response output while a reversion of response resurgence for two of
simultaneously suppressing the magnitude of three subjects. Finally, withholding reinforce-
target response recovery. ment produced a recency effect for two of three
Figures 3 through 8 show a second-by-second subjects. From a clinical standpoint, this out-
analysis of responding during the resurgence come is desirable because alternative responses
phases of each subject. These data were used to are often the most recently reinforced responses
determine the order in which previously extin- in the response class before lapses in procedural
guished responses were emitted. Top panels show fidelity occur.
a timeline of responding, with a summary of the These results are promising and indicate that
total number of each response emitted in there may be several reasons to teach clients
parentheses to the right. Bottom panels show multiple alternative responses during the train-
the same data depicted in a cumulative record. ing stages of DRA. First, doing so may decrease
Data from the test segments of this analysis show the relative amount of problem behavior
that a reversion, or the recovery of previously resurgence when functional reinforcers are
extinguished responses in the reverse order in temporarily withheld. In addition, doing so
which they were reinforced and extinguished, could increase the probability that appropriate
occurred for two of three subjects (Jenny and alternative responses resurge before problem
Andrea). behavior. This outcome may decrease the
Figure 9 shows each recovering response as a probability of the resurgence of problem
percentage of all relapses that occurred during behavior altogether, because one alternative
the resurgence phase for each subject. For all response may recruit reinforcement more effec-
three subjects, the first reinforced and extin- tively than another, thus inoculating individuals
guished response (the target response) occupied against lapses in treatment fidelity by giving
the smallest percentage of total relapse. For two them the means to continue to recruit functional
of three subjects (Andrea and Jenny), the last reinforcers appropriately before they revert to
previously reinforced and extinguished response problem behavior. Should a care provider
772 JOSEPH M. LAMBERT et al.

Figure 3. Second-by-second analysis of responding in control component of Mary’s resurgence test. The graph in the
top panel depicts a timeline of responding (across sessions). Numbers in parentheses to the right of the timeline depict the
sum total of each response emitted. The graph in the bottom panel shows the same data depicted in a cumulative record.

reinforce one response in a class at any point study was how quickly target response resurgence
during extinction, subject response patterns was eliminated in both control and test
would likely mirror those seen during the components for all of our subjects. Our results
elimination phase of this experiment (i.e., support the assertion made by St. Peter Pipkin
near-exclusive response allocation toward the et al. (2010) that fidelity to target response
reinforced response) and problem behavior may extinction is more vital to the effectiveness of an
not occur. intervention than the reinforcement of alter-
Despite the fact that more target response native responses. Had we reinforced target
resurgence was observed in control conditions behavior during the resurgence phase, one might
that reflect traditional DRA-with-extinction assume that target responding would have
procedures, another promising result of this returned to baseline rates of responding (as was
 SERIAL ALTERNATIVE RESPONSE TRAINING 773

Figure 4. Second-by-second analysis of responding in test components of Mary’s resurgence test. The graph in the top
panel depicts a timeline of responding (across sessions). Numbers in parentheses to the right of the timeline depict the sum
total of each response emitted. The graph in the bottom panel shows the same data depicted in a cumulative record.

the case each time a new response was trained reinforcement in Context A and we place it on
and reinforced during the elimination phase). extinction in Context B, when we return our
Because we conducted each component (con- client to Context A and then observe a recovery of
trol and test) of our multiple schedule in a the extinguished response when no other variables
different room, we could not assume that the in Context A have changed, the recovery can be
recovery of target responding during the resur- characterized as renewal and is a function of
gence test was resurgence without first ruling out transitioning between contexts (cf. Bouton,
renewal. Renewal occurs when a previously Todd, Vurbic, & Winterbauer, 2011; Nakajima,
extinguished response reemerges as a function Tanaka, Urushihara, & Imada, 2000).
of a context change (Welker & McAuley, 1978). To parse the effects of context change
For example, if a client’s response has a history of (renewal) from the effects of alternative response
774 JOSEPH M. LAMBERT et al.

Figure 5. Second-by-second analysis of responding in control components of Jenny’s resurgence test. The graph in the
top panel depicts a timeline of responding (across sessions). Numbers in parentheses to the right of the timeline depict the
sum total of each response emitted. The graph in the bottom panel shows the same data depicted in a cumulative record.

extinction (resurgence), we only identified target with those context changes across time. The
response recovery as resurgence if the highest rate results of our resurgence analysis are shown in
of responding during the resurgence phase was Table 2. According to our criteria, the target
higher than the highest rate of responding during response resurged in the control component for
any of the last five sessions of the elimination all three subjects and resurged in the test
phase. We selected the final five sessions of the component for two of the three subjects (Mary
elimination phase to account for each subject’s and Andrea).
learning history. That is, we wanted to account A few procedural limitations should be noted.
for the effect that context change had on First, we reinforced only one alternative response
response recovery while we also accounted for at any given time during the serial training
decreases in that effect attributable to experience program. Thus, if a previously trained alternative
 SERIAL ALTERNATIVE RESPONSE TRAINING 775

Figure 6. Second-by-second analysis of responding in test components of Jenny’s resurgence test. The graph in the top
panel depicts a timeline of responding (across sessions). Numbers in parentheses to the right of the timeline depict the sum
total of each response emitted. The graph in the bottom panel shows the same data depicted in a cumulative record.

response occurred during a subsequent response’s to produce the outcomes found in this study.
baseline, that response contacted extinction Furthermore, because the control component
before the resurgence phase. An analogue to consisted of a single alternative response and the
this is not likely to be found in most clinical test component consisted of three alternative
settings in which care providers are typically responses, it is impossible to establish whether
taught to reinforce all topographies of appro- the observed effects were a function of the serial
priate behavior whenever they occur. In addi- training program, a larger response class, or some
tion, our study did not systematically evaluate combination of the two. Because a serial DRA
how long an alternative response needs to be program calls for the establishment and sub-
reinforced before it can be placed on extinction, sequent extinction of multiple appropriate
or how long it can remain on extinction, in order alternative responses in populations in which
776 JOSEPH M. LAMBERT et al.

Figure 7. Second-by-second analysis of responding in control components of Andrea’s resurgence test. The graph in the
top panel depicts a timeline of responding (across sessions). Numbers in parentheses to the right of the timeline depict the
sum total of each response emitted. The graph in the bottom panel shows the same data depicted in a cumulative record.

appropriate behavior is often limited, it is target response in one component influenced

important to verify the necessity of the serial target responding in the alternative component.
training portion of the intervention before it is It should also be noted that the rate of occurrence
implemented with applied populations. Future of Alternative Responses A1 and B3 (i.e., the
researchers may wish to conduct a component responses placed on extinction for the first
analysis of this intervention. time during the resurgence phase) increased
Another limitation of our study was that we substantially during the final two sessions of
used the same target response in both control and Jenny’s resurgence phase. Although these
test components for each of our subjects. Thus, it responses decreased considerably during the first
was possible that contingencies contacted by the three sessions of extinction, their recovery during
 SERIAL ALTERNATIVE RESPONSE TRAINING 777

Figure 8. Second-by-second analysis of responding in test components of Andrea’s resurgence test. The graph in the top
panel depicts a timeline of responding (across sessions). Numbers in parentheses to the right of the timeline depict the sum
total of each response emitted. The graph in the bottom panel shows the same data depicted in a cumulative record.

the final two sessions is peculiar. Because we did contingency shifted to a new response, it is
not conduct additional extinction sessions for possible that the observed increase in the rate of
Jenny (like we did for Andrea), we cannot say responding for Responses A1 and B3 (the first
with certainty that control of these responses did time in the experiment in which reinforcement
not transfer to unprogrammed consequences and was completely withheld) was an extinction
that they truly contacted extinction during this burst (for a discussion of extinction bursts, see
phase. However, considering the efficiency with Lerman, Iwata, & Wallace, 1999).
which the programmed reinforcer increased all Finally, we did not include an unreinforced
other responses when targeted and the rapidity control response in either component of our
in which these responses decreased when the evaluation. Thus, it is not possible for us to
778 JOSEPH M. LAMBERT et al.

individuals to emit more than one alternative

response in a serial DRA program. However, we
obtained these results by targeting arbitrary
responses over which we could control reinforce-
ment histories. It is difficult to predict the
generality of these results if the responses had
more extensive histories of reinforcement, as
with problem behavior treatment situations.
Mace and Critchfield (2010) noted that
translational research has historically served
three functions. First, it has been a method to
Figure 9. The percentage of total response recovery replicate in humans the effects originally
occupied by responses reinforced and extinguished first, produced with nonhumans. Second, it has
second, and last before the resurgence test as a percentage of been used to develop laboratory models of
all relapses that occurred during the resurgence test for each specific human problems to facilitate their
subject.
resolution. Finally, it has been used to study
uniquely human behavior. Given our method,
determine whether the recovery of previously
goals, and results, this study is translational
extinguished responses was resurgence or
because it accomplishes the first two functions
whether it was the function of some other
noted by Mace and Critchfield. That is, we
behavioral operation evoked by extinction, such
replicated response resurgence (an effect that was
as extinction-induced variability (Neuringer,
originally produced in nonhumans) with hu-
Kornell, & Olufs, 2001).
mans. In addition, we demonstrated the use of a
Notwithstanding these limitations, our results
low-stakes model for evaluating the effect of an
provide preliminary evidence to suggest that it
intervention whose results have treatment
may be possible to decrease relative response
implications. If our results can be replicated in
allocation to problem behavior when appropriate
applied contexts, then serial DRA may be one
behavior contacts extinction by teaching
way to empower individuals with disabilities who
receive treatment for problem behavior to recruit
reinforcement even when initial appropriate
Table 2 attempts are not reinforced.
Resurgence Analysis

Highest rate of responding

in a single session REFERENCES
Last five Resurgence Borrero, C. S. W., & Borrero, J. C. (2008). Descriptive and
Subject (elimination) test Resurgence? experimental analyses of potential precursors to
problem behavior. Journal of Applied Behavior Analysis,
Mary (control) 0 4.2 yes
Mary (test) 0 0.2 yes 41, 83–96. doi: 10.1901/jaba.2008.41-83
Jenny (control) 0 0.2 yes Bouton, M. E., Todd, T. P., Vurbic, D., & Winterbauer,
Jenny (test) 0 0 no N. E. (2011). Renewal after the extinction of free
Andrea (control) 0 0.6 yes operant behavior. Learning & Behavior, 39, 57–67. doi:
Andrea (test) 0 4 yes 10.3758/s13420-011-0018-6
Bruzek, J. L., Thompson, R. H., & Peters, L. C. (2009).
Note. Response recovery was identified as resurgence if Resurgence of infant caregiving responses. Journal of
the highest rate of responding in the component of the the Experimental Analysis of Behavior, 92, 327–343. doi:
resurgence test was higher than the highest rate of 10.1901/jeab.2009-92-327
responding in the same component during the final five Cleland, B. S., Foster, T. M., & Temple, W. (2000).
sessions of the elimination condition. Resurgence: The role of extinction. Behavioural
 SERIAL ALTERNATIVE RESPONSE TRAINING 779

Processes, 52, 117–129. doi: 10.1016/S0376-6357(00) Modification, 32, 804–827. doi: 10.1177/
00131-5 0145445508317943
da Silva, S. P., Maxwell, M. E., & Lattal, K. A. (2008). Leitenberg, H., Rawson, R. A., & Mulick, J. A. (1975).
Concurrent resurgence and behavioral history. Journal Extinction and reinforcement of alternative behavior.
of the Experimental Analysis of Behavior, 90, 313–331. Journal of Comparative and Physiological Psychology, 88,
doi: 10.1901/jeab.2008.90-313 640–652. doi: 10.1037/h0076418
Dempsey, C. M., Iwata, B. A., Fritz, J. N., & Lerman, D. C., Iwata, B. A., & Wallace, M. D. (1999). Side
Rolider, N. U. (2012). Observer training revisited: effects of extinction: Prevalence of bursting and
A comparison of in vivo and video instruction. aggression during the treatment of self-injurious
Journal of Applied Behavior Analysis, 45, 827–832. behavior. Journal of Applied Behavior Analysis, 32,
doi: 10.1901/jaba.2012.45-827 1–8. doi: 10.1901/jaba.1999.32-1
Doughty, A. H., Cash, J. D., Finch, E. A., Holloway, C., & Lieving, G. A., Hagopian, L. P., Long, E. S., & O’Connor, J.
Wallington, L. K. (2010). Effects of training history on (2004). Response-class hierarchies and resurgence of
resurgence in humans. Behavioural Processes, 83, severe problem behavior. The Psychological Record, 54,
340–343. doi: 10.1016/j.beproc.2009.12.001 621–634.
Fisher, W., Piazza, C. C., Bowman, L. G., Hagopian, L. P., Lieving, G. A., & Lattal, K. A. (2003). Recency,
Owens, J. C., & Slevin, I. (1992). A comparison of repeatability, and reinforcer retrenchment: An exper-
two approaches for identifying reinforcers for persons imental analysis of resurgence. Journal of the Exper-
with severe and profound disabilities. Journal of imental Analysis of Behavior, 80, 217–233. doi:
Applied Behavior Analysis, 25, 491–498. doi: 10.1901/jeab.2003.80-217
10.1901/jaba.1992.25-491 Mace, F. C., & Critchfield, T. S. (2010). Translational
Fritz, J. N., Iwata, B. A., Hammond, J. L., & Bloom, S. E. research in behavior analysis: Historical traditions and
(2013). Experimental analysis of precursors to severe imperative for the future. Journal of the Experimental
problem behavior. Journal of Applied Behavior Analysis, Analysis of Behavior, 93, 293–312. doi: 10.1901/
46, 101–129. doi: 10.1002/jaba.27 jeab.2010.93-293
Hagopian, L. P., Fisher, W. W., Sullivan, M. T., Acquisto, J., Magee, S. K., & Ellis, J. (2000). Extinction effects during
& LeBlanc, L. A. (1998). Effectiveness of functional the assessment of multiple problem behaviors. Journal
communication training with and without extinction of Applied Behavior Analysis, 33, 313–316. doi:
and punishment: A summary of 21 inpatient cases. 10.1901/jaba.2000.33-313
Journal of Applied Behavior Analysis, 31, 211–235. doi: Mechner, F., & Jones, L. (2011). Effects of sequential aspects
10.1901/jaba.1998.31-211 of learning history. Mexican Journal of Behavior Analysis,
Harding, J. W., Wacker, D. P., Berg, W. K., Barretto, A., 37, 109–138. doi: 10.5514/rmac.v37.i1.24688
Winborn, L., & Gardner, A. (2001). Analysis of Najdowski, A. C., Wallace, M. D., Ellsworth, C. L.,
response class hierarchies with attention-maintained MacAleese, A. N., & Cleveland, J. M. (2008).
problem behaviors. Journal of Applied Behavior Functional analyses and treatment of precursor
Analysis, 34, 61–64. doi: 10.1901/jaba.2001.34-61 behavior. Journal of Applied Behavior Analysis, 41,
Harding, J. W., Wacker, D. P., Berg, W. K., Winborn- 97–105. doi: 10.1901/jaba.2008.41-97
Kemmerer, L., Lee, J. F., & Ibrahimovic, M. (2009). Nakajima, S., Tanaka, S., Urushihara, K., & Imada, H.
Analysis of multiple manding topographies during (2000). Renewal of extinguished lever-press responses
functional communication training. Education and upon return to the training context. Learning and
Treatment of Children, 32, 21–36. doi: 10.1353/ Motivation, 31, 416–431. doi: 10.1006/lmot.2000.1064
etc.0.0045 Neuringer, A., Kornell, N., & Olufs, M. (2001). Stability
Horner, R. H., & Day, H. M. (1991). The effects of and variability in extinction. Journal of Experimental
response efficiency on functionally equivalent com- Psychology: Animal Behavior Processes, 27, 79–94. doi:
peting behaviors. Journal of Applied Behavior Analysis, 10.1037//0097-7403.27.1.79
24, 719–732. doi: 10.1901/jaba.1991.24-719 Podlesnik, C. A., & Shahan, T. A. (2009). Behavioral
Kelley, M. E., Lerman, D. C., & Van Camp, C. M. (2002). momentum and relapse of extinguished operant
The effects of competing reinforcement schedules on responding. Learning & Behavior, 37, 357–364. doi:
the acquisition of functional communication. Journal 10.3758/LB.37.4.357
of Applied Behavior Analysis, 35, 59–63. doi: 10.1901/ Podlesnik, C. A., & Shahan, T. A. (2010). Extinction,
jaba.2002.35-59 relapse, and behavioral momentum. Behavioural Proc-
Lalli, J. S., Mace, F. C., Wohn, T., & Livezey, K. (1995). esses, 84, 400–411. doi: 10.1016/j.beproc.2010.02.001
Identification and modification of a response-class Rawson, R. A., Leitenberg, H., Mulick, J. A., & Lefebvre,
hierarchy. Journal of Applied Behavior Analysis, 28, M. F. (1977). Recovery of extinction responding in rats
551–559. doi: 10.1901/jaba.1995.28-551 following discontinuation of reinforcement of alter-
Langdon, N. A., Carr, E. G., & Owen-DeSchryver, J. S. native behavior: A test of two explanations. Animal
(2008). Functional analysis of precursors for serious Learning & Behavior, 5, 415–420. doi: 10.3758/
problem behavior and related intervention. Behavior BF03209589
780 JOSEPH M. LAMBERT et al.

Reed, P., & Morgan, T. A. (2006). Resurgence of response during treatment with functional communication
sequences during extinction in rats shows a primacy training. Journal of Applied Behavior Analysis, 42,
effect. Journal of the Experimental Analysis of Behavior, 145–160. doi: 10.1901/jaba.2009.42-145
86, 307–315. doi: 10.1901/jeab.2006.20-05 Vollmer, T. R., Roane, H. S., Ringdahl, J. E., & Marcus,
Richman, D. M., Wacker, D. P., Asmus, J. M., Casey, S. D., B. A. (1999). Evaluating treatment challenges with
& Andelman, M. (1999). Further analysis of problem differential reinforcement of alternative behavior.
behavior in response class hierarchies. Journal of Journal of Applied Behavior Analysis, 32, 9–23.
Applied Behavior Analysis, 32, 269–283. doi: doi: 10.1901/jaba.1999.32-9
10.1901/jaba.1999.32-269 Wacker, D. P., Harding, J. W., Morgan, T. A., Berg, W. K.,
Rooker, G. W., Jessel, J., Kurtz, P. F., & Hagopian, L. P. Schieltz, K. M., Lee, J. F., & Padilla, Y. C. (2013). An
(2013). Functional communication training with and evaluation of resurgence during functional communi-
without alternative reinforcement and punishment: An cation training. The Psychological Record, 63, 3–20. doi:
analysis of 58 applications. Journal of Applied Behavior 10.11133/j.tpr.2013.63.1.001
Analysis, 46, 708–722. doi: 10.1002/jaba.76 Welker, R. L., & McAuley, K. (1978). Reductions in
Shahan, T. A., & Sweeney, M. M. (2011). A model of resistance to extinction and spontaneous recovery as
resurgence based on behavioral momentum theory. a function of changes in transportational and
Journal of the Experimental Analysis of Behavior, 95, contextual stimuli. Animal Learning & Behavior, 6,
91–108. doi: 10.1901/jeab.2011.95-91 451–457. doi: 10.3758/BF03209643
Shirley, M. J., Iwata, B. A., Kahng, S., Mazaleski, J. L., & Winborn, L., Wacker, D. P., Richman, D. M., Asmus, J., &
Lerman, D. C. (1997). Does functional communica- Geier, D. (2002). Assessment of mand selection for
tion training compete with ongoing contingencies of functional communication training packages.
reinforcement? An analysis during response acquisition Journal of Applied Behavior Analysis, 35, 295–298.
and maintenance. Journal of Applied Behavior Analysis, doi: 10.1901/jaba.2002.35-295
30, 93–104. doi: 10.1901/jaba.1997.30-93 Winborn-Kemmerer, L., Ringdahl, J. E., Wacker, D. P., &
Smith, R. G., & Churchill, R. M. (2002). Identification of Kitsukawa, K. (2009). A demonstration of individual
environmental determinants of behavior disorders preference for novel mands during functional commu-
through functional analysis of precursor behaviors. nication training. Journal of Applied Behavior Analysis,
Journal of Applied Behavior Analysis, 35, 125–136. 42, 185–189. doi: 10.1901/jaba.2009.42-185
doi: 10.1901/jaba.2002.35-125 Winborn-Kemmerer, L., Wacker, D. P., Harding, J.,
St. Peter Pipkin, C., Vollmer, T. R., & Sloman, K. N. Boelter, E., Berg, W., & Lee, J. (2010). Analysis of
(2010). Effects of treatment integrity failures during mand selection across different stimulus conditions.
differential reinforcement of alternative behavior: A Education and Treatment of Children, 33, 49–64.
translational model. Journal of Applied Behavior doi: 10.1353/etc.0.0086
Analysis, 43, 47–70. doi: 10.1901/jaba.2010.43-47 Worsdell, A. S., Iwata, B. A., Hanley, G. P., Thompson,
Sweeney, M. M., & Shahan, T. A. (2013). Behavioral R. H., & Kahng, S. (2000). Effects of continuous and
momentum and resurgence: Effects of time in intermittent reinforcement for problem behavior
extinction and repeated resurgence tests. Learning & during functional communication training. Journal of
Behavior, 41, 414–424. doi: 10.3758/s13420-013- Applied Behavior Analysis, 33, 167–179. doi: 10.1901/
0116-8 jaba.2000.33-167
Tiger, J. H., Hanley, G. P., & Bruzek, J. (2008). Functional
communication training: A review and practical guide.
Behavior Analysis in Practice, 1, 16–23. Received May 16, 2014
Volkert, V. M., Lerman, D. C., Call, N. A., & Trosclair- Final acceptance May 26, 2015
Lasserre, N. (2009). An evaluation of resurgence Action Editor, Joel Ringdahl