CLINICAL TRANSLATION OF ANIMAL MODELS OF TREATMENT RELAPSE
DUNCAN PRITCHARD1,2, MARGUERITE HOERGER2, F. CHARLES MACE3, HEATHER PENNEY1,AND BRIAN HARRIS1
1ARAN HALL SCHOOL2BANGOR UNIVERSITY
3NOVA SOUTHEASTERN UNIVERSITY
Behavioral Momentum Theory (BMT) has inspired animal models of treatment relapse. We translated themodels of reinstatement and resurgence into clinical procedures to test whether relapse tests wouldreplicate behavior pattern found in basic research. Following multiple schedule baseline reinforcement ofa 16-year-old males problem behavior at equal rates by two therapists, treatment was introduced using avariable-interval, variable-time (VI VT) schedule arrangement with therapists delivering reinforcers atdifferent rates. Despite the differing rates of VI VT reinforcers, the treatment produced comparablereductions in problem behavior. Following successful treatment, the two therapists discontinued treatmentand resumed reinforcement of problem behavior at equal rates that constituted a reinstatement of baselineconditions. As predicted by BMT, reinstatement resulted in an immediate return of high rates of problembehavior but was 2.6 times higher for the therapist using the higher rate VI VT treatment. A secondtreatment phase was implemented followed by a test of resurgence in a single extended extinction sessionconducted separately for each therapist. The unequal VI VT treatment rates by therapists resulted in 2.1times greater responding in the resurgence test for the therapist who implemented the higher rate VI VTprocedure. These results are consistent with basic research studies and BMT.Key words: behavioral momentum, reinstatement, resurgence, treatment relapse
The fundamental finding from BehavioralMomentum Theory (BMT) research is thatbehavior is more persistent during disruptionin a context correlated with greater reinforce-ment (Nevin, 1992; Nevin & Grace, 2000; Nevin,Tota, Torquato & Shull, 1990). This generalfunctional relationship has attracted the atten-tion of translational researchers because of theimplications basic BMT research has for com-mon clinical treatments for behavior disorderssuch as differential reinforcement of alternativebehavior (DRA) and variable- and fixed-timereinforcer deliveries (VT, FT).
For example, Ahearn, Clark, Gardenier,Chung and Dube (2003) found that VT accessto toys reduced automatically reinforced ste-reotypic behavior relative to a no-interactionbaseline for three children with autistic disor-der. However, when stereotypic behavior wasdisrupted by providing children continuousaccess to an activity that competed with stereo-typed responses, stereotypy occurred more
often following the VT treatment comparedto sessions following the no-interaction base-line. BMT predicts this finding due to thehigher rate of reinforcement in the VTtreatment relative to baseline. Mace et al.(2010) reported similar findings for the effectsDRA can have on the resistance to extinction ofproblem behaviors that were reinforced bysocially mediated consequences (attention andfood). Thus, although treatments based onalternative reinforcement effectively reduceproblem behavior, they can paradoxicallyincrease its resistance to extinction (cf. Nevinet al., 1990).
Treatment relapse is a generic term used inseveral medical and mental health fields. Itrefers to the recurrence of a condition that hasbeen previously treated effectively. Podlesnikand Shahan (2009) examined three paradigmsof treatment relapse (reinstatement, resurgenceand renewal) with pigeons that are conceptual-ized within the framework of BMT. A commonbaseline procedure was used to test each of therelapse models. Pigeons were provided equal VI120-s response-dependent food in a two-compo-nent multiple schedule. Time-contingent foodwas added to one of the components on avariable time (VT) 20-s schedule. The tworelapse paradigms relevant to the currentresearch are reinstatement and resurgence.
This research was conducted as partial fulfillment of thefirst authors Ph.D. degree in psychology at BangorUniversity. Preparation of this article was supported inpart by NICHD 064576.Address correspondence to: Duncan Pritchard, Aran Hall
School, Rhydymain, Dolgellau, Gwynedd, LL40 2AR. Email:firstname.lastname@example.org: 10.1002/jeab.87
JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2014, 101, 442449 NUMBER 3 (MAY)
In tests of the reinstatement paradigm,following extinction of baseline responding,food was presented response-independently infour sessions and response-dependently foranother four sessions, with session blocks beingseparated by a second phase of extinction. Foodwas presented independently of key pecks at 2and 8 s into the first presentation of eachcomponent in a session and later contingenton the first two responses in each component ofthe session; food was withheld for the remainderof the session. The reinstatement procedure wasan analog of the resumption of reinforcerdeliveries that had previously maintained prob-lem behavior. Although reinforcer rates duringthe response-independent and response-depen-dent tests were identical, the magnitude ofrelapse was greater in the component withadded VT reinforcers during baseline.For the resurgence model, extinction of the
target baseline key in both components wascombined with equal VI 30-s DRA for a secondkey and continued until responding on thetarget baseline key reduced below 10% ofbaseline. The resurgence test discontinuedreinforcement for all responses, but the re-sponse of interest was a resumption of respond-ing on the target baseline key in bothcomponents. Despite ongoing extinction ofthe target and alternative responses, baselinekey responding resumed in both componentsbut to a greater extent in the component withthe added VT reinforcers in baseline. Thisprocedure models treatment relapse followingdiscontinuation of an effective DRA treatment.The clinical value of these animal models of
treatment relapse depends on the extent towhich they predict clinically relevant humanbehavior exposed to treatment challengescommonly encountered in clinical settings.Only one clinical translational study has exam-ined the BMT reinstatement paradigm. DeL-eon, Willams, Gregory and Hagopian (2005)reported clinical data consistent with thereinstatement model of relapse. Problem be-havior was reinforced by attention on an FR1schedule during baseline. Following baselinereinforcement, all attention was withheld forproblem behavior until problem behavior waseliminated for four consecutive sessions. Atten-tion was reintroduced on a FT 60-s schedule forthree sessions and problem behavior immedi-ately returned as predicted by the reinstatementparadigm tested by Podlesnik and Shahan
(2009). However, DeLeon et al. did not arrangedifferent rates of reinforcement in a multipleschedule to demonstrate that reinstatementeffects were greater in the component correlat-ed with a higher rate of reinforcement.We are aware of only one clinical study that
approximates a test of one of the BMTparadigms of treatment relapse. MacDonald,Ahearn, Parry-Cruwys, and Bancroft (2013)conducted a clinical translation of the resur-gence paradigm of relapse described by Pod-lesnik and Shahan (2009) with some variationfrom the basic research. In the first phase of thestudy, problem behavior was alternately rein-forced on a CRF schedule and a variable-ratio(VR) 3 schedule in separate components of amixed schedule of reinforcement. Each rein-forcement component was followed by extinc-tion for all responses. The obtained rates ofreinforcement expressed as a ratio of intermit-tent to continuous reinforcement were 2.30,1.66, 1.41 and 1.8 for the four participants.Results showed that the transition from the CRFschedule to extinction resulted in substantiallyhigher rates of problem behavior than thetransition from VR 3 to extinction. Although theprocedure has no direct parallels with theanimal model of resurgence, MacDonald et al.(2013) did demonstrate that relatively higherrates of reinforcement for clinically significantproblem behavior can result in higher responserates during extinction.The present study aimed to translate the
animal models of reinstatement and resurgenceinto clinical procedures that closely approximatethose used in laboratory studies with nonhumans(e.g., Podlesnik & Shahan, 2009, 2010; Sweeney& Shahan, 2013). Both models were testedfollowing a multiple schedule arrangement of aVI VT treatment at different rates, therebyholding constant a 4-to-1 relative reinforcementratio in the schedule components. Reinstate-ment was tested by a return to contingentreinforcement of problem behavior and resur-gence was tested by extinction following success-ful treatment.
Participant, Setting and MaterialsThe participant was a 16 year-old male with
severe intellectual disability and severe aggres-sive and disruptive behavior. All experimentalsessions were conducted in a 10m by 6m living
TRANSLATION OF TREATMENT RELAPSE MODELS 443
area that contained two sofas, two tables withbench seats and an enclosed television. Theparticipant had access to preferred items such aspicture books and toys throughout all sessions.One of two therapists and one or two datacollectors were present for all sessions. Theprimary data collector was blind to the experi-mental hypotheses.
Target Behaviors, Data Collection andInterobserver Agreement
Participant target behaviors were: (a) aggres-sion, defined as slapping, biting, kicking, head-butting, hitting the therapist with an object,spitting on the therapist, forceful pushing, andattempted eye gouging; and (b) disruption,defined as throwing objects and tearing paper.Staff attention took two forms. During thefunctional analysis baseline and reinstatementphases, staffmade disapproving comments (e.g.,stop that., that hurts., dont tear things up.)contingent on occurrences of aggression anddisruption. During treatment, therapists re-sponded to demands for attention or providedVT attention in the form of pleasant interactionwith the participant for approximately 10 s.
Two independent data collectors recordedcounts of aggression and disruption in continu-ous 10-s intervals. Staff attention was recordedusing a 10-s partial-interval procedure. Interob-server agreement was balanced across studyphases and calculated using the exact agreementmethod on a point-by-point or interval-by-inter-val method (Cooper, Heron & Heward, 2007)for an average of 59.3%of the sessions for Phases1a to 4 and during the entire session for Phase 5.Mean total agreement on occurrences andnonoccurrences across all phases was 89.0%(range, 73.6% to 100%).
ProcedureThe study was composed of five successive
phases. The conditions in each phase werearranged in a multiple schedule with mostcomponents presented in a strictly alternatingsequence, with exceptions noted below. Twodifferent therapists served as discriminativestimuli for the schedule components. Sessionswere conducted on consecutive days and allsessions except extinction in Phase 5 were10min in duration and separated by a 5minperiod without adult attention.
Phase 1Functional analysis baseline. Previ-ous clinical work with this participant indicated
that his problem behavior was reinforced byattention (Mace et al., 2008). Separate attentionand control conditions were conducted bydifferent therapists (three sessions each forthe attention condition and three and twocontrol sessions conducted by Therapists 1 and2, respectively). Both therapists in the attentioncondition only interacted with the participantcontingent on problem behavior on a quasiran-dom VI 60-s schedule of reinforcement withouta limited hold. In the control condition, thetherapist provided near continuous interactionwith the participant and provided no instruc-tional demands. Mean obtained rates of atten-tion in the attention condition provided byTherapist 1 were 46.0/hr and 48.0/hr forTherapist 2. No attention was provided contin-gent on problem behavior by either therapist inthe control condition (see Table 1).
Phases 2 and 4VI VT treatment provided atdifferent rates of reinforcement. Therapist 1and Therapist 2 provided attention at differentrates in a quasirandom VI VT schedule oftreatment. Each time reinforcers were sched-uled to be available, the therapist approachedthe participant and pointed to a photograph ofthe therapist and participant interacting alongwith a verbal prompt (would you like to talk?serving as an SD for communication). Commu-nication responses resulted in contingent atten-tion for 10 s (i.e., VI attention). If the participantfailed to respond to the prompt, the therapistprovided verbal and physical attention for asimilar duration (i.e., VT attention). Therapist 1provided attention on a VI VT 30-s schedule andTherapist 2 provided the same treatment on a VIVT 120-s schedule. No interaction occurredduring the interreinforcement intervals. Ob-tained rates of attention averaged 120/hr forTherapist 1 and 30/hr for Therapist 2 in bothtreatment phases. The VT component of thetreatment permitted scheduled and obtainedreinforcement rates to be equivalent. Treatmentsessions continued in phase 2 and phase 4 untilaverage rates of problem behavior for boththerapists were 8.5% and 17% of the initialfunctional analysis baseline, respectively.
Phase 3Reinstatement of equal rates ofreinforcement. Following the initial treatmentphase, the VI VT intervention was discontinuedand both therapists resumed reinforcement ofaggression and disruption on equal VI 60-sschedules of reinforcement. In the attentioncondition, the mean rates of attention provided
444 DUNCAN PRITCHARD et al.
by Therapist 1 were 46.8/hr and 48.0/hr,respectively. In the control condition, no atten-tion was provided contingent on problembehavior by either therapist (see Table 1).Phase 5Extinction test of resurgence. A
single 74-min session of extinction followed thesecond treatment phase for each therapist. Thetherapist stood within 1m of the participant atall times and did not look at or talk to theparticipant during the entire session. Therapist1 conducted the first extinction session andTherapist 2 conducted the second session fourhours later.
Figure 1 shows the participants rates ofaggression and disruption during the functionalanalysis baseline, VI VT treatments in the initialand replication phases, and reinstatement ofbaseline reinforcement of problem behavior asimplemented by the two therapists. Meanresponses per min by phase and therapist arepresented in Table 1. The functional analysisbaseline showed clear differentiation betweenproblem behavior in the attention condition(Phase 1a) and the control condition (Phase1b)for both therapists, thus confirming that prob-lem behavior was maintained by therapistattention. The VI VT intervention producedsimilar reductions in problem behavior acrosstherapists in both treatment phases (M 2.7/min for Therapist 1 and 1.9/min for Therapist 2reduced from 8.3/min in the functionalanalysis...