JOURNAL OF APPLIED BEHAVIOR ANALYSIS
2016, 49, 900–914
NUMBER
4 (WINTER)
USE OF A LATENCY-BASED DEMAND ASSESSMENT TO IDENTIFY POTENTIAL DEMANDS FOR FUNCTIONAL ANALYSES NATHAN A. CALL MARCUS AUTISM CENTER, CHILDREN’S HEALTHCARE OF ATLANTA, AND EMORY UNIVERSITY SCHOOL OF MEDICINE
SARAH J. MILLER MARCUS AUTISM CENTER, CHILDREN’S HEALTHCARE OF ATLANTA, AND LOUISIANA STATE UNIVERSITY
JOSLYN CYNKUS MINTZ, JOANNA LOMAS MEVERS, AND MINDY C. SCHEITHAUER MARCUS AUTISM CENTER, CHILDREN’S HEALTHCARE OF ATLANTA, AND EMORY UNIVERSITY SCHOOL OF MEDICINE
JULIE E. ESHELMAN MARCUS AUTISM CENTER AND CHILDREN’S HEALTHCARE OF ATLANTA
AND
GRACIE A. BEAVERS GEORGIA STATE UNIVERSITY
Unlike potential tangible positive reinforcers, which are typically identified for inclusion in functional analyses empirically using preference assessments, demands are most often selected arbitrarily or based on caregiver report. The present study evaluated the use of a demand assessment with 12 participants who exhibited escape-maintained problem behavior. Participants were exposed to 10 demands, with aversiveness measured by average latency to the first instance of problem behavior. In subsequent functional analyses, results of a demand condition that included the demand with the shortest latency to problem behavior resulted in identification of an escape function for 11 of the participants. In contrast, a demand condition that included the demand with the longest latency resulted in identification of an escape function for only 5 participants. The implication of these findings is that for the remaining 7 participants, selection of the demand for the functional analysis without using the results of the demand assessment could have produced a false-negative finding. Key words: demands, functional analysis, negative reinforcement
Functional analysis (FA) methodology, which consists of the explicit manipulation of antecedent variables and consequences for a targeted behavior, has become the gold standard in the assessment and treatment of problem behavior (Hanley, Iwata, & McCord, 2003). Functional analysis has advanced the field by allowing researchers and practitioners Correspondence concerning this article should be addressed to Nathan Call, Marcus Autism Center, 1920 Briarcliff Rd., Atlanta, Georgia 30329 (e-mail: nathan.
[email protected]). doi: 10.1002/jaba.341
to identify the reinforcers that maintain problem behavior. This information can then be used to inform treatment selection for that behavior. Function-based treatments have been demonstrated to be more effective than treatments that are not based on function (Campbell, 2003; Heyvaert, Saenen, Campbell, Maes, & Onghena, 2014). Furthermore, there is evidence that FAs are more reliable and valid than indirect or descriptive assessments when identifying the function of problem behavior (Zarcone, Rodgers, Iwata, Rourke, & Dorsey, 1991).
900
DEMAND ASSESSMENT The superiority of FAs over other functional assessment methods could potentially be attributed to the procedures employed to maximize internal validity, which augment the relative influence of the environmental variables that are responsible for evoking and maintaining problem behavior and diminish the role of extraneous variables. An important aspect of these procedures involves control of the stimuli that are present in various conditions. Through systematic manipulation, researchers and clinicians can observe the unique influences of these stimuli on behavior. In particular, it has become commonplace to evaluate an individual’s preferences for tangible items included in control conditions or conditions designed to evaluate the role of positive reinforcement in maintaining problem behavior (Hanley et al., 2003). Although the preference assessment methodologies used to identify items that may be positive reinforcers for inclusion in an FA may vary, most involve dependent measures in the form of behaviors such as selection responses, interaction with items, or vocalizations in response to item presentation (e.g., Piazza, Fisher, Hagopian, Bowman, & Toole, 1996). Preference assessments are an improvement over simply interviewing caregivers for reports about an individual’s preferences or using a standard set of leisure items for participants because they utilize direct observation of behavior to inform selection (Fisher & Mazur, 1997; Northup, George, Jones, Broussard, & Vollmer, 1996; Piazza et al., 1996). An apparent inequity with respect to how potential positive and negative reinforcers have been identified for evaluation in FA test conditions is apparent in articles published in the Journal of Applied Behavior Analysis between 1982 and 2013. Ninety-three articles describe an FA that included at least one test condition designed to evaluate the role of preferred tangible items as reinforcers for problem behavior and provided details of how those items were identified. Of these articles, 69 (74%) reported
901
using some form of systematic and direct preference assessment for selecting stimuli to be included in those test conditions. The remaining articles reported the use of caregiver report to identify tangible items to be evaluated as potential positive reinforcers. In contrast, of the 81 articles in which experimenters conducted a test condition for a potential escape function and provided details of how specific demands were selected, only nine (11%) reported using a direct assessment of demands. The remaining articles that reported procedures for selection of demands identified them via indirect means (i.e., based on the participant’s individualized education plan or caregiver report). Thus, although it has become common practice to conduct preference assessments before an FA, direct assessments of potential negative reinforcers are not as common. This imbalance has potential significance because arbitrary selection of demands could result in a failure to identify an escape function for problem behavior when it actually exists (Kodak, Northup, & Kelley, 2007). For this reason, demand assessments have been developed to identify which demands serve as an establishing operation (EO) for escape for a given individual. For example, Roscoe, Rooker, Pence, and Longworth (2009) selected 12 tasks from each participant’s individualized education plan and continuously presented each for three 5-min sessions. The authors measured the rates of compliance and problem behavior for each demand. Subsequent FAs included two demand conditions: one with demands that produced higher levels of compliance, lower levels of problem behavior, or both (high-p condition) and a second with demands that produced lower levels of compliance, higher levels of problem behavior, or both (low-p condition). For three of four participants, differentiation between the control and escape conditions was evident only for the low-p condition, whereas function could not be determined for the high-p condition. These results indicated that, for some individuals, not
902
NATHAN A. CALL et al.
all demands serve equally as an EO for escape. Thus, assessment of demands before an FA may be necessary to evaluate a hypothesized escape function adequately. In a similar study, Call, Pabico, and Lomas (2009) evaluated a demand assessment that used latency to problem behavior as a measure of a demand’s aversive properties. Caregivers completed the Negative Reinforcement Rating Scale (NRRS; Zarcone, Crosland, Fisher, Worsdell, & Herman, 1999) to identify categories of demands that they considered often or always to bother their child. A clinician then helped the participants’ caregivers to identify 10 specific demands that fell in these categories. Each demand was then presented in an individual session that lasted up to 10 min using a three-step progressive prompting procedure, followed by two additional series of demand sessions. Participants received brief praise for compliance, followed by re-presentation of the demand. However, if problem behavior occurred at any point, the session was terminated. This assessment resulted in a hierarchy of demands according to their mean latencies to problem behavior across the three series. Demands that produced the longest average latency to problem behavior were considered less aversive (LA) demands, whereas demands that produced the shortest average latency to problem behavior were considered highly aversive (HA) demands. The experimenters then used the LA demands and HA demands in separate escape conditions in a subsequent FA that also included test conditions for positive reinforcement and a control condition. Results of the FAs showed that for one of the two participants, both demand conditions resulted in identification of an escape function, whereas for the other only the HA demand condition did so. For this second participant, the LA demand produced results that were undifferentiated from the control condition. This finding is important because the participant’s caregiver had previously identified the LA demand as
likely to evoke problem behavior. Thus, the common practice of selecting demands based on caregiver report would have resulted in a failure to identify an escape function for this participant if only the LA demand had been selected for inclusion in the FA. Despite the development and apparent utility of these demand assessments, neither appears regularly in the literature on FAs at this time, despite the fact that a condition to evaluate the role of escape from demands is typically included in FAs. In fact, reviews of FA outcomes have shown that escape from demands is one of the most common, if not the modal, function of some topographies of problem behavior (Hanley, Iwata, & McCord, 2003; Iwata, Pace, Cowdery, & Miltenberger, 1994). Given the findings of Roscoe et al. (2009) and Call et al. (2009), it is likely that the inclusion of demand assessments could increase the sensitivity of FA conditions designed to evaluate whether escape from demands maintains problem behavior. This increased sensitivity has the potential to demonstrate the presence of an escape function that may otherwise remain undetected when demands are selected for inclusion via less systematic means. Similarly, escape conditions that include demands selected via a direct assessment may demonstrate greater differentiation between test and control conditions. Thus, demand assessments have the potential to improve the efficiency of FAs, because increased differentiation could allow clinicians to detect an escape function in fewer sessions. Because inclusion of a demand assessment in the study by Call et al. (2009) produced differing results across the two demand conditions for only one participant, it remains unclear whether this result is typical. Thus, the purpose of the current study was to evaluate whether inclusion of a demand assessment to inform the selection of demands for the FA would increase the sensitivity of the FA to detect an escape function.
DEMAND ASSESSMENT METHOD Participants and Setting All of the individuals who attended a daytreatment program for the assessment and treatment of problem behavior over a 14month period, and whose caregivers provided informed consent, served as participants. Each individual participated in a demand assessment to identify demands for two demand conditions in a subsequent multielement FA. Individuals were included in the current study if the clinical team determined that their problem behavior was maintained by escape from demands based on one or both of the demand conditions (see below). A total of 14 participants met these criteria, out of a total of 26 participants who completed the demand assessments. Two of these 14 participants were removed from the current data set because no interobserver agreement data were collected during the demand assessment. Data from the remaining 12 participants were included in subsequent analyses (see Table 1 for details regarding these participants). Of these participants, 75% were male. Their average age was 11.7 years (range, 6 to 21), and 75% had been diagnosed with autism spectrum disorders. The most common targeted problem behavior was aggression (91.7% of participants), followed by disruptive behavior (66.7%) and self-injurious behavior (50%).
903
With one exception (Participant 12), all sessions were conducted in therapy rooms (4 m by 5 m) equipped with one-way observation panels that separated an adjoining observation room, which allowed unobtrusive observation. Therapy rooms contained materials required for each specific experimental condition (e.g., table, chairs, demand materials, preferred items). Eight to 16 sessions were conducted daily. For Participant 5, her FA began in a session room but shifted to a playroom. Response Measurement and Reliability The dependent variable for the demand assessment consisted of latency to problem behavior, defined as the time elapsed from the start of the session to the first occurrence of any of the problem behaviors that were targeted for intervention for that participant after the delivery of the demand (Call et al., 2009). During the FA, the rate of each participant’s targeted problem behavior served as the dependent variable. Problem behaviors were identified by each participant’s caregiver and then operationally defined individually (see Table 1 for the specific topographies of problem behavior exhibited by each participant). Observers collected data using desktop or laptop computers equipped with software that allowed collection of both frequency and duration measures. To determine interobserver
Table 1 Participant Demographics Participant 1 2 3 4 5 6 7 8 9 10 11 12
Gender
Age
Diagnosis
Topography of problem behavior
M M M M M F M M M F M F
11 8 6 18 9 9 10 10 8 21 15 15
ADHD, disruptive behavior disorder ASD ASD ASD ASD ASD ASD ASD ASD ASD Global developmental delay PDD
Aggression, disruptive behavior Aggression, disruptive behavior, self-injurious behavior Aggression, disruptive behavior, self-injurious behavior Self-injurious behavior, aggression Self-injurious behavior, aggression, disruptive behavior Negative vocalizations, Elopement Aggression, disruptive behavior Aggression, disruptive behavior, elopement Aggression, disruptive behavior, self-injurious behavior Aggression, spitting, dropping, screaming Self-injurious behavior, aggression Aggression, disruptive behavior, vomiting
904
NATHAN A. CALL et al.
agreement, a second observer simultaneously but independently recorded data during 28% (range, 10% to 88%) of demand assessment sessions and 38% (range, 17% to 53%) of FA sessions across participants. Observer records for both the demand assessment and FA sessions were compared by separating each session into 10-s intervals. Total agreement percentages were calculated by dividing the smaller number of responses recorded in each interval by the larger number of responses. These quotients were then averaged and converted to a percentage. During the demand assessment across all participants, the mean percentage agreement was 96% (range, 93% to 98%) for latency to problem behavior. During the FA, the mean percentage of agreement across participants was 97% (range, 91% to 99.9%). Procedure Indirect assessment. A caregiver completed the NRRS (Zarcone et al., 1999) for each participant. The NRRS is a brief questionnaire on which respondents rate several categories of potential negative reinforcers on a Likert-type scale. Potential responses range from 1 (does not bother child) to 4 (always bothers child). A therapist was available to answer caregiver questions. After completion of the NRRS, the therapist prompted the caregiver to identify 10 specific problematic demands that fell within categories they had rated 4. Demand assessment. Demands were assessed in sessions that lasted up to 10 min. Each session evaluated one of the 10 demands identified from the indirect assessment. All 10 demands were assessed in a randomized sequence before the second and third series of demand-assessment sessions were conducted. The sequence of demands was rerandomized for each series. All participants completed three series of sessions for each demand. For each session, the participant began seated at a table, on which the demand materials were
placed, or standing, if the demand required moving around the room (e.g., picking up litter and placing it in a receptacle). The therapist stood beside the participant and said, “[Participant’s name], it’s time to do some work.” The therapist then guided the participant to complete a demand using three-step, least-to-most prompting. That is, the therapist initially instructed the participant to complete the demand using a vocal prompt. If compliance did not occur, a model prompt was delivered in which the therapist demonstrated the target response. If compliance still did not occur, the therapist physically guided the participant to complete the demand. A 5-s interprompt interval gave the participant the opportunity to comply. The participant received brief praise for independent compliance, followed by representation of the demand. If physical guidance was required, another trial was presented immediately and the therapist did not deliver praise. This prompting procedure was repeated continuously for the duration of the session. If no targeted problem behavior occurred before the 10-min session limit was reached, the session for that demand was assigned a value of 600 s. If targeted problem behavior did occur, the session ended and the next session began immediately. The latency to the first instance of problem behavior was averaged across all sessions for each demand. Demands were then ranked in order of average latency to the first instance of problem behavior. The demand that produced the shortest average latency was labeled the highly aversive (HA) demand, and the demand that produced the longest average latency was labeled the less aversive (LA) demand. If two demands tied for shortest or longest latency, one of them was selected randomly for inclusion in the subsequent FA. Functional analysis. We conducted a modified FA based on procedures described by Iwata, Dorsey, Slifer, Bauman, and Richman (1982/1994) with each participant. All FAs were conducted using a multielement design
DEMAND ASSESSMENT and included at least attention, demand, and toy-play (control) conditions. For some but not all participants, the FA also included tangible and alone or ignore conditions. Additional conditions were included if indirect assessments or parent report indicated that problem behavior may have been maintained by tangible items or automatic reinforcement. Two demand conditions were included for each participant: a demand (HA) condition and a demand (LA) condition. Functional analysis sessions lasted 10 min, and condition order within each series was randomized. Before the FA, a pairedstimulus preference assessment (Fisher et al., 1992) was conducted with each participant to identify high-, moderate-, and low-preference items for inclusion in some of the FA conditions, as described below. Toy play (control). A therapist delivered continuous access to a moderately preferred item and attention. Any targeted or nontargeted problem behaviors that occurred during this condition were neutrally blocked and ignored. This condition served as the control condition for the FA. Attention. Immediately before the session, participants were given high-quality attention for approximately 2 min, after which the therapist moved away and read a magazine or otherwise ignored him or her. During the session, the participant was given a a low-preference item. Contingent on the occurrence of targeted problem behavior, the therapist provided the participant with attention for either 20 or 30 s (depending on the participant) in the form of vocal verbal disapproval (e.g., “I don’t like it when you hit me”). Tangible. Immediately before the session, the participant was given access to a highpreference item for approximately 2 min. At the commencement of the session, the therapist restricted access to the item. Contingent on the occurrence of targeted problem behavior, the therapist provided access to the high-preference item for 20 or 30 s (depending on the
905
participant). After the reinforcement interval had elapsed, the therapist again restricted access to the item. Ignore. A therapist was present in an otherwise empty room but ignored all targeted and nontargeted behavior. Alone. The participant remained alone in an empty therapy room. There were no programmed consequences for targeted or nontargeted problem behavior. Demand (LA). Participants were prompted to complete the LA demand. For each presentation of the demand, the therapist used the three-step, least-to-most prompting procedure to ensure that the participant completed the LA demand. After the occurrence of targeted problem behavior, the therapist made a statement such as, “Okay, you don’t have to [complete the demand],” and gave the participant a break from the LA demand for 20 or 30 s (depending on the participant). After the break, the LA demand was re-presented. Demand (HA). This condition was identical to the demand (LA) condition, with the exception that the therapist prompted the participant to complete the HA demand. RESULTS Figure 1 depicts the results of the demand assessment for all participants. The data for each participant are presented in order of shortest to longest average latencies for each demand but do not reflect the order in which each demand was presented. Overall, results varied across participants. Participants 4, 5, and 7 had relatively long latencies to problem behavior with little differentiation for the majority of demands. For these participants, at least eight of the 10 demands had average latencies greater than 400 s (6 min 40 s), and some demands never evoked problem behavior (range, 0 to 4 demands per participant). It appears that for these participants only a few demands were aversive enough to evoke problem behavior
NATHAN A. CALL et al.
906 600
01
02
03
04
05
07
08
09
10
Average Latency to Problem Behavior (seconds)
400
200
0 600
06
400
200
0 600
11
12
1 2 3 4 5 6 7 8 9 10 Demands
1 2 3 4 5 6 7 8 9 10 Demands
1 2 3 4 5 6 7 8 9 10 Demands
400
200
0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Demands Demands
Figure 1. Demand-assessment data for the 12 participants who were treated for an escape function.
within less than 5 min of presentation (e.g., Participant 4, Demand 1). Conversely, Participant 9 displayed relatively short latencies to problem behavior for the majority of the demands, with little differentiation between demands. For this participant, all demands resulted in latencies of less than 240 s (4 min). For the remaining eight participants, latencies to problem behavior varied across demands. For these participants, results ranged from very short latencies for some demands to the absence of problem behavior for others. A systematic hierarchy of latencies was found for some participants (e.g., Participant 6), with relatively continuous progression from shorter to longer. For others, latencies to problem behavior for the various demands were less hierarchical, with several demands that produced similar
latencies, whereas other demands produced shorter or longer latencies (e.g., Participant 7). Figures 2, 3, and 4 present the data from the control condition and both demand conditions from the FA for all 12 participants (see Supporting Information for results from all FA conditions). For five participants (Participants 1, 2, 3, 4, and 5), the FA resulted in clear differentiation between the rate of problem behavior in the toy-play condition and the demand (HA) condition but not the demand (LA) condition. That is, there was no overlap between the data paths of the toy-play and demand (HA) conditions, whereas problem behavior either did not occur in the demand (LA) condition (Participants 2, 3, and 4) or there was significant overlap between the toy-play and demand (LA) conditions (Participants 1 and 5). Results
DEMAND ASSESSMENT
907
2
1
1
0 10
2
20
30 2
Targeted Problem Behavior(s) per Minute
1
0 2
2
4
6
8
10
12
3
1
0
5
10
15
20
2
4
1
0 5 4
10
Session Room
15 Playroom
20 5
2
0 5
10 Toy Play
15
20 Session
Demand (HA)
25
30
35
40
Demand (LA)
Figure 2. Graphs of demand (HA), demand (LA), and toy-play conditions for Participants 1 through 5.
for Participants 6 and 7 showed a similar pattern, although the results were slightly less clear (i.e., there was some overlap between the demand (HA) condition and the other conditions). Thus, for these seven participants, the results of an FA that included only the demand from the demand (LA) condition might have failed to identify an escape function, whereas
including the demand from the demand (HA) condition likely would have done so. Participant 8 exhibited the opposite pattern from the other participants. That is, problem behavior occurred at a rate that showed differentiation between the toy-play and demand (LA) conditions but not the demand (HA) condition. Thus, the results of an FA that
NATHAN A. CALL et al.
908 12
6 8
4
0 10
6
20
30
Targeted Problem Behavior(s) per Minute
7
3
0 6
5
10
15
20 8
10
20
30
40 9
4
2
0 6
3
0 10
20
30
40
50
Sessions Toy Play
Demand (HA)
Demand (LA)
Figure 3. Graphs of demand (HA), demand (LA), and toy-play conditions for Participants 6 through 9.
included only the demand from the demand (HA) condition would likely have failed to identify an escape function, whereas including the demand from the demand (LA) condition most likely would have done so. Results for the remaining four participants (9, 10, 11, and 12) were less clear. For example, the demand (HA) condition conducted with Participant 9 exhibited a clear escape
function. However, even though problem behavior was observed less frequently and at lower rates in the demand (LA) condition, the fact that it occurred in three of eight sessions and never occurred in the toy-play condition could also suggest the possibility of an escape function. For Participant 10, an initial phase that included reinforcement of both negative
DEMAND ASSESSMENT 1.5
Elope, Drop, SIB, DIS, AGG
909
Drop, SIB, DIS, AGG
10
1
Targeted Problem Behavior(s) per Minute
0.5
0 30
20
60 11
10
0 24
30 AGG, Elope, SIB, DIS
60 AGG, Elope
12
12
0 15 Toy Play
Sessions Demand (HA)
30
45
Demand (LA)
Figure 4. Graphs of demand (HA), demand (LA), and toy-play conditions for Participants 10 through 12.
vocalizations and elopement resulted in clear differentiation between the rate of problem behavior in the toy-play and both demand conditions, although there was greater differentiation in the demand (LA) condition and a decreasing trend in the demand (HA) condition. However, all of the problem behavior observed in this phase consisted of elopement. Because negative vocalizations were one of the primary reasons for this participant’s referral, elopement was placed on extinction (Richman, Wacker, Asmus, Casey, & Andelman, 1999), and data in the second phase represents only negative vocalizations. In the second phase, negative vocalizations were
observed in five of the seven sessions of the demand (HA) condition and three of seven sessions of the demand (LA) condition. Thus, it could be concluded that the demand (LA) condition resulted in a relatively clear demonstration of an escape function for elopement but a less clear escape function for negative vocalizations. In contrast, the demand (HA) condition produced consistent negative vocalizations that were differentiated from the toy-play condition, suggesting an escape function. Elopement also occurred consistently in the first phase of this condition when it resulted in escape from demands, but the decreasing trend across sessions complicates the
NATHAN A. CALL et al.
910
interpretation. Therefore, for this participant it may be most reasonable to conclude that both demand conditions resulted in identification of at least one escape function, albeit for different behaviors. For Participants 11 and 12, the demand (HA) condition resulted in higher overall rates of problem behavior than the demand (LA) condition. However, Participant 11 engaged in problem behavior inconsistently across the sessions of the demand (HA) condition, whereas after he began to engage in problem behavior in the demand (LA) condition he did so in every session thereafter. For Participant 12, an initial phase of the FA included reinforcement for aggression, elopement, self-injurious behavior (SIB), and disruption, but, similar to Participant 10, the observed problem behavior consisted almost exclusively of SIB and disruption. Thus, a subsequent phase discontinued reinforcement for these behaviors, and aggression and elopement were observed more consistently in both demand conditions. It is important to note that for both of these participants the scale on the y axis extends to 20 and 24 responses per minute, respectively, to accommodate the elevated rates of problem behavior that occurred in the demand (HA) condition. This scale may obscure the differentiation that exists between the rate of problem behavior in the toy-play and demand (LA) conditions. Therefore, one could conclude that both the demand (HA) and demand (LA) conditions produced rates of problem behavior that were sufficiently differentiated from those observed in the toy-play condition to allow identification of an escape function. DISCUSSION We implemented a demand assessment in which the latency to the first instance of problem behavior served as a measure of aversiveness of tasks nominated by participants’ caregivers. Results of this demand assessment were then incorporated into two demand
conditions of a subsequent FA. Comparison of the rate of problem behavior in each demand condition to that observed in the toy-play condition resulted in identification of an escape function based on the results of the demand (HA) condition but not the demand (LA) condition for seven of the 12 participants. One participant exhibited the opposite pattern; that is, an escape function was identified based on the results of the demand (LA) but not the demand (HA) condition. The remaining four participants engaged in problem behavior that resulted in identification of an escape function based on results of both the demand (HA) and demand (LA) conditions. Thus, the demand condition that incorporated the task identified as highly aversive by the demand assessment was more likely to result in a determination that problem behavior served an escape function (91.6% of participants) than the one that included the task identified as least aversive (41.6% of participants). This difference in the results of the demand conditions was observed even though these conditions differed only in terms of the demands presented. Despite these overall findings, some participants’ results constituted a clearer demonstration of an escape function than others. However, with one exception (i.e., Participant 8), each participant engaged in more problem behavior in the demand (HA) condition than in the demand (LA) condition. That is, even when the results were equivocal or when both demand conditions resulted in identification of an escape function, there was a larger difference between the rate of problem behavior in the toy-play and the demand (HA) conditions than between the toy-play and the demand (LA) conditions. Therefore, even in cases in which an escape condition that includes a LA demand could allow an identification of an escape function, there may be advantages to using a more aversive demand in terms of efficiency. For example, it is possible that fewer sessions would be needed to reach a conclusion
DEMAND ASSESSMENT that problem behavior is maintained by escape from demands. This potential reduction in assessment duration, combined with other potential advantages outlined below, may outweigh the extra time required to complete a demand assessment. These results extend the findings of previous studies on the use of demand assessments (Call et al., 2009; Roscoe et al., 2009) by confirming that selection of demands for the escape condition of an FA arbitrarily, or based solely on indirect reports, increases the risk of failing to identify an escape function when one could otherwise be detected. Call et al. (2009) and Roscoe et al. (2009) presented two procedures that showed promise in reducing the probability of this result. The current study extended these procedures by applying them across a wide range of participants and referral concerns. Furthermore, the Call et al. study included only a single participant who demonstrated a pattern that suggests that use of the results of the demand assessment to select demands increased the sensitivity of the escape condition of the FA, whereas the current study demonstrated that such a finding is not unusual. That is, for eight of the 12 participants in this study (66.7%), the LA demand did not result in identification of an escape function in the subsequent FA, but the HA demand condition did. This finding is striking, given that the LA demands were identified as aversive using the same structured caregiver interview, a methodology that appears to be considered best practice in the field and is more stringent than was used in the majority of FAs in our review of the literature. Therefore, it seems reasonable to conclude that some of the more common FA practices regarding selection of demands have the potential to mask the presence of an escape function more readily than previously thought. In addition to demonstrating that task aversiveness may affect FA outcomes, the results of the current study showed substantial variability in the results of the demand assessment, even
911
within participants for whom problem behavior was maintained by escape from demands. For three participants, mean latencies to problem behavior were relatively long (≥400 s) for most of the demands, with an average of two demands never evoking problem behavior. Conversely, one participant displayed mean latencies to problem behavior that were relatively short (≤240 s) for all demands. Finally, responding for the remaining eight participants (66.7%) varied substantially across demands in terms of the average latency to problem behavior, suggesting a continuum of aversiveness. We investigated whether there were patterns in the demand-assessment results for all participants, including those for whom problem behavior was not determined to be maintained by escape. Although some patterns emerged (e.g., on average, participants whose problem behavior was later shown to be maintained by automatic reinforcement had the shortest average latency to problem behavior in the demand assessment across all demands), no results reached the level of statistical significance. However, future research examining patterns of demand-assessment results and their relation to results of FAs and treatment may be warranted. Preceding an FA with a demand assessment that discontinued demands contingent on problem behavior may have affected the subsequent FA results. For example, participants may have demonstrated more immediate discrimination of the contingencies in the escape condition of the FA as a result of the history of negative reinforcement in the demand assessment. Demand-assessment sessions were designed to terminate after the first instance of problem behavior to avoid placing problem behavior on extinction before conducting the FA. The influence of negatively reinforcing problem behavior during the demand assessment on the subsequent FA may have been mitigated by the fact that the termination of the demand being assessed in that session was generally followed by the immediate initiation
912
NATHAN A. CALL et al.
of the next randomly selected demand. However, the influence of preceding a FA with a demand assessment is not well understood at this point and may be another area for additional study. Similarly, all participants in this study simultaneously experienced an FA with two demand conditions, and it is unknown how these two conditions may have influenced one another. The results of this study raise questions about the relative efficiency of these procedures; the inclusion of a demand assessment certainly added to the overall assessment time and delayed initiation of the FA. On the other hand, as discussed above, greater differentiation between the escape and control conditions of an FA may decrease the number of sessions required to identify an escape function. Furthermore, mistakenly ruling out an escape function when a less aversive demand is selected for inclusion in an FA has the potential to introduce additional inefficiencies. For example, selection of an ineffective treatment, or failure to implement a potentially effective treatment, is likely to be costly in terms of time and resources. It might be possible to strike a balance between the potential advantages of including a demand assessment and the increased time required to do so. For example, future studies could evaluate whether three series of demand presentations are required in a demand assessment or whether fewer series would suffice to reliably identify HA and LA demands. If the number of series in the demand assessment could be decreased while maintaining accuracy, doing so would greatly increase the efficiency of this assessment. Similarly, decreasing the number of demands evaluated for a participant would also decrease the time required to complete the demand assessment. The extent to which these findings extend to FAs generally depends on the degree to which the specific FA procedures are similar to the format of FA evaluated in this study. For
example, it is not uncommon for the escape condition to include more than one demand, often in a rotating or randomized fashion. The fact that the demand conditions conducted in this study each repeatedly presented a single demand may have produced a stronger EO for escape than if there had been variability in the demand. However, inclusion of multiple demands could have confounded the results by making it unclear which demands served as a motivating operation for escape. Despite this rationale, the degree to which inclusion of HA demands in an FA based on results of a demand assessment influences FA results may differ for FAs that employ escape conditions with more than one demand. Similarly, the degree to which aversiveness, as measured by the demand assessment, affects FA results may differ for other formats of FA, such as the latency-based FA (Thomason-Sassi, Iwata, Neidert, & Roscoe, 2011). This study evaluated the use of demand-assessment results within a multielement FA, because it is more commonly used and well established in the literature (Hanley et al., 2003). However, future research is necessary to determine whether these findings apply to FAs that differ significantly in methodology from the format used here. Differentiating between individuals for whom only one or a few demands evoke problem behavior and those for whom a wide range of demands evoke problem behavior may have other clinical applications. For example, certain treatment approaches may be better suited to individuals for whom only one or two demands are sufficiently aversive to evoke escapemaintained problem behavior. For these individuals, treatments that establish tolerance of the specific demand or that focus on identifying the aversive dimension of the specific demands and modifying them to decrease the EO for escape may be ideal. For example, a demand assessment could be an important part of the process of identifying the reason why a demand is aversive, such as when escape behavior occurs
DEMAND ASSESSMENT because the demand requires a difficult skill or one that is not in the individual’s repertoire. However, such an approach may be less effective with individuals for whom escape from most or all demands maintains problem behavior. For these individuals, it may be necessary to teach generalized compliance across a wide range of demands. Such differential approaches to the treatment of escape maintained problem behavior could be made possible through the use of demand assessments such as the one presented here. In addition, treatments for escape-maintained problem behavior frequently include differential reinforcement (e.g., differential reinforcement of alternative behavior, DRA) procedures in which problem behavior is placed on extinction and compliance is reinforced on a fixed-ratio 1 schedule before the schedule of reinforcement is thinned. It may be possible to enhance this approach by integrating the findings of a demand assessment. Treatment could start by using DRA to replace problem behavior with compliance with a less aversive task, followed by more aversive tasks. This approach might be quicker or produce less problem behavior than starting with the most aversive demand. This approach could be useful for particularly dangerous or destructive escape-maintained problem behavior for which an extinction burst cannot be tolerated. Similarly, the results of a demand assessment could be useful in skillacquisition programming by allowing clinicians to yoke the quality or magnitude of reinforcement to the aversiveness of tasks, thereby maximizing efficiency in instruction while high rates of compliance and learning are maintained. In the current study, data were not collected on compliance with demands during the demand assessment. Future research could be conducted to evaluate whether any relation exists between compliance and the aversiveness of a demand as measured by the demand assessment. Given the findings from this study, future research that examines the use of demand assessments to
913
address these types of applications, in addition to their use in FAs, is warranted. REFERENCES Call, N. A., Pabico, R. S., & Lomas, J. E. (2009). Use of latency to problem behavior to evaluate demands for inclusion in functional analyses. Journal of Applied Behavior Analysis, 42, 723–728. doi: 10.1901/ jaba.2009.42-723 Campbell, J. M. (2003). Efficacy of behavioral interventions for reducing problem behavior in persons with autism: A quantitative synthesis of single-subject research. Research in Developmental Disabilities, 24, 120–138. doi: 10.1016/S0891-4222(03)00014-3 Fisher, W. W., & Mazur, J. E. (1997). Basic and applied research on choice responding. Journal of Applied Behavior Analysis, 30, 387–410. doi: 10.1901/ jaba.1997.30-387 Fisher, W., Piazza, C. C., Bowman, L. G., Hagopian, L. P., Owens, J. C., & Slevin, I. (1992). A comparison of two approaches for identifying reinforcers for persons with severe and profound disabilities. Journal of Applied Behavior Analysis, 25, 491–498. doi: 10.1901/jaba.1992.25-491 Hanley, G. P., Iwata, B. A., & McCord, B. E. (2003). Functional analysis of problem behavior: A review. Journal of Applied Behavior Analysis, 36, 147–185. doi: 10.1901/jaba.2003.36-147 Heyvaert, M., Saenen, L., Campbell, J. M., Maes, B., & Onghena, P. (2014). Efficacy of behavioral interventions for reducing problem behavior in persons with autism: An updated quantitative synthesis of singlesubject research. Research in Developmental Disabilities, 35, 2463–2476. doi: 10.1016/j. ridd.2014.06.017 Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1994). Toward a functional analysis of self-injury. Journal of Applied Behavior Analysis, 27, 197–209. doi: 10.190/ jaba.1994.27-197 (Reprinted from Analysis and Intervention in Developmental Disabilities, 2, 3–20, 1982) Iwata, B. A., Pace, G. M., Cowdery, G. E., & Miltenberger, R. G. (1994). What makes extinction work: An analysis of procedural form and function. Journal of Applied Behavior Analysis, 27, 131–144. doi: 10.1901/jaba.1994.27-131 Kodak, T., Northup, J., & Kelley, M. E. (2007). An evaluation of the types of attention that maintain problem behavior. Journal of Applied Behavior Analysis, 40, 167–171. doi: 10.1901/jaba.2007.43-06 Northup, J., George, T., Jones, K., Broussard, C., & Vollmer, T. R. (1996). A comparison of reinforcer assessment methods: The utility of verbal and pictorial choice procedures. Journal of Applied Behavior
914
NATHAN A. CALL et al.
Analysis, 29, 201–212. doi: 10.1901/ jaba.1996.29-201 Piazza, C. C., Fisher, W. W., Hagopian, L. P., Bowman, L. G., & Toole, L. (1996). Using a choice assessment to predict reinforcer effectiveness. Journal of Applied Behavior Analysis, 29, 1–9. doi: 10.1901/ jaba.1996.29-1 Richman, D. M., Wacker, D. P., Asmus, J. M., Casey, S. D., & Andelman, M. (1999). Further analysis of problem behavior in response class hierarchies. Journal of Applied Behavior Analysis, 32, 269–283. doi: 10.1901/jaba.1999.32-269 Roscoe, E. M., Rooker, G. W., Pence, S. T., & Longworth, L. J. (2009). Assessing the utility of a demand assessment for functional analysis. Journal of Applied Behavior Analysis, 42, 819–825. doi: 10.1901/jaba.2009.42-819 Thomason-Sassi, J. L., Iwata, B. A., Neidert, P. L., & Roscoe, E. M. (2011). Response latency as an index
of response strength during functional analyses of problem behavior. Journal of Applied Behavior Analysis, 44, 51–67. doi: 10.1901/jaba.2011.44-51 Zarcone, J. R., Crosland, K., Fisher, W. W., Worsdell, A. S., & Herman, K. (1999). A brief method for conducting a negative-reinforcement assessment. Research in Developmental Disabilities, 20, 107–124. doi: 10.1016/S0891-4222(98)00036-5 Zarcone, J. R., Rodgers, T. A., Iwata, B. A., Rourke, D. A., & Dorsey, M. F. (1991). Reliability analysis of the Motivation Assessment Scale: A failure to replicate. Research in Developmental Disabilities, 12, 349–360. doi: 10.1016/0891-4222 (91)90031-M Received May 27, 2015 Final acceptance January 20, 2016 Action Editor, Einar Ingvarsson
