<%BANNER%>

Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2010-08-31.

Permanent Link: http://ufdc.ufl.edu/UFE0022622/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2010-08-31.
Physical Description: Book
Language: english
Creator: Fritz, Jennifer
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Psychology -- Dissertations, Academic -- UF
Genre: Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Jennifer Fritz.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Iwata, Brian.
Electronic Access: INACCESSIBLE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022622:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022622/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2010-08-31.
Physical Description: Book
Language: english
Creator: Fritz, Jennifer
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Psychology -- Dissertations, Academic -- UF
Genre: Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Jennifer Fritz.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Iwata, Brian.
Electronic Access: INACCESSIBLE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022622:00001


This item has the following downloads:


Full Text





EXPERIMENTAL ANALYSIS OF PRECURSORS TO PROBLEM BEHAVIOR


By

JENNIFER N. FRITZ

















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2008


































2008 Jennifer N. Fritz


































To Patti, David, Heather, Helen, and Carl









ACKNOWLEDGMENTS

This research was supported in part by a grant from the Florida Agency on Persons with

Disabilities. I especially thank Jen Hammond and Sarah Bloom for their thoughtful comments

and suggestions throughout the course of the study, as well as Kathryn Jann, Zachariah Sims,

Ashley Greenwald, Barbara Tomlian, Lisa Smalheiser, and Alex Avelino for the many hours

they assisted with scoring videos and various other aspects of the project. I also thank Carrie

Dempsey, Natalie Rolider, and Charles Nowell for their assistance with overseeing subjects. I

sincerely thank Timothy Vollmer, Donald Stehouwer, and Stephen Smith for their guidance and

helpful suggestions on this and other projects. Finally, I would like to extend my deepest

appreciation to Brian Iwata for all of the time, dedication, and support he provided to this project

and to me throughout my career.










TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ..............................................................................................................4

LIST OF TABLES .................................................................. ................. ........7

LIST OF FIGURES ................................... ............ .. .............................8

ABSTRACT ............................ .......................................................9

CHAPTER

1 IN TR OD U CTION ................... .......................................... .. .. ..... ................

Functional A analysis of Problem B ehavior......................................................................... 11
Operant Response Classes .................. ................ ...... .... .................... 12
Problem Behavior as an Operant Class ........................................ .................. 14

2 STUDY 1: EMPIRICAL IDENTIFICATION OF PRECURSORS..................... ..........19

M ethod ................... ................... .................19.........
Subjects and Setting ............................................... ..... ... 19
Procedures ............................................ ...............19
Response M easurement and Reliability .............................. ............... 19
Caregiver Interview ................... .......... .. ................... ....................20
Precursor A ssessm ent ............... ................................................................ ................2 1
Probability Analyses.................. ................. .........24
Results and Discussion .................................................. .... ..............26
Analysis of Precursor Behaviors .............................................................. ............... 26
Comparison of Caregiver Report and Precursor Data...................................29

3 STUDY 2: FUNCTIONAL ANALYSIS OF PRECURSOR AND PROBLEM
BEHAVIORS .............................................................. 35

Method ...................................... ....................................35
Subjects and Setting .................................................................35
Response M easurement and Reliability .............................. ...............36
Procedures ................................................38
Results and Discussion .................................................. ........ 39

4 STUDY 3: PRECURSOR ASSESSMENT AS THE BASIS FOR INTERVENTION .........48

M ethod ...................................... 1............................
Subjects and Setting ..................................................51
Response M easurement and Reliability .............................. ............... 51










Procedures ............................................................53
Baseline ...................................................... ........53
Continuous N CR ................. .......... .. .............. 53
NCR Schedule Thinning Plus DRA............... .............. ................... 54
Results and Discussion .................................................... .. ..............54

5 DISCUSSION ......................................................... ................ ........64

LIST O F R EFEREN CE S ........................................................................79
















































6









LIST OF TABLES

Table page

2-1 Subject characteristics........................................................................................... ........74

2-2 Precursor checklist........................... ...... .............. 75

2-3 Probability analysis formulas..........................................76

2-4 Precursors reported by caregivers vs. assessment-identified precursors. .......................77









LIST OF FIGURES


Figure page

2-1 Precursor-assessment results for Liv, Chuck, Billy, and Amanda..................................31

2-2 Precursor-assessment results for Kelly, George, Amy, and Sammy ..............................32

2-3 Precursor-assessment results for Renee, Curtis, Gerald, and Adam............... ...............33

2-4 Precursor-assessment results for Donald, Leigh, Guy, and Kevin .................................34

3-1 Results of the independent functional analyses for Renee, Curtis, Gerald, and Adam .....45

3-2 Results of the independent functional analyses for Donald, Leigh, Guy, and Kevin........46

3-3 Proportional distribution of precursor responses observed during the precursor FA ........47

4-1 Results of the precursor FA for Amanda in Study 3.................. .... .................59

4-2 Treatment results for Amanda in Study 3 .............. ....................60

4-3 Results of the precursor FA for Sammy in Study 3 .....................................61

4-4 Treatment for behavior maintained by negative reinforcement for Sammy ......................62

4-5 Treatment for behavior maintained by positive reinforcement for Sammy .....................63









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

EXPERIMENTAL ANALYSIS OF PRECURSORS TO PROBLEM BEHAVIOR

By

Jennifer N. Fritz

August 2008

Chair: Brian A. Iwata
Major: Psychology

Standard functional analyses require the repeated observation of a target behavior to

determine behavioral function, but this method can prove problematic in the case of severe or

dangerous behaviors. Previous studies have shown, however, that individuals who engage in

problem behaviors sometimes engage in both mild and severe forms and that severe behaviors

are observed less frequently when reinforcement is delivered contingent upon the more mild

behaviors. Studies also have shown that functional analyses of mild behaviors that reportedly

precede severe behaviors can (a) be members of the same operant response class and (b) reduce

the number of severe topographies that are observed during the assessment. These mild

behaviors (i.e., precursors) are typically identified via caregiver verbal report or informal, direct

observations, but it is possible that precursors could exist even when they are not readily

identifiable. Therefore, we developed a checklist to identify precursors via videotaped trials in

Study 1, and results showed that the empirical method of identifying precursors successfully

identified at least 1 precursor for all 16 subjects. Separate functional analyses then were

conducted of precursor and severe problem behaviors for 8 subjects in Study 2, and

correspondence between outcomes was observed in 7 cases. Furthermore, few studies have

evaluated treatments for severe problem behavior based upon the results of precursor









assessments. Therefore, we evaluated a sequential treatment consisting of a dense schedule of

noncontingent reinforcement (NCR) followed by NCR schedule thinning plus differential

reinforcement of alternative behavior to reduce precursors, increase appropriate behavior, and

maintain low rates of severe behavior. Results showed that this treatment strategy was effective

for behaviors maintained by positive reinforcement and negative reinforcement.









CHAPTER 1
INTRODUCTION

Functional Analysis of Problem Behavior

Experimental approaches to the assessment of problem behaviors (such as self-injurious

behavior, aggression, property destruction, tantrums, stereotypy, etc.) have been reported in

isolated studies since the 1960s, primarily for the purpose of demonstrating that a particular

contingency could exacerbate problem behavior (e.g., Lovaas & Simmons, 1969). Since then,

several approaches to identifying sources of reinforcement that maintain problem behavior have

been developed and systematically evaluated, but the approach with the most empirical validity

is the functional or experimental analysis (see Iwata, Kahng, Wallace, & Lindberg, 2000, for a

recent review).

Iwata, Dorsey, Slifer, Bauman, and Richman (1982/1994) were the first to develop a

general experimental model for identifying which of several common sources of reinforcement

maintained a particular problem behavior self-injurious behavior or SIB. They created a series

of four conditions, three of which involved manipulation of antecedent and consequent events

that formed contingencies previously shown to maintain SIB, plus a control. The conditions used

in that study were: alone (antecedent event: austere environment; consequent event: none; test

for automatic reinforcement), social disapproval (antecedent event: no attention; consequent

event: statement of disapproval; test for positive reinforcement), academic demand (antecedent

event: tasks presented; consequent event: break from work; test for negative reinforcement, and

unstructured play (antecedent events: no demands, noncontingent attention, leisure materials

available; consequent event: none; control condition). Differential responding in the form of

higher rates of SIB in one of the three test conditions (or high rates across all conditions)

identified the source of reinforcement maintaining subjects' SIB.









The functional analysis (FA) approach described by Iwata et al. (1982/1994) has been

replicated in hundreds of studies and has become the standard method for assessing a wide range

of problem behaviors (Hanley, Iwata, & McCord, 2003). In addition, the results of such an

analysis can be used to design an intervention in which the reinforcement contingency is

manipulated to reduce problem behavior and also to increase appropriate, alternative behaviors

(Carr, Coriaty, & Dozier, 2000). An inherent limitation of all FA methods, however, is the

explicit arrangement of conditions that increase the frequency of potentially dangerous behavior.

Although such increases may be temporary and may present risks no greater than those already

posed by the problem behavior, strategies that minimize risk during assessment would be

beneficial to both researchers and clinicians. One promising approach is to assess mild behaviors

that are members of the same operant response class as the severe problem behavior.

Operant Response Classes

As developed by Skinner, the concept of the operant response class has far-reaching

implications for understanding the etiology and maintenance of complex human behavior. In its

current usage, the operant refers to a class (or variety) of responses that can differ

topographically but that are maintained by a common class of consequences reinforcerss)

(Catania, 1973; Dews, 1966; Skinner, 1953).

The formation of response classes is essential for performing a variety of complex

behaviors. The unifying principle of all of these behaviors is that they produce the same

outcome, but, in the specific instance, the behaviors can vary on a large number of topographical

dimensions. For example, one could exhibit any number of responses in order to obtain a snack.

The individual might first look in the pantry by sliding open the door, scan the shelves for

preferred snack foods, and select chips, which requires pulling the bag in opposing directions to

access the chips. If chips were not available, the individual might then check the refrigerator,









which involves pulling open the door (a topographically dissimilar response that produces a

similar outcome as the pantry example visual access to food items), scanning the shelves for

snacks, and selecting a container of fresh fruit. Opening the container involves very different

responses than opening a bag of chips; the lid must be pulled from the corner in a vertical

direction, whereas the chip bag required pulling both sides in a horizontal direction. The ultimate

result of opening both containers is the same in both cases, however tangible access to food

items.

An understanding of how these response classes develop is essential for ensuring that an

individual can function competently in various facets of daily life; thus, not surprisingly, the

development of operant response classes has received extensive attention in applied research. For

example, researchers have evaluated strategies for teaching various responses involved in

developing imitation skills (e.g., Peterson, 1968; Werts, Caldwell, & Wolery, 1996; Young,

Krantz, McClannahan, & Poulson, 1994), academic skills (e.g., Bonfiglio, Daly, Martens, Lin, &

Corsaut, 2004; Cuvo, Ashley, Marso, Zhang, & Fry, 1995; Rosenbaum & Breiling, 1976), social

skills (e.g., Barton & Ascione, 1979; Charlop & Milstein, 1989; Krantz & McClannahan, 1998;

Reeve, Reeve, Townsend, & Poulson, 2007), self-help skills (e.g., Day & Horner, 1989; Nutter

& Reid, 1978; Pierce & Schriebman, 1994), and various other socially important behaviors.

Problems arise, however, when one attempts to identify a particular response as a member

of a particular response class. Early on, Skinner acknowledged this problem and posited that

such an analysis cannot be "an act of arbitrary subdividing, and we cannot define the concepts of

stimulus and response quite as simply as 'parts of behavior and environment' without taking

account of the natural lines of fracture along which behavior and environment actually break"

(Skinner, 1938; p. 33). In other words, operant behavior must be analyzed over time to identify









the environmental events that determine its occurrence. Functional analyses are used for this

purpose during the assessment of problem behaviors, and behaviors (however dissimilar

topographically) that are maintained by the same source of reinforcement are identified as

members of a common response class.

Problem Behavior as an Operant Class

Within an operant response class, covariation among member responses has been

documented extensively in applied research with respect to adaptive and problem behaviors (e.g.,

Koegel & Covert, 1972; Parrish, Cataldo, Kolko, Neef, & Egel, 1986; Sprague & Horner, 1992).

Additionally, in some cases, an individual might allocate responding in such a way that members

of a given response class generally are exhibited in a hierarchical order; in other words, some

behaviors are more probable than other behaviors.

The identification of response hierarchies with respect to the assessment and treatment of

problem behavior has received increased attention in recent work (e.g., Borrero & Borrero, 2008;

Harding et al., 2001; Lalli et al., 1995; Smith & Churchill, 2002). Results of these studies

showed that mild behaviors sometimes occurred before severe problem behavior and that mild

and severe problem behaviors were members of the same response class. As previously

mentioned, standard FAs are sometimes contraindicated when the topography of problem

behavior poses risks due to its severity, and these studies showed that one promising assessment

approach to minimizing risk is to assess behaviors that predict occurrences of the target problem

behavior (precursors). This approach minimizes risk because if mild and severe problem

behaviors are members of the same operant response class, reinforcement contingencies arranged

for behaviors that occur before the severe behavior could result in reduced rates of severe

behaviors.









Such an analysis involves several steps: (a) identifying potential precursor responses, (b)

verifying that these responses do, in fact, predict the occurrence of the target, and (c) determining

whether precursors are members of the same response class as the target. Recent studies have

provided evidence for the validity of this type of analysis by focusing on one or more of these

steps in the assessment of severe problem behaviors.

After observing informally that an individual tended to engage in stereotypy (finger

waving) prior to eye poking, Hagopian, Paclawskyj, and Kuhn (2005) calculated several

conditional and unconditional probabilities to verify the correlation between the behaviors and

found that stereotypy actually was predictive of eye poking. They also examined cumulative

records of responding and observed a temporal contiguity between stereotypy and eye poking.

Thus, results of both analyses suggested that stereotypy was a precursor to eye poking. Borrero

and Borrero (2008) used similar procedures to identify response-response relations after

informally observing that two individuals tended to emit loud vocalizations before engaging in

severe problem behaviors (i.e., SIB, aggression, or property destruction). They conducted

observations in the subjects' classroom on the occurrence of both vocalizations and problem

behavior. Results of probability analyses showed that vocalizations and problem behavior were

highly correlated, and results of lag-sequential analyses showed that vocalizations were most

likely to occur immediately preceding an instance of problem behavior. Taken together, the

results of these studies show that correlational analyses, and conditional probability analyses in

particular, are useful for determining which behaviors in an individual's repertoire are predictive

of severe problem behaviors.

Studies also have demonstrated via experimental analyses that less severe behaviors can be

members of the same response class as more severe behaviors. For example, Lalli, Mace, Wohn,









and Livezey (1995) observed that the screams, aggression, and SIB exhibited by a young girl

appeared to be maintained by negative reinforcement and often occurred in a particular sequence.

Subsequently, escape from demands was provided contingent upon individual behaviors (SIB,

aggression, and screams, respectively) while the other 2 behaviors were placed on extinction.

They found that when the reinforcement contingency was placed on the last behavior in the

sequence (SIB), all behaviors tended to occur in a hierarchical order (i.e., screams, aggression,

then SIB). Conversely, when the reinforcement contingency was placed on behaviors occurring

earlier in the sequence (screams or aggression), behaviors that tended to occur later were

observed rarely. These data indicated that screaming (a relatively innocuous problem behavior)

predicted the occurrence of more severe problem behaviors (aggression and SIB) and that all

behaviors were members of the same response class.

In similar studies, Borrero and Borrero (2008); Richman, Wacker, Asmus, Casey, and

Andelman (1999); and Smith and Churchill (2002) determined through direct observation and/or

parental interviews that less severe problem behaviors apparently preceded the occurrence of the

most severe form of problem behavior. Richman et al. conducted a functional analysis in which

reinforcers were provided contingent upon all topographies of problem behavior and observed

higher rates of less severe problem behaviors and near-zero rates of more severe problem

behaviors. When the less severe problem behaviors were placed on extinction, increases were

observed in the mores severe problem behaviors, thus demonstrating that all behaviors were

maintained by the same source of reinforcement. Other studies have found similar effects when

extinction is applied to the most commonly occurring response during assessment (e.g., Harding

et al., 2001; Magee & Ellis, 2000). Smith and Churchill, and Borrero and Borrero, conducted









independent functional analyses of the precursor and target problem behaviors and found that the

precursor and target behaviors were, in fact, members of the same response class.

Results of these studies are important in demonstrating that less severe behaviors may

predict the occurrence of target problem behaviors and that programming reinforcement

contingencies for less severe behaviors might decrease the rate of the severe behaviors during

assessment. The extent to which individuals who engage in problem behaviors also exhibit

precursors is unknown, however. Therefore, the primary purpose of this study was to determine

whether precursor behaviors typically precede the occurrence of problem behaviors.

In addition, a limitation common to all studies was that no systematic method was used to

initially identify the precursor behaviors. Potential precursors were defined in those studies based

on caregiver verbal report or informal observations conducted by the experimenters prior to

assessment (Borrero & Borrero, 2008; Hagopian et al., 2005; Smith & Churchill; 2002);

however, no other systematic procedures for selecting precursors were described. These

anecdotal sources may provide useful information, but correlations between precursor and target

responses, to the extent that they exist, also should be readily observable and quantifiable. Also,

it is possible that precursors (a) might exist even when caregivers cannot identify them, (b) may

be different than those reported, or (c) are not readily detected during informal observations.

Thus, by using informal methods of precursor identification, it is possible that an important step

in the analysis of precursors could be based on inaccurate information or limited sampling of

client behavior.

Finally, numerous instances of the target problem behavior were observed before the

relation between precursor and target responses was determined, thereby making the procedure

difficult to use in situations for which it is ideally designed the assessment of severe problem









behavior. Therefore, a second purpose of this study was to evaluate a method for identifying

precursors that (a) was based solely on direct observation and (b) minimized the number of

occurrences of the target problem behavior required to identify the precursors. Conditional

probability analyses were used to determine which of several potential precursors were

predictive of target behavior (Study 1), and independent functional analyses of precursors and

target problem behaviors then were conducted to verify that all behaviors were members of the

same response class (Study 2). Finally, only one published study to date has developed an

intervention based upon the results of precursor analyses only (Najadowski, Wallace, Ellsworth,

MacAleese, & Cleveland, 2008). Thus, the third purpose of this study was to determine whether

(a) an effective intervention could be designed based upon the results of precursor analyses alone

and (b) the sequential introduction of noncontingent reinforcement (NCR) followed by NCR

schedule thinning plus differential reinforcement of alternative behavior (DRA) would be

effective in reducing precursors while maintaining low rates of severe problem behavior.









CHAPTER 2
STUDY 1: EMPIRICAL IDENTIFICATION OF PRECURSORS

Method

Subjects and Setting

Sixteen individuals diagnosed with developmental disabilities who engaged in problem

behavior participated in Study 1. Subject characteristics (age, diagnostic classification, and

definition of the target problem behavior) are listed in Table 2-1. All sessions were conducted in

an observation room at a day program for adults or in designated areas of a classroom at a special

education school.

Procedures

Study 1 was conducted in 3 phases: (a) caregiver interview, (b) structured observations to

identify precursors (precursor assessment), and (c) probability analysis to select precursors. Next,

a trial-based, precursor assessment was conducted, in which conditions known to evoke problem

behaviors were presented sequentially until 10 instances of the target behavior were observed. A

probability analysis then was used to select precursors. This analysis consisted of comparing

several conditional and unconditional probability calculations to determine which responses that

predicted the occurrence of the target behavior. Finally, results obtained from the empirical

precursor analysis were compared with precursors reported by caregivers to determine the degree

of correspondence between the two sets of data.

Response Measurement and Reliability

Because potential precursors were unknown prior to assessment, all trials were videotaped

and were scored later by two observers using a checklist. Responses were grouped

topographically in the checklist as: (a) vocalizations, (b) facial expressions, (c) postures, (d)

repetitive motor movements, (e) locomotion, (f) object manipulation, and (g) other problem









behaviors. Examples of possible response topographies were listed within each category, and

additional space was provided to allow observers to write in behaviors that were observed but

not included on the checklist (Table 2-2). All responses that were included in the topographical

definition of the target behavior or could be considered mild forms of the target (e.g., pushing the

therapist when the target was aggression) were excluded as potential precursors. The precursor

assessment was conducted in two phases: (a) potential precursor topographies were identified

and operationally defined and (b) potential precursors were scored as occurrence or

nonoccurrence in all assessment trials.

Interobserver agreement was assessed by having a second observer score the occurrence of

precursors and the target behaviors during all assessment trials. After scoring a trial, observers'

data records were compared. If scoring discrepancies were found, the observers discussed the

operational definitions, watched the video, and/or modified the operational definitions of

precursors, then rescored the trialss. This process was repeated until 100% interobserver

agreement was achieved for the occurrence of all precursors and targets during the precursor

assessment.

Caregiver Interview

Prior to the precursor assessment, an experimenter conducted an open-ended interview

with a caregiver for each subject in an attempt to identify potential precursors. Caregivers were

either a parent or teacher who had known the subject for at least 6 months and who had observed

instances of the target problem behavior. During the interview, the caregiver first was asked to

identify the subject's most severe class of problem behavior (SIB, aggression, or property

destruction), which was selected as the target problem behavior for assessment during

subsequent phases. If the target behavior was identified previously, caregivers were asked if they

had observed the occurrence of problem behavior and if they could identify situations in which









problem behavior was likely to occur. Caregivers then were asked if they had observed any

behaviors that tended to precede the target behavior (i.e., if they could identify any potential

precursors). The experimenter noted any responses that were mentioned and clarified any vague

descriptions. For example, if a caregiver reported that the individual "got upset" before engaging

in the target behavior, the experimenter asked the caregiver to describe "getting upset" in greater

detail in an attempt to identify observable responses that might function as precursors.

Precursor Assessment

The precursor assessment consisted of discrete trials in which antecedent conditions that

might serve as establishing operations (EOs; Michael, 1982) for the target behavior were

presented and were similar to the attention and demand conditions of a functional analysis (Iwata

et al., 1982/1994). If caregivers reported that the target behavior was likely to occur when

preferred items were removed or access to items was denied or if the experimenters observed

that the target behavior occurred under these conditions, a tangible condition also was included

in the assessment. Given that many problem behaviors are maintained by positive reinforcement

(access to attention or preferred items) or negative reinforcement (escape from demands),

presenting these conditions presumably increased the likelihood of observing the target behavior

in a relatively short period of time. If the target behavior was observed in a trial, the consequence

relevant to the antecedent condition (attention, escape, and/or access to leisure items) was

delivered. A trial was terminated following the occurrence of the target behavior or after 5 min in

which the target behavior did not occur (described in more detail below), whichever came first.

During attention trials, the therapist did not interact with the subject, unless the target

behavior was observed, at which time the therapist delivered a reprimand (e.g., "Don't do that,

you will hurt yourself.") and gentle physical contact. The therapist continued to interact with the

subject (e.g., rubbing the subject's back, talking about preferred topics, etc.) until the target









behavior was not observed for 30 s. Once the target behavior was not observed for 30 s or if the

target behavior was not observed in 5 min, a demand trial was conducted.

During demand trials, the therapist presented instructions to complete tasks appropriate to

the subject's functioning level. The therapist used a 3-step prompting procedure (vocal

instruction, model, physical guidance) but terminated the instructional sequence and moved

away from the subject contingent upon the first occurrence of the target behavior. The next trial

began once the target behavior was not observed for 30 s or if the target behavior was not

observed in 5 min. If a tangible condition was included in the assessment, it was conducted

following the demand trial. If a tangible condition was not included, another attention trial was

conducted.

During tangible trials, the therapist allowed the subject brief (1-2 min) access to preferred

items and then removed the items. Contingent upon the target behavior, the items were returned

to the subject. Once the target behavior was not observed for 30 s or if the target behavior was

not observed in 5 min, the toys again were removed, and another attention trial was conducted.

The assessment was considered complete after 10 instances of the target behavior were

observed, except for Adam and Amy. Only 7 trials with the target problem behavior were

included in Adam's precursor assessment due to an oversight. Amy engaged in the target

problem behavior during every one of the first 10 trials of the precursor assessment, thus

precluding some of the probability calculations. Therefore, 3 play trials were conducted in which

Amy had noncontingent access to preferred leisure and edible items, as well as the therapist's

attention. The target problem behavior was not observed during these play trials, and Amy's

precursor assessment was considered complete with 10 trials containing the target behavior and 3

play trials in which the target behavior was not observed. In general, the total duration of trials in









which the target behavior was not observed was approximately equal to or greater than the total

duration of trials in which the target behavior was observed. In addition, if the target behavior

was observed during a trial, the next trial was not conducted until the target behavior had not

occurred for 30 s to reduce the likelihood that the subject would engage in multiple, consecutive

instances of the target behavior. Presumably, it was unlikely that the subject could engage in

precursor behaviors while exhibiting a burst of target behaviors; thus, the requisite 30 s of the

absence of the target behavior increased the likelihood of observing potential precursor behaviors

during the assessment.

The above method differed from those used by Hagopian et al. (2005) and Borrero and

Borrero (2008), in which numerous instances of the target behavior were observed before the

relation between precursors and target behavior was established. For example, Borrero and

Borrero required a minimum of 45 target behaviors before the descriptive analysis was

considered complete. Similarly, 18 eye pokes were depicted in the cumulative records of the

Hagopian et al. study, which represented only 2 of 31 assessment sessions that were conducted.

By using a trial-based format and by restricting the number of target behaviors to 10 occurrences,

we hoped to have an adequate sample from which to identify precursors but to greatly limit the

frequency of problem behavior.

All trials were videotaped for subsequent data collection. When the assessment was

complete, two observers watched the videos and used the checklist to mark any potential

precursor topographies observed in trials in which the target problem behavior occurred. The

checklist contained examples of a wide range of possible behavioral topographies, as well as

space for observers to record behaviors that were not listed. Thus, all behaviors that occurred in

trials in which the target behavior was observed were scored to identify responses that had the









potential to predict the target behavior. Responses that occurred after the target behavior were

not scored, however.

The checklist contained examples of response topographies within 7 general categories of

behavior, including vocalizations, facial expressions, postures, locomotion, repetitive motor

movements, object manipulation, and other problem behaviors that differed from the target

behavior. The observers compared the topographies marked on each checklist and developed

operational definitions of all potential precursors.

Finally, two observers recorded the presence or absence of behaviors during all trials using

a binary scoring code (1 = occurrence of precursors and/or the target behavior within a trial, 0 =

nonoccurrence). Following each trial, the observers compared their data records. In the event of

any discrepancies, the observers watched the video together and discussed the observed

behaviors. The observers then rescored the trial until 100% agreement was attained for each

precursor and the target behavior.

Probability Analyses

The purpose of the probability analysis was to determine which behaviors predicted the

occurrence of the target behavior (i.e., precursors) in a quantitative manner. Several probabilities

were calculated based on all trials of the precursor assessment. The probability of the target

behavior given the precursor [p(T|P,)] was calculated by dividing the number of trials in which

that precursor and the target behavior were observed by the total number of trials in which that

precursor was observed. The probability of the target given the absence of the precursor

[p(T-|~P)] was calculated by dividing the number of trials in which the target behavior was

observed but the precursor was not by the total number of trials in which that precursor was not

observed. The unconditional probability of the target [p(T)] was calculated by dividing the









number of trials in which the target behavior was observed by the total number of trials in the

assessment.

Similar calculations were performed to determine probabilities for each of the precursors.

The probability of that precursor given the target behavior [p(P|,T)] was calculated by dividing

the number of trials in which both the precursor and target behavior occurred by the number of

trials in which the target behavior occurred. The probability of the precursor given the absence of

the target behavior [p(PI-T)] was calculated by dividing the number of trials in which the

precursor was observed but the target behavior was not by the total number of trials in which the

target behavior was not observed. Finally, the unconditional probability of the precursor [p(P,)]

was determined by dividing the number of trials in which the precursor was observed by the total

number of trials in the assessment. (Formulas for each of the probabilities are listed in Table 2-

3.)

The relative probability values for each response were compared to select the precursors.

First, the probability of the target behavior given each potential precursor was compared to (a)

the probability of the target behavior given the absence of each precursor and (b) the

unconditional probability of the target behavior. Next, the probability of each precursor given the

target behavior was compared to (a) the probability of each precursor given the absence of the

target behavior and (b) the unconditional probability of each precursor. Behaviors were selected

as precursors if they satisfied both of the following criteria. First, the probability of the target

behavior given the precursor was higher than the probability of the target behavior given the

absence of the precursor and the unconditional probability of the target behavior, orp(TIP,) >

p(T-~Pn) and p(T|Pn) >p(T). Second, the probability of the precursor given the target behavior

was higher than the probability of the precursor given the absence of the target behavior and the









unconditional probability of the precursor, orp(P|,T) >p(PnIT) and p(P|,T) >p(Pn). If numerous

potential precursors were observed, some response topographies were combined if the responses

(a) met the criteria for classification as either precursors or non-precursors and (b) could be

described succinctly based upon similar topographical features (e.g., "crawl", "run", and "climb"

were combined into "move around room" for Amy).

Results and Discussion

Analysis of Precursor Behaviors

Results of the precursor assessments are shown in Figures 2-1 through 2-4. The top panel

of each subject's graph shows the probability analysis for the target behavior. In these panels, the

dark, solid gray bars show the probability of the target behavior given that precursor, the striped

bars show the probability of the target behavior given the absence of that precursor, and the

horizontal line that bisects each bar shows the unconditional probability of the target behavior.

The bottom panel of each subject's graph shows the probability analysis for the potential

precursors. In these panels, the light, solid gray bars show the probability of that precursor given

the target behavior, the striped bars show the probability of that precursor given the absence of

the target behavior, and the small, horizontal lines that bisect each bar show the unconditional

probability of that precursor.

Figure 2-1 shows results for Liv, Billy, Chuck, and Amanda. Three precursors were

identified for Liv. All 3 of her precursors were highly correlated with the occurrence of target

problem behavior: Property destruction always occurred in trials in which the precursor was

observed (p(T|P) = 1.0), although the probability of the target behavior given the absence of the

precursors (p(T-|~P)) also was high. In addition, Liv's precursors never occurred in trials in which

the target behavior was not observed (i.e., p(P-|~T) = 0). Billy always engaged in the target

behavior in trials in which at least 1 of the 3 selected precursors also occurred. In addition, these









precursors never occurred in trials in which the target behavior was not observed. Chuck's

precursors were somewhat less predictive than Liv's and Billy's precursors in that the target

behavior was not always observed following the selected precursors, and his precursors were

sometimes observed in the absence of the target behavior. Three precursors were identified for

Amanda. Hand postures and stretching did not occur very often, but when they did they only

occurred in trials with the target behavior. Reaching for the therapist was not as predictive of the

target behavior, as the target was not always observed in trials with this response, and the

response sometimes occurred in trials without the target. Amanda's caregivers reported that she

seemed to engage in higher rates of SIB (the target behavior) in the presence of food. Therefore,

food items were included in the tangible condition, and signing for food also met the precursor

selection criteria.

Figure 2-2 shows results for Kelly, George, Amy, and Sammy. Three precursors were

selected for Kelly. She always engaged in the target behavior in trials in which she put her hand

inside her clothes, although this behavior was observed in few trials. Mouthing her fingers and

toes, as well as whining, were observed more frequently, and the target behavior occurred more

often in trials in which these responses were observed. She also frequently engaged in these

precursors in trials in which the target was observed compared to trials in which the target was

not observed. Six precursors were identified for George. Like Amanda, he also engaged in

appropriate behavior (i.e., signs such as "more," "play," etc.) in trials in which the target

behavior was observed. He did not, however, engage in the target behavior in all trials in which

any of the precursors were observed, and he sometimes engaged in each of the precursors in

trials in which the target was not observed. Ten precursors were identified for Amy. She always

engaged in the target behavior in trials in which 9 of these behaviors occurred, and she never









engaged in these precursors in trials in which the target behavior was not observed. Mouthing

objects also was highly predictive of the target behavior; however, the target behavior occurred

in trials in which this behavior was not observed and mouthing objects also occurred in trials in

which the target behavior was not observed. Six precursors were identified for Sammy. Tugging

on the therapist's shirt occurred in only 1 trial of the assessment in which the target also was

observed. His other precursors occurred more frequently, but the target behavior frequently

occurred in trials in which the precursor was not observed, and the precursors occurred in trials

in which the target behavior was not observed.

Figure 2-3 shows results for Renee, Curtis, Gerald, and Adam. Only 1 precursor was

selected for Renee, although this behavior was not highly predictive of the target behavior. In

other words, the precursor frequently occurred in trials in which the target behavior was not

observed, and the target behavior frequently occurred in trials in which the precursor was not

observed. Two precursors were selected for Curtis. Leg scratching only was observed during 1

trial of the precursor assessment, but it occurred during a trial in which aggression also was

observed; thus, the behavior met the precursor selection criteria. Blocking the therapist from

touching items also was selected, and it occurred more often overall, even though aggression

occurred during several trials in which blocking the therapist was not observed. Four precursors

were identified for Gerald. The target behavior always occurred in trials in which three of the

responses were observed, and those behaviors never occurred in trials without the target

behavior. The 4th precursor occurred more frequently; however, the target behavior sometimes

occurred in trials in which the behavior was not observed, and the precursor often occurred in

trials without the target behavior. Six precursors were identified for Adam, all of which were

strongly predictive of aggression. In other words, Adam always engaged in the target behavior in









trials in which the precursors were observed, and he never engaged in the precursors in trials in

which the target behavior was not observed.

Figure 2-4 shows results for Donald, Leigh, Guy, and Kevin. Six precursors were

identified for both Donald, although the target behavior only occurred in all trials in which 2 of

the precursors were observed, and those precursors never occurred in trials in which the target

behavior was not observed. Six precursors also were identified for Leigh; the target always

occurred in trials in which 4 of those behaviors were observed, and those precursors never

occurred in trials without the target behavior. Sixteen precursors were identified for Guy,

although the target behavior only always occurred in the same trials as 8 of these precursors, and

those same precursors never occurred in trials without the target behavior. The other precursor

occurred somewhat more frequently; however, those precursors were less predictive of the target

behavior. Twelve precursors were selected for Kevin, and the target always occurred in trials in

which these behaviors were observed. His precursors were never observed in trials in which the

target behavior was not observed.

In summary, results show that each of the 16 subjects engaged in behaviors that were

predictive of the occurrence of their target problem behaviors. The number of identified

precursors ranged from 1 precursor (Renee) to 16 precursors (Guy).

Comparison of Caregiver Report and Precursor Data

When the precursors reported by caregivers were compared to those identified by the

precursor assessment (Table 2-4), caregivers for 6 of 16 subjects (Billy, Chuck, Renee, Leigh,

Amanda, and Sammy) were unable to report any precursors whatsoever. Results for the

remaining 10 subjects showed that caregivers reported only 10 of the 90 precursors identified via

the precursor assessment (approximately 12%). Furthermore, caregivers of 7 subjects (George,

Amy, Curtis, Gerald, Adam, Donald, and Kelly) reported additional potential precursors that









differed from those identified via the precursor assessment. Even though it was possible for the

subject to engage in the caregiver-reported precursors during the precursor assessment, either the

behaviors were never observed (George, Amy, Curtis, Gerald, Adam, and Donald) or they

occurred but did not predict the occurrence of the target problem behavior based upon the results

of the probability analysis (Curtis, Gerald, and Kelly). Finally, the precursor assessment

identified precursors for all subjects that were not reported by caregivers, ranging from 1 (Renee)

to 16 (Guy). Therefore, results of this analysis suggest that caregivers are relatively inaccurate in

identifying precursors.












Billy


Spcn1p)

-pmn

o p(PIT)
X p(PI-T)
- p(P)


s -
g ?

1- b


Amanda
1.0- Chuck 1.0

0.8- 0.8-





0.2- 0.2-

in (n). n


POTENTIAL PRECURSORS


Figure 2-1. Precursor-assessment results for Liv, Chuck, Billy, and Amanda. The top and
bottom graphs for each subject show probabilities for target behavior and precursor
behaviors, respectively.


InlB0l1



























1.0

0.8

0.6

0.4-



0.0

U >


Sammy
1.0

0.8- *

0.6-


* p(TIP)
Q p(TI-P)
- p(T)

o p(PIT)
Sp(I-T)
- P(P)


......mM raMM


0,4-


1.0- 1.0

0.8- 0.8

0.6- 0.

0.4

0.2- 0.




Figure 2-2. Precursor-assessment results for Kelly George Amy and Sammy
nE j E >0






POTENTIAL PRECURSORS


Figure 2-2. Precursor-assessment results for Kelly, George, Amy, and Sammy


Kelly
1.0-

0.8-1 ~ fY ~ 0,8

0.6-ggSS~iS~~i~ 0.

OAI- ~ III~~~ 0.4-

0.2- h Y~ ~III IIIY 0.2-


00-i

























1.0-

0.8-


0.6




Oil


L p(TIP)

- p(4T)
-p( T)

I p(p-T)


BrY aba E "9 i
E


Adam










on


0.4


02- 0

Z a
U, -B 4J


Figure 2-3. Precursor-assessment results for Renee, Curtis, Gerald, and Adam


onI


POTENTIAL PRECURSORS


Renee


gi~s
x
a


h I MMR












Donald


10 *

0.6



02

on


Lcisgb
1I *

0-81111 fJ ^
"~I qjn MEl^^~a


I.0

0.8



0 5



I. .1


ii,-.-*
S i 0 E
rT~ i2*v
d~ UT. SO
Cl E


Kevin


GUY






04

.1 ,




r. S

0.6-





ili



P 3^ ^ P a-
0.r
C
tC t-.ny


0.6-



02 -

2
Z ,T -ll n l jn
< St^' -'
-0 :2 r i"|n|

C>
WX :2 &' f
I3 I |
CLir 13


POTENTIAL PRECURSORS


Figure 2-4. Precursor-assessment results for Donald, Leigh, Guy, and Kevin


Sp(TIP)



O p(FT)
S p(PI-T)


5 r -5i


C, -^
0v









CHAPTER 3
STUDY 2: FUNCTIONAL ANALYSIS OF PRECURSOR AND PROBLEM BEHAVIORS

Although the precursor assessment conducted in Study 1 identified behaviors that

predicted (i.e., were correlated with) the target problem behavior, the extent to which these

behaviors were members of the same response class remained unknown. Therefore, independent

functional analyses (FAs) were conducted for a subset of subjects from Study 1 to identify the

function of the identified precursors as well as the function of the target behavior. Unlike Smith

and Churchill (2002) and Borrero and Borrero (2008), we conducted the functional analysis of

the precursors (precursor FA) first to determine whether the function of the precursors matched

the function of the target behavior (i.e., the target behavior was observed at the highest rates in

the same condition of the target FA as precursors had been observed during the precursor FA)

and to perhaps limit the number of occurrences of the target behavior. If precursors and target

behaviors were determined to be members of the same response class, and if the majority of

subjects did not engage in the target behavior during the precursor FA, this information might

prove useful to clinicians during assessment and treatment of severe forms of problem behavior

by minimizing risk to the subject and/or therapist.

Method

Subjects and Setting

Eight individuals from Study 1 (Renee, Curtis, Gerald, Adam, Donald, Leigh, Guy, and

Kevin) participated in Study 2. Subjects were selected based on results of their precursor

assessments, which identified different numbers of precursors across subjects. Therefore, FAs

were conducted on a variety of response topographies identified as precursors for these subjects,

ranging from 1 (Renee) to 12 (Kevin) precursors in the initial assessments. All sessions were









conducted in an observation room at a day program for adults or in designated areas of a

classroom at a special education school.

Response Measurement and Reliability

One precursor, covering her eyes, was included in the precursor FA for Renee. Two

precursors (scratching his leg and blocking the therapist from moving) were included for Curtis.

Four precursors were included for Gerald (flicking his lips, grimacing, hitting the therapist, and

jerking his head). Although 6 precursors were identified for Adam, only 5 precursors (slouching,

saying "No," putting paper in his mouth, turning away from the therapist, and pushing materials

away) were included in his precursor FA. Grimacing was excluded because Adam frequently

turned his head such that observers were unable to view his facial expression. Six precursors

were included for Donald (hand flapping, pressing his hand to his mouth, snarling, clapping his

hands, vocalizing negatively, and moving toward objects), as well as for Leigh (covering her

eyes, resting her head, saying "Yeah yeah yeah," guiding the therapist around the room,

stomping or shuffling her feet, moving her hands in circles by her side, and motioning toward

objects). Numerous precursors were identified in Study 1 for Guy; however, only the 8

precursors (flopping on the ground, swing his body side to side, stomping his feet, biting objects,

throwing objects, pushing materials away, crumpling paper, and cursing) that most strongly

predicted the occurrence of the target behavior were included in his precursor FA. Finally,

twelve precursors (saying "No," grunting, dropping to or scooting on the floor, waving his arms

in the air, pulling the therapist's arm, saying "Good boy," holding his knees, smiling, fidgeting,

stacking chairs, head butting the therapist, knocking on the table) were included in Kevin's

precursor FA because all of the precursors were highly predictive of his target behavior.

Observers recorded the frequency of (a) target behaviors, (b) precursors, and (c) therapist

interactions and their delivery of consequences during continuous, 10-s intervals for each subject









using a handheld PDA. The target problem behavior was the most severe form of behavior

reported by caregivers during the initial interview and/or direct observation by the

experimenters. Operational definitions of the target problem behaviors for each subject are listed

in Table 2-1.

Interobserver agreement was assessed by having a second observer independently collect

data during at least 25% of sessions. Proportional agreement percentages were calculated for

each response by comparing the two observers' recorded frequencies for all responses in each

10-s interval. The smaller number of responses was divided by the larger number of responses in

each interval with a disagreement, the fractions were summed across all intervals, and the total

was added to the total number of agreement intervals in the session. The sum was divided by the

total number of intervals in the session and multiplied by 100% to yield reliability scores for

each measure. Mean reliability scores were as follows: Renee, 97.5% for target behaviors (range,

93.6% to 100%), 97.9% for precursors (range, 90.8% to 100%), and 96.5% for therapist

responses (range, 91% to 100%); Curtis, 99.4% for target behaviors (range, 92.5% to 100%),

99.3% for precursors (range, 93.3% to 100%), and 98.9% for therapist responses (range, 91.6%

to 100%); Gerald, 100% for target behaviors, 99.8% for precursors (range, 94.7% to 100%), and

97.5% for therapist responses (range, 83.3% to 100%); Adam, 99.9% for target behaviors (range,

99.2% to 100%), 99.4% for precursors (range, 89.3% to 100%), and 92.6% for therapist

responses (range, 76.4% to 100%); Donald, 99.8% for target behaviors (range, 98.3% to 100%),

99.6% for precursors (range, 91.7% to 100%), and 97.9% for therapist responses (range, 90% to

100%); Leigh, 98.7% for target behaviors (range, 90.3% to 100%), 98.7% for precursors (range,

88.6% to 100%), and 95.6% for therapist responses (range, 80% to 100%); Guy, 98.2% for target

behaviors (range, 86.1% to 100%), 99.9% for precursors (range, 98.3% to 100%), and 97.7% for









therapist responses (range, 91.7% to 100%); and Kevin, 97.6% for target behaviors (range, 90%

to 100%), 99.4% for precursors (range, 88.3% to 100%), and 92.7% for therapist responses

(range, 78.3% to 100%).

Procedures

Two independent FAs were conducted using procedures similar to those described by

Iwata et al. (1982/1994). During the precursor FA, consequences were delivered for the

occurrence of precursor behaviors but not for occurrences of the target problem behavior(s) (i.e.,

the target behavior was ignored). During the target FA, consequences were delivered for

occurrences of the target behavior(s) but not for occurrences of the precursor behaviors.

Attention, play, and demand conditions were included in all FAs. An alone or ignore condition

was not included if the target behavior was aggression, and a tangible condition was included if

caregivers indicated that the subject tended to engage in problem behavior when preferred items

were removed or access to preferred items was denied.

During attention sessions, the subject had access to 2-3 moderately preferred toys

identified via a paired-stimulus (Fisher et al., 1992) or a multiple-stimulus (DeLeon & Iwata,

1996) preference assessment. At the start of session, the therapist told the subject, "I have some

work to do, but you can play with these toys, if you'd like." The therapist then sat next to but did

not interact with the subject. Contingent upon each occurrence a precursor (in the precursor FA)

or target (in the target FA) behavior, the therapist delivered a brief reprimand (e.g., "Stop doing

that; that's not nice!") and gentle physical contact (e.g., placed a hand on the subject's arm).

During play sessions, the subject had access to 2-3 highly preferred toys (identified in the

preference assessment), and the therapist interacted with the subject at least every 30 s or any

time the subject initiated interaction. No consequences were delivered following occurrences of

either precursor or target behaviors.









During demand sessions, the therapist continuously presented learning trials appropriate to

the subject's functioning level using a 3-step prompting sequence and delivered praise following

compliance. Contingent upon each instance of a precursor (in the precursor FA) or target (in the

target FA) behavior, the therapist removed the work materials and provided a 30-s break from

the task.

If the target behavior was SIB or property destruction, an alone or ignore condition was

included in the FAs. During alone sessions, the subject was seated alone in a room without any

materials. If an ignore condition was conducted, the subject was seated in an area of the room

away from all other individuals, and no consequences were delivered contingent upon any

behaviors emitted by the subject.

If caregivers reported that problem behaviors occurred when preferred items were removed

or access to these items was denied, a tangible condition was included in the FAs. At the start of

the session, the therapist removed all toys and remained near the subject. If the subject initiated

interaction with the therapist during this condition, the therapist briefly responded to the subject

(e.g., quickly answered a question) then terminated interactions (e.g., "We can talk later.").

Contingent upon the occurrence of a precursor (in the precursor FA) or target (in the target FA)

behavior, the therapist provided access to the toys for 30 s.

Results and Discussion

Figure 3-1 shows results of the independent FAs for Renee, Curtis, Gerald, and Adam.

Renee engaged in higher rates of eye covering during the demand condition of the precursor FA

(aggression also was observed in this condition, although the rate of aggression was much lower

than the rate of eye covering except during one session). During the target FA (aggression),

aggression maintained only in the demand condition throughout the assessment. (Renee also

continued to engage in eye covering in the demand condition of the target FA, as well as during









the play condition.) These results indicated that Renee's eye covering and aggression were

maintained by negative reinforcement (i.e., escape from academic tasks).

Curtis engaged in higher rates of precursors in the tangible condition of the precursor FA

(zero instances of aggression were observed). During the target FA, Curtis engaged in higher

rates of aggression in the tangible condition (rates of precursors also were observed). Thus,

Curtis' precursors and aggression were maintained by positive reinforcement (i.e., access to

preferred leisure items).

Gerald engaged in higher rates of precursors in the demand condition of the precursor FA

(SIB was observed during two demand sessions). During the target FA, Gerald engaged in higher

rates of SIB (as well as precursors) in the demand condition. These results indicate that Gerald's

precursors and SIB were maintained by negative reinforcement (i.e., escape from academic

tasks).

Adam engaged in higher rates of precursors during the demand condition of the precursor

FA (aggression was never observed). During the target FA, Adam engaged in higher rates of

aggression (and precursors) in the demand condition. These results indicate that Adam's

precursors and aggression were maintained by negative reinforcement (i.e., escape from

academic tasks).

Figure 3-2 shows the results of the independent functional analyses for Donald, Leigh,

Guy, and Kevin. Donald engaged in higher rates of precursors in the tangible condition of the

precursor FA (aggression was never observed). During the target FA, Donald engaged in higher

rates of aggression (and precursors) in the tangible condition. These results indicate that

Donald's precursors and aggression were maintained by positive reinforcement (i.e., access to

preferred items).









Leigh engaged in higher rates of precursors in the tangible condition of the precursor FA (a

higher rate of aggression was observed in one attention session). During the target FA, Leigh

engaged in high rates of SIB in the tangible and demand conditions (her precursors also occurred

initially during these conditions, but did not maintain in the demand condition). Thus, results of

the precursor FA indicated that precursors were maintained by positive reinforcement (access to

tangible items), whereas results of the target FA indicated that SIB was maintained by both

positive reinforcement (access to tangible items) and negative reinforcement (escape from

academic tasks). In other words, the precursor FA was effective in identifying one of two sources

of reinforcement that maintained the target behavior.

Guy engaged in higher rates of precursors in the demand condition of the precursor FA

(somewhat lower rates of aggression also were observed in this condition). During the target FA,

Guy engaged in higher rates of aggression (and precursors) in the demand condition. These

results indicate that Guy's precursors and aggression were maintained by negative reinforcement

(i.e., escape from academic tasks).

Finally, Kevin engaged in higher rates of precursors in the demand condition of the

precursor FA (and even higher rates of property destruction). During the target FA, Kevin

engaged in higher rates of property destruction (and precursors) during the demand condition.

These results indicate that Kevin's precursors and property destruction were maintained by

negative reinforcement (i.e., escape from academic tasks); however, the precursor FA was not

effective in reducing rates of his target problem behavior.

Precursors were found to be members of the same response class as the target problem

behavior for 7 of the 8 subjects. The 8th subject's (Leigh's) precursors were maintained by 1 of 2

sources of reinforcement that also maintained her target problem behavior. In addition, the









precursor FA eliminated instances of the target problem behavior for 3 subjects (Curtis, Adam,

and Donald), and resulted in low rates of the target problem behavior for 4 subjects (Renee,

Gerald, Donald, and Leigh), but did not reduce rates of the target problem behavior for 1 subject

(Kevin). A within-session analysis of Kevin's data revealed that during demand sessions, Kevin

engaged in the target behavior before precursors prior to reinforcement delivery in 100%, 57%,

and 33% of cases, respectively. It is possible that a recent history of reinforcement for the target

behavior during the precursor assessment might have shifted response allocation toward the

target behavior, which extinguished over subsequent demand sessions. Therefore, although the

precursor assessment results indicated that some responses predicted the occurrence of the target

behavior, Kevin engaged in high rates of all behaviors, and his response allocation seemed to

shift depending on which topographies contacted the reinforcement contingency. In fact, during

the last 5 min of the third demand session of the precursor FA, Kevin simply said "No" and

received escape without engaging in other response topographies.

It is interesting to note that subjects did not engage in all of their selected precursors during

the precursor FA, except Renee, for whom only 1 precursor was identified. Therefore, the

function of each subject's precursors was determined based on a subset of responses selected as

precursors in Study 1. Proportions of observed precursors during the precursor FA for each

subject are shown in Figure 3-3. The function of Curtis' precursors was determined based solely

on occurrences of blocking the therapist. Gerald engaged in only 2 of 4 precursors (grimacing

and head movements). Adam engaged in 4 of 5 precursors (saying "No", turning away,

slouching, pushing materials away), and Donald engaged in only 2 of 6 precursors (negative

vocalizations and reaching for objects). Leigh engaged in 3 of 6 precursors (hand circles, chin

down, and stomping/shuffling her feet) during the precursor FA. Guy engaged in 6 of 8









precursors (biting objects, pushing materials away, cursing, body swinging, throwing objects,

and stomping), and Kevin engaged in 6 of 12 precursors (saying "No," grunting, arm waving,

table knocking, fidgeting with pants, and placing his hands on his knees).

The function of unobserved precursors remains unknown. It is possible that either (a) the

precursor assessment simply yielded a high number of false alarms or (b) these behaviors are

members of the same response class as the target problem behavior but were not observed

because other precursors contacted the reinforcement contingency and maintained. Some

precursors emerged in the same condition as the target behavior during the target FA for 4

subjects (Donald, Leigh, Guy, and Kevin). For example, Donald engaged in snarling during the

tangible condition of target FA, which could have emerged as a function of extinction for more

frequently occurring precursors. Three of Leigh's precursors (covering her eyes, resting her

head, and moving her hands in circles) occurred at higher rates only in the tangible condition

during the target FA. Guy began to flop on the floor during the demand condition of the target

FA, and Kevin began to drop to/scoot on the floor during the demand condition of the target FA.

Therefore, this provides some evidence that those precursor topographies might be maintained

by the same source of reinforcement as the target behavior and emerged as a function of

extinction of other precursors. Additional analyses would be required, however, to verify this

possibility, which was beyond the scope of the present study.

Further Comparison of Caregiver Report and Precursor Data: The comparison of

precursors reported by caregivers and those identified by the precursor assessment conducted in

Study 1 indicated that caregivers only reported 10 of 90 precursors identified by the precursor

assessment (approximately 12%). Given that many precursors identified by the precursor

assessment were not observed during the functional analyses in Study 2, it seemed possible that









caregivers might have identified precursors that were not observed during the precursor

assessment but that were observed in the functional analyses. Therefore, precursors reported by

caregivers and those observed during the functional analyses were compared for subjects in

Study 2 to determine the extent to which caregivers reported precursors that emerged in the same

conditions) as the target behavior during the functional analyses. Results showed that caregivers

reported 7 of 31 precursors that were observed during the functional analysis (approximately

23%). These results suggest that caregivers were only slightly more accurate in reporting

precursors than was concluded during the comparison conducted in Study 1.
































10 15 20 25 30


5 10 15
SESSIONS


20 25 io


61


4
3]
2


5 10 IS
SESSIONS


A-- **" d


20 25 30


Figure 3-1. Results of the independent functional analyses for Renee, Curtis, Gerald, and Adam

in Study 2. The top and bottom graphs for each subject show rates of the precursors

and target behaviors, respectively. The left panel of each set of graphs shows results

of the precursor FA; the right panel of each set of graphs shows results of the target

FA.


I-








I-


10 15 20 25

















121 Pumr FA


7
0


-' i



3:









25





0
a

a:




EUI


8




r


10
SESSIONS


PAcursor FA


TargeFA



Ltlft


8-










10 20 3X 40 50 60(
6-

4-


































5 10 15 20 S
7,
6-


4-







3-

1-


5 10 15 20 25
SESSIONS


Figure 3-2. Results of the independent functional analyses for Donald, Leigh, Guy, and Kevin


o" ^ 9 n ==T ___


3i


20


~P










1/1 1/2 2/4 4/5 2/6 3/6 6/8 6/12


S0.8


P 0.6


Z 0.4- :


O 0.2












number of selected precursors for each subject.
E -o -


SUBJECT

Figure 3-3. Proportional distribution of precursor responses observed during the precursor FA
(each section of a bar graph represents a different precursor) for subjects in Study 2.
Numbers above each bar show the number of precursors observed out of the total
number of selected precursors for each subject.









CHAPTER 4
STUDY 3: PRECURSOR ASSESSMENT AS THE BASIS FOR INTERVENTION

Given that the precursor assessment was effective in identifying precursors for all subjects

(Study 1) and that these behaviors typically were members of the same response class as the

severe problem behavior (Study 2), the combination of precursor assessment and precursor FA

seemed to be a promising basis for making conclusions about the function of severe problem

behavior while minimizing risk. This information presumably could then be used to design an

effective reinforcement-based intervention to specifically demonstrate reductions in precursors

while maintaining low rates of severe problem behavior.

Very few studies, however, have shown that an intervention aimed at reducing less severe

behaviors can maintain low rates of more severe behaviors. Results of a study by Shukla and

Albin (1996), however, showed that one individual engaged in mild and severe problem

behaviors during demand situations. During treatment, the participant was prompted to sign

"break" following an instance of mild problem behavior, which was effective in reducing rates of

all problem behaviors and increasing rates of communication. The procedure maintained a close

temporal contiguity between the mild behavior and reinforcer delivery, however, which could

result in adventitious reinforcement of mild behaviors for some individuals. In addition,

extinction was not programmed for the severe problem behavior, which might not be effective in

reducing rates of severe problem behaviors for others.

Only one published study to date has designed an intervention based upon the results of a

precursor FA. Najadowski et al. (2008) conducted a precursor FA for 3 subjects and observed

that all subjects' precursors were maintained by positive reinforcement (access to attention or

preferred items). Severe problem behavior was never observed during the assessment. The

intervention was similar to that used by Shukla and Albin (1996) and consisted of differential









reinforcement of alternative behavior (appropriate requests) and extinction for both precursors

and severe problem behavior. These procedures resulted in a reduction in precursors, an increase

in appropriate requests, and zero instances of severe problem behavior (except during one

session for one subject during a reversal to baseline). Therefore, although these results are very

promising with respect to reducing risks posed by severe problem behaviors, the severe problem

behavior either never occurred throughout assessment or treatment (2 subjects) or occurred in

one session only (1 subject), thus limiting the conclusions that can be drawn with respect to the

function of the severe problem behavior. In addition, subjects were prompted to engage in

appropriate requests following instances of precursors, which could prove problematic for some

individuals (as described above).

Another promising approach to decreasing problem behaviors, while strengthening

appropriate forms of communication, is a combination of noncontingent reinforcement (NCR)

and differential reinforcement of alternative behavior (DRA) procedures. Marcus and Vollmer

(1996) evaluated this treatment for 2 individuals and found that the initial, continuous NCR

schedule eliminated problem behaviors; however, appropriate communication (alternative

behavior) only emerged as the NCR schedule was thinned. Because dense NCR schedules of

reinforcement were implemented only briefly for both subjects, it was unclear if appropriate

communication might have emerged eventually under the dense NCR schedule. Therefore, Goh,

Iwata, and DeLeon (2000) evaluated the NCR plus DRA intervention and provided subjects with

extended exposure to the dense NCR schedule with DRA in place. They found that appropriate

communication did not emerge under the dense NCR schedule and only emerged when the NCR

schedule was thinned. Rates of problem behavior generally remained low in both studies under

the combined NCR plus DRA intervention. These results suggest an alternative treatment









strategy in which NCR might be used initially to suppress problem behaviors, and DRA could be

added to treatment during NCR schedule thinning to produce increases in appropriate, alternative

behaviors.

A review of 26 studies using NCR (9 studies) and/or DRA (18 studies) interventions was

conducted to determine the relative effectiveness of each intervention in quickly suppressing

problem behaviors. Studies were included in the analysis if: (a) NCR was delivered continuously

throughout sessions, (b) DRA was implemented using an FR-1 schedule of reinforcement, (c)

functional reinforcers were used during NCR or DRA, (d) no other treatment components were

included (e.g., response blocking, timeout, etc.), and (e) extinction was programmed for problem

behavior. Results revealed that continuous schedules of NCR were more effective than DRA in

eliminating problem behaviors during the first session of treatment compared to DRA. More

specifically, NCR resulted in zero instances of problem behavior during the first session in

67.57% of cases, whereas DRA resulted in zero instances of problem behavior during the first

session in only 35.94% of cases. Additionally, NCR was more effective than DRA in reducing

overall rates of problem behaviors during treatment: The mean reduction in problem behavior

under NCR was 91.76%, whereas the mean reduction under DRA was 81.16%. One noted

disadvantage of NCR is that the procedure does not specifically teach the individual an

appropriate means of obtaining reinforcers (Carr et al., 2000; Vollmer, Iwata, Zarcone, Smith &

Mazaleski, 1993); however, the results of this review suggest that dense schedules of NCR might

prove advantageous for suppressing severe problem behaviors before teaching an appropriate

form of communication.

The purpose of this Study 3 was to determine whether an effective treatment could be

designed based on the results of precursor assessments alone while (a) obtaining indirect









evidence regarding the function of severe problem behavior (i.e., the target behavior) and (b)

maintaining low rates of the target behavior throughout assessment and treatment. In general,

treatment consisted of a sequence beginning with continuous NCR, in which the reinforcer was

identified from a precursor FA. Subsequently, the NCR schedule was thinned, and DRA was

introduced for an alternative behavior that served the same function as the precursor behaviors.

Method

Subjects and Setting

Because the intervention strategy consisted of sequential introduction of social reinforcers

noncontingentt followed by contingent), only individuals whose precursor FA suggested that the

behaviors were maintained by social reinforcement were included in this study. Two individuals

from Study 1 (Amanda and Sammy) who engaged in severe problem behavior, who had not

participated in Study 2, and who were not participating in other projects participated in Study 3.

Three other individuals exhibited problem behavior (precursors and/or the target behavior) that

appeared to be maintained by automatic reinforcement and, thus, were not included. All sessions

were conducted in a classroom at a special education school.

Response Measurement and Reliability

Amanda's target behavior was SIB and was defined as hitting her face or head with her

hand or fist. Her precursors included reaching for the therapist, a hand posture (i.e., placing her

elbow on the table with her wrist bent), and stretching (i.e., leaning back in her chair with her

arms stretched above her head). The response selected to be strengthened as an appropriate,

alternative form of communication for reinforcement (i.e., mand) was signing "food," which was

already in her repertoire and was shown to be predictive of SIB during the precursor assessment

(Study 1).









Sammy's target behavior was aggression and was defined as hitting, kicking, biting, head

butting, or throwing objects that hit others. His precursors included tugging on the

experimenter's shirt, mouth movements (i.e., opening and closing his mouth without making

noise), climbing on furniture, running across the room, throwing or pushing furniture, and angry

vocalizations (i.e., growling or guttural sounds). The responses selected to be strengthened as

appropriate, alternative forms of communication for reinforcement (i.e., mands) were signing

"break" (during treatment for behavior maintained by negative reinforcement) or signing "play"

(during treatment for behavior maintained by positive reinforcement).

Data were collected and interobserver agreement was calculated as previously described

for the precursor assessment (described in Study 1) and precursor FA (Study 2). In addition, the

same data collection methods and interobserver agreement measures were used in treatment

conditions as in the precursor FA, and interobserver agreement was assessed during at least 25%

of sessions in each condition of the precursor FA and in each condition of treatment for both

subj ects.

During the precursor FA, mean interobserver agreement scores for Amanda were 99.5%

for precursors (range, 92% to 100%) and 99.2% for SIB (range, 93.6% to 100%). Mean

agreement scores for Sammy were 98.9% for precursors (range, 88.1% to 100%) and 98.2% for

aggression (range, 90.3% to 100%). During treatment, mean interobserver agreement scores for

Amanda were 98% for precursors (range, 80% to 100%), 98.4% for SIB (range, 92.5% to 100%),

and 99.1% for mands (range, 79.9% to 100%). During treatment for Sammy's behaviors

maintained by negative reinforcement, mean interobserver agreement scores were 98.2% for

precursors (range, 75% to 100%), 97.4% for aggression (range, 81.3% to 100%), and 96.6% for

mands (range, 85% to 100%). During treatment for Sammy's behaviors maintained by positive









reinforcement, mean interobserver agreement scores for Sammy were 99.4% for precursors

(range, 95% to 100%), 100% for aggression, and 98.3% for mands (range, 93.3% to 100%).

Procedures

Precursors identified in Study 1 for each subject were included in a precursor FA in which

consequences were provided following precursors only (i.e., no consequences were provided

following the target behavior). Conditions of the FA were the same as described in Study 2 and

consisted of attention, demand, play, and tangible. An ignore condition also was included for

Amanda to rule out the possibility that SIB was maintained by automatic reinforcement.

Treatment was based upon the results of the precursor FA only (i.e., the function of the

target behavior was inferred from response patterns during this assessment) and generally

consisted of baseline, continuous NCR, and NCR schedule thinning plus DRA. All sessions were

10 min in duration.

Baseline

Baseline was identical to the condition of the precursor FA in which the highest rates of

precursors were observed. These were the tangible (Amanda and Sammy) and demand (Sammy)

conditions. Consequences were delivered following precursors only (i.e., a small piece of food

for Amanda, 30-s escape from academic tasks for Sammy in the first treatment, or 30-s access to

toys for Sammy in the second treatment). No consequences were delivered following the target

behavior or appropriate communication.

Continuous NCR

The reinforcer shown to maintain precursors during the precursor FA was delivered freely

and noncontingently throughout each session. No consequences were delivered following

precursors, the target behavior, or appropriate communication in this condition. During treatment

for behaviors maintained by positive reinforcement, subjects had continuous access to highly









preferred food (Amanda) or leisure items (Sammy). During treatment for behaviors maintained

by negative reinforcement, no demands were placed on Sammy throughout session.

NCR Schedule Thinning Plus DRA

Once low, stable rates of precursors and the target behavior were observed under

continuous NCR, the NCR schedule was thinned by removing one 10-s interval of NCR per min

using procedures similar to those described by Goh et al. When precursors were observed at rates

less than 80% of baseline rates and rates of the target behavior were low, schedule thinning

progressed by removing another 10-s interval of NCR per min. At the start of this condition,

DRA was implemented during intervals in which NCR was not provided by physically

prompting the participant to engage in appropriate communication. The prompts then were

systematically delayed (i.e., the therapist waited a few additional seconds before prompting

appropriate communication) to allow the subject to engage in appropriate communication

independently. Prompts were removed when the subject began to exhibit independent

appropriate communication consistently. Contingent upon appropriate communication

(independent or prompted) the therapist delivered the reinforcer. No consequences were

delivered following the target behavior. Additional treatment components were implemented as

needed and are described in the results section for each subject.

Results and Discussion

Results of the precursor FA showed that Amanda's precursors were maintained by positive

reinforcement in the form of access to preferred food items (Figure 4-1). Amanda also engaged

in SIB at increasing rates in the tangible condition, even though no consequences were provided

for this behavior. These results suggested that her target problem behavior was likely maintained

by the same source of reinforcement as precursors.









During treatment (Figure 4-2), Amanda engaged in moderate rates of precursors during

baseline (mean, 3.1 rpm). She also engaged in moderate rates of SIB (mean, 1.1 rpm) and low

rates of mands (mean, 0.2 rpm), even though no consequences were provided for these

behaviors. When continuous NCR was implemented, Amanda did not exhibit any precursor

behaviors, and rates of SIB and mands were low (means, 0.1 and 0.3 rpm, respectively). During

NCR schedule thinning plus DRA, Amanda engaged in variable and increasing rates of

precursors (mean, 1.1 rpm), SIB (mean, 0.2 rpm), and independent mands (mean, 1.6 rpm). It

appeared that a response hierarchy was developing in which Amanda engaged in precursors (and

sometimes SIB) followed shortly by independent mands. Therefore, it seemed possible that

precursors and SIB might have been adventitiously reinforced as a result of the close temporal

contiguity between those behaviors and the delivery of reinforcement for mands. Other factors

might have accounted for the development of this particular response hierarchy, including the

presumably low effort necessary to engage in precursors (primarily reaching toward the

therapist), a recent history of reinforcement for precursors during baseline, and possibly a recent

history of reinforcement for SIB outside of the experimental setting. Therefore, response

blocking was added to the treatment to disrupt the development of a response hierarchy, and

blocked responses were scored and included in the session rate. This intervention resulted in

decreasing rates of precursors (mean, 0.7 rpm), near-zero rates of SIB (mean, 0.03 rpm), and

increased rates of independent mands (mean, 3.3 rpm). Next, a reversal to baseline was

conducted and resulted in increased rates of precursors (mean, 3.0 rpm), low rates of SIB (mean,

0.3 rpm), and variable rates of mands (mean, 2.0 rpm). A return to NCR thinning plus DRA and

response blocking resulted in decreasing rates of precursors (mean, 0.4 rpm), zero rates of SIB,

and high rates of independent mands (mean, 4.0 rpm). By the end of this phase, Amanda was









engaging primarily in independent mands, which seemed to preclude the need for NCR and

response blocking. Therefore, DRA alone was evaluated and resulted in low rates of precursors

(mean, 0.1 rpm), low rates of SIB (mean, 0.2 rpm), and similar rates of independent mands as in

the preceding phase (mean, 4.1 rpm).

Results of the precursor FA showed that Sammy's precursors (Figure 4-3) were maintained

by both positive reinforcement (access to preferred leisure items) and negative reinforcement

(escape from demands). In addition, Sammy engaged in increasing rates of aggression in the

demand condition, and he only engaged in aggression during the first session of the tangible

condition. These results suggested that the identified function of precursors likely matched the

function of his target problem behavior.

Two treatments were evaluated for Sammy: the first treatment targeted problem behaviors

maintained by negative reinforcement (Figure 4-4), and the second treatment targeted problem

behaviors maintained by positive reinforcement (Figure 4-5). During treatment for problem

behaviors maintained by negative reinforcement, Sammy engaged in moderate rates of

precursors during baseline (mean, 3.1 rpm), low rates of aggression (mean, 0.4 rpm), and low

rates of independent mands (mean, 0.1 rpm). When continuous NCR was implemented, Sammy

engaged in low rates of precursors (mean, 0.3 rpm), near-zero rates of aggression (mean, 0.03

rpm), and zero independent mands. During NCR schedule thinning plus DRA, he engaged in

variable rates of precursors (mean, 1.6 rpm), variable, increasing rates of aggression (mean, 0.8

rpm), and increasing rates of independent mands (mean, 1.2 rpm). Like Amanda, it seemed that

Sammy was exhibiting a response hierarchy in which he engaged in precursors as the

experimenter approached to deliver a demand and, when escape was not provided for the

precursors, he engaged in an independent mand (or sometimes aggression). Therefore, a type of









change-over delay was added to treatment in which independent mands were prevented

immediately following a precursor behavior, and Sammy was physically guided to complete the

demand issued by the experimenter. When he had not engaged in a precursor behavior for 5 s,

Sammy was permitted to mand for escape and a 30-s break was provided at that time. This

resulted in an initial burst in precursors, which decreased over subsequent sessions (mean, 1.7

rpm), decreasing rates of aggression (mean, 0.5 rpm), and steady rates of independent mands

(mean, 1.6 rpm). A reversal to baseline then was conducted, and rates of precursors increased

(mean, 2.2 rpm), rates of aggression were near zero (mean, 0.1 rpm), and low rates of mands

were observed (mean, 0.5 rpm). NCR thinning plus DRA with the change-over delay again was

implemented and resulted in decreasing rates of precursors (mean, 0.7 rpm), low rates of

aggression (mean, 0.3 rpm), and increased rates of mands (mean, 1.5 rpm). By the end of the

condition, NCR seemed unnecessary and, thus, was removed in the final phase. DRA plus the

change-over delay alone resulted in decreasing rates of precursors to near zero (mean, 0.5 rpm),

near-zero rates of aggression (mean, 0.3 rpm), and steady rates of independent mands (mean, 1.6

rpm).

During treatment for problem behaviors maintained by positive reinforcement (access to

preferred leisure items), Sammy engaged in moderate rates of precursors during baseline (mean,

2.1 rpm), near-zero rates of aggression (mean, 0.03 rpm), and zero independent mands. During

continuous NCR, Sammy engaged in near-zero rates of precursors (mean, 0.2 rpm) and zero

instances of aggression and mands. When NCR schedule thinning plus DRA was implemented,

Sammy engaged in decreasing rates of precursors (mean, 0.3 rpm), near-zero rates of aggression

(mean, 0.01 rpm), and increasing rates of independent mands (mean, 1.3 rpm). A return to

baseline resulted in increasing rates of precursors (mean, 1.6 rpm), near-zero rates of aggression









(mean, 0.03 rpm), and decreasing rates of mands (mean, 0.6 rpm). When NCR schedule thinning

plus DRA was again implemented, precursors decreased (mean, 0.8 rpm), aggression occurred at

higher rates during the first session but remained at zero for all subsequent sessions (mean, 0.2

rpm), and moderate rates of mands were observed (mean, 1.4 rpm). The NCR component was

removed in the final phase, and Sammy engaged in decreasing rates of precursors (mean, 0.3

rpm), zero rates of aggression, and increasing rates of mands (mean, 1.5 rpm).














Amanda


,
O 3-








2 -


7~L U...


5 10


SESSIONS


Results of the precursor FA for Amanda in Study 3.


Attention
Ignore

i i-


Tangible




Play
Demand


Figure 4-1.













BL NCR


NCR THINNING
+ DRA


NCR THINNING +
DRA NCR THINNING +
+ BLOCKING BL DRA + BLOCKING


Treatment results for Amanda in Study 3.


DRA


5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

SESSIONS


Figure 4-2.












5-



3-

2-




0-
5-




2-


01

0-






Figure 4-3.


SESSIONS


Results of the precursor FA for Sammy in Study 3.











NCR
I I NCRThinning+ DRA


7-

6-
Precui
5-

4-

3-
0





04
8-

7-

6-

S5-
Z
O
S4-

3-

< 2-



0-1
5




Figure 4-4.


SESSIONS

Treatment for behavior maintained by negative reinforcement for Sammy.









NCR Thinning+
DRA


- -- .. 1.-


NCR Thinning+
DRA DRA





Crnrn


\ Sammy-


0*-*-*-**-*-*


SESSIONS


Figure 4-5. Treatment for behavior maintained by positive reinforcement for Sammy.


NCR


BL





Precursors


Mands


AGG

U.


:L.









CHAPTER 5
DISCUSSION

The present studies examined the relation between precursor and problem behavior in three

stages: empirical identification and selection of precursor responses (Study 1), response-class

analysis of precursor and problem behavior (Study 2), and evaluation of treatment based on the

functional analysis of precursor behavior (Study 3). Taken together, results indicated that

precursor behaviors are both common and readily identifiable, that they often are members of the

same class as problem behavior, and that they may be used as the basis for developing effective

interventions.

Study 1 evaluated an objective yet brief method for identifying precursors to problem

behavior, and results of the assessment indicated that all 16 subjects engaged in at least one

precursor response. In addition, the assessment required very few instances (10 or fewer) of the

severe problem behavior to identify precursors; thus, the trial-based precursor assessment seems

to be a viable method of assessing severe problem behavior while minimizing the risks posed by

dangerous topographies. The fact that problem behavior often is preceded by precursors suggests

that problem behavior may simply be the terminal response in a hierarchy that begins with

mildly annoying, disruptive behaviors (e.g., negative vocalizations, pushing materials away, etc.)

or appropriate behaviors that are not reinforced (e.g., saying "No" or signing "food"). If so, it is

surprising that caregivers rarely were able to identify precursors. In fact, caregivers for only 6

subjects were able to report potential precursors, and the reported precursors matched the

behaviors identified by the precursor assessment in approximately 12% of cases. When the

precursors reported by caregivers were compared to those actually observed during the

functional analyses for subjects in Study 2, correspondence only increased to 23% of cases. It is

possible that caregivers are not as attentive when problem behaviors are not occurring and miss









the occurrence of precursors. These behaviors still might become members of the same response

class as problem behavior if precursors are followed quickly by problem behavior. Therefore,

these behaviors might become members of the same response class as a result of the close

temporal contiguity between the precursors and problem behavior (Catania, 1971).

Correspondence between caregiver report and results of the precursor assessment might have

been higher, however, if caregivers had been given access to the precursor checklist during the

interview. The presence of example response topographies might have prompted caregivers to

report potential precursors in cases in which none were reported or additional response

topographies that were not reported independently.

There are, however, some limitations of using a trial-based method to identify precursor

behaviors. First, the procedure may not be practical for very low-rate problem behavior, because

the duration of trials may be too brief to evoke the target problem behavior. Wallace and Iwata

(1999) compared results from functional analysis sessions based on 5-, 10-, and 15-min durations

and found that some individuals did not engage in the target problem behavior until session

conditions had been in effect longer than 5 min. Therefore, the 5-min trials of the precursor

assessment might not prove to be a useful assessment method for some individuals' problem

behavior. In fact, precursor assessments could not be completed for 3 potential subjects because

their problem behavior was seen rarely; they subsequently were assessed during functional

analyses with extended session durations. This methodology also might not be useful for very

high-rate problem behavior, because short inter-response times for the target behavior would

reduce the likelihood of observing other behaviors that could be identified as precursors. This

problem was encountered with Amy and required the inclusion of play trials to provided periods

of time in which the target behavior was not observed in order to calculate some probabilities









calculations (e.g., the probability of the target behavior given the absence of the precursor and

the probability of the precursor given the absence of the target behavior).

Results of Study 2 verified that responses selected from the precursor assessment were

members of the same response class as the target problem behavior for 7 of 8 subjects (precursor

and target behaviors matched for one of two functions for the 8th subject). In addition, the

precursor FAs eliminated the occurrence of the severe problem behavior for 3 subjects and

reduced rates of the severe behavior for 4 other subjects. Taken together, these findings are

significant in validating a combined precursor assessment and precursor functional analysis as a

viable method for identifying contingencies that maintain severe problem behavior.

Because not all precursors actually were observed for each subject during the precursor

FA, it is unclear whether the unobserved precursors were members of the same response class as

the target behavior. One possibility is that subjects simply allocated responding toward

topographies that initially contacted the reinforcement contingency, whereas the other precursors

extinguished but were members of the same response class as the target behavior. Indirect

evidence of this can be seen in cases in which previously unobserved precursors emerged in the

same condition as the target problem behavior during the target FA. Previous research has

demonstrated this type of finding (e.g., Magee & Ellis, 2000; Richman, Wacker, Asmus, Casey,

& Andelman, 1999), in which placing the most frequently observed response topographies on

extinction resulted in increased rates of other topographies of problem behavior, and similar

effects have been shown with respect to increases in adaptive behaviors (Grow, Kelley, Roane,

& Shillingsburg, 2008). More specifically, Grow et al. showed that appropriate but infrequent

forms of communication might emerge when problem behaviors are placed on extinction,

although the extinction bursts observed using this method of alternative response selection might









preclude its use with severe problem behavior. Given the results of these studies, it is possible

that the selective extinction of observed precursors might have clarified the results of the current

study; however, because the function of observed precursors matched the function of the target

behavior in almost all cases, additional analyses seemed unnecessary given the purpose of this

study.

Another potentially influential variable in determining subjects' response allocation toward

particular precursor topographies is the relative response effort required to emit some

topographies compared to others. In fact, many of the identified precursors did not appear to

require much effort (e.g., negative vocalizations), which incidentally seem to be identified as

precursors frequently in previous research (Borrero & Borrero, 2008; Lalli et al., 1995;

Najadowski et al., 2008; Smith & Churchill, 2002). No specific procedures were used to

determine the relative effort required for precursors or the target problem behavior in this study,

however, although previous research has shown that effort can influence response allocation

toward adaptive and problem behaviors. For example, Homer and Day (1991) compared the

effects of teaching a high-effort (full sentence signs) versus low-effort (single-word sign)

functional, alternative responses for one individual whose problem behavior was maintained by

negative reinforcement. They found that the individual engaged in low rates of problem behavior

and high rates of communication only when the alternative response required less effort.

Therefore, subjects in the current study could have engaged in the less effortful behaviors, which

maintained following reinforcement during the precursor FA, thus precluding the occurrence of

other precursors.

An alternative explanation for unobserved response topographies during the precursor FA

is that the current methodology simply yields a high rate of false alarms when identifying









precursors. The criterion for including a potential precursor in the probability analyses was

simply its occurrence within a trial, and interpretations of the probability analysis results were

based on relative (rather than absolute) values of conditional and unconditional probabilities for

each potential precursor. This method was used because it seemed to be a simple, conservative

method for identifying responses that preceded and were correlated with the target behavior, but

it might have resulted in the selection of precursors that did not occur frequently before the target

behavior (false alarms). Additionally, the termination criterion for the precursor assessment was

10 trials in which the target problem behavior was observed; therefore, a high rate of false alarms

might have resulted as a function of the brevity of the assessment. No attempt was made to

standardize the number of 5-min trials in which the target behavior was not observed for most

subjects (except for Amy, who engaged in property destruction during the first 10 trials of the

assessment); however, the total duration of trials 1i iithvit the target problem behavior was nearly

equal to the total duration of trials i/ i/h the target behavior for all subjects in the current study.

Future research might evaluate different precursor selection or assessment-termination criteria in

an attempt to clarify these results. For example, one might (a) examine relative frequencies of

precursors in trials with and without the target behavior, (b) select responses that tend to occur

closer in time before the target behavior, or (c) apply different interpretative rules to analyze the

probability data. More specifically, future research might examine more stringent selection

criteria by selecting only responses with high probabilities of the target given the precursor and

vice versa (e.g., a probability of 0.6 or higher). Another option would be to conduct more trials in

which the target problem behavior is unlikely to occur to, resulting in a larger sample of behavior

for determining the probability of observing potential precursors in the absence of the target

problem behavior.









It also is important to note that only half of the relation between precursors and the target

behavior was examined in the current study by determining the function of responses that were

predictive of the target behavior. It remains unknown whether responses that were not predictive

of the target behavior were maintained by a different source of reinforcement. It is possible that

any behavior that contacted the reinforcement contingencies might maintain and be effective in

predicting the function of the target problem behavior, even though these behaviors might not

necessarily be observed in typical settings. Renee's data most closely approximate this

possibility in that only 1 precursor to aggression was selected, and it did not appear to strongly

predict the occurrence of the target behavior. Results of independent FAs, however, showed that

both behaviors were members of the same response class. It seems unlikely that responses that

are not predictive of the target behavior would occur under similar conditions as the target

behavior, contact the reinforcement contingency, and maintain, however, given that antecedent

conditions (EOs) were specifically arranged during the precursor assessment to evoke the target

problem behavior. Presumably, other responses that are sensitive to that source of reinforcement

also would be observed, and behaviors maintained by a different source of reinforcement would

be less likely to be observed. In this way, responses that are likely members of the same response

class as the target problem behavior would probably be identified during the precursor

assessment. Future research might compare the results of functional analyses of behaviors that do

not predict the target problem behavior to the results of a functional analysis of the target

behavior to determine the extent to which these non-predictive behaviors are maintained by the

same or different sources of reinforcement.

Finally, it is highly unlikely that a precursor FA would be effective in reducing instances

of the target problem behavior if it were maintained by automatic reinforcement because









arranging social consequences for precursors would not be expected to affect the rate of the

target behavior. This would not, however, necessarily preclude the development of an effective

treatment for behavior maintained by automatic reinforcement based upon the results of

precursor assessments. For example, Hagopian et al. (2005) were able to design treatment

following an assessment of precursors by blocking stereotypy (hand flapping) that predicted the

occurrence of SIB (eye poking) for one individual. This subject's SIB was maintained by

automatic reinforcement, and blocking the precursor (stereotypy) was shown to be more

effective in reducing both stereotypy and SIB than blocking SIB alone. Therefore, these results

suggest that precursor analyses per se may have some clinical utility regardless of the function of

problem behavior and even if rates of the target problem behavior do not decrease during

assessment.

Results of Study 3 showed that effective reinforcement-based interventions can be

designed based on the results of precursor analyses only. Although severe problem behavior was

not eliminated during the precursor FA or baseline, lower rates of the target behavior were

observed relative to rates of precursors. Therefore, if reinforcement contingencies had been

placed on the target behavior (i.e., no reinforcement for precursors) during the FA and baseline,

it is probable that higher rates of the target behavior would have been exhibited by both subjects.

The effects of continuous NCR replicated the results of previous studies (Goh et al., 2000;

Marcus & Vollmer, 1996) in that nearly all responding (precursor and target behavior) was

suppressed under these conditions, and subjects did not emit the appropriate alternative response

(mand). As the DRA component was introduced while the NCR schedule was systematically

thinned, both subjects acquired the mand; however, these procedures were not effective in

reducing precursors while maintaining low rates of the target behavior in 2 of 3 cases. The









addition of a response blocking component (Amanda) or a change-over delay (Sammy, treatment

for behavior maintained by negative reinforcement) was effective in reducing rates of precursors

and target behaviors, while mands maintained under the DRA contingency. When low rates of

precursors and target behaviors were attained, the additional treatment components (NCR and

blocking for Amanda and NCR for Sammy) were removed and similar effects on all behaviors

were observed. (The change-over delay component remained in Sammy's final treatment

package due to the severity of his aggression, although it was rarely implemented during the final

sessions of the evaluation.) Therefore, both subjects allocated responding toward mands under

conditions that would typically be encountered in their classrooms (i.e., Amanda could sign to

receive food and Sammy could sign to receive a break from work). It is also interesting to note

that no additional treatment components were necessary to reduce Sammy's precursors and

maintain low rates of the target behavior during the second intervention evaluation. In fact, the

second evaluation was completed in approximately /4 of the number of sessions required to

complete the first evaluation.

As in Study 2, not all precursors identified via the trial-based precursor assessment actually

were observed during the precursor FA in Study 3. For example, the function of Amanda's

precursors was determined primarily by the occurrence of reaching toward the therapist (i.e.,

hand postures and stretching were observed rarely). The positive reinforcement function of

Sammy's precursors was determined solely by the occurrence of angry vocalizations, and the

negative reinforcement function was determined by the occurrence of climbing, angry

vocalizations, mouth movements, and moving furniture. Two of Sammy's precursors (running

across the room and tugging on the therapist's shirt) were never observed. As previously

discussed, it is possible that modifications to the methods of data analysis and/or precursor









selection criteria might lead to better predictions with respect to which response topographies are

likely members of the same response class as the target behavior and, thus, would be exhibited

during the precursor FA.

It also seems possible that improvements in the precursor assessment methodology might

have resulted in greater initial behavioral reductions under NCR schedule thinning plus DRA.

For example, if responses other than the selected precursors also predicted the occurrence of

severe problem behavior, their inclusion in treatment could have prevented the occurrence of the

target behavior. This is because the greater the number of responses that occur before severe

problem behavior and contact the reinforcement contingency, the less likely severe problem

behavior would be emitted. Anecdotally, it did not appear that either subject exhibited other

"precursor" behaviors that were not identified via the precursor assessment. In fact, the precursor

exhibited most frequently by Amanda during treatment was reaching toward the therapist, and

the precursor exhibited most frequently by Sammy was angry vocalizations.

An alternative explanation for the initial poor treatment effects with NCR schedule

thinning plus DRA is that some of the precursors selected for inclusion during treatment actually

were members of a different response class than the target behavior. For example, some response

topographies that were selected as precursors via the precursor assessment actually were not

observed during the precursor FA; therefore, the function of these "precursors" was unknown.

Given that the responses were selected suggests that they occurred at sufficient rates in general to

be detected, and it could have been mere coincidence that the responses occurred in trials in

which the target behavior also was observed. If this were true, these response topographies could

have been exhibited during treatment and detected simply as a function of extended observation

periods. Anecdotally, Sammy engaged in some precursor topographies in the absence of the









establishing operation (i.e., when demands were not presented or when preferred items were not

removed), which suggests that these behaviors might have been maintained by a different source

of reinforcement. Alternatively, features of the environment might have been discriminative for

the presentation of demands or removal of preferred items, thus evoking some of the precursors

in the momentary absence of the establishing operation. These possibilities remain speculative as

additional analyses of selected precursors were beyond the scope of the present study.

In summary, the current series of studies demonstrates a method of analyzing precursor

behavior and for progressing from assessment to treatment of severe problem behaviors while

minimizing risk posed by those behaviors. Other methods of potentially reducing risk during

assessment include the use of protective equipment (Le & Smith, 2002), a different dependent

variable such as latency to problem behavior (Thomason, Iwata, Neidert, & Roscoe, in press),

and brief session durations (Wallace & Iwata, 1999). The advantage of precursor analyses is that

reinforcement contingencies are not placed on severe problem behaviors, thus decreasing the

likelihood that severe behaviors would occur at high rates during the assessment and/or continue

to occur following the assessment period. In addition, Study 1 provides a new method of

empirically identifying precursor behaviors, and results of Study 2 verified that the identified

precursors typically are members of the same response class as severe problem behavior. Finally,

the sequential introduction of NCR and NCR schedule thinning plus DRA appears to be a viable

treatment option for shifting response allocation from problem behavior to appropriate behavior

while maintaining low rates of severe problem behavior and reducing risk. The results of Study 3

also indicate that this intervention strategy is appropriate for problem behavior maintained by

positive and/or negative reinforcement and that NCR can be gradually thinned such that

appropriate behavior maintains under DRA contingencies only.









Table 2-1. Subiect characteristics


Name
Liv

Billy

Chuck
Kelly
George


Age Classification
10 Down syndrome

15 Down & Kleinfelter's
syndromes
14 Arthrogryposis syndrome
10 Seizure disorder & retinopathy
9 Autism


Amy 3 Down syndrome


Renee 15 Angelman's syndrome

Curtis 13 Autism

Gerald 19 Cerebral palsy, MR (level
unspecified)
Adam 11 Prader-Willi syndrome

Donald 14 Autism, seizure disorder

Leigh 13 Trainable mentally handicapped
& language impaired
Guy 12 Autism

Kevin 54 Severe MR, seizure disorder


Amanda
Sammy


Autism, profound MR
Deaf, learning disabilities


Definition of Target Problem Behaviors
Property destruction (throwing items and
knocking over furniture)
Clothing destruction (ripping, tearing, or
unraveling socks)
SIB (head hitting)
SIB (self biting)
Aggression (hitting, kicking, pinching, and
biting)
Property destruction (throwing objects,
tearing materials from walls, and
destroying materials)
Aggression (hair pulling, hitting, and
pushing)
Aggression (hitting, kicking, biting, and
head butting)
SIB (hand biting)

Aggression (hitting, kicking, biting, and
throwing objects that hit people)
Aggression (hitting, kicking, biting, and
head butting)
SIB (chin hitting and banging)

Aggression (hitting, kicking, biting, and
head butting)
Property destruction (throwing furniture,
pounding on walls, and destroying or
throwing materials)
SIB (face and head hitting)
Aggression (hitting, kicking, biting, head
butting, and throwing objects that hit
people)









Table 2-2. Precursor checklist


Category
Vocalizations
Facial Expressions
Postures
Locomotion
Repetitive Motor
Movements
Object Manipulation
Other Problem Behaviors
Self-injurious Behavior

Aggression
Property Destruction


Examples
Screaming, laughing, cursing, squealing, requests
Smiling, grimacing, frowning, surprised
Slouching, dropping, head down, standing
Walking, running, jumping
Fidgeting, tapping fingers, tapping feet, stomping, hand flapping,
head movements, hair twirling, nail picking, clapping
Playing with objects, tapping pencil, twirling objects

Head banging, head hitting, skin picking, body hitting, self-biting,
hair pulling
Hitting, kicking, grabbing, head butting, biting, scratching
Breaking objects, knocking over furniture, banging objects,
throwing objects, hitting surfaces, kicking surfaces









Table 2-3. Probability analysis formulas
Probability Type
Conditional probability of the target behavior (T)
given the precursor (P,)
Conditional probability of the target behavior
given the absence of the precursor
Unconditional probability of the target behavior

Conditional probability of the precursor given the
target behavior
Conditional probability of the precursor given the
absence of the target behavior
Unconditional probability of the precursor


Formula
# trials with P, that also contain T
# trials with P,
p(T )# trials containing T but not P
# trials not containing P)
# trials containing T
p(T) -
total # trials
# trials with T that also contain P,
p(P, I T)= -
# trials with T
# trials containing P but not T
p(P# trials T)not containing P
# trials not containing P,
# trials containing P,
total # trials









Table 2-4. Precursors reported by caregivers vs. assessment-identified precursors. Italicized
precursors were behaviors identified by both caregivers and the precursor assessment.


Subject


Liv

Billy


Caregiver-Reported
Precursor(s)
Make a cry screech noise

NONE


Chuck NONE

Kelly Run away

George Yell, diru 1r items, tip over
chairs
Amy Laugh


Renee NONE
Curtis Whine, repeat phrases,
grimace
Gerald Scream, hit head or ear

Adam Vocalize negatively, put
head down, change entire
facial expression/make
faces at others, roll eyes,
yell, tongue click
Donald Bruxism




Leigh NONE



Guy Drop to ground, roll on
floor, curse, scream


Assessment-Identified
Precursor(s)
Vocalize positively, flap hands,
mouth objects
Cross legs, pull up pants/touch leg,
rub glasses
Hit surfaces, grab tongue, bounce
hands on face
Whine, mouth fingers, place hands
in clothes
Yell, ithr i- objects, sign, rub head,
swing arms, bang surfaces
Manipulate objects, make noises,
touch face, move around room,
move repetitively, hand on foot, say
"Mine," put face in object, bend at
waist, mouth object
Cover eyes
Scratch leg, block therapist from
objects
Flick lips, grimace, hit others, move
head
Say "No, slouch, grimace, turn
away, put paper in mouth, push
materials away



Flap hands, put hand to mouth,
snarl, clap hands, vocalize
negatively, move to objects

Cover eyes, rest head, say "Yeah
yeah," guide therapist,
stomp/shuffle, circle hands
Flop (includes rolling), curse,
vocalize negatively (includes
screaming), swing body, stomp, bite
objects, throw objects, push
materials away, crumple paper, bite
hand, bang head, grimace, shake
head "No," hit surfaces, slouch,
make requests


Precursors Observed
in FA
N/A

N/A

N/A

N/A

N/A

N/A


Cover eyes
Block therapist from
obj ects
Grimace, move head

Say "No," slouch,
turn away, push
materials away



Flap hands, snarl,
clap hands, vocalize
negatively, move to
obj ects
Cover eyes, chin
down, stomp/shuffle,
circle hands
Flop, curse, swing
body, bite objects,
throw objects, stomp,
push materials away









Table 2-4. Continued
Kevin Say "No "


Amanda NONE

Sammy NONE


Say "No, grunt, drop/scoot on
floor, wave arms, pull therapist's
arm, say "Good boy," hold knees,
smile, fidget, stack chairs, hit with
head, knock on table
Hand posture, reach for therapist,
stretch
Angry vocalizations, run, climb,
mouth movements, move furniture,
tug on therapist's shirt


Say "No, grunt,
drop/scoot on floor,
wave arms, hold
knees, fidget, knock
on table


N/A

N/A









LIST OF REFERENCES


Barton, E. J. & Ascione, F. R. (1979). Sharing in preschool children: Facilitation, stimulus
generalization, response generalization, and maintenance. Journal ofApplied Behavior
Analysis, 12, 417-430.

Bonfiglio, C. M., Daly, E. J., III, Martens, B. K., Lin, L. R., & Corsaut, S. (2004). An
experimental analysis of reading interventions: Generalization across instructional
strategies, time, and passages. Journal ofApplied Behavior Analysis, 37, 111-114.

Borrero, C. S. W. & Borrero, J. C. (2008). Descriptive and experimental analyses of potential
precursors to problem behavior. Journal ofApplied Behavior Analysis, 41, 83-96.

Carr, J. E., Coriaty, S., & Dozier, C. L. (2000). Current issues in the function-based treatment of
aberrant behavior in individuals with developmental disabilities. In J. Austin & J. E. Carr
(Eds.). Handbook ofApplied Behavior Analysis (pp. 91-112). Reno, NV: Context Press.

Carr, J. E., Coriaty, S., Wilder, D. A., Gaunt, B. T., Dozier, C. L., Britton, L. N., Avina, C., &
Reed, C. L. (2000). A review of noncontingentt" reinforcement as treatment for the
aberrant behavior of individuals with developmental disabilities. Research in
Developmental Disbilities, 21, 377-391.

Catania, A. C. (1971). Reinforcement schedules: The role of responses preceding the one that
produces the reinforcer. Journal of the Experimental Analysis ofBehavior, 15, 271-287.

Catania, A. C. (1973). The concept of the operant in the analysis of behavior. Behaviorism, 1,
103-116.

Charlop, M. H. & Milstein, J. P. (1989). Teaching autistic children conversational speech using
video modeling. Journal ofApplied Behavior Analysis, 22, 275-285.

Cuvo, A. J., Ashley, K. M., Marso, K. J., Zhang, B. L., & Fry, T. A. (1995). Effect of response
practice variables on learning spelling and sight vocabulary. Journal ofApplied Behavior
Analysis, 28, 155-173.

Day, H. M. & Horner, R. H. (1989). Building response classes: A comparison of two procedures
for teaching generalized pouring to learners with severe disabilities. Journal of Applied
Behavior Analysis, 22, 223-229.

DeLeon, I. G. & Iwata, B. A. (1996). Evaluation of a multiple-stimulus presentation format for
assessing reinforcer preferences. Journal ofApplied Behavior Analysis, 29, 519-533.

Dews, P. B. (1966). The effect of multiple SA periods on responding on a fixed-interval
schedule: V. Effect of periods of complete darkness and of occasional omissions of food
presentations. Journal of the Experimental Analysis ofBehavior, 9, 573-578.









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 ofApplied Behavior Analysis, 25, 491-498.

Goh, H., Iwata, B. A., & DeLeon, I. G. (2000). Competition between noncontingent and
contingent reinforcement schedules during response acquisition. Journal ofApplied
Behavior Analysis, 33, 195-205.

Grow, L. L., Kelley, M. E., Roane, H. S., & Shillingsburg, M. A. (2008). Utility of extinction-
induced response variability for the selection of mands. Journal ofApplied Behavior
Analysis, 41, 15-24.

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

Hagopian, L. P., Paclawskyj, T. R., & Kuhn, S. C. (2005). The use of conditional probability
analysis to identify a response chain leading to the occurrence of eye poking. Research in
Developmental Disabilities, 26, 393-397.

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

Horner, R. H. & Day, H. M. (1991). The effects of response efficiency on functionally equivalent
competing behaviors. Journal ofApplied Behavior Analysis, 24, 719-732.

Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1994). Toward a
functional analysis of self-injury. Journal ofApplied Behavior Analysis, 27, 197-209.
Reprinted from Analysis and Intervention in Developmental Disabilities, 2, 3-20, 1982.

Iwata, B. A., Kahng, S., Wallace, M. D., & Lindberg, J. L. (2000). The functional analysis model
of behavioral assessment. In J. Austin & J. E. Carr (Eds.), Handbook of applied behavior
analysis (pp. 61-89). Reno, NV: Context Press.

Koegel, R. L. & Covert, A. (1972). The relationship of self-stimulation to learning in autistic
children. Journal ofApplied Behavior Analysis, 5, 381-387.

Krantz, P. J. & McClannahan, L. E. (1998). Social interaction skills for children with autism: A
script-fading procedure for beginning readers. Journal ofApplied Behavior Analysis, 31,
191-202.

Lalli, J. S., Mace, F. C., Wohn, T., & Livezey, K. (1995). Identification and modification of a
response-class hierarchy. Journal ofApplied Behavior Analysis, 28, 551-559.

Le, D. D. & Smith, R. G. (2002). Functional analysis of self-injury with and without protective
equipment. Journal ofDevelopmental and Physical Disabilities, 14, 277-290.









Lovaas, 0. I. & Simmons, J. Q. (1969). Manipulation of self-destruction in three retarded
children. Journal ofApplied Behavior Analysis, 2, 143-157.

Magee, S. K. & Ellis, J. (2000). Extinction effects during the assessment of multiple problem
behaviors. Journal ofApplied Behavior Analysis, 33, 313-316.

Marcus, B. A. & Vollmer, T. R. (1996). Combining noncontingent reinforcement and differential
reinforcement schedules as treatment for aberrant behavior. Journal of Applied Behavior
Analysis, 29, 43-51.

Michael, J. L. (1982). Distinguishing between discriminative and motivational functions of
stimuli. Journal of the Experimental Analysis ofBehavior, 37, 149-155.

Najadowski, A. C., Wallace, M. D., Ellsworth, C. L., MacAleese, A. N., & Cleveland, J. M.
(2008). Functional analyses and treatment of precursor behavior. Journal ofApplied
Behavior Analysis, 41, 97-105.

Nutter, D., & Reid, D. H. (1978). Teaching retarded women a clothing selection skill using
community norms. Journal ofApplied Behavior Analysis, 11, 475-487.

Parrish, J. M., Cataldo, M. F., Kolko, D. J., Neef, N. A., & Egel, A. L. (1986). Experimental
analysis of response covariation among compliant and inappropriate behaviors. Journal
ofApplied Behavior Analysis, 19, 241-254.

Peterson, R. F. (1968). Some experiments on the organization of a class of imitative behaviors.
Journal ofApplied Behavior Analysis, 1, 225-235.

Pierce, K. L. & Schreibman, L. (1994). Teaching daily living skills to children with autism in
unsupervised settings through pictorial self-management. Journal ofApplied Behavior
Analysis, 27, 471-481.

Reeve, S. A., Reeve, K. F., Townsend, D. B., & Poulson, C. L. (2007). Establishing a
generalized repertoire of helping behavior in children with autism. Journal of Applied
Behavior Analysis, 40, 123-136.

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 ofApplied Behavior
Analysis, 32, 269-283.

Rosenbaum, M. S. & Breiling, J. (1976). The development and functional control of reading-
comprehension behavior. Journal ofApplied Behavior Analysis, 9, 323-333.

Shukla, S. & Albin, R. W. (1996). Effects of extinction alone and extinction plus functional
communication training on covariation of problem behaviors. Journal ofApplied
Behavior Analysis, 29, 565-568.

Skinner, B. F. (1938). The behavior oforganisms. Acton: Copley Publishing Group.









Skinner, B. F. (1953). Science and human behavior. New York: Macmillan.


Smith, R. G. & Churchill, R. M. (2002). Identification of environmental determinants of
behavior disorders through functional analysis of precursor behaviors. Journal of Applied
Behavior Analysis, 35, 125-135.

Sprague, J. R. & Horner, R. H. (1992). Covariation within functional response classes:
Implications for treatment of severe problem behavior. Journal of Applied Behavior
Analysis, 25, 735-745.

Thomason, J. L., Iwata, B. A., Neidert, P. L., & Roscoe, E. M. (in press). Evaluation of response
latency as the index of problem behavior during functional analyses. Journal of Applied
Behavior Analysis.

Vollmer, T. R., Iwata, B. A., Zarcone, J. R., Smith, R. G., & Mazaleski, J. L. (1993). The role of
attention in the treatment of attention-maintained self-injurious behavior: Noncontingent
reinforcement and differential reinforcement of other behavior. Journal of Applied
Behavior Analysis, 26, 9-21.

Wallace, M. D. & Iwata, B. A. (1999). Effects of session duration on functional analysis
outcomes. Journal ofApplied Behavior Analysis, 32, 175-183.

Werts, M. G., Caldwell, N. K., & Wolery, M. (1996). Peer modeling of response chains:
Observational learning by students with disabilities. Journal ofApplied Behavior
Analysis, 29, 53-66.

Young, J. M., Krantz, P. J., McClannahan, L. E., & Poulson, C. L. (1994). Generalized imitation
and response-class formation in children with autism. Journal ofApplied Behavior
Analysis, 27, 685-697.









BIOGRAPHICAL SKETCH

I completed my Bachelor of Science degree at the University of Florida in 2001 then

accepted a position on the inpatient Neurobehavioral Unit at the Kennedy Krieger Institute in

Baltimore, MD. There I was responsible for the assessment and treatment of severe problem

behaviors, such as self-injury, aggression, and property destruction. I returned to the University

of Florida in 2003 to pursue a doctoral degree in psychology and specializing in behavior

analysis. During my graduate training, I have been involved in research projects on refining

behavioral assessment methods, evaluating treatments for problem behaviors, comparing

methods of training observers, and evaluating the effects of varying reinforcement parameters on

performance. I also served as coordinator of an outpatient clinic for individuals diagnosed with

autism, provided behavioral services to students and teachers within a special education setting,

and served as teaching assistant and primary instructor for introductory courses in applied

behavior analysis. Following graduation, I will join the faculty at the University of Houston -

Clear Lake as an assistant professor in applied behavior analysis within the psychology program.





PAGE 1

1 EXPERIMENTAL ANALYSIS OF PRECURSORS TO PROBLEM BEHAVIOR By JENNIFER N. FRITZ A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Jennifer N. Fritz

PAGE 3

3 To Patti, David, Heather, Helen, and Carl

PAGE 4

4 ACKNOWLEDGMENTS This research was supported in part by a gr ant from the Florida Agency on Persons with Disabilities. I especially th ank Jen Hammond and Sarah Bloo m for their thoughtful comments and suggestions throughout the course of the study, as well as Kathryn Jann, Zachariah Sims, Ashley Greenwald, Barbara Tomlian, Lisa Smal heiser, and Alex Avelino for the many hours they assisted with scoring videos and various other aspects of the project. I also thank Carrie Dempsey, Natalie Rolider, and Charles Nowell for their assistance with overseeing subjects. I sincerely thank Timothy Vollmer, Donald Stehouwer, and Stephen Smith for their guidance and helpful suggestions on this a nd other projects. Finally, I w ould like to extend my deepest appreciation to Brian Iwata for all of the time, dedication, and s upport he provided to this project and to me throughout my career.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES.........................................................................................................................8 ABSTRACT.....................................................................................................................................9 CHAP TER 1 INTRODUCTION..................................................................................................................11 Functional Analysis of Problem Behavior.............................................................................. 11 Operant Response Classes......................................................................................................12 Problem Behavior as an Operant Class .................................................................................. 14 2 STUDY 1: EMPIRICAL IDENTI FICATION OF PRECURSORS ....................................... 19 Method....................................................................................................................................19 Subjects and Setting........................................................................................................ 19 Procedures..................................................................................................................... ..19 Response Measurement and Reliability.......................................................................... 19 Caregiver Interview.........................................................................................................20 Precursor Assessment...................................................................................................... 21 Probability Analyses........................................................................................................ 24 Results and Discussion......................................................................................................... ..26 Analysis of Precursor Behaviors..................................................................................... 26 Comparison of Caregiver Re port and Precursor Data ..................................................... 29 3 STUDY 2: FUNCTIONAL ANALYSIS OF PRE CURSOR AND PROBLEM BEHAVIORS.........................................................................................................................35 Method....................................................................................................................................35 Subjects and Setting........................................................................................................ 35 Response Measurement and Reliability.......................................................................... 36 Procedures..................................................................................................................... ..38 Results and Discussion......................................................................................................... ..39 4 STUDY 3: PRECURSOR ASSESSMENT AS THE BASIS FOR INTERVENTION ......... 48 Method....................................................................................................................................51 Subjects and Setting........................................................................................................ 51 Response Measurement and Reliability.......................................................................... 51

PAGE 6

6 Procedures..................................................................................................................... ..53 Baseline....................................................................................................................53 Continuous NCR...................................................................................................... 53 NCR Schedule Thinning Plus DRA.........................................................................54 Results and Discussion......................................................................................................... ..54 5 DISCUSSION.........................................................................................................................64 LIST OF REFERENCES...............................................................................................................79

PAGE 7

7 LIST OF TABLES Table page 2-1 Subject characteristics.................................................................................................... ....74 2-2 Precursor checklist........................................................................................................ .....75 2-3 Probability analysis formulas............................................................................................. 76 2-4 Precursors reported by caregivers vs assessm ent-identified precursors...........................77

PAGE 8

8 LIST OF FIGURES Figure page 2-1 Precursor-assessment results fo r Liv, Chuck, Billy and Amanda..................................... 31 2-2 Precursor-assessment results for Kelly, George, Amy, and Sammy................................. 32 2-3 Precursor-assessment results for Renee, Curtis, Gerald, and Adam .................................. 33 2-4 Precursor-assessment results fo r Donald, Leigh, Guy, and Kevin .................................... 34 3-1 Results of the independent functional anal yses for Renee, Curtis, Gerald, and Ada m..... 45 3-2 Results of the independent functional an alyses for Donald, Leigh, Guy, and Kevin ........ 46 3-3 Proportional distribution of precursor responses observed during the precursor FA ........ 47 4-1 Results of the precursor FA for Amanda in Study 3. ........................................................59 4-2 Treatm ent results for Amanda in Study 3.......................................................................... 60 4-3 Results of the precursor FA for Sa mmy in Study 3.......................................................... 61 4-4 Treatm ent for behavior maintained by negative reinforcement for Sammy...................... 62 4-5 Treatm ent for behavior maintained by positive reinforcement for Sammy..................... 63

PAGE 9

9 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EXPERIMENTAL ANALYSIS OF PREC URSORS TO PROBLEM BEHAVIOR By Jennifer N. Fritz August 2008 Chair: Brian A. Iwata Major: Psychology Standard functional analyses require the rep eated observation of a target behavior to determine behavioral function, but this method can prove problematic in the case of severe or dangerous behaviors. Previous studies have sh own, however, that individuals who engage in problem behaviors sometimes engage in both mild and severe forms and that severe behaviors are observed less frequently when reinforcement is delivered contingent upon the more mild behaviors. Studies also have s hown that functional analyses of mild behaviors that reportedly precede severe behaviors can (a) be members of the same operant response class and (b) reduce the number of severe topographies that are observed during the assessment. These mild behaviors (i.e., precursors) are t ypically identified via ca regiver verbal report or informal, direct observations, but it is possible th at precursors could exist even when they are not readily identifiable. Therefore, we deve loped a checklist to identify precu rsors via videotaped trials in Study 1, and results showed that the empirical method of identifying precursors successfully identified at least 1 precursor for all 16 subj ects. Separate functiona l analyses then were conducted of precursor and severe problem behaviors for 8 subjects in Study 2, and correspondence between outcomes was observed in 7 cases. Furthermore, few studies have evaluated treatments for severe problem beha vior based upon the results of precursor

PAGE 10

10 assessments. Therefore, we evaluated a sequentia l treatment consisting of a dense schedule of noncontingent reinforcement (NCR) followed by NCR schedule thinni ng plus differential reinforcement of alternative behavior to reduc e precursors, increase appropriate behavior, and maintain low rates of severe behavior. Results s howed that this treatment strategy was effective for behaviors maintained by positive rein forcement and negative reinforcement.

PAGE 11

11 CHAPTER 1 INTRODUCTION Functional Analysis of Problem Behavior Experim ental approaches to the assessment of problem behaviors (such as self-injurious behavior, aggression, property dest ruction, tantrums, stereotypy, et c.) have been reported in isolated studies since the 1960s, primarily for th e purpose of demonstrating that a particular contingency could exacerbate problem behavior (e.g., Lovaas & Simmons, 1969). Since then, several approaches to identifying sources of reinforcement that ma intain problem behavior have been developed and systematically evaluated, but the approach with the most empirical validity is the functional or experimental analysis (see Iwata, Kahng, Wallace, & Lindberg, 2000, for a recent review). Iwata, Dorsey, Slifer, Bauman, and Richma n (1982/1994) were the first to develop a general experimental model for identifying whic h of several common sources of reinforcement maintained a particular problem be havior self-injurious behavior or SIB. They created a series of four conditions, three of wh ich involved manipulation of antecedent and consequent events that formed contingencies previously shown to ma intain SIB, plus a control. The conditions used in that study were: alone (ante cedent event: austere environment; consequent event: none; test for automatic reinforcement), social disapprova l (antecedent event: no attention; consequent event: statement of disapproval; test for positiv e reinforcement), academic demand (antecedent event: tasks presented; consequent event: break fr om work; test for negative reinforcement, and unstructured play (antecedent events: no demands noncontingent attention, leisure materials available; consequent event: none ; control condition). Differentia l responding in the form of higher rates of SIB in one of the three test conditions (or high rates across all conditions) identified the source of reinforcement maintaining subjects' SIB.

PAGE 12

12 The functional analysis (FA) approach desc ribed by Iwata et al. (1982/1994) has been replicated in hundreds of studies and has become the standard method for assessing a wide range of problem behaviors (Hanley, Iwata, & McCor d, 2003). In addition, the results of such an analysis can be used to design an interventi on in which the reinforcement contingency is manipulated to reduce problem behavior and also to increase appropriate, alternative behaviors (Carr, Coriaty, & Dozier, 2000). An inherent limitation of all FA me thods, however, is the explicit arrangement of conditions that increase the frequency of potentially dangerous behavior. Although such increases may be te mporary and may present risks no greater than those already posed by the problem behavior, strategies that minimize risk during assessment would be beneficial to both researchers and clinicians. One promising approach is to assess mild behaviors that are members of the same operant res ponse class as the severe problem behavior. Operant Response Classes As developed by Skinner, the concept of the operant response cl ass has far-reaching im plications for understanding the etiology and ma intenance of complex human behavior. In its current usage, the operant refers to a cla ss (or variety) of responses that can differ topographically but that are maintained by a common class of consequences (reinforcers) (Catania, 1973; Dews, 1966; Skinner, 1953). The formation of response classes is esse ntial for performing a variety of complex behaviors. The unifying principl e of all of these behaviors is that they produce the same outcome, but, in the specific in stance, the behaviors can vary on a large number of topographical dimensions. For example, one could exhibit any nu mber of responses in order to obtain a snack. The individual might first look in the pantry by sliding open the door, scan the shelves for preferred snack foods, and select chips, which re quires pulling the bag in opposing directions to access the chips. If chips were not available, th e individual might then check the refrigerator,

PAGE 13

13 which involves pulling open the door (a topogra phically dissimilar response that produces a similar outcome as the pantry example visual access to food items), scanning the shelves for snacks, and selecting a container of fresh fruit. Opening the cont ainer involves very different responses than opening a bag of chips; the lid mu st be pulled from the corner in a vertical direction, whereas the chip bag required pulling bot h sides in a horizontal direction. The ultimate result of opening both containers is the same in both cases, however tangible access to food items. An understanding of how these response classes develop is essential for ensuring that an individual can function competently in various f acets of daily life; thus, not surprisingly, the development of operant response classes has receiv ed extensive attention in applied research. For example, researchers have evaluated strategi es for teaching various responses involved in developing imitation skills (e.g., Peterson, 1968; Werts, Caldwell, & Wolery, 1996; Young, Krantz, McClannahan, & Poulson, 1994), academic skills (e.g., Bonfiglio, Daly, Martens, Lin, & Corsaut, 2004; Cuvo, Ashley, Marso, Zhang, & Fry, 1995; Rosenbaum & Breiling, 1976), social skills (e.g., Barton & Ascione, 1979; Charlop & M ilstein, 1989; Krantz & McClannahan, 1998; Reeve, Reeve, Townsend, & Poulson, 2007), sel f-help skills (e.g., Day & Horner, 1989; Nutter & Reid, 1978; Pierce & Schriebman, 1994), and va rious other socially important behaviors. Problems arise, however, when one attempts to identify a particular response as a member of a particular response cla ss. Early on, Skinner acknowledged this problem and posited that such an analysis cannot be "an act of arbitrary subdivi ding, and we cannot define the concepts of stimulus and response quite as simply as parts of behavior and envi ronment' without taking account of the natural lines of fracture along wh ich behavior and environment actually break" (Skinner, 1938; p. 33). In other words, operant beha vior must be analyzed over time to identify

PAGE 14

14 the environmental events that determine its occurrence. Functional analyses are used for this purpose during the assessment of problem beha viors, and behaviors (however dissimilar topographically) that are maintained by the same source of reinforcem ent are identified as members of a common response class. Problem Behavior as an Operant Class W ithin an operant response class, cova riation among member responses has been documented extensively in applie d research with respect to ad aptive and problem behaviors (e.g., Koegel & Covert, 1972; Parrish, Cataldo, Kolko, Neef, & Egel, 1986; Sprague & Horner, 1992). Additionally, in some cases, an individual might allocate responding in such a way that members of a given response class generally are exhibited in a hierarchical order; in other words, some behaviors are more probabl e than other behaviors. The identification of response hi erarchies with respect to th e assessment and treatment of problem behavior has received increased attent ion in recent work (e.g., Borrero & Borrero, 2008; Harding et al., 2001; Lalli et al ., 1995; Smith & Churchill, 2 002). Results of these studies showed that mild behaviors sometimes occurred before severe problem be havior and that mild and severe problem behaviors were members of the same re sponse class. As previously mentioned, standard FAs are sometimes contraindicated when the topography of problem behavior poses risks due to its severity, and these studies showed that one promising assessment approach to minimizing risk is to assess behaviors that predict o ccurrences of the target problem behavior (precursors). This approach minimizes risk because if mild and severe problem behaviors are members of the same operant respon se class, reinforcement contingencies arranged for behaviors that occur before the severe beha vior could result in reduced rates of severe behaviors.

PAGE 15

15 Such an analysis involves several steps: (a ) identifying potential precursor responses, (b) verifying that these responses do, in fact, predict the o ccurrence of the target and (c) determining whether precursors are members of the same re sponse class as the targ et. Recent studies have provided evidence for the validity of this type of analysis by focusing on one or more of these steps in the assessment of severe problem behaviors. After observing informally that an individua l tended to engage in stereotypy (finger waving) prior to eye poking, Hagopian, Paclawskyj, and Kuhn (2005) calculated several conditional and unconditional probab ilities to verify the correlation between the behaviors and found that stereotypy actually was predictive of eye poking. They also examined cumulative records of responding and observed a temporal contiguity between stereotypy and eye poking. Thus, results of both analyses suggested that stereotypy was a precursor to eye poking. Borrero and Borrero (2008) used similar procedures to identify response-re sponse relations after informally observing that two i ndividuals tended to emit loud vocalizations before engaging in severe problem behaviors (i.e., SIB, aggres sion, or property destru ction). They conducted observations in the subjects' classroom on the occurrence of both vocalizations and problem behavior. Results of probability analyses showed that vocalizations and problem behavior were highly correlated, and resu lts of lag-sequential analyses show ed that vocalizations were most likely to occur immediately preceding an instan ce of problem behavior. Taken together, the results of these studies show that correlationa l analyses, and conditional probability analyses in particular, are useful for determining which behavi ors in an individual's re pertoire are predictive of severe problem behaviors. Studies also have demonstrated via experimental analyses that less severe behaviors can be members of the same response class as more seve re behaviors. For example, Lalli, Mace, Wohn,

PAGE 16

16 and Livezey (1995) observed that the screams, aggression, and SIB exhibited by a young girl appeared to be maintained by negative reinforcemen t and often occurred in a particular sequence. Subsequently, escape from demands was provided contingent upon individual behaviors (SIB, aggression, and screams, respectively) while th e other 2 behaviors were placed on extinction. They found that when the reinfo rcement contingency was placed on the last behavior in the sequence (SIB), all behaviors tended to occur in a hierarchical order (i.e., screams, aggression, then SIB). Conversely, when the reinforcement c ontingency was placed on behaviors occurring earlier in the sequence (screams or aggression), behaviors that tended to occur later were observed rarely. These data indicated that scream ing (a relatively innocuous problem behavior) predicted the occurrence of mo re severe problem behaviors (a ggression and SIB) and that all behaviors were members of the same response class. In similar studies, Borrero and Borrero (2008); Richman, Wacker, Asmus, Casey, and Andelman (1999); and Smith and Churchill (2002) determined through direct observation and/or parental interviews that less severe problem be haviors apparently preceded the occurrence of the most severe form of problem be havior. Richman et al. conducted a functional analysis in which reinforcers were provided contingent upon all t opographies of problem behavior and observed higher rates of less severe problem behaviors and near-zero rates of more severe problem behaviors. When the less severe problem beha viors were placed on extinction, increases were observed in the mores severe pr oblem behaviors, thus demonstr ating that all behaviors were maintained by the same source of reinforcement. Other studies have f ound similar effects when extinction is applied to the most commonly occurring response during assessment (e.g., Harding et al., 2001; Magee & Ellis, 2000). Smith and Chur chill, and Borrero and Borrero, conducted

PAGE 17

17 independent functional analyses of the precursor and target problem behaviors and found that the precursor and target behaviors were, in f act, members of the same response class. Results of these studies are important in de monstrating that less severe behaviors may predict the occurrence of target problem be haviors and that programming reinforcement contingencies for less severe behaviors might d ecrease the rate of the severe behaviors during assessment. The extent to which individuals who engage in problem behaviors also exhibit precursors is unknown, however. Therefore, the pr imary purpose of this study was to determine whether precursor behaviors typically prec ede the occurrence of problem behaviors. In addition, a limitation common to all studies was that no systematic method was used to initially identify the precursor be haviors. Potential precursors were defined in those studies based on caregiver verbal report or informal observations conducted by the experimenters prior to assessment (Borrero & Borrero, 2008; Hagopi an et al., 2005; Smith & Churchill; 2002); however, no other systematic procedures for selecting precursors were described. These anecdotal sources may provide useful information, but correlations betwee n precursor and target responses, to the extent that they exist, also should be readily observable and quantifiable. Also, it is possible that precursors (a) might exist ev en when caregivers cannot identify them, (b) may be different than those reporte d, or (c) are not readily detect ed during informal observations. Thus, by using informal methods of precursor identification, it is po ssible that an important step in the analysis of precursors could be based on inaccurate information or limited sampling of client behavior. Finally, numerous instances of the target pr oblem behavior were observed before the relation between precursor and ta rget responses was determine d, thereby making the procedure difficult to use in situations for which it is id eally designed the assessment of severe problem

PAGE 18

18 behavior. Therefore, a second purpose of this study was to evaluate a method for identifying precursors that (a) was based solely on direct observation and (b) minimized the number of occurrences of the target problem behavior required to identify the precursors. Conditional probability analyses were used to determine which of several potential precursors were predictive of target behavior (Study 1), and inde pendent functional analyses of precursors and target problem behaviors then were conducted to verify that all behaviors were members of the same response class (Study 2). Finally, only one published study to date has developed an intervention based upon the results of precursor an alyses only (Najadowsk i, Wallace, Ellsworth, MacAleese, & Cleveland, 2008). Thus, the third pu rpose of this study was to determine whether (a) an effective intervention could be designed ba sed upon the results of precursor analyses alone and (b) the sequential introduction of noncontin gent reinforcement (NCR) followed by NCR schedule thinning plus differential reinforcemen t of alternative behavior (DRA) would be effective in reducing precursors while mainta ining low rates of severe problem behavior.

PAGE 19

19 CHAPTER 2 STUDY 1: EMPIRICAL IDENTI FICATION OF PRECURSORS Method Subjects and Setting Sixteen individuals diagnosed with d evelopmental disabiliti es who engaged in problem behavior participated in Study 1. Subject characteristics (age, diagnostic classification, and definition of the target problem behavior) are lis ted in Table 2-1. All sessions were conducted in an observation room at a day program for adults or in designated areas of a classroom at a special education school. Procedures Study 1 was conducted in 3 phases: (a) caregiver interview, (b ) structured observations to identify precursors (precursor assessm ent), and (c ) probability analysis to select precursors. Next, a trial-based, precursor assessment was conducte d, in which conditions known to evoke problem behaviors were presented sequentially until 10 instan ces of the target behavior were observed. A probability analysis then was used to select pr ecursors. This analysis consisted of comparing several conditional and unconditional probability calculations to determine which responses that predicted the occurrence of th e target behavior. Finally, resu lts obtained from the empirical precursor analysis were compared with precursor s reported by caregivers to determine the degree of correspondence between the two sets of data. Response Measurement and Reliability Because potential precursors were un known prior to assessment, all trials were videotaped and were scored later by two observers using a checklist. Responses were grouped topographically in the checklist as: (a) vocalizati ons, (b) facial expressi ons, (c) postures, (d) repetitive motor movements, (e) locomotion, (f) object manipulation, and (g) other problem

PAGE 20

20 behaviors. Examples of possibl e response topographies were listed within each category, and additional space was provided to allow observers to write in behaviors that were observed but not included on the checklist (Table 2-2). All responses that were included in the topographical definition of the target behavior or could be c onsidered mild forms of the target (e.g., pushing the therapist when the target was aggression) were excluded as potential precursors. The precursor assessment was conducted in two phases: (a) pote ntial precursor topograp hies were identified and operationally defined and (b) potential precursors were scored as occurrence or nonoccurrence in all assessment trials. Interobserver agreement was assessed by havi ng a second observer scor e the occurrence of precursors and the target behaviors during all asse ssment trials. After scoring a trial, observers' data records were compared. If scoring discre pancies were found, the observers discussed the operational definitions, watched the video, and/ or modified the operational definitions of precursors, then rescored the trial(s). This process was repeated until 100% interobserver agreement was achieved for the occurrence of al l precursors and target s during the precursor assessment. Caregiver Interview Prior to the precursor assessm ent, an expe rimenter conducted an open-ended interview with a caregiver for each subject in an attempt to identify potential precu rsors. Caregivers were either a parent or teacher who had known the subj ect for at least 6 months and who had observed instances of the target problem behavior. During the interview, the caregiver first was asked to identify the subject's most se vere class of problem behavior (SIB, aggression, or property destruction), which was selected as the targ et problem behavior for assessment during subsequent phases. If the target behavior was identified previously, caregivers were asked if they had observed the occurrence of prob lem behavior and if they coul d identify situations in which

PAGE 21

21 problem behavior was likely to occur. Caregive rs then were asked if they had observed any behaviors that tended to precede the target beha vior (i.e., if they c ould identify any potential precursors). The experimenter noted any responses that were mentioned and clarified any vague descriptions. For example, if a caregiver reported that the individual "got upset" before engaging in the target behavior, the experi menter asked the caregiver to describe "getting upset" in greater detail in an attempt to identify observable responses that might function as precursors. Precursor Assessment The precursor assessm ent consisted of discrete trials in which antecedent conditions that might serve as establishing operations (EOs; Michael, 1982) for the target behavior were presented and were similar to the attention and demand conditions of a functional analysis (Iwata et al., 1982/1994). If caregivers re ported that the target behavi or was likely to occur when preferred items were removed or access to items was denied or if the experimenters observed that the target behavior occu rred under these cond itions, a tangible condition also was included in the assessment. Given that many problem behaviors are maintained by positive reinforcement (access to attention or preferred items) or negative reinforcement (escape from demands), presenting these conditions presumably increased the likelihood of observi ng the target behavior in a relatively short period of tim e. If the target behavior was obse rved in a trial, the consequence relevant to the antecedent c ondition (attention, escape, and/or access to leisure items) was delivered. A trial was terminated following the occurr ence of the target behavior or after 5 min in which the target behavior did not occur (described in more detail below), whichever came first. During attention trials, the therapist did not interact with the subject, unless the target behavior was observed, at which time the ther apist delivered a reprimand (e.g., "Don't do that, you will hurt yourself.") and gentle p hysical contact. The therapist c ontinued to interact with the subject (e.g., rubbing the subject's back, talking about preferred topics, etc.) until the target

PAGE 22

22 behavior was not observed for 30 s. Once the target behavior was not observed for 30 s or if the target behavior was not observed in 5 min, a demand trial was conducted. During demand trials, the therapist presented instructions to complete tasks appropriate to the subject's functioning level. The therapist used a 3-step prompting procedure (vocal instruction, model, physical gu idance) but terminated the in structional sequence and moved away from the subject contingent upon the first occurrence of the target behavior. The next trial began once the target behavior was not observed for 30 s or if the target behavior was not observed in 5 min. If a tangible condition was included in the assessment, it was conducted following the demand trial. If a tangible conditio n was not included, another attention trial was conducted. During tangible trials, the therapist allowed th e subject brief (1-2 min) access to preferred items and then removed the items. Contingent up on the target behavior, th e items were returned to the subject. Once the target behavior was not observed for 30 s or if the target behavior was not observed in 5 min, the toys again were re moved, and another attention trial was conducted. The assessment was considered complete after 10 instances of the ta rget behavior were observed, except for Adam and Amy. Only 7 tria ls with the target pr oblem behavior were included in Adam's precursor assessment due to an oversight. Amy engaged in the target problem behavior during every one of the first 10 trials of the precursor assessment, thus precluding some of the probability calculations. Therefore, 3 play trials were conducted in which Amy had noncontingent access to preferred leisure and edible items, as well as the therapist's attention. The target pr oblem behavior was not observed dur ing these play trials, and Amy's precursor assessment was considered complete with 10 trials containing the target behavior and 3 play trials in which the target behavior was not ob served. In general, the total duration of trials in

PAGE 23

23 which the target behavior was not observed was a pproximately equal to or greater than the total duration of trials in which the target behavior was observed. In addition, if the target behavior was observed during a trial, the next trial was not conducted until the target behavior had not occurred for 30 s to reduce the likelihood that th e subject would engage in multiple, consecutive instances of the target behavior Presumably, it was unlikely that the subject could engage in precursor behaviors while exhibiti ng a burst of target behaviors; thus, the requisite 30 s of the absence of the target behavior increased the likelihood of observi ng potential precu rsor behaviors during the assessment. The above method differed from those used by Hagopian et al. ( 2005) and Borrero and Borrero (2008), in which numerous instances of the target behavior were observed before the relation between precursors and target behavior was established. For example, Borrero and Borrero required a minimum of 45 target beha viors before the descriptive analysis was considered complete. Similarly, 18 eye pokes were depicted in the cumu lative records of the Hagopian et al. study, which repr esented only 2 of 31 assessment sessions that were conducted. By using a trial-based format and by restricting the number of target beha viors to 10 occurrences, we hoped to have an adequate sample from whic h to identify precursors but to greatly limit the frequency of problem behavior. All trials were videotaped for subsequent data collection. When the assessment was complete, two observers watched the videos an d used the checklist to mark any potential precursor topographies observed in trials in which the target problem behavior occurred. The checklist contained examples of a wide range of possible beha vioral topographies, as well as space for observers to record behaviors that were not listed. Thus, all beha viors that occurred in trials in which the target behavior was observed were scored to identify responses that had the

PAGE 24

24 potential to predict the target be havior. Responses that occurred after the target behavior were not scored, however. The checklist contained examples of response topographies within 7 general categories of behavior, including vocalizations facial expressions, postures, locomotion, repetitive motor movements, object manipulation, and other proble m behaviors that differed from the target behavior. The observers compared the topographi es marked on each checklist and developed operational definitions of all potential precursors. Finally, two observers recorded the presence or absence of behaviors during all trials using a binary scoring code (1 = occurr ence of precursors and/or the targ et behavior within a trial, 0 = nonoccurrence). Following each trial, the observers co mpared their data reco rds. In the event of any discrepancies, the observ ers watched the video together and discussed the observed behaviors. The observers then rescored the trial until 100% agreement was attained for each precursor and the target behavior. Probability Analyses The purpose of the probability analysis was to determ ine which behaviors predicted the occurrence of the target behavior (i.e., precursors) in a quantitativ e manner. Several probabilities were calculated based on all trials of the precu rsor assessment. The probability of the target behavior given the precursor [ p (T|P n )] was calculated by dividing the number of trials in which that precursor and the target be havior were observed by the total number of trials in which that precursor was observed. The probability of the target given the absence of the precursor [ p (T|~P n )] was calculated by dividing the number of tr ials in which the target behavior was observed but the precursor was not by the total numb er of trials in which that precursor was not observed. The unconditional probability of the target [ p (T)] was calculated by dividing the

PAGE 25

25 number of trials in which the ta rget behavior was observed by the total number of trials in the assessment. Similar calculations were performed to determ ine probabilities for each of the precursors. The probability of that precursor given the target behavior [ p (P n |T)] was calculated by dividing the number of trials in which both the precursor and target behavior oc curred by the number of trials in which the target behavior occurred. The probability of the precursor given the absence of the target behavior [ p (P n |~T)] was calculated by dividing the number of trials in which the precursor was observed but the targ et behavior was not by the total number of trials in which the target behavior was not observed. Finally, the unconditional probability of the precursor [ p (P n )] was determined by dividing the number of trials in which the precursor was observed by the total number of trials in the assessment. (Formulas fo r each of the probabilities are listed in Table 23.) The relative probability values fo r each response were compared to select the precursors. First, the probability of the ta rget behavior given each potential precursor was compared to (a) the probability of the target behavior given the absence of each precursor and (b) the unconditional probability of the target behavior. Next, the probability of each precursor given the target behavior was compared to (a) the probabi lity of each precursor given the absence of the target behavior and (b) the unconditional probability of each precursor. Behaviors were selected as precursors if they sati sfied both of the following criteria. Fi rst, the probability of the target behavior given the precursor was higher than th e probability of the target behavior given the absence of the precursor and the unconditional probability of the target behavior, or p (T|P n ) > p (T|~P n ) and p (T|P n ) > p (T). Second, the probability of the precursor given the target behavior was higher than the probability of the precursor given the absence of the target behavior and the

PAGE 26

26 unconditional probability of the precursor, or p (P n |T) > p (P n |~T) and p (P n |T) > p (P n ). If numerous potential precursors were obser ved, some response topographies were combined if the responses (a) met the criteria for classification as either precursors or non-precursors and (b) could be described succinctly based upon similar topographical features (e .g., "crawl", "run", and "climb" were combined into "move around room" for Amy). Results and Discussion Analysis of Precursor Behaviors Results of the precursor assessm ents are show n in Figures 2-1 through 2-4. The top panel of each subject's graph shows the probability analys is for the target behavior. In these panels, the dark, solid gray bars show the probability of the target behavior given that precursor, the striped bars show the probability of the target behavior given the absence of that precursor, and the horizontal line that bisects each bar shows the unconditional probability of the target behavior. The bottom panel of each subject's graph shows the probability analysis for the potential precursors. In these panels, the light, solid gray bars show the probability of that precursor given the target behavior, the striped bars show the pr obability of that precursor given the absence of the target behavior, and the small, horizontal lines that bisect each bar show the unconditional probability of that precursor. Figure 2-1 shows results for Liv, Billy, C huck, and Amanda. Three precursors were identified for Liv. All 3 of her precursors were highly correlated with th e occurrence of target problem behavior: Property destru ction always occurred in tria ls in which the precursor was observed (p(T|P) = 1.0), although the probability of the target behavior given the absence of the precursors (p(T|~P)) also was high. In addition, Liv's precursors never occurred in trials in which the target behavior was not observed (i.e., p(P| ~T) = 0). Billy always engaged in the target behavior in trials in which at least 1 of the 3 selected precursor s also occurred. In addition, these

PAGE 27

27 precursors never occurred in trials in which th e target behavior was not observed. Chuck's precursors were somewhat less predictive than Li v's and Billy's precursors in that the target behavior was not always observed following the selected precursors, a nd his precursors were sometimes observed in the absence of the target behavior. Three precursor s were identified for Amanda. Hand postures and stretching did not occu r very often, but when they did they only occurred in trials with the targ et behavior. Reaching for the therap ist was not as predictive of the target behavior, as the target was not always observed in trials with this response, and the response sometimes occurred in tria ls without the target. Amanda's caregivers reported that she seemed to engage in higher rates of SIB (the ta rget behavior) in the pr esence of food. Therefore, food items were included in the tangible condition, and signing for food also met the precursor selection criteria. Figure 2-2 shows results for Kelly, George Amy, and Sammy. Th ree precursors were selected for Kelly. She always enga ged in the target behavior in trials in which she put her hand inside her clothes, although this behavior was observed in few trials. Mouthing her fingers and toes, as well as whining, were observed more freq uently, and the target behavior occurred more often in trials in which these responses were observed. She also frequently engaged in these precursors in trials in which the target was obser ved compared to trials in which the target was not observed. Six precursors were identified for George. Like Amanda, he also engaged in appropriate behavior (i.e ., signs such as "more," "play," etc. ) in trials in which the target behavior was observed. He did not, however, engage in the target be havior in all trials in which any of the precursors were observed, and he sometimes engaged in each of the precursors in trials in which the target was not observed. Te n precursors were identified for Amy. She always engaged in the target behavior in trials in which 9 of these behaviors occurred, and she never

PAGE 28

28 engaged in these precursors in trials in which the target be havior was not observed. Mouthing objects also was highly predictive of the target behavior; however, the target behavior occurred in trials in which this behavior was not observe d and mouthing objects also occurred in trials in which the target behavior was not observed. Six precursors were identified for Sammy. Tugging on the therapist's shirt occurred in only 1 trial of the assessment in which the target also was observed. His other precursors occurred more fr equently, but the target behavior frequently occurred in trials in which the precursor was not observed, and the precurso rs occurred in trials in which the target behavior was not observed. Figure 2-3 shows results for Renee, Curtis, Gerald, and Adam. Only 1 precursor was selected for Renee, although this behavior was no t highly predictive of th e target behavior. In other words, the precursor frequently occurred in trials in which the ta rget behavior was not observed, and the target behavior frequently occu rred in trials in which the precursor was not observed. Two precursors were selected for Cu rtis. Leg scratching only was observed during 1 trial of the precursor assessment, but it occurre d during a trial in whic h aggression also was observed; thus, the behavior met the precursor se lection criteria. Blocki ng the therapist from touching items also was selected, and it occu rred more often overall, even though aggression occurred during several trials in which blocki ng the therapist was not observed. Four precursors were identified for Gerald. The target behavior al ways occurred in trials in which three of the responses were observed, and those behaviors never occurred in trials without the target behavior. The 4 th precursor occurred more frequently; however, the target behavior sometimes occurred in trials in which the behavior was not observed, and the precursor often occurred in trials without the target behavi or. Six precursors were identifie d for Adam, all of which were strongly predictive of aggression. In other words, Adam always enga ged in the target behavior in

PAGE 29

29 trials in which the precursors were observed, and he never engaged in the precursors in trials in which the target behavior was not observed. Figure 2-4 shows results for Donald, Le igh, Guy, and Kevin. Six precursors were identified for both Donald, although the target behavi or only occurred in a ll trials in which 2 of the precursors were observed, and those precursors never occurred in trials in which the target behavior was not observed. Six pr ecursors also were identified for Leigh; the target always occurred in trials in which 4 of those beha viors were observed, and those precursors never occurred in trials without the target behavi or. Sixteen precursors we re identified for Guy, although the target behavior only always occurred in the same tr ials as 8 of these precursors, and those same precursors never occurred in trials without the target behavior. The other precursor occurred somewhat more frequen tly; however, those precursors were less predictive of the target behavior. Twelve precursors were se lected for Kevin, and the target always occurred in trials in which these behaviors were observed. His precursors were never observed in trials in which the target behavior was not observed. In summary, results show that each of the 16 subjects engaged in behaviors that were predictive of the occurrence of their target problem behaviors. The number of identified precursors ranged from 1 precursor (Renee) to 16 precursors (Guy). Comparison of Caregiver Report and Precursor Data When the precursors reported by caregivers were com pared to those identified by the precursor assessment (Table 2-4), caregivers for 6 of 16 subjects (Billy, Chuck, Renee, Leigh, Amanda, and Sammy) were unable to report any precursors whatsoever. Results for the remaining 10 subjects showed that caregivers repo rted only 10 of the 90 precursors identified via the precursor assessment (approximately 12%). Fu rthermore, caregivers of 7 subjects (George, Amy, Curtis, Gerald, Adam, Donald, and Kelly) reported additional poten tial precursors that

PAGE 30

30 differed from those identified via the precursor assessment. Even though it was possible for the subject to engage in the caregi ver-reported precursors during the precursor assessment, either the behaviors were never observed (George, Amy, Curtis, Gerald, Adam, and Donald) or they occurred but did not pred ict the occurrence of the target pr oblem behavior based upon the results of the probability analysis (Curtis, Geral d, and Kelly). Finally, th e precursor assessment identified precursors for all subjects that were not reported by caregivers, ranging from 1 (Renee) to 16 (Guy). Therefore, results of th is analysis suggest that caregivers are relatively inaccurate in identifying precursors.

PAGE 31

31 Figure 2-1. Precursor-assessment results fo r Liv, Chuck, Billy, and Amanda. The top and bottom graphs for each subject show probabi lities for target behavior and precursor behaviors, respectively.

PAGE 32

32 Figure 2-2. Precursor-assessment results for Kelly, George, Amy, and Sammy

PAGE 33

33 Figure 2-3. Precursor-assessment results for Renee, Curtis, Gerald, and Adam

PAGE 34

34 Figure 2-4. Precursor-assessment resu lts for Donald, Leigh, Guy, and Kevin

PAGE 35

35 CHAPTER 3 STUDY 2: FUNCTIONAL ANALYSIS OF PRE CURSOR AND PROBLEM BEHAVIORS Although the precursor assessment conducted in Study 1 identified behaviors that predicted (i.e., were correlated with) the target problem behavior, the extent to which these behaviors were members of the same response class remained unknown. Therefore, independent functional analyses (FAs) were conducted for a subset of subjects from Study 1 to identify the function of the identified precursors as well as the function of the target behavior. Unlike Smith and Churchill (2002) and Borrero and Borrero (2008), we conducte d the functional analysis of the precursors (precursor FA) first to determin e whether the function of the precursors matched the function of the target behavior (i.e., the target behavior was observed at the highest rates in the same condition of the target FA as precursors had been observed during the precursor FA) and to perhaps limit the number of occurrences of the target beha vior. If precursors and target behaviors were determined to be members of th e same response class, and if the majority of subjects did not engage in the target behavior during the precur sor FA, this information might prove useful to clinicians duri ng assessment and treatment of severe forms of problem behavior by minimizing risk to the subject and/or therapist. Method Subjects and Setting Eight individuals from Study 1 (Renee, Curt is, Gerald, Adam, Donald, Leigh, Guy, and Kevin) participated in Study 2. Subjects were se lected based on results of their precursor assessments, which identified different numbers of precursors across subjects. Therefore, FAs were conducted on a variety of response topographi es identified as precursors for these subjects, ranging from 1 (Renee) to 12 (Kevin) precursors in the initial assessments. All sessions were

PAGE 36

36 conducted in an observation room at a day program for adults or in designated areas of a classroom at a special education school. Response Measurement and Reliability One precursor, covering her eyes, was includ ed in the precursor FA for Renee. Two precurso rs (scratching his leg and blocking the therapist from movi ng) were included for Curtis. Four precursors were included for Gerald (flickin g his lips, grimacing, hitting the therapist, and jerking his head). Although 6 pr ecursors were identified for Adam, only 5 precursors (slouching, saying "No," putting paper in his mouth, turning away from the th erapist, and pushing materials away) were included in his pr ecursor FA. Grimacing was excluded because Adam frequently turned his head such that observers were unabl e to view his facial e xpression. Six precursors were included for Donald (hand flapping, pressi ng his hand to his mouth, snarling, clapping his hands, vocalizing negatively, and moving toward objects), as well as for Leigh (covering her eyes, resting her head, saying "Yeah yeah y eah," guiding the therap ist around the room, stomping or shuffling her feet, moving her hands in circles by her side, and motioning toward objects). Numerous precursors were identifi ed in Study 1 for Guy; however, only the 8 precursors (flopping on the ground, swing his body side to side, stomping his feet, biting objects, throwing objects, pushing materials away, crumpl ing paper, and cursing) that most strongly predicted the occurrence of the target behavior were include d in his precursor FA. Finally, twelve precursors (saying "No," grunting, dropping to or scooting on the floor, waving his arms in the air, pulling the therapist's arm, sa ying "Good boy," holding his knees, smiling, fidgeting, stacking chairs, head butting th e therapist, knocking on the tabl e) were included in Kevin's precursor FA because all of the precursors we re highly predictive of his target behavior. Observers recorded the frequenc y of (a) target behaviors, (b) precursors, and (c) therapist interactions and their delivery of consequences during continuous, 10-s intervals for each subject

PAGE 37

37 using a handheld PDA. The target problem beha vior was the most seve re form of behavior reported by caregivers during the initial in terview and/or direct observation by the experimenters. Operational definitions of the targ et problem behaviors for each subject are listed in Table 2-1. Interobserver agreement was assessed by havi ng a second observer i ndependently collect data during at least 25% of sessions. Proportional agreement pe rcentages were calculated for each response by comparing the two observers' r ecorded frequencies for all responses in each 10-s interval. The smaller number of responses was divided by the larger number of responses in each interval with a disagreement, the fractions were summed across all intervals, and the total was added to the total number of agreement in tervals in the session. Th e sum was divided by the total number of intervals in the session and multiplied by 100% to yield reliability scores for each measure. Mean reliability scores were as fo llows: Renee, 97.5% for target behaviors (range, 93.6% to 100%), 97.9% for precursors (range 90.8% to 100%), a nd 96.5% for therapist responses (range, 91% to 100%); Curtis, 99.4% for target behaviors (range, 92.5% to 100%), 99.3% for precursors (range, 93.3% to 100%), an d 98.9% for therapist responses (range, 91.6% to 100%); Gerald, 100% for target behaviors, 99. 8% for precursors (range, 94.7% to 100%), and 97.5% for therapist responses (range, 83.3% to 10 0%); Adam, 99.9% for target behaviors (range, 99.2% to 100%), 99.4% for precursors (range 89.3% to 100%), a nd 92.6% for therapist responses (range, 76.4% to 100%); Donald, 99.8% for target beha viors (range, 98.3% to 100%), 99.6% for precursors (range, 91.7% to 100%), and 97.9% for therapist responses (range, 90% to 100%); Leigh, 98.7% for target behaviors (range 90.3% to 100%), 98.7% for precursors (range, 88.6% to 100%), and 95.6% for therapist respons es (range, 80% to 100%); Guy, 98.2% for target behaviors (range, 86.1% to 100%), 99.9% for pr ecursors (range, 98.3% to 100%), and 97.7% for

PAGE 38

38 therapist responses (range, 91.7% to 100%); and Kevin, 97.6% for target behaviors (range, 90% to 100%), 99.4% for precursors (range, 88.3% to 100%), and 92.7% for therapist responses (range, 78.3% to 100%). Procedures Two independent FAs were conducted using pr ocedures sim ilar to those described by Iwata et al. (1982/1994). During the precursor FA, consequences were delivered for the occurrence of precursor behaviors but not for occu rrences of the target pr oblem behavior(s) (i.e., the target behavior was ignored ). During the target FA, cons equences were delivered for occurrences of the target behavior(s) but not for occurrences of the precursor behaviors. Attention, play, and demand conditions were incl uded in all FAs. An alone or ignore condition was not included if the target behavior was a ggression, and a tangible condition was included if caregivers indicated that the subj ect tended to engage in problem behavior when preferred items were removed or access to preferred items was denied. During attention sessions, the subject had access to 2-3 moderately preferred toys identified via a paired-stimulus (Fisher et al., 1992) or a multiple-stim ulus (DeLeon & Iwata, 1996) preference assessment. At the start of session, the th erapist told the subj ect, "I have some work to do, but you can play with these toys, if you'd like. The therapist then sat next to but did not interact with the subject. Contingent upon ea ch occurrence a precursor (in the precursor FA) or target (in the target FA) be havior, the therapist delivered a brief reprimand (e.g., "Stop doing that; that's not nice!") and ge ntle physical contact (e.g., placed a hand on the subject's arm). During play sessions, the subjec t had access to 2-3 highly prefe rred toys (identified in the preference assessment), and the ther apist interacted with the subj ect at least every 30 s or any time the subject initiated interact ion. No consequences were deliv ered following occurrences of either precursor or target behaviors.

PAGE 39

39 During demand sessions, the therapist continuousl y presented learning trials appropriate to the subject's functioning level us ing a 3-step prompting sequence and delivered praise following compliance. Contingent upon each instance of a precu rsor (in the precursor FA) or target (in the target FA) behavior, the therap ist removed the work materials and provided a 30-s break from the task. If the target behavior was SIB or property destruction, an alone or ignore condition was included in the FAs. During alone sessions, the s ubject was seated alone in a room without any materials. If an ignore condition was conducted, the subject was seated in an area of the room away from all other individuals, and no conse quences were delivered contingent upon any behaviors emitted by the subject. If caregivers reported that problem behaviors occurred when preferred items were removed or access to these items was denied, a tangible cond ition was included in the FAs. At the start of the session, the therapist removed all toys and rema ined near the subject. If the subject initiated interaction with the ther apist during this conditio n, the therapist briefly responded to the subject (e.g., quickly answered a question) then terminat ed interactions (e.g., "We can talk later."). Contingent upon the occurrence of a precursor (in th e precursor FA) or target (in the target FA) behavior, the therapist provided access to the toys for 30 s. Results and Discussion Figure 3-1 shows results of the independent FAs for Renee, Curtis, Gerald, and Ada m. Renee engaged in higher rates of eye covering during the demand condition of the precursor FA (aggression also was observed in this condition, although the rate of aggression was much lower than the rate of eye covering except during one session). During the ta rget FA (aggression), aggression maintained only in the demand cond ition throughout the assessment. (Renee also continued to engage in eye covering in the dema nd condition of the target FA, as well as during

PAGE 40

40 the play condition.) These results indicated th at Renee's eye covering and aggression were maintained by negative reinforcement (i.e., escape from academic tasks). Curtis engaged in higher rates of precursors in the tangible condition of the precursor FA (zero instances of aggression we re observed). During the target FA, Curtis engaged in higher rates of aggression in the tangible condition (ra tes of precursors also were observed). Thus, Curtis' precursors and aggression were mainta ined by positive reinforcement (i.e., access to preferred leisure items). Gerald engaged in higher rates of precursors in the demand condition of the precursor FA (SIB was observed during two dema nd sessions). During the target FA, Gerald engaged in higher rates of SIB (as well as precursor s) in the demand condition. These re sults indicate that Gerald's precursors and SIB were maintained by negative reinforcement (i.e., escape from academic tasks). Adam engaged in higher rates of precursors during the demand condition of the precursor FA (aggression was never observed). During the ta rget FA, Adam engaged in higher rates of aggression (and precursors) in the demand cond ition. These results indicate that Adam's precursors and aggression were maintained by negative reinforcement (i.e., escape from academic tasks). Figure 3-2 shows the results of the independent functional an alyses for Donald, Leigh, Guy, and Kevin. Donald engaged in higher rates of precursors in the ta ngible condition of the precursor FA (aggression was ne ver observed). During the target FA, Donald engaged in higher rates of aggression (and precursors) in the ta ngible condition. These re sults indicate that Donald's precursors and aggression were mainta ined by positive reinforcement (i.e., access to preferred items).

PAGE 41

41 Leigh engaged in higher rates of precursors in the tangible co ndition of the precursor FA (a higher rate of aggression was observed in one at tention session). During the target FA, Leigh engaged in high rates of SIB in the tangible and demand conditions (her pr ecursors also occurred initially during these conditions, but did not main tain in the demand condition). Thus, results of the precursor FA indicated that precursors were maintained by positive reinforcement (access to tangible items), whereas results of the target FA indicated that SIB was maintained by both positive reinforcement (access to tangible items) and negative reinforcement (escape from academic tasks). In other words, the precursor FA was effective in identifying one of two sources of reinforcement that maintained the target behavior. Guy engaged in higher rates of precursors in the demand condition of the precursor FA (somewhat lower rates of aggression also were ob served in this condition) During the target FA, Guy engaged in higher rates of aggression (a nd precursors) in the demand condition. These results indicate that Guy's precursors and aggression were maintained by negative reinforcement (i.e., escape from academic tasks). Finally, Kevin engaged in hi gher rates of precursors in the demand condition of the precursor FA (and even higher rates of propert y destruction). During the target FA, Kevin engaged in higher rates of property destruc tion (and precursors) during the demand condition. These results indicate that Kevin's precursor s and property destruction were maintained by negative reinforcement (i.e., escape from academic tasks); however, the precursor FA was not effective in reducing rates of his target problem behavior. Precursors were found to be members of the same response class as the target problem behavior for 7 of the 8 subjects. The 8 th subject's (Leigh's) precursors were maintained by 1 of 2 sources of reinforcement that also maintained her target problem be havior. In addition, the

PAGE 42

42 precursor FA eliminated instances of the target problem behavior for 3 subjects (Curtis, Adam, and Donald), and resulted in low rates of the target problem behavior for 4 subjects (Renee, Gerald, Donald, and Leigh), but di d not reduce rates of the target problem behavior for 1 subject (Kevin). A within-session analys is of Kevin's data revealed that during demand sessions, Kevin engaged in the target behavior before precursors prior to rein forcement delivery in 100%, 57%, and 33% of cases, respectively. It is possible that a recent history of reinforcement for the target behavior during the precursor as sessment might have shifted response allocation toward the target behavior, which extinguished over subs equent demand sessions. Therefore, although the precursor assessment results indicated that some responses predicted the o ccurrence of the target behavior, Kevin engaged in high rates of all beha viors, and his response allocation seemed to shift depending on which topographies contacted the reinforcement contingency. In fact, during the last 5 min of the third demand session of th e precursor FA, Kevin simply said "No" and received escape without engaging in other response topographies. It is interesting to note that s ubjects did not engage in all of their selected precursors during the precursor FA, except Renee, for whom only 1 precursor was identified. Therefore, the function of each subject's precursors was determined based on a subset of responses selected as precursors in Study 1. Proportions of observed precursors during the precursor FA for each subject are shown in Figure 3-3. Th e function of Curtis' precursors was determined based solely on occurrences of blocking the therapist. Gerald engaged in only 2 of 4 precursors (grimacing and head movements). Adam engaged in 4 of 5 precursors (saying "No", turning away, slouching, pushing materials away), and Donald engaged in only 2 of 6 precursors (negative vocalizations and reaching for objects). Leigh engaged in 3 of 6 precursors (hand circles, chin down, and stomping/shuffling her feet) during the precursor FA. Guy engaged in 6 of 8

PAGE 43

43 precursors (biting objects, pushing materials aw ay, cursing, body swingi ng, throwing objects, and stomping), and Kevin engaged in 6 of 12 precursors (saying "No," grunting, arm waving, table knocking, fidgeting with pants, a nd placing his hands on his knees). The function of unobserved precu rsors remains unknown. It is po ssible that either (a) the precursor assessment simply yielded a high number of false alarms or (b) these behaviors are members of the same response cl ass as the target problem be havior but were not observed because other precursors contacted the reinforcement contingency and maintained. Some precursors emerged in the same condition as the target behavior during the target FA for 4 subjects (Donald, Leigh, Guy, and Kevin). For ex ample, Donald engaged in snarling during the tangible condition of target FA, which could have emerged as a function of extinction for more frequently occurring precursors. Three of Le igh's precursors (covering her eyes, resting her head, and moving her hands in circles) occurred at higher rates only in the tangible condition during the target FA. Guy began to flop on the floor during the demand co ndition of the target FA, and Kevin began to drop to/scoot on the floo r during the demand conditi on of the target FA. Therefore, this provides some evidence that t hose precursor topographies might be maintained by the same source of reinforcement as the ta rget behavior and emer ged as a function of extinction of other precursors. Additional analyses would be required, however, to verify this possibility, which was beyond the scope of the present study. Further Comparison of Caregiver Report and Precursor Data : The comparison of precursors reported by caregivers and those iden tified by the precursor assessment conducted in Study 1 indicated that caregiver s only reported 10 of 90 precursors identified by the precursor assessment (approximately 12%). Given that many precursors identified by the precursor assessment were not observed during the functional analyses in Study 2, it seemed possible that

PAGE 44

44 caregivers might have identified precursors that were not observed during the precursor assessment but that were observed in the functio nal analyses. Therefore, precursors reported by caregivers and those obser ved during the functional analyses were compared for subjects in Study 2 to determine the extent to which caregivers reported precursors that emerged in the same condition(s) as the target behavi or during the functional analyses. Results showed that caregivers reported 7 of 31 precursors that were observed during the functional analysis (approximately 23%). These results suggest that caregivers were only slightly more accurate in reporting precursors than was concluded during the comparison conducted in Study 1.

PAGE 45

45 Figure 3-1. Results of the indepe ndent functional analyses for Re nee, Curtis, Gerald, and Adam in Study 2. The top and bottom graphs for each subject show rates of the precursors and target behaviors, respectively. The left panel of each set of graphs shows results of the precursor FA; the right panel of each set of graphs shows results of the target FA.

PAGE 46

46 Figure 3-2. Results of the inde pendent functional analyses for Donald, Leigh, Guy, and Kevin

PAGE 47

47 Figure 3-3. Proportional di stribution of precursor responses observed during the precursor FA (each section of a bar graph represents a different precursor) for subjects in Study 2. Numbers above each bar show the number of precursors observed out of the total number of selected precursors for each subject.

PAGE 48

48 CHAPTER 4 STUDY 3: PRECURSOR ASSESSMENT AS THE BASIS FOR INTERVENTION Given that the precu rsor assessment was effec tive in identifying precursors for all subjects (Study 1) and that these behavi ors typically were members of the same response class as the severe problem behavior (Study 2), the combination of precursor assessm ent and precursor FA seemed to be a promising basis for making conc lusions about the function of severe problem behavior while minimizing risk. This information presumably could then be used to design an effective reinforcement-based intervention to sp ecifically demonstrate reductions in precursors while maintaining low rates of severe problem behavior. Very few studies, however, have shown that an intervention aimed at reducing less severe behaviors can maintain low rates of more seve re behaviors. Results of a study by Shukla and Albin (1996), however, showed that one individual engaged in mild and severe problem behaviors during demand situations. During treatm ent, the participant was prompted to sign "break" following an instance of mild problem beha vior, which was effective in reducing rates of all problem behaviors and increa sing rates of communication. The procedure maintained a close temporal contiguity between the mild behavior and reinforcer delive ry, however, which could result in adventitious reinforcement of mild behaviors for some individuals. In addition, extinction was not programmed for the severe probl em behavior, which might not be effective in reducing rates of severe pr oblem behaviors for others. Only one published study to date has designed an intervention base d upon the results of a precursor FA. Najadowski et al. (2008) conducte d a precursor FA for 3 subjects and observed that all subjects' precursors were maintained by positive reinforcement (access to attention or preferred items). Severe problem behavior was never observed during the assessment. The intervention was similar to that used by Shukla and Albin (1996) and consisted of differential

PAGE 49

49 reinforcement of alternative behavior (appropria te requests) and extinc tion for both precursors and severe problem behavior. These procedures resu lted in a reduction in precursors, an increase in appropriate requests, and zer o instances of severe proble m behavior (except during one session for one subject during a reversal to base line). Therefore, although these results are very promising with respect to reducing risks posed by severe problem behaviors, the severe problem behavior either never occurred throughout assessment or treatment (2 subjects) or occurred in one session only (1 subject), thus limiting the conclusions that can be drawn with respect to the function of the severe problem behavior. In ad dition, subjects were prompted to engage in appropriate requests following instances of prec ursors, which could prove problematic for some individuals (as de scribed above). Another promising approach to decreasi ng problem behaviors, while strengthening appropriate forms of communication, is a combin ation of noncontingent reinforcement (NCR) and differential reinforcement of alternative be havior (DRA) procedures. Marcus and Vollmer (1996) evaluated this treatment for 2 individua ls and found that the in itial, continuous NCR schedule eliminated problem behaviors; how ever, appropriate communication (alternative behavior) only emerged as the NCR schedule was thinned. Because dense NCR schedules of reinforcement were implemented only briefly fo r both subjects, it was unc lear if appropriate communication might have emerged eventually u nder the dense NCR schedule. Therefore, Goh, Iwata, and DeLeon (2000) evaluated the NCR plus DRA intervention and provided subjects with extended exposure to the dense NCR schedule with DRA in place. They found that appropriate communication did not emerge under the dense NCR schedule and only emerged when the NCR schedule was thinned. Rates of problem behavior generally remained low in both studies under the combined NCR plus DRA intervention. Thes e results suggest an alternative treatment

PAGE 50

50 strategy in which NCR might be used initially to suppress problem behaviors, and DRA could be added to treatment during NCR schedule thinning to produce increases in appropriate, alternative behaviors. A review of 26 studies using NCR (9 studies) and/or DRA (18 studies) interventions was conducted to determine the relative effectiven ess of each intervention in quickly suppressing problem behaviors. Studies were included in the analysis if: (a) NCR was delivered continuously throughout sessions, (b) DRA was implemented usi ng an FR-1 schedule of reinforcement, (c) functional reinforcers were used during NCR or DRA, (d) no othe r treatment components were included (e.g., response blocking, timeout, etc.), and (e) extinction was programmed for problem behavior. Results revealed that continuous schedules of NCR were more effective than DRA in eliminating problem behaviors during the first session of treatment compared to DRA. More specifically, NCR resulted in zer o instances of problem behavi or during the first session in 67.57% of cases, whereas DRA resulted in zero in stances of problem behavior during the first session in only 35.94% of cases. Additionally, N CR was more effective than DRA in reducing overall rates of problem behavi ors during treatment: The mean reduction in problem behavior under NCR was 91.76%, whereas the mean reduction under DRA was 81.16%. One noted disadvantage of NCR is that th e procedure does not specifical ly teach the individual an appropriate means of obtaining reinforcers (Carr et al., 2000; Vo llmer, Iwata, Zarcone, Smith & Mazaleski, 1993); however, the results of this revi ew suggest that dense schedules of NCR might prove advantageous for suppressi ng severe problem be haviors before teaching an appropriate form of communication. The purpose of this Study 3 was to determin e whether an effective treatment could be designed based on the results of precursor as sessments alone while (a) obtaining indirect

PAGE 51

51 evidence regarding the function of severe problem behavior (i.e ., the target behavior) and (b) maintaining low rates of the target behavior throughout assessment and treatment. In general, treatment consisted of a sequen ce beginning with continuous NCR, in which the reinforcer was identified from a precursor FA. Subsequent ly, the NCR schedule was thinned, and DRA was introduced for an alternative behavior that served the same function as the precursor behaviors. Method Subjects and Setting Because the intervention strategy con sisted of sequential intro duction of social reinforcers (noncontingent followed by continge nt), only individuals whose precu rsor FA suggested that the behaviors were maintained by social reinforcem ent were included in th is study. Two individuals from Study 1 (Amanda and Sammy) who engaged in severe problem behavior, who had not participated in Study 2, and who were not particip ating in other projects participated in Study 3. Three other individuals exhibited problem behavior (precursors and/ or the target behavior) that appeared to be maintained by automatic reinforc ement and, thus, were not included. All sessions were conducted in a classroom at a special education school. Response Measurement and Reliability Am anda's target behavior was SIB and was defined as hitting her face or head with her hand or fist. Her precursors included reaching for the therapist, a hand posture (i.e., placing her elbow on the table with he r wrist bent), and stretching (i.e., l eaning back in her chair with her arms stretched above her head). The response sele cted to be strengthened as an appropriate, alternative form of communication for reinforc ement (i.e., mand) was signing "food," which was already in her repertoire and was shown to be predictive of SIB during the precursor assessment (Study 1).

PAGE 52

52 Sammy's target behavior was aggression and was defined as hitting, kicking, biting, head butting, or throwing objects that hit others His precursors included tugging on the experimenter's shirt, mouth movements (i.e., opening and closing his mouth without making noise), climbing on furniture, running across the room, throwing or pushing furniture, and angry vocalizations (i.e., growling or guttural sounds). Th e responses selected to be strengthened as appropriate, alternative forms of communication for reinforcement (i.e., mands) were signing "break" (during treatment for behavior maintained by negative reinforcement) or signing "play" (during treatment for behavior main tained by positive reinforcement). Data were collected and interobserver agreem ent was calculated as previously described for the precursor assessment (described in Study 1) and precursor FA (Study 2). In addition, the same data collection methods and interobserver agreement measures were used in treatment conditions as in the precursor FA, and interobser ver agreement was assessed during at least 25% of sessions in each condition of the precursor FA and in each condition of treatment for both subjects. During the precursor FA, mean interobserve r agreement scores for Amanda were 99.5% for precursors (range, 92% to 100%) and 99.2% for SIB (range, 93.6% to 100%). Mean agreement scores for Sammy were 98.9% for pr ecursors (range, 88.1% to 100%) and 98.2% for aggression (range, 90.3% to 100%). During treatmen t, mean interobserver agreement scores for Amanda were 98% for precursors (range, 80% to 100%), 98.4% for SIB (range, 92.5% to 100%), and 99.1% for mands (range, 79.9% to 100%). During treatment for Sammy's behaviors maintained by negative reinforcement, mean in terobserver agreement scores were 98.2% for precursors (range, 75% to 100%), 97.4% for a ggression (range, 81.3% to 100%), and 96.6% for mands (range, 85% to 100%). During treatment for Sammy's behaviors maintained by positive

PAGE 53

53 reinforcement, mean interobserver agreement scores for Sammy were 99.4% for precursors (range, 95% to 100%), 100% for aggression, a nd 98.3% for mands (range, 93.3% to 100%). Procedures Precurso rs identified in Study 1 for each subject were included in a precursor FA in which consequences were provided following precursor s only (i.e., no consequences were provided following the target behavior). C onditions of the FA were the same as described in Study 2 and consisted of attention, demand, play, and tangibl e. An ignore condition also was included for Amanda to rule out the possibility that SI B was maintained by automatic reinforcement. Treatment was based upon the re sults of the precursor FA onl y (i.e., the function of the target behavior was inferred from response pa tterns during this assessment) and generally consisted of baseline, continuous NCR, and NCR schedule thinning plus DR A. All sessions were 10 min in duration. Baseline Baseline was identical to the c ondition of the precursor FA in which the highest rates of precursors were observed. These were the tang ible (Am anda and Sammy) and demand (Sammy) conditions. Consequences were delivered followi ng precursors only (i.e., a small piece of food for Amanda, 30-s escape from academic tasks for Sammy in the first treatment, or 30-s access to toys for Sammy in the second treatment). No consequences were delivered following the target behavior or appropr iate communication. Continuous NCR The reinforcer shown to m aintain precursors during the precursor FA was delivered freely and noncontingently throughout each session. No consequences were delivered following precursors, the target behavior or appropriate comm unication in this condition. During treatment for behaviors maintained by positive reinforcemen t, subjects had continuous access to highly

PAGE 54

54 preferred food (Amanda) or leisure items (Sammy ). During treatment for behaviors maintained by negative reinforcement, no demands we re placed on Sammy throughout session. NCR Schedule Thinni ng Plus DRA Once low, stable rates of precursors and the target behavior were observed under continuous NCR, the NCR schedul e was thinned by removing one 10s interval of NCR per min using procedures similar to those described by Goh et al. When precursors were observed at rates less than 80% of baseline rates and rates of the target behavi or were low, schedule thinning progressed by removing another 10-s interval of NCR per min. At the start of this condition, DRA was implemented during intervals in which NCR was not pr ovided by physically prompting the participant to engage in appr opriate communication. The prompts then were systematically delayed (i.e., the therapist wait ed a few additional seconds before prompting appropriate communication) to allow the subj ect to engage in a ppropriate communication independently. Prompts were removed when the subject began to exhibit independent appropriate communication consistently. Contingent upon appropriate communication (independent or prompted) the therapist delivered the reinfo rcer. No consequences were delivered following the target behavior. Additio nal treatment components were implemented as needed and are described in the results section for each subject. Results and Discussion Results of the precursor FA showed that Amanda's precursors were maintained by positive reinforcement in the form of access to preferre d food items (Figure 4-1). Amanda also engaged in SIB at increasing rates in the tangible condition, ev en though no consequences were provided for this behavior. These results suggested that her target proble m behavior was likely maintained by the same source of rein forcement as precursors.

PAGE 55

55 During treatment (Figure 4-2), Amanda engage d in moderate rates of precursors during baseline (mean, 3.1 rpm). She also engaged in moderate rates of SIB (mean, 1.1 rpm) and low rates of mands (mean, 0.2 rpm), even though no consequences were provided for these behaviors. When continuous NCR was implemen ted, Amanda did not exhibit any precursor behaviors, and rates of SIB and mands were low (means, 0.1 and 0.3 rpm, respectively). During NCR schedule thinning plus DRA, Amanda e ngaged in variable and increasing rates of precursors (mean, 1.1 rpm), SIB (mean, 0.2 rpm), and independent mands (mean, 1.6 rpm). It appeared that a response hierar chy was developing in which Ama nda engaged in precursors (and sometimes SIB) followed shortly by independent mands. Therefore, it seemed possible that precursors and SIB might have been adventitiously reinforced as a result of the close temporal contiguity between those behavi ors and the delivery of reinfo rcement for mands. Other factors might have accounted for the development of this particular response hi erarchy, including the presumably low effort necessary to engage in precursors (primarily reaching toward the therapist), a recent history of reinforcement fo r precursors during baseline, and possibly a recent history of reinforcement for SIB outside of th e experimental setting. Therefore, response blocking was added to the treatment to disrupt the development of a response hierarchy, and blocked responses were scored and included in the session rate. This intervention resulted in decreasing rates of precursors (mean, 0.7 rpm), near-zero rates of SIB (mean, 0.03 rpm), and increased rates of independent mands (mean, 3.3 rpm). Next, a revers al to baseline was conducted and resulted in increased rates of pr ecursors (mean, 3.0 rpm), low rates of SIB (mean, 0.3 rpm), and variable rates of mands (mean, 2.0 rpm). A return to NCR thinning plus DRA and response blocking resulted in decreasing rates of precursors (mean, 0.4 rpm), zero rates of SIB, and high rates of independent mands (mean, 4.0 rp m). By the end of this phase, Amanda was

PAGE 56

56 engaging primarily in independent mands, which seemed to preclude the need for NCR and response blocking. Therefore, DRA alone was evalua ted and resulted in low rates of precursors (mean, 0.1 rpm), low rates of SIB (mean, 0.2 rpm), and similar rates of independent mands as in the preceding phase (mean, 4.1 rpm). Results of the precursor FA s howed that Sammy's precursors (F igure 4-3) were maintained by both positive reinforcement ( access to preferred leisure items ) and negative reinforcement (escape from demands). In addition, Sammy engaged in increasing rates of aggression in the demand condition, and he only engaged in aggression during the first session of the tangible condition. These results suggested that the identified function of precursors likely matched the function of his target problem behavior. Two treatments were evaluated for Sammy: the first treatment targeted problem behaviors maintained by negative reinforcement (Figure 44), and the second treatment targeted problem behaviors maintained by positive reinforcement (Figure 4-5). During treatment for problem behaviors maintained by negative reinforcem ent, Sammy engaged in moderate rates of precursors during baseline (mean, 3.1 rpm), low rates of aggression (mean, 0.4 rpm), and low rates of independent mands (mean, 0.1 rpm). Wh en continuous NCR was implemented, Sammy engaged in low rates of precursors (mean, 0.3 rpm), near-zero rates of aggression (mean, 0.03 rpm), and zero independent mands. During NCR schedule thinning plus DRA, he engaged in variable rates of precursors (mean, 1.6 rpm), va riable, increasing rates of aggression (mean, 0.8 rpm), and increasing rates of independent mands (mean, 1.2 rpm). Like Amanda, it seemed that Sammy was exhibiting a response hierarchy in which he enga ged in precursors as the experimenter approached to deliver a dema nd and, when escape was not provided for the precursors, he engaged in an independent mand (o r sometimes aggression). Therefore, a type of

PAGE 57

57 change-over delay was added to treatment in which independent mands were prevented immediately following a precursor behavior, and Sammy was physically guided to complete the demand issued by the experimenter. When he had not engaged in a precursor behavior for 5 s, Sammy was permitted to mand for escape and a 30 -s break was provided at that time. This resulted in an initial burst in precursors, which decreased over subsequent sessions (mean, 1.7 rpm), decreasing rates of aggression (mean, 0.5 rpm), and steady rates of independent mands (mean, 1.6 rpm). A reversal to baseline then wa s conducted, and rates of precursors increased (mean, 2.2 rpm), rates of aggression were near zero (mean, 0.1 rpm), and low rates of mands were observed (mean, 0.5 rpm). NCR thinning plus DRA with the change-over delay again was implemented and resulted in decreasing rate s of precursors (mean, 0.7 rpm), low rates of aggression (mean, 0.3 rpm), and increased rates of mands (mean, 1.5 rpm). By the end of the condition, NCR seemed unnecessary and, thus, was removed in the final phase. DRA plus the change-over delay alone resulted in decreasing rates of precursors to near zero (mean, 0.5 rpm), near-zero rates of aggression (mean, 0.3 rpm), an d steady rates of independent mands (mean, 1.6 rpm). During treatment for problem behaviors maintained by positive reinforcement (access to preferred leisure items), Sammy engaged in mode rate rates of precursors during baseline (mean, 2.1 rpm), near-zero rates of aggression (mea n, 0.03 rpm), and zero independent mands. During continuous NCR, Sammy engaged in near-zero rates of precursors (mean, 0.2 rpm) and zero instances of aggression and mands. When NCR schedule thinning plus DRA was implemented, Sammy engaged in decreasing rates of precursors (mean, 0.3 rpm), near-zero rates of aggression (mean, 0.01 rpm), and increasing rates of independent mands (mean, 1.3 rpm). A return to baseline resulted in increasing ra tes of precursors (mean, 1.6 rpm) near-zero rates of aggression

PAGE 58

58 (mean, 0.03 rpm), and decreasing rates of mands (mean, 0.6 rpm) When NCR schedule thinning plus DRA was again implemented, precursors decr eased (mean, 0.8 rpm), aggression occurred at higher rates during the first session but remained at zero for all subsequent sessions (mean, 0.2 rpm), and moderate rates of mands were obs erved (mean, 1.4 rpm). Th e NCR component was removed in the final phase, and Sammy engage d in decreasing rates of precursors (mean, 0.3 rpm), zero rates of aggression, and incr easing rates of ma nds (mean, 1.5 rpm).

PAGE 59

59 Figure 4-1. Results of the precu rsor FA for Amanda in Study 3.

PAGE 60

60 Figure 4-2. Treatment results for Amanda in Study 3.

PAGE 61

61 Figure 4-3. Results of the pr ecursor FA for Sammy in Study 3.

PAGE 62

62 Figure 4-4. Treatment for behavior mainta ined by negative reinforcement for Sammy.

PAGE 63

63 Figure 4-5. Treatment for behavior maintained by positive reinforcement for Sammy.

PAGE 64

64 CHAPTER 5 DISCUSSION The presen t studies examined the relation betw een precursor and problem behavior in three stages: empirical identif ication and selection of precursor responses (Study 1), response-class analysis of precursor and problem behavior (Stu dy 2), and evaluation of treatment based on the functional analysis of precursor behavior (St udy 3). Taken together, re sults indicated that precursor behaviors are both common and readily id entifiable, that they often are members of the same class as problem behavior, an d that they may be used as th e basis for developing effective interventions. Study 1 evaluated an objective yet brief me thod for identifying precursors to problem behavior, and results of the assessment indicated that all 16 subjects engaged in at least one precursor response. In addition, the assessment re quired very few instances (10 or fewer) of the severe problem behavior to identify precursors; thus, the trial-based pr ecursor assessment seems to be a viable method of assessing severe prob lem behavior while mini mizing the risks posed by dangerous topographies. The fact that problem behavior often is preceded by precursors suggests that problem behavior may simply be the termin al response in a hierar chy that begins with mildly annoying, disruptive behaviors (e.g., negative vocalizations, pushing materials away, etc.) or appropriate behaviors that ar e not reinforced (e.g., saying "No" or signing "food"). If so, it is surprising that caregivers rarely were able to id entify precursors. In f act, caregivers for only 6 subjects were able to report potential precu rsors, and the reported precursors matched the behaviors identified by the precu rsor assessment in approximat ely 12% of cases. When the precursors reported by caregivers were comp ared to those actually observed during the functional analyses for subjects in Study 2, correspondence only increased to 23% of cases. It is possible that caregivers are not as attentive when problem beha viors are not occurring and miss

PAGE 65

65 the occurrence of precursors. Th ese behaviors still might become members of the same response class as problem behavior if precursors are foll owed quickly by problem behavior. Therefore, these behaviors might become members of the same response class as a result of the close temporal contiguity between the precurso rs and problem beha vior (Catania, 1971). Correspondence between caregiver report and results of the precu rsor assessment might have been higher, however, if caregivers had been gi ven access to the precursor checklist during the interview. The presence of example response top ographies might have prompted caregivers to report potential precursors in cases in which none were reported or additional response topographies that were no t reported independently. There are, however, some limitations of usi ng a trial-based method to identify precursor behaviors. First, the procedure may not be practical for very low-rate problem behavior, because the duration of trials may be too brief to evoke the target problem behavior. Wallace and Iwata (1999) compared results from functional analysis sessions based on 5-, 10-, and 15-min durations and found that some individuals did not engage in the target problem behavior until session conditions had been in effect longer than 5 min. Therefore, the 5-min trials of the precursor assessment might not prove to be a useful asse ssment method for some individuals' problem behavior. In fact, precursor asse ssments could not be completed for 3 potential subjects because their problem behavior was seen rarely; they subsequently were assessed during functional analyses with extended session durations. This methodology also might not be useful for very high-rate problem behavior, beca use short inter-response times fo r the target behavior would reduce the likelihood of observing ot her behaviors that could be id entified as precursors. This problem was encountered with Amy and required the inclusion of play trials to provided periods of time in which the target behavior was not obs erved in order to calculate some probabilities

PAGE 66

66 calculations (e.g., the probability of the target behavior given th e absence of the precursor and the probability of the precursor given th e absence of the target behavior). Results of Study 2 verified that responses se lected from the precursor assessment were members of the same response clas s as the target problem behavior for 7 of 8 subjects (precursor and target behaviors matched for one of two functions for the 8 th subject). In addition, the precursor FAs eliminated the occurrence of the severe problem behavior for 3 subjects and reduced rates of the severe behavior for 4 ot her subjects. Taken toge ther, these findings are significant in validating a combined precursor as sessment and precursor functional analysis as a viable method for identifying contingencies that maintain severe problem behavior. Because not all precursors actually were obs erved for each subject during the precursor FA, it is unclear whether the unobserved precursors were members of the same response class as the target behavior. One possibi lity is that subjects simply allocated responding toward topographies that initially contacted the reinfor cement contingency, whereas the other precursors extinguished but were members of the same response class as the target behavior. Indirect evidence of this can be seen in cases in which previously unobs erved precursors emerged in the same condition as the target problem behavior during the target FA. Pr evious research has demonstrated this type of finding (e.g., Mag ee & Ellis, 2000; Richman, Wacker, Asmus, Casey, & Andelman, 1999), in which placing the most frequently observed response topographies on extinction resulted in increased rates of other topographies of problem behavior, and similar effects have been shown with respect to increas es in adaptive behavior s (Grow, Kelley, Roane, & Shillingsburg, 2008). More specif ically, Grow et al. showed th at appropriate but infrequent forms of communication might emerge when problem behaviors are placed on extinction, although the extinction bursts obser ved using this method of alternative response selection might

PAGE 67

67 preclude its use with severe prob lem behavior. Given the results of these studies, it is possible that the selective extinction of observed precursors might have clar ified the results of the current study; however, because the function of observed precursors matched the function of the target behavior in almost all cases, additional analys es seemed unnecessary given the purpose of this study. Another potentially influential variable in dete rmining subjects' response allocation toward particular precursor topographi es is the relative response e ffort required to emit some topographies compared to others. In fact, many of the identified precursors did not appear to require much effort (e.g., negative vocalizations), which incidentally seem to be identified as precursors frequently in previous research (Borrero & Borrero, 2008; Lalli et al., 1995; Najadowski et al., 2008; Smith & Churchill, 2002). No specific procedures were used to determine the relative effort requ ired for precursors or the target prob lem behavior in this study, however, although previous research has shown th at effort can influence response allocation toward adaptive and problem behaviors. For example, Horner and Day (1991) compared the effects of teaching a high-effort (full sentence signs) versus low-effort (single-word sign) functional, alternative responses for one individual whose prob lem behavior was maintained by negative reinforcement. They found that the individual engaged in low rates of problem behavior and high rates of communication only when th e alternative response required less effort. Therefore, subjects in the curren t study could have engaged in th e less effortful behaviors, which maintained following reinforcement during the pr ecursor FA, thus precluding the occurrence of other precursors. An alternative explanation for unobserved resp onse topographies during the precursor FA is that the current methodology simply yields a high rate of false alarms when identifying

PAGE 68

68 precursors. The criterion for including a potentia l precursor in the proba bility analyses was simply its occurrence within a trial, and interpre tations of the probability analysis results were based on relative (rather than ab solute) values of conditional and unconditional probabilities for each potential precursor. This method was used b ecause it seemed to be a simple, conservative method for identifying responses that preceded and were correlated with th e target behavior, but it might have resulted in the sel ection of precursors that did not o ccur frequently be fore the target behavior (false alarms). Additi onally, the termination criterion for the precursor assessment was 10 trials in which the target prob lem behavior was observed; therefor e, a high rate of false alarms might have resulted as a function of the brevit y of the assessment. No attempt was made to standardize the number of 5-min trials in which the target behavior was not observed for most subjects (except for Amy, who engaged in propert y destruction during the first 10 trials of the assessment); however, the total duration of trials without the target problem behavior was nearly equal to the total duration of trials with the target behavior for all subjects in the current study. Future research might evaluate different precurso r selection or assessment-termination criteria in an attempt to clarify these results. For example, one might (a) examine relative frequencies of precursors in trials with and wit hout the target behavior, (b) select responses that tend to occur closer in time before the target behavior, or (c) appl y different interpretative rules to analyze the probability data. More specifically, future rese arch might examine more stringent selection criteria by selecting only responses with high proba bilities of the target given the precursor and vice versa (e.g., a probability of 0.6 or higher). Another option would be to conduct more trials in which the target problem behavior is unlikely to o ccur to, resulting in a larg er sample of behavior for determining the probability of observing pote ntial precursors in the absence of the target problem behavior.

PAGE 69

69 It also is important to note that only half of the relation be tween precursors and the target behavior was examined in the current study by de termining the function of responses that were predictive of the target behavior. It remains unknown whether responses that were not predictive of the target behavior were maintained by a differe nt source of reinforcement. It is possible that any behavior that contacted the reinforcement c ontingencies might maintain and be effective in predicting the function of the target problem behavior, ev en though these behaviors might not necessarily be observed in typical settings. Renee's data most clos ely approximate this possibility in that only 1 precursor to aggressi on was selected, and it did not appear to strongly predict the occurrence of the target behavior. Re sults of independent FAs, however, showed that both behaviors were members of the same response class. It seems unlikely that responses that are not predictive of the target behavior woul d occur under similar conditions as the target behavior, contact the reinforcem ent contingency, and maintain, however, given that antecedent conditions (EOs) were specifically arranged during the precursor assessment to evoke the target problem behavior. Presumably, othe r responses that are sensitive to that source of reinforcement also would be observed, and beha viors maintained by a different source of reinforcement would be less likely to be observed. In this way, responses that are like ly members of the same response class as the target problem behavior would probably be id entified during the precursor assessment. Future research might compare the resu lts of functional analyses of behaviors that do not predict the target problem be havior to the results of a func tional analysis of the target behavior to determine the extent to which th ese non-predictive behaviors are maintained by the same or different sources of reinforcement. Finally, it is highly unlikely that a precursor FA would be ef fective in reducing instances of the target problem behavior if it were maintained by automatic reinforcement because

PAGE 70

70 arranging social consequences fo r precursors would not be expected to affect the rate of the target behavior. This would not, however, necessar ily preclude the develo pment of an effective treatment for behavior maintained by automa tic reinforcement based upon the results of precursor assessments. For example, Hagopian et al. (2005) were able to design treatment following an assessment of precursors by blocking stereotypy (hand flappi ng) that predicted the occurrence of SIB (eye poking) for one individual. This subject's SIB was maintained by automatic reinforcement, and blocking the pr ecursor (stereotypy) was shown to be more effective in reducing both stereotypy and SIB th an blocking SIB alone. Therefore, these results suggest that precursor analyses pe r se may have some clinical util ity regardless of the function of problem behavior and even if rates of the target problem be havior do not decrease during assessment. Results of Study 3 showed that effective reinforcement-based interventions can be designed based on the results of precursor anal yses only. Although severe problem behavior was not eliminated during the precursor FA or baseline, lower rates of the target behavior were observed relative to rates of precursors. Theref ore, if reinforcement contingencies had been placed on the target behavior (i.e., no reinforcem ent for precursors) during the FA and baseline, it is probable that higher rates of the target beha vior would have been exhibited by both subjects. The effects of continuous NCR replicated the re sults of previous studies (Goh et al., 2000; Marcus & Vollmer, 1996) in that nearly all re sponding (precursor and target behavior) was suppressed under these conditions, a nd subjects did not emit the a ppropriate alternative response (mand). As the DRA component was introduced while the NCR schedule was systematically thinned, both subjects acquired the mand; however, these proce dures were not effective in reducing precursors while maintaining low rates of the target behavior in 2 of 3 cases. The

PAGE 71

71 addition of a response blocking component (Amanda ) or a change-over delay (Sammy, treatment for behavior maintained by negativ e reinforcement) was effective in reducing rates of precursors and target behaviors, while mands maintained under the DRA contingency. When low rates of precursors and target behaviors were attaine d, the additional treatme nt components (NCR and blocking for Amanda and NCR for Sammy) were removed and similar effects on all behaviors were observed. (The change-over delay compone nt remained in Sammy's final treatment package due to the severity of his aggression, although it was rarely implemented during the final sessions of the evaluation.) Therefore, both s ubjects allocated responding toward mands under conditions that would typically be encountered in their classrooms (i.e., Amanda could sign to receive food and Sammy could sign to receive a break from work). It is also interesting to note that no additional treatment components were necessary to reduce Sammy's precursors and maintain low rates of the target behavior during the second intervention evaluation. In fact, the second evaluation was completed in approximately of the number of sessions required to complete the first evaluation. As in Study 2, not all precursors identified vi a the trial-based precursor assessment actually were observed during the precurs or FA in Study 3. For example, the function of Amanda's precursors was determined primarily by the occu rrence of reaching toward the therapist (i.e., hand postures and stretching were observed rare ly). The positive rein forcement function of Sammy's precursors was determined solely by th e occurrence of angry vocalizations, and the negative reinforcement function was determ ined by the occurrence of climbing, angry vocalizations, mouth movements, and moving fu rniture. Two of Sammy's precursors (running across the room and tugging on th e therapist's shirt) were ne ver observed. As previously discussed, it is possible that m odifications to the methods of da ta analysis and/or precursor

PAGE 72

72 selection criteria might lead to better predictions with respect to which response topographies are likely members of the same response class as the target behavior and, t hus, would be exhibited during the precursor FA. It also seems possible that improvements in the precursor assessment methodology might have resulted in greate r initial behavioral reductions unde r NCR schedule thinning plus DRA. For example, if responses other than the select ed precursors also pred icted the occurrence of severe problem behavior, their inclusion in trea tment could have prevented the occurrence of the target behavior. This is because the greater the number of responses that occur before severe problem behavior and contact th e reinforcement contingency, the less likely severe problem behavior would be emitted. Anecdotally, it did no t appear that either subject exhibited other "precursor" behaviors that were not identified via the precursor a ssessment. In fact, the precursor exhibited most frequently by Amanda during trea tment was reaching toward the therapist, and the precursor exhibited most frequen tly by Sammy was angry vocalizations. An alternative explanation for the initial poor treatment effects with NCR schedule thinning plus DRA is that some of the precursors selected for inclusion during treatment actually were members of a different response class than the target behavior. For example, some response topographies that were selected as precursors via the precursor assessment actually were not observed during the precursor FA ; therefore, the f unction of these "precursors" was unknown. Given that the responses were selected suggests that they occurred at sufficient rates in general to be detected, and it could have been mere coinci dence that the responses occurred in trials in which the target behavior also was observed. If this were true, these re sponse topographies could have been exhibited during treatment and detected simply as a function of extended observation periods. Anecdotally, Sammy engaged in some pr ecursor topographies in the absence of the

PAGE 73

73 establishing operation (i.e., when demands were not presented or when preferred items were not removed), which suggests that these behaviors might have been maintained by a different source of reinforcement. Alternatively, features of th e environment might have been discriminative for the presentation of demands or removal of prefer red items, thus evoking some of the precursors in the momentary absence of the establishing operation. These possibilities remain speculative as additional analyses of selected precursors were beyond the scope of the present study. In summary, the current series of studies demonstrates a method of analyzing precursor behavior and for progressing from assessment to treatment of severe problem behaviors while minimizing risk posed by those behaviors. Othe r methods of potentially reducing risk during assessment include the use of protective equi pment (Le & Smith, 2002), a different dependent variable such as latency to problem behavior (Thomason, Iwata, Neidert, & Roscoe, in press), and brief session durations (Wallace & Iwata, 1999) The advantage of precurs or analyses is that reinforcement contingencies are not placed on se vere problem behaviors, thus decreasing the likelihood that severe behaviors wo uld occur at high rates during th e assessment and/or continue to occur following the assessment period. In addition, Study 1 provides a new method of empirically identifying precursor be haviors, and results of Study 2 verified that the identified precursors typically are members of the same res ponse class as severe pr oblem behavior. Finally, the sequential introduction of NCR and NCR schedul e thinning plus DRA app ears to be a viable treatment option for shifting response allocation fro m problem behavior to appropriate behavior while maintaining low rates of severe problem be havior and reducing ris k. The results of Study 3 also indicate that this intervention strategy is appropriate for problem behavior maintained by positive and/or negative reinforcement and that NCR can be graduall y thinned such that appropriate behavior maintain s under DRA contingencies only.

PAGE 74

74 Table 2-1. Subject characteristics Name Age Classification Definiti on of Target Problem Behaviors Liv 10 Down syndrome Property destruction (throwing items and knocking over furniture) Billy 15 Down & Kleinfelter's syndromes Clothing destruction (ripping, tearing, or unraveling socks) Chuck 14 Arthrogryposis syndrome SIB (head hitting) Kelly 10 Seizure disorder & retinopathy SIB (self biting) George 9 Autism Aggression (h itting, kicking, pinching, and biting) Amy 3 Down syndrome Property destruction (throwing objects, tearing materials from walls, and destroying materials) Renee 15 Angelman's syndrome Aggre ssion (hair pulling, hitting, and pushing) Curtis 13 Autism Aggression (hitting, kicking, biting, and head butting) Gerald 19 Cerebral palsy, MR (level unspecified) SIB (hand biting) Adam 11 Prader-Willi syndrome Aggr ession (hitting, kicking, biting, and throwing objects that hit people) Donald 14 Autism, seizure disorder Aggression (hitting, kicking, biting, and head butting) Leigh 13 Trainable mentally handicapped & language impaired SIB (chin hitting and banging) Guy 12 Autism Aggression (h itting, kicking, biting, and head butting) Kevin 54 Severe MR, seizure disorder Pr operty destruction (throwing furniture, pounding on walls, and destroying or throwing materials) Amanda 18 Autism, profound MR SIB (face and head hitting) Sammy 6 Deaf, learning disabilities A ggression (hitting, kicking, biting, head butting, and throwing objects that hit people)

PAGE 75

75 Table 2-2. Precursor checklist Category Examples Vocalizations Screaming, laughi ng, cursing, squealing, requests Facial Expressions Smiling, grimacing, frowning, surprised Postures Slouching, dropping, head down, standing Locomotion Walking, running, jumping Repetitive Motor Movements Fidgeting, tapping fingers, tapping feet, stomping, hand flapping, head movements, hair twirling, nail picking, clapping Object Manipulation Playing with obj ects, tapping pencil, twirling objects Other Problem Behaviors Self-injurious Behavior Head banging, head hitting, skin picking, body hitting, self-biting, hair pulling Aggression Hitting, kicking, grabbin g, head butting, biting, scratching Property Destruction Breaking objects, knocking over furniture, banging objects, throwing objects, hitting surfaces, kicking surfaces

PAGE 76

76 Table 2-3. Probability analysis formulas Probability Type Formula Conditional probability of the target behavior (T) given the precursor (P n ) p (T|P n ) #trials with P n that also contain T # trials with P n Conditional probability of the target behavior given the absence of the precursor p (T|~P n )= #trials containing T but not P n # trials not containing P n Unconditional probability of the target behavior p (T)= #trials containing T total #trials Conditional probability of the precursor given the target behavior p (P n |T) #trials with T that also contain P n # trials with T Conditional probability of the precursor given the absence of the target behavior p (P n |~T)= #trials containing P n but not T # trials not containing P n Unconditional probability of the precursor p (P n )= #trials containing P n total # trials

PAGE 77

77 Table 2-4. Precursors reported by caregivers vs. assessment-identi fied precursors. Italicized precursors were behaviors id entified by both caregivers and the precursor assessment. Subject Caregiver-Reported Precursor(s) Assessment-Identified Precursor(s) Precursors Observed in FA Liv Make a cry/screech noise Vocalize positively flap hands, mouth objects N/A Billy NONE Cross legs, pull up pants/touch leg, rub glasses N/A Chuck NONE Hit surfaces, grab tongue, bounce hands on face N/A Kelly Run away Whine, mouth fingers, place hands in clothes N/A George Yell throw items tip over chairs Yell throw objects sign, rub head, swing arms, bang surfaces N/A Amy Laugh Manipulate objects, make noises, touch face, move around room, move repetitively, hand on foot, say "Mine," put face in object, bend at waist, mouth object N/A Renee NONE Cover eyes Cover eyes Curtis Whine, repeat phrases, grimace Scratch leg, block therapist from objects Block therapist from objects Gerald Scream, hit head or ear Flick lips, grimace, hit others, move head Grimace, move head Adam Vocalize negatively put head down change entire facial expression / make faces at others roll eyes, yell, tongue click Say "No," slouch grimace turn away, put paper in mouth, push materials away Say "No," slouch, turn away, push materials away Donald Bruxism Flap hands, put hand to mouth, snarl, clap hands, vocalize negatively, move to objects Flap hands, snarl, clap hands, vocalize negatively, move to objects Leigh NONE Cover eyes, rest head, say "Yeah yeah," guide therapist, stomp/shuffle, circle hands Cover eyes, chin down, stomp/shuffle, circle hands Guy Drop to ground roll on floor curse scream Flop ( includes rolling), curse vocalize negatively (includes screaming), swing body, stomp, bite objects, throw objects, push materials away, crumple paper, bite hand, bang head, grimace, shake head "No," hit surfaces, slouch, make requests Flop, curse, swing body, bite objects, throw objects, stomp, push materials away

PAGE 78

78 Table 2-4. Continued Kevin Say "No" Say "No," grunt, drop/scoot on floor, wave arms, pull therapist's arm, say "Good boy," hold knees, smile, fidget, stack chairs, hit with head, knock on table Say "No," grunt, drop/scoot on floor, wave arms, hold knees, fidget, knock on table Amanda NONE Hand posture, reach for therapist, stretch N/A Sammy NONE Angry vocalizations, run, climb, mouth movements, move furniture, tug on therapist's shirt N/A

PAGE 79

79 LIST OF REFERENCES Barton, E. J. & Ascione, F. R. (1979). Sharing in preschool children: Facilitation, stimulus generalization, response generalization, and m aintenance. Journal of Applied Behavior Analysis, 12, 417-430. Bonfiglio, C. M., Daly, E. J., III, Martens, B. K., Lin, L. R., & Corsaut, S. (2004). An experimental analysis of reading interventions: Generalization across instructional strategies, time, and passages Journal of Applied Behavior Analysis, 37, 111-114. Borrero, C. S. W. & Borrero, J. C. (2008). Descri ptive and experimental analyses of potential precursors to problem behavior. Journal of Applied Behavior Analysis, 41, 83-96. Carr, J. E., Coriaty, S., & Dozier C. L. (2000). Current issues in the function-base d treatment of aberrant behavior in individuals with developmental disabilities In J. Austin & J. E. Carr (Eds.). Handbook of Applied Behavior Analysis (pp. 91-112). Reno, NV: Context Press. Carr, J. E., Coriaty, S., Wilder, D. A., Gaunt, B. T., Dozier, C. L., Britton, L. N., Avina, C., & Reed, C. L. (2000). A review of "nonconti ngent" reinforcement as treatment for the aberrant behavior of individuals with developmental disabilities. Research in Developmental Disbilities, 21, 377-391. Catania, A. C. (1971). Reinforcement schedules: The role of responses preceding the one that produces the reinforcer. Journal of the Experimental Analysis of Behavior, 15 271-287. Catania, A. C. (1973). The concept of th e operant in the analysis of behavior. Behaviorism, 1 103-116. Charlop, M. H. & Milstein, J. P. (1989). Teachin g autistic children conve rsational speech using video modeling. Journal of Applied Behavior Analysis, 22 275-285. Cuvo, A. J., Ashley, K. M., Marso, K. J., Zhang, B. L., & Fry, T. A. (1995). Effect of response practice variables on learning spelling and sight vocabulary. Journal of Applied Behavior Analysis, 28 155-173. Day, H. M. & Horner, R. H. (1989). Building response classes: A comparison of two procedures for teaching generalized pouring to le arners with severe disabilities. Journal of Applied Behavior Analysis, 22, 223-229. DeLeon, I. G. & Iwata, B. A. (1996). Evaluation of a multiple-stimulus presentation format for assessing reinforcer preferences. Journal of Applied Behavior Analysis, 29 519-533. Dews, P. B. (1966). The effect of multiple SA periods on responding on a fixed-interval schedule: V. Effect of periods of complete darkness and of occasional omissions of food presentations. Journal of the Experimental Analysis of Behavior, 9, 573-578.

PAGE 80

80 Fisher, W., Piazza, C. C., Bowman, L. G., Hagopi an, 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. Goh, H., Iwata, B. A., & DeLeon, I. G. (2000). Competition between noncontingent and contingent reinforcement schedu les during response acquisition. Journal of Applied Behavior Analysis, 33 195-205. Grow, L. L., Kelley, M. E., Roane, H. S., & Sh illingsburg, M. A. (2008). Utility of extinctioninduced response variability for the selection of mands. Journal of Applied Behavior Analysis, 41, 15-24. 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. Hagopian, L. P., Paclawskyj, T. R., & Kuhn, S. C. (2005). The use of conditional probability analysis to identify a re sponse chain leading to th e occurrence of eye poking. Research in Developmental Disabilities, 26, 393-397. Harding, J. W., Wacker, D. P., Berg, W. K., Barretto, A., Winborn, L., & Gardner, A. (2001). Analysis of response class hi erarchies with atte ntion-maintained problem behaviors. Journal of Applied Behavior Analysis, 34, 61-64. Horner, R. H. & Day, H. M. (1991). The effects of response efficiency on functionally equivalent competing behaviors. Journal of Applied Be havior Analysis, 24, 719-732. Iwata, B. A., Dorsey, M. F., Slifer, K. J., Ba uman, K. E., & Richman, G. S. (1994). Toward a functional analysis of self-injury Journal of Applied Behavior Analysis, 27 197-209. Reprinted from Analysis and Intervention in De velopmental Disabilities, 2 3-20, 1982. Iwata, B. A., Kahng, S., Wallace, M. D., & Lindberg, J. L. (2000). The functional analysis model of behavioral assessment. In J. Austin & J. E. Carr (Eds.), Handbook of applied behavior analysis (pp. 61-89). Reno, NV: Context Press. Koegel, R. L. & Covert, A. (1972). The relations hip of self-stimulation to learning in autistic children. Journal of Applied Behavior Analysis, 5, 381-387. Krantz, P. J. & McClannahan, L. E. (1998). Social interaction skills for children with autism: A script-fading procedure for beginning readers. Journal of Applied Behavior Analysis, 31 191-202. Lalli, J. S., Mace, F. C., Wohn, T., & Livezey, K. (1995). Identification and modification of a response-class hierarchy. Journal of Applied Behavior Analysis, 28, 551-559. Le, D. D. & Smith, R. G. (2002). Functional analys is of self-injury with and without protective equipment. Journal of Developmental and Physical Disabilities, 14, 277-290.

PAGE 81

81 Lovaas, O. I. & Simmons, J. Q. (1969). Manipu lation of self-destructi on in three retarded children. Journal of Applied Behavior Analysis, 2 143-157. Magee, S. K. & Ellis, J. (2000). Extinction e ffects during the assessment of multiple problem behaviors. Journal of Applied Behavior Analysis, 33, 313-316. Marcus, B. A. & Vollmer, T. R. (1996). Combin ing noncontingent reinforcement and differential reinforcement schedules as treatment for aberrant behavior. Journal of Applied Behavior Analysis, 29, 43-51. Michael, J. L. (1982). Distinguishing between discriminative and motivational functions of stimuli. Journal of the Experimental Analysis of Behavior, 37, 149-155. Najadowski, A. C., Wallace, M. D., Ellsworth, C. L., MacAleese, A. N., & Cleveland, J. M. (2008). Functional analyses and tr eatment of precursor behavior. Journal of Applied Behavior Analysis, 41, 97-105. Nutter, D., & Reid, D. H. (1978). Teaching reta rded women a clothing selection skill using community norms. Journal of Applied Behavior Analysis, 11 475-487. Parrish, J. M., Cataldo, M. F., Kolko, D. J., N eef, N. A., & Egel, A. L. (1986). Experimental analysis of response covariation among compliant and inappropriate behaviors. Journal of Applied Behavior Analysis, 19 241-254. Peterson, R. F. (1968). Some experiments on the or ganization of a class of imitative behaviors. Journal of Applied Behavior Analysis, 1, 225-235. Pierce, K. L. & Schreibman, L. (1994). Teaching daily living skills to ch ildren with autism in unsupervised settings through pictorial self-management. Journal of Applied Behavior Analysis, 27, 471-481. Reeve, S. A., Reeve, K. F., Townsend, D. B., & Poulson, C. L. (2007). Establishing a generalized repertoire of helping behavior in children with autism. Journal of Applied Behavior Analysis, 40 123-136. 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. Rosenbaum, M. S. & Breiling, J. (1976). The development and functiona l control of readingcomprehension behavior. Journal of Applied Behavior Analysis, 9, 323-333. Shukla, S. & Albin, R. W. (1996). Effects of extinction alone and extinction plus functional communication training on covariat ion of problem behaviors. Journal of Applied Behavior Analysis, 29 565-568. Skinner, B. F. (1938). The behavior of organisms. Acton: Copley Publishing Group.

PAGE 82

82 Skinner, B. F. (1953). Science and human behavior. New York: Macmillan. Smith, R. G. & Churchill, R. M. (2002). Iden tification of environmental determinants of behavior disorders through functional analysis of precursor behaviors. Journal of Applied Behavior Analysis, 35, 125-135. Sprague, J. R. & Horner, R. H. (1992). Cova riation within functi onal response classes: Implications for treatment of severe problem behavior. Journal of Applied Behavior Analysis, 25, 735-745. Thomason, J. L., Iwata, B. A., Neidert, P. L., & Roscoe, E. M. (in press). Evaluation of response latency as the index of problem be havior during functional analyses. Journal of Applied Behavior Analysis Vollmer, T. R., Iwata, B. A., Zarcone, J. R., Smith, R. G., & Mazal eski, J. L. (1993). The role of attention in the treatment of attention-mainta ined self-injurious behavior: Noncontingent reinforcement and differential reinforcement of other behavior. Journal of Applied Behavior Analysis, 26 9-21. Wallace, M. D. & Iwata, B. A. (1999). Eff ects of session duration on functional analysis outcomes. Journal of Applied Behavior Analysis, 32, 175-183. Werts, M. G., Caldwell, N. K., & Wolery, M. (1996). Peer modeli ng of response chains: Observational learning by students with disabilities. Journal of Applied Behavior Analysis, 29 53-66. Young, J. M., Krantz, P. J., McClannahan, L. E ., & Poulson, C. L. (1994). Generalized imitation and response-class formation in children with autism. Journal of Applied Behavior Analysis, 27 685-697.

PAGE 83

83 BIOGRAPHICAL SKETCH I com pleted my Bachelor of Science degree at the University of Florida in 2001 then accepted a position on the inpatient Neurobehavior al Unit at the Kennedy Krieger Institute in Baltimore, MD. There I was responsible for the assessment and treatment of severe problem behaviors, such as self-injury, aggression, and property destruction. I returned to the University of Florida in 2003 to pursue a doctoral degree in psychology a nd specializing in behavior analysis. During my graduate training, I have been involved in resear ch projects on refining behavioral assessment methods, evaluating trea tments for problem behaviors, comparing methods of training observers, and evaluating the effects of varying reinforcement parameters on performance. I also served as coordinator of an outpatient clinic for i ndividuals diagnosed with autism, provided behavioral services to students and teachers within a special education setting, and served as teaching assistant and primary in structor for introductory courses in applied behavior analysis. Following gra duation, I will join the faculty at the University of Houston Clear Lake as an assistant professor in applied behavior analysis within the psychology program.