The next four chapters will describe how the research findings from the experimental analysis of behavior can be translated into applied assessment and intervention techniques. The first step in the process of conducting an applied behavioral intervention is to clearly identify the behaviors that need to be changed. These behaviors must be described in ways that can be systematically observed and measured on a more or less continuous basis. Additionally, the relationship between these behaviors and their environmental determinants must be detailed, in this chapter the assessment techniques used to identify a behavior and its environmental determinants will be described.
Identifying the behavior to be changed is only the first step. The behavior must also be measured over time in such a manner that the impact of the intervention on the behavior can be causally evaluated. Applied behavior analysts have designed a number of experimental methods that can be used to examine the effectiveness of behavioral interventions on selected behaviors. These research methods will be described in Chapter 8.
Following this, in Chapters 9 and 10 a series of applied behavioral intervention techniques will be described that are derived from the operant principles of reinforcement and punishment. These techniques can be used to increase appropriate behavior or decrease inappropriate behavior. Applied interventions often involve combinations of these intervention techniques to decrease maladaptive behaviors while concurrently increasing adaptive behaviors. Issues related to the selection of interventions based on prior assessment of the target behavior will be discussed in these chapters.
Behavioral assessment involves selecting and defining the behavior to be changed. This behavior is typically referred to as the target behavior. Additionally, the context or environment in which the behavior occurs must also be clearly identified. Environmental conditions may be operationalized as establishing operations (sometimes called setting events), antecedent stimuli, and consequences or reinforcing stimuli. The relationship between these environmental conditions and the behavior of interest is presented in Figure 7.1. Information from assessment is essential prior to selecting intervention strategies that may increase adaptive behavior or decrease maladaptive behavior.
Figure 7.1 A model of the behavior and environmental conditions to be assessed. Note that here the terms "Antecedent stimuli", "Behavior", and "Consequence stimuli" are used in ways that are exactly equivalent to the terms "discriminative stimulus" or SD, "response" or R, and reinforcing stimulus or S+, in Chapter 4. The terms used here give us the "ABC of behavior analysis".
In Chapter 2, we described a two-term relationship between an operant response class and the consequent or reinforcing stimulus as a step towards a definition of operant conditioning, in Chapter 4, we recognized the limitations of that simple formula by adding in the antecedent or discriminative stimulus to make the well-known three-term relationship, of SD, R and S+, our definition. In Figure 7.1, there are four terms with the addition of establishing operations. For reasons that are explained below, this does not really make a "four-term relationship", but establishing operations are none the less necessary conditions for operant conditioning to occur. Establishing operations are social or biological conditions of satiation or deprivation that may affect the evocative power of a discriminative stimulus and/or the reinforcing power of a consequence (Carr & Smith, 1995; Michael, 1982; 1993). For example, candies may be powerful reinforcers for increasing appropriate behavior (such as cleaning up the toy room) with some children, but the effectiveness of such a reinforcer may be related to the level of food deprivation at any given point in time. Hence, contingent candy (that is, presentations of candy made contingent upon appropriate behavior) may produce high levels of room cleaning prior to dinner but low levels following dinner. O'Reilly (1995) investigated a day-care situation where a person with severe intellectual disabilities sometimes showed a great deal of aggression. He demonstrated that this occurred on days following loss of sleep during the previous night, and that sleep deprivation was correlated with high levels of escapemaintained aggression during demanding tasks. Sleep deprivation acted as an establishing operation to enhance the aversiveness of demanding tasks and increase the power of escape as a reinforcer. Establishing operations include all the social and biological conditions necessary for the three-term relationship of SD, R and S+ to be effective, and thus while of great importance, should not be construed as a specific fourth term with respect to the other three.
Antecedent stimuli are those environmental conditions that occur prior to the performance of a behavior and predict specific consequences. In other words, behavior has been reinforced in the presence of these stimuli and therefore these stimuli acquire a discriminative function. Chapter 4 gave a detailed description of stimulus control and Chapter 9 will describe applied interventions with this technology. Much of our everyday behavior is under stimulus control. A knock at the door or a telephone ring predicts that opening the door or answering the phone will reveal a person or a voice at the other end. In our earlier example, the presence of a parent may predict that cleaning the room will result in candy for the child. Cleaning the toy room in the presence of the babysitter will not result in a candy reward and therefore the room remains untidy. The presence of a parent is therefore a discriminative stimulus for room cleaning.
Behavior refers to the responses of an individual or group. These responses must be observable by at least one individual. Behaviors in our previous examples include, picking up toys and placing them in appropriate containers, hitting and pinching staff, picking up the phone, and answering the door.
Consequence stimuli are those events that follow the performance of the behavior. They are also contingent upon the performance of the behavior. In other words, these consequences are only available to the person following the performance of that behavior and are otherwise not available. Contingent consequences can either increase (reinforce) or decrease (punish) the probability of a behavior. Chapter 2 provided a detailed description of the functional properties of positive reinforcement, and Chapter 5 described negative reinforcement and punishment, Chapters 9 and 10 will describe corresponding applied interventions with these processes.
In summary, a comprehensive behavioral assessment must provide unambiguous and measurable information about the behavior to be decreased or increased and the context in which the behavior occurs. The context is operationalized in terms of how establishing operations, antecedent stimuli, and consequence stimuli enter into a functional relationship with the behavior of interest.
Prior to beginning an assessment it is important that the behavior identified as in need of change is of social relevance for the client. It is usually self-evident that the desired change in the target behavior will increase positive reinforcement and minimize punishment for that person. On occasions, however, the reasons for selecting a specific target behavior may not be entirely clear or appropriate. For example, Taylor, O'Reilly, and Lancioni (1996) assessed a child with autism who exhibited severe aggressive behavior in the classroom. His teacher requested a behavioral program to decrease the aggression. Following further observation it was noted that the child's aggression only occurred when he was not engaged in academic activities. The behavioral program eventually focused on teaching the teacher to engage the student continuously in academic activities which resulted in significant reductions in classroom aggression. This revised target behavior eliminated the need for the use of punishment procedures with the child.
There are no hard and fast rules for selecting appropriate target behaviors. It may be helpful for the behavior analyst to ask a number of questions about the proposed target behaviors prior to a formal assessment. These questions, coupled with a rationale for each, are presented below.
Will increasing and/or decreasing this behavior result in positive outcomes for the person? It is important that a behavior program results in a positive outcome for the individual. Developing a list of these potential outcomes often helps to clarify and justify the development of a behavioral program.
Who will be the primary beneficiary of this program? This is especially relevant when the referring agent is somebody other than the potential client. In such cases it may be important to do an ecological assessment of the client's environment (for example, the family home or the classroom). Such ecological assessments may reveal that the primary motive for the behavioral program may be to maximize reinforcers for the carers and not for the client. In these cases alternative target behaviors should be selected. Is the behavior appropriate or typical for somebody of that age? This is an important consideration when selecting behaviors to decrease and increase. For example, head-banging (i.e., rhythmic hitting of head against solid stationary objects) is a behavior that is seen in approximately 15% of infants up to 24 months of age (deLissovoy, 1964). The vast majority of these children do not continue to head-bang into early childhood and do not require intervention (Oliver, 1995). Extensive behavioral interventions are typically warranted for those cases that do persist into later childhood. Occasionally, teachers may choose rehabilitation goals for persons with developmental disabilities based on developmental age and not on chronological age. For example, toy-play with infant materials may be the planned outcome for adults with developmental disabilities. Such goals are not appropriate, because if they are achieved they will only serve to exclude these people from activities with same-age non-handicapped peers in normal life settings.
Once there is agreement regarding the need for a behavioral intervention the target behavior must be carefully defined. Initially, the presenting behavior is typically defined in general terms such as personality characteristics, and summary labels in the form of statements such as "I want to lose weight", "My child is too aggressive", or "This student is withdrawn". These global terms must be translated into operational definitions. An operational definition is a description of the subject matter in terms of how it will be measured. In applied behavioral analysis an operational definition is a description of overt, observable behavior.
This is an essential step in any behavioral program for several reasons. First, it clarifies to all stakeholders (that is, the client, their family, teachers etc.) the precise nature of the behavior to be changed. If stakeholders disagree with the focus of treatment, then new target behaviors may be selected and unnecessary interventions may therefore be forestalled. Second, it allows for ongoing evaluation of the target behavior throughout the intervention process. If desired change is not occurring then the intervention can be altered. Third, it allows for dissemination and replication of findings within the scientific community.
Hawkins and Dobes (1975) provide three guidelines for developing operational definitions of target behaviors:
Thus, the operational definition should be objective, referring to observable behaviors and environmental events. It should also be clear or unambiguous to the extent that other observers couid accurately paraphrase it. Finally, the definition should be complete and identify what should and should not be included as an instance of the target behavior.
An example of an operational definition of behavior arising from the statement "My child is too aggressive" might be, "Striking siblings in a forceful manner with an open hand or closed fist but excluding touching siblings with tips of fingers or with an open hand in a nonforceful manner." This operational definition identifies the problematic behaviors in objective terms and delineates examples of behavior to be included and excluded from the definition. In the case of the withdrawn student in the examples cited at the beginning of this section, it is most probable that the target of the intervention will be to increase appropriate social interactions in the school setting. One potential target behavior might be to increase verbal interactions and might be defined as, "A verbal statement that begins a conversation, changes a topic, or provides instruction to take some action".
It is often the case that the focus of a behavioral intervention will be with individual and discrete target behaviors such as those described in the last section. In other instances, an intervention may be designed to develop a complex sequence of behaviors. In such cases it is not possible to produce a simple operational definition. Such complex behaviors must be broken down into their component parts. Each component behavior must be discrete and essential for completing the task sequence. The process by which a complex sequence of behaviors is broken down and the outcome of such a process (that is, the set of behaviors) is called a task analysis.
Developing a task analysis can be a complex affair. The first step consists of identifying the essential behaviors involved in performing the task sequence. This can be achieved by observing and recording the behaviors of individuals who are skilled in performing that task. For example, Cuvo, Leaf, and Borakove (1978) taught janitorial skills (sweep and mop the floor, clean the stool, clean the mirror, etc.) to persons with intellectual disabilities. The task analyses for these skills were developed by observing janitors perform these tasks.
Alternatively, task analyses can be developed by consulting individuals with expertise in a given area. These people can identify the important behaviors and how they are to be performed. For example, O'Reilly and Cuvo (1989) taught self-treatment of cold symptoms (identify symptoms and over-the-counter medication to treat symptoms; recognize when to consult a physician) to a woman who had suffered a severe brain injury due to cardiac arrest. The task analysis was developed with input from a group of physicians.
Finally, specialist texts may be consulted. For example, O'Reilly, Green, and Braunling-McMorrow (1990) taught adolescents who suffered brain injuries from vehicle accidents to amend potential hazards in their daily living arrangements. Common home hazards that often led to injury were identified from publications available from the American National Safety Council. Task analyses were developed for each of these situations by observing non-disabled adults dealing with these situations.
Once the essential behaviors of a task are identified they must be broken down into a series of discrete and trainable behaviors. The size or number of behaviors in each step of the task analysis must be determined by the skill level of the person to be taught. Individuals with low levels of functioning will require a task to be broken into small component behaviors. A summarized series of generic behaviors may be required for individuals with higher levels of functioning. Once the task has been tailored to the person then training is initiated to teach each individual component behavior until the entire task sequence is performed correctly.
An example of a task analysis to teach grocery shopping skills to adults with mild intellectual disabilities is illustrated below (Taylor & O'Reilly, 1997). The purpose of the intervention was to enhance independence for these individuals in ordinary life settings. The task analysis was developed by observing non-disabled individuals engaged in grocery shopping in local community stores. The level of difficulty of each step of the task analysis was determined by consulting staff who worked on a daily basis with these individuals. These skills were then successfully trained in local grocery stores through repeated practice, prompts and reinforcement. This example clearly illustrates how a behavior that will enhance the quality of life for an individual (through increasing independence), can be operationalized into a series of objective behaviors (task analysis), and successfully taught to individuals using applied behavioral techniques.
In another example of a task analysis, Friman, Finney, Glasscock, Weigel, and Christophersen (1986) taught a group of adult males to examine themselves for testicular cancer. Testicular cancer is one of the leading causes of death in males between the ages of 15 to 40. Early identification
Table 7.1 Steps of the Supermarket Shopping Task Analysis (Taylor & O'Reilly, 1997)
| Step | Description |
| 1. | Walks from car to supermarket |
| 2. | Enters the supermarket through the correct door |
| 3. | Lifts a basket |
| 4. | Looks at shopping list |
| 5. | Looks on shelves for item |
| 6. | Puts item in basket or picks up item |
| 7. | Looks at list for next item |
| 8. | Looks on shelves for item |
| 9. | Puts item in basket or picks up item |
| 10. | Checks list to see that both items are in the basket |
| 11. | Goes to correct checkout (i.e., express checkout) |
| 12. | Takes place in line |
| 13. | Behaves appropriately in line (i.e. moves forward when line moves) |
| 14. | Puts contents of basket on counter |
| 15. | Replaces basket |
| 16. | Pays for items using next dollar strategy |
| 17. | Waits for change |
| 18. | Packs sack |
| 19. | Picks up sack |
| 20. | Exits store through correct door |
| 21. | Returns to car |
and medical treatment of this disease can lead to recovery. The behaviors required for correct self-examination were identified through consultation with urologists and various professional materials on testicular cancer. The task analysis of testicular self-examination behaviors is presented in Table 7.2. The participants were then successfully taught to perform these behaviors.
The development of a successful behavioral intervention rests initially on the selection of an appropriate target behavior and the correct operationalization of this behavior. The importance of this assessment process cannot be overstated. Interventions are tailored to the type of behavior to be changed. Also, the success of the intervention is measured by continuous assessment of changes in the target behavior. These issues will be discussed again in Chapters 8, 9 and 10.
Once the target behavior has been identified and operationalized the next step is to select an appropriate measurement strategy for that behavior. The measurement strategies used in applied behavior analysis provide a direct assessment of the target behavior in the criterion or real-life setting before, during, and after the implementation of treatment procedures. The measurement procedures used in applied behavior analysis usually consist of an assessment of discrete occurrences of the target response or the amount of time the response occurs. It is important that the measurement procedure selected be both sensitive and practical. A sensitive measurement system is one that produces an accurate and complete picture of the target
Table 7.2 Task Analysis for Testicular Self-Examination (Friman, Finney, Glasscock, Weigel, & Christopherson, 1986)
| 1. | Gently pulls scrotum so that it hangs freely. |
| 2. | Uses fingers and thumbs of both hands to isolate and examine one testicle. |
| 3. | Locates the soft tender mass (the epididymis and spermatic cord) on top of and extending behind the testicle. |
| 4. | Rotates the entire surface areas of the testicle between fingers and thumbs. |
| 5. | Uses fingers and thumbs to isolate and examine the other testicle. |
| 6. | Locates the soft tender mass on top of and extending behind the testicle. |
| 7. | Rotates the entire surface area of the testicle between fingers and thumbs. |
behavior. For example, if the interest lies in increasing or decreasing the duration of a behavior it is important that a measurement of duration of the target response is used. Similarly, if the frequency of the behavior is of interest then a frequency measurement system should be adopted. A practical measurement system is one that is usable from the point of view of the observer. Often, the applied behavior analyst must balance sensitivity with practicality when choosing a measurement procedure. Whether a measurement system is usable depends on the applied context. For example, it would be difficult for a teacher with a normal classroom responsibility of some 30 students to conduct duration measurements of "out-of-seat behavior" (that is, time spent away from the seat the student should be occupying) for an individual student. Such a context may require less time consuming assessment protocol that may not be as sensitive to the behavior of interest (such as momentary time sampling, discussed below). Some of the most common measurement protocols used include frequency or event recording, duration and latency, and time sampling methods.
Frequency or event recording requires a tally or count of the number of times that the target behavior occurs during a given period of time. Frequency recording is most often used with discrete behaviors of a constant duration. A discrete response is one that has a clear beginning and end. Discrete responses allow the observer to separate each instance or occurrence of the behavior. Each instance of the behavior should take a similar amount of time to perform. If different instances of the behavior are of different durations then an alternative measurement strategy that is sensitive to the temporal dimension of the target behavior should be used. For example, certain ongoing behaviors such as conversing with others might not be suitable for assessment using frequency measures. If a student plays with one peer for 10 seconds and another peer for 5 minutes then these would be counted as two instances of talking. The temporal dimension of these two instances of playing would be lost if a frequency count was used.
Frequency measures have been used to assess a variety of behaviors in applied settings. For example, Stark, Knapp, Bowen, Powers, Jelalian, Evans, Passero, Mulvihill, and Novell (1993) successfully taught parents to implement a behavioral program to increase caloric intake for three malnourished children with cystic fibrosis. One of the measures of the programs effectiveness was the number of bites of food each child ate at dinner. Other examples of behaviors measured using frequency protocol include social skills (such as initiating conversations), aggression (verbal and physical abuse), attendance (such as appointment keeping at pediatric outpatient clinics).
Frequency measures can be expressed as the number of times a target behavior occurred (for example, number of families that attended outpatient clinic). Number should not be used as an expression if the observation times themselves differ from session to session. In such instances rate of behavior is a more appropriate expression. Rate of response can be calculated by dividing the total frequency of the target response by the number of minutes for that particular observation. Response rate per minute is therefore comparable across observation sessions of different durations.
Frequency or event recording has a number of obvious advantages. First, it is an easy recording system to use. It merely requires a tally of ongoing behavior. Various recording devices such as wrist counters and hand-tally digital counters are available that can facilitate recording Second, because they are a direct measure of the amount of behavior, frequency or event measures are sensitive to changes in behavior resulting from contingency manipulation (Kazdin, 1994).
Duration and latency are time-based methods of measurement. These measurement systems are used when the temporal dimension of the response is of interest. Measures of duration are particularly useful with continuous behaviors or with discrete behaviors of such high rate that it would be difficult to measure each response accurately using frequency or event recording. Continuous behaviors such as free play or conversations are behaviors that can occur for extended periods of time and usually happen for different lengths of time during each occurrence. Some individuals with severe behavior disorders emit high rates of discrete behaviors such as head hitting, and eye poking. Counting each occurrence of such behavior would be difficult and unreliable. One alternative approach is to count the amount of time or duration the individual engages in such behavior.
Duration measures can be conducted using either a duration per occurrence or a total duration procedure. For duration per occurrence, the observer measures the duration of each instance of the behavior during an observation session. A total duration procedure measures the total duration of responding during an observation session. Duration per occurrence provides a temporal and numerical assessment of behavior. For example, Gaylord-Ross, Haring, Breen, and Pitts-Conway (1984) measured the duration and number of social interactions between students in a school setting. Such a protocol provides a measure of how many interactions each student engaged in and how long these interactions were. As the purpose of the intervention was to increase the quality and quantity of interactions between students a duration per occurrence measure was more sensitive to the goals of the program than frequency or total duration measures. De Luca and Holborn (1992) used a total duration measure to examine the impact of a variable-ratio reinforcement schedule with changing criteria on amount of time spent pedaling stationary exercise bicycles with obese and non-obese boys. The total amount of time spent pedaling during observation sessions was of interest in this intervention. The use of a duration per occurrence protocol would have provided additional and unnecessary information for the purposes of this study. It is therefore important to consider the purpose of the planned intervention (that is, which dimensions of the behavior are to be increased or decreased) in addition to the nature of the target behavior when selecting a measurement system.
Latency is an additional measure based on the time dimension of responding. Instead of measuring the duration of a response when it occurs, latency is a measure of the length of time between the presentation of an initiating stimulus and the occurrence of the target response. Latency should be used when the major issue is the length of time between an opportunity to perform a behavior and the initiation of that behavior. For example, latency can be used when an individual is too slow at following directions (for example, beginning a classroom exercise, or complying with parental requests). The main purpose of an intervention in such cases may be to decrease the latency between the request and performance of the task. Latency to responding may also be used as a measure when performing the behavior can be dangerous to the individual concerned. For example, O'Donoghue and O'Reilly (1996) examined the function of serious self-destructive behavior for an adult with autism and developmental disabilities. Self-injury was so severe (including tearing eyelids, and banging head on hard surfaces) that the individual was physically restrained throughout the day. During assessment procedures a latency protocol (amount of time from the initiation of a task to engagement in the first episode of self-injury) was used as it was dangerous and unethical to allow the individual to engage in the problematic behavior for an extended period of time.
Typically, a stopwatch is used to record duration or latency during observation sessions. For total duration the stopwatch is activated as the behavior begins and is stopped as the behavior ends. This is repeated during the observation session without resetting the stopwatch. A total duration is then available at the end of the observation session. For latency, the stopwatch is activated once the stimulus is presented and is stopped as the target behavior occurs. Duration is usually expressed as the cumulative duration the person engaged in the behavior during a session. If observation session times differ across days then total duration should be reported as percentage of time ([duration of target behavior/total time of observation session] x 100 = %). Percentage of time allows for the direct comparison of the target behavior across observation sessions of different durations. Reporting practices are similar for latency measures. Duration per occurrence is not affected by observation session length and is simply reported or expressed as the duration for each occurrence of the behavior.
Interval recording procedures are the most frequently used measurement protocol in applied behavior analysis. Interval measures can be used with discrete, continuous, and high rate behaviors. These procedures are used to record the number of time intervals within an observation session that the target behavior occurred. Each observation session is divided into brief time intervals of equal size. The target behavior is recorded as occurring or not occurring during each interval. Multiple occurrences of the target behavior within each interval are not scored separately. However, multiple target behaviors can be scored within each interval. Interval recording therefore does not provide an estimate of frequency but of occurrence per interval. Additionally, target behaviors for multiple individuals can be recorded during an observation session. For example, if an observation session is 10 minutes long and each observation interval is 15 seconds, the session is divided into forty 15 second units. The specific interval size should provide enough time for the observer to accurately observe and record the behavior. To facilitate accurate recording a data sheet is usually developed with a box for each interval (see Figure 7.2). During the observation each box (interval) is filled with a symbol to indicate whether the target behavior occurred or did not occur. Data are subsequently reported as percentage of intervals in which the target behavior occurred during each observation session (number of intervals in which behavior occurred divided by total number of intervals multiplied by 100). The data for Figure 7.2 show that the target behavior occurred for 50% of the intervals ([8/16] x 100 = 50%).
Figure 7.3 presents a data sheet to record multiple behaviors for an individual. In this case a symbol for each behavior is included in each interval box. The observer circles each behavior that occurs during an interval. This data sheet indicates that the student spent the majority of the time out of his seat and talking with other students during the observation. Each behavior (i.e., out of seat, talking to other students, talking to teacher) is presented separately as percentage of intervals. A data sheet for recording the behavior of multiple individuals is presented in Figure 7.4. When collecting interval observations with multiple individuals an observer usually focuses on one individual for each interval. For example, the first individual is observed for the first 15 second interval followed by the second person for the second interval. This sequence is continued until each person is observed for one interval. The sequence is then repeated across individuals.
There are two types of interval recording protocol. The most frequently used procedure is called partial-interval recording. This protocol requires
Figure 7.2 Example of interval recording.
Figure 7.3 Example of interval recording using multiple behaviors.
Figure 7.4 Example of interval recording with multiple persons.
the observer to record the target behavior as present if it occurs at any time during an observation interval. Alternatively, a whole-interval recording procedure can be used. Whole-interval recording requires that the behavior be present throughout the entire interval for it to be recorded as occurring. The whole-interval measurement protocol is therefore more sensitive to the duration of occurrence of the target behavior during an observation session. Partial-interval procedures tend to overestimate the occurrence of the behavior whereas whole-interval protocol tend to underestimate behavior. Again, the type of interval protocol chosen will be determined by the nature of the target behavior and the goals of the proposed intervention program. For example, Allen, Loiben, Allen, and Stanley (1992) evaluated the effectiveness of a dentist-implemented intervention (escape contingent on brief periods of cooperative behavior) on levels of disruptive behavior for 4 children who were receiving restorative dental treatment. Body movement, crying, moaning, and complaining were recorded for the children using a 15 second partial interval recording protocol. Lagomarcino, Reid, Ivancic, and Faw (1984) evaluated the effectiveness of an intervention procedure including stimulus prompts and contingent positive consequences to teach dance skills to persons with severe developmental disabilities. Observation protocol consisted of observing each of two dancers individually for one-minute sessions. Appropriate dancing was measured using a 10-second whole-interval procedure during the 1-minute observations.
An alternative interval measurement procedure is called momentary time sampling. Momentary time sampling differs from the other interval protocol in that target behaviors are recorded as occurring or not occurring immediately after the specified time intervals. Recording devices (i.e., data sheets) and protocol (i.e., observation sessions are divided into specified time intervals) are similar to the other interval measurement systems. Momentary time sampling is easy to use especially when and observer is also involved in concurrent tasks. For example, teachers often use a momentary time sampling procedure as it does not interfere with the teaching routine. The end of an interval can be signaled by a device such as a tape recorder. At that point the teacher observes the target student and records whether or not the behavior of interest was occurring. The teacher can then continue with their teaching duties until the end of the next interval.
In summary, once a target behavior is operationalized it is necessary to select an appropriate measurement system for that behavior. The target behavior must be measured over time in order to provide an evaluation of the behavior before, during, and after the intervention. Selecting a measurement system for a target behavior requires consideration of a number of factors. In particular, the measurement system must be sensitive and practical. Measurement systems are typically based on the number of occurrences or duration of the target behavior. The most frequently used measurement systems have been outlined here (these are frequency, duration, and interval recording). Specific variations of these measurement procedures can be found in the behavioral literature (e.g., physiological measures such as heart rate; permanent products of behavior such as amount of weight lost) but all address either a frequency or duration dimension of the target behavior.
The purpose of assessment is to identify the extent to which the target behavior is performed before, during, and after the intervention. In addition to operationally defining the target behavior and selecting an appropriate measurement strategy, it is important to clarify when observations of the behavior in the applied setting will be conducted. Observation sessions must be conducted in a manner that will produce a representative sample of the target behavior. Behavior typically fluctuates over time. Observations should be conducted in a manner that will account for these fluctuations and produce an accurate picture of the overall rate of the behavior. Additionally, the strategy for conducting observations will also be influenced by the overall goal of the proposed intervention, A number of factors should be considered when deciding to conduct observations. First, the number of times that observations are to be conducted should be considered. It is preferable to conduct observations each day and during those periods of the day when the target behavior is most probable. This is typically not feasible because of time constraints on observers. Generally, the more observations the better with at least one observation daily. Second, the length of time for which each observation is to be conducted must be decided. It is preferable to observe behavior for a period of time that will produce data that is representative of performance for the period of interest. For example, if a child is engaged in problem behavior throughout the school day it would be preferable to observe for an extended period of time during the day rather than conduct a brief observation at the beginning of the day. Finally, it is important to consider when the observations should be conducted. Often the behavior of interest will occur during a particular time period (for example, bedtime tantrums, or aggression during lunch period at school). Observations should therefore be conducted during these critical time periods. If behavior occurs throughout the day then multiple observations during each day are preferable.
Another essential criterion of an applied behavioral investigation is that the target behavior be reliably assessed. Reliability generally means that two observers can concurrently but independently observe the target behavior and agree on its occurrence and nonoccurrence. Reliability or interobserver agreement is therefore an assessment of the consistency and accuracy with which the target behavior is measured before, during, and after the intervention. High levels of interobserver agreement allow the therapist to conclude that patterns of responding (e.g., percentage of intervals that the target behavior occurs across days) reflect actual client performance. Low levels of interobserver agreement may imply that patterns of client responding reflect observation biases and not actual performance. Observer bias typically occurs when an observers' definition of the target behavior changes over time. For example, an observer may see improvements in the target behavior during an intervention phase because he or she expects this to occur. Interobserver reliability is therefore a check or control for observer bias. There is no single rule about how often interobserver reliability should be conducted during an applied behavioral intervention, it is generally accepted that observations of the target behavior by another independent observer should be conducted during baseline {or prior to intervention), intervention, and follow-up phases (or after intervention). Interobserver agreement measurement is typically required for a minimum of 20 percent of all observations in research studies. An acceptable level of agreement between observers typically ranges between 80 to 100 percent. An agreement of less than 80 percent indicates that there is an unacceptable level of error occurring with the recording protocol. Such low agreement can indicate that the target behavior is not clearly defined or that observers are not adequately trained to record the behavior. These potential biases should be identified and remedied prior to conducting baseline observations of the target behavior. This can be accomplished by conducting reliability observations and checking agreement levels on the target behavior prior to formal baseline observations.
The protocols for estimating agreement differ depending on the measurement procedure used to assess the target behavior. Methods for estimating frequency and interval reliability will be discussed as these protocols can be adapted for all other measurement systems (e.g., duration). If frequency of the target behavior is the measurement system used then interobserver agreement is assessed using a percentage frequency agreement between observers. This percentage agreement measures the degree to which both observers agree regarding the occurrence of the target behavior. Interobserver agreement is calculated by dividing the smaller frequency by the larger frequency and multiplying by 100. For example, the number of times a student hits other students may be a target for intervention in a classroom. Two observers independently count the number of times the student hits others during an observation session in the school. By the end of the observation period one observer has counted 10 hits while the other observer has counted 8 hits. Percentage agreement for this observation was therefore 80 percent ([8/10] x 100). It is important to note that this form of reliability reflects agreement on the total number of responses and not on any specific response. It is impossible to determine whether observers agree on each specific response. Such agreement estimates must therefore be treated with caution as they may conceal disagreement on individual target responses.
Interobserver agreement with interval recording methods is usually calculated on the basis of the percentage of intervals in which two observers agree on the occurrence of the target behavior. An agreement is scored if both observers agree on the occurrence of the target behavior during the same interval. A disagreement is scored if one observer records the behavior as occurring during an interval and the other observer does not. Interobserver agreement is calculated by dividing the number of intervals which both observers agreed the behavior occurred (agreements) by the number of intervals that they did not agree the behavior occurred (disagreements) plus the number of agreements and multiplying by 100. For example, if two observers recorded behavior for 20 10-second intervals and agreed on the occurrence of the behavior for 15 intervals and disagreed on 5 intervals, overall agreement would be [15/(5 + 15)] x 100 = 75%. While this is the generally accepted method for calculating interval agreement, some investigators have questioned whether agreement should be confined to intervals where both observers record an occurrence of the behavior. Agreement can also be extended to intervals where both observers record a nonoccurrence of the behavior. The inclusion of nonoccurrences as well as occurrences in the calculation of interval reliability percentages inflates reliability estimates beyond the level obtained when occurrences alone are calculated. The more conservative estimate of interobserver agreement is to use occurrence agreement percentages only.
At the beginning of this chapter a model was presented that identified the important parameters to be considered when conducting an assessment of behavior. In addition to operationalizing the target behavior, this model emphasized an assessment of the environmental stimuli (both antecedent and consequent stimuli) that entered into functional relationships with that behavior. It is no accident that the majority of this chapter focuses on the strategies used to develop a clear identification of the target behavior. Until recently, applied behavior analysts have primarily been interested in isolating target behaviors and selecting interventions to increase or decrease the frequency of these behaviors. The behavior of applied behavior analysts themselves (in terms of the interventions they have selected) seems to have been controlled by the structure or topography of the target behaviors rather than by the operant function of these behaviors. For example, the target behavior might be "eye gouging", without reference to its function for the individual. This approach contradicts a fundamental premise of behavioral analysis which describes behavior primarily in terms of its function (see Chapter 1). These points do not detract from the importance of systematically identifying and operationalizing target behaviors. However, an analysis of the environmental determinants of target behaviors from a functional perspective has often been missing from the practice of behavioral assessment. The assessment of severe aberrant behavior is one notable exception to this. Over the last decade the majority of published studies that have examined aberrant behavior have conducted a functional assessment or analysis of the environmental determinants of target behaviors and subsequently matched a treatment to maintaining contingencies. It is therefore instructive to examine the assessment strategies developed in this particular area of applied behavior analysis.
The term functional assessment has been used to describe a variety of systematic procedures to determine antecedent and consequent variables which occasion and maintain aberrant behaviors. Functional assessment typically involves a process whereby target behaviors are defined by interviewing significant others and are subsequently observed in naturalistic contexts (in those contexts where the behavior has been described as being problematic). This form of assessment reveals correlational information regarding establishing/discriminative conditions and consequences for the target behavior. Assessment procedures may also involve the systematic manipulation of hypothesized controlling variables to empirically demonstrate causal relationships. This latter assessment technique is typically referred to as a functional analysis (Axelrod, 1987; Iwata, Dorsey, Slifer, Bauman, & Richman, 1982).
A knowledge of controlling variables derived from such assessment protocol assists in the development of effective treatment procedures in at least three ways (Lennox & Miltenburger, 1989). First, assessment may identify reinforcing consequences (positive or negative) contingent on target behavior performance which can subsequently be eliminated or prevented (Carr, Newsom, & Binkoff, 1980). This means that instead of the planned intervention "inventing" new contingencies to modify problem target behavior, it can change those that already control it. Second, an assessment may identify motivational (Michael, 1982; 1993) and/or discriminative (Skinner, 1935) conditions that evoke the target behavior, and by removing or altering these conditions the behavior may be prevented (Dunlap, Kern-Dunlap, Clarke, & Robbins, 1991). For example, if sleep deprivation leads to inappropriate aggression, the aggression can be eliminated by ensuring that suffcient sleep takes place (O'Reilly, 1995). Finally, such pre-intervention assessments may allow the practitioner to identify more efficient and socially appropriate responses to access similar consequences as the problematic behaviors (Carr & Durand, 1985). That is, if the client engages in the problem target behavior because it is their only way of obtaining attention from caregivers, for example, that attention may be provided contingent upon other, more acceptable, behavior.
The remainder of this chapter will outline specific examples of functional assessment and functional analysis strategies, and examine their strengths and weaknesses.
The major functional assessment protocols include behavioral interviews and direct observation methods such as scatterplots and ABC assessments.
Behavioral interviews rely on subjective verbal reports to identify the nature of the aberrant behavior and the environmental conditions that are controlling it (Cone, 1987). Those who are interviewed (such as parents, teachers, and others) should be in daily contact with the client and therefore be in a position to describe events as they have witnessed them in the past and to draw conclusions about the causes of an individual's behavior. There are three main objectives of a behavioral interview: 1) operationalization of the behavior(s) – what is it?; 2) identification of those physical and environmental factor(s) predictive of the aberrant behavior(s) - when does it occur?; 3) identification of the potential functions of behavior(s) in terms of their maintaining consequences– what reinforces it?
To achieve these outcomes a complete interview should include questions which probe the informant about the topography of the behavior, the situations in which it does and does not occur, and the typical reactions of others in response to the aberrant behavior. In essence, the behavioral interview attempts to review a large number of potential variables and narrow the focus to those that appear to be of some importance in generating and maintaining the undesirable behavior. A number of behavioral rating scales, checklists, and questionnaires are commercially available and can be used to guide the interview process. For example, one of the most frequently used instruments is the Motivation Assessment Scale (Durand, 1990) which provides a specific description of the targeted problem behavior and attempts to isolate one of four possible reasons for this behavior: positive reinforcement in the form of attention, positive reinforcement through access to materials, negative reinforcement through escape, or sensory reinforcement. Some examples of questions used in the Motivation Assessment Scale are presented in Table 7.3. These questions assess whether the aberrant behavior serves a sensory function (i.e., automatic positive or negative reinforcement); a negative reinforcement function (i.e., escape from tasks); and a positive reinforcement function (i.e., access to attention) respectively.
Table 7.3 Examples of Questions Used in the Motivation Assessment Scale (Durand, 1990)
| 1. | Would the behavior occur continuously, over and over, if this person was left alone for long periods of time? (For example, several hours). |
| 14. | Does the behavior stop occurring shortly after (one to five minutes) you stop working or making demands on this person? |
| 15. | Does this person seem to do the behavior to get you to spend some time with him or her? |
There are a number of advantages to the interview approach, including ease of application, cost and efficiency (administration takes only a brief period). On the reverse side of the coin, there are a number of inherent difficulties. Such methods do not allow for direct access to the relevant behaviors and their controlling variables and are therefore subject to a variety of difficulties including faulty recollection of events, observer bias, observer expectation etc. (Kazdin, 1994). As such, information gained through these methods may provide unreliable estimates of behavior and lead to invalid conclusions about its controlling variables.
A more objective and systematic approach to assessment involves first hand observation of an individual's behavior in environmental contexts that are relevant to the problem. The individual is observed in their typical daily routine in as many settings and across as much time per day as is possible for a minimum period of 3-5 days (O'Neill, Horner, Albin, Storey, & Sprague, 1990). Notably, there is little, if any, control exerted over the environmental conditions during assessment. Such direct observations should be based upon information gleaned from the interview process (i.e., behaviors and situations that have been identified as problematic are observed). The process is usually carried out by those parents, teachers, and support staff who already work with the individual and is conducted in a manner that does not require extensive training on their part. Two general classes of descriptive analyses have been forwarded in the literature and each will be discussed.
This observation method attempts to evaluate the immediate antecedent and consequent events surrounding the target behavior and assess the extent to which these specific events may be related to the occurrence of behavior. This assessment usually entails a narrative account of directly observed behavior and temporally related environmental events (Bijou, Peterson, & Ault, 1968).
Those working with the individual exhibiting the aberrant behavior write brief descriptors of what occurs immediately prior to and following the target behavior. Such accounts are usually recorded on an ABC or sequence analysis chart (Sulzer-Azaroff & Meyer, 1977). Although the procedure is relatively easy to learn, it requires extensive effort to implement (Pyles & Bailey, 1990). Further, such a procedure often leads to subjective interpretation of events rather than objective descriptions (Lerman & Iwata, 1993).
To overcome such difficulties a number of approaches have been recommended. It is essential that observers are aware of the temporal parameters involved in an ABC assessment to combat the temptation to record global environmental events that are far removed from the target behavior. In a practical measure to overcome this difficulty of subjective interpretation, Pyles and Bailey (1990) developed the "inappropriate record form" which lists already-specified preceding and consequential events coupled with the problem behavior. Observers are therefore cued to record certain antecedent and consequent events upon the occurrence of the target behavior. A less formal measure is simply to train observers not to infer motivation from an observation but to describe events clearly and accurately (Lennox & Miltenburger, 1989).
The most recently-developed and simplest direct observation method is the scatterplot assessment which records temporal distributions of behavior (Touchette, MacDonald, & Langer, 1985). Observers are trained to record the time of day of the occurrence of each instance of the target behavior on a grid that identifies time of day on the ordinate (usually in 30 min segments) and consecutive days on the abscissa. As the behaviors are repeatedly observed and plotted, correlations between particular times of day and differential rates of behavior can become evident. This data allows for more detailed observational analyses (such as ABC assessments) during those time periods in which the behavior has been identified as most probable. An example of a scatterplot data sheet used by staff to identify the temporal distribution of "scratching other clients" by an adult with severe mental disabilities in a group home is presented in Figure 7.5. This scatterplot shows that the target behavior clusters around certain time periods during the day when the client was required to engage in task related activities. These results would imply that the aggressive scratching behavior might be maintained by negative reinforcement contingencies (i.e., escape from demanding activities).
Direct observation methods have a number of advantages. They allow direct access to problem behavior in the natural environment and therefore are more objective in that they reflect current behavior and not recall of past observations. Like most procedures, direct observations have a number of limitations. Relative to indirect methods such procedures are time consuming. More important perhaps is that these procedures do not necessarily reveal functional relationships (Iwata, Vollmer, & Zarcone, 1990). For example, it may be difficult to identify the consequences of behavior maintained by intermittent reinforcement schedules (Lerman & Iwata, 1993).
Figure 7.5 Example of a scatterplot grid over a five-day period.
Experimental analyses of behavior constitute the final means of conducting an assessment of aberrant behavior. The most distinguishing feature of this method of analysis lies in its direct and systematic manipulation of variables that potentially maintain the aberrant behavior (Iwata et al., 1990). Of the many assessment techniques to evolve from the literature in recent years, functional or experimental analyses have perhaps been used most frequently. This approach has been used successfully in the analysis and treatment of such behavior problems as stereotypic behavior (Durand & Carr, 1987; Sturmey, Carisen, Crisp, & Newton, 1988), disruption (Carr & Durand, 1985), aggression (Slifer, Ivancic, Parrish, Page, & Burgio, 1986), pica (Mace & Knight, 1986), and self-injurious behavior (O'Reilly, 1996).
This form of analysis is important for many reasons. First, it emphasizes the importance of gaining information about the contingencies maintaining behavior rather than merely describing the topographical features (e.g., biting or hitting). It also explains how topographically similar behaviors can serve different functions for a given individual. For example, one individual may engage in self-injurious behavior to gain access to attention and their behavior may be maintained by attention functioning as positive reinforcement. On the other hand, another individual's self-injury may be negatively reinforced and serve to escape from an aversive situation. It is through a realization of these different functions of topographically similar behaviors that researchers have recognized the need to develop highly individualized treatment programs that are tailored to the specific function of aberrant behavior.
Methods of conducting such an in-depth analysis of the functions of behavior are a relatively recent advance, with the degree of rigor and sophistication of the different methods varying. The control necessary to adequately demonstrate functional relationships in an experimental analysis is often difficult to obtain in the natural environment. Functional relationships are therefore often verified in an analogue setting which approximates the natural environment. Once the conditions that control the behavior are identified, these contingencies can then be manipulated in the natural environment. Iwata et al., (1990) describe this model as involving at least one condition (experimental) in which the variable of interest is present and another condition (control) in which the variable is absent. These conditions are then alternated in a multielement or reversal design while the behavior of interest is observed. A complete description of these experimental designs is found in Chapter 8.
There are two variations of this model that can be found in the literature. One approach involves demonstration of the effects of a single hypothetical controlling variable (such as attention from a care-giver) on a particular behavior. An early example of this method was conducted by Lovaas and Simmons (1969) in which a client who exhibited self-injury was exposed to several conditions differing on the variable of attention (social deprivation, non-contingent delivery of attention, and social attention contingent on occurrences of self-injurious behavior), and demonstrated that self-injury was higher during the contingent attention condition.
More recent research has shown that aberrant behavior may be multiply controlled (that is, simultaneously influenced by a number of reinforcers) and therefore a second model has developed in which several variables are compared to determine behavioral function. Iwata et al., (1994) presented an epidemiological analysis of 152 cases that used a multielement design format to compare four analogue conditions to assess the function of self-injury. These analogue conditions assessed the impact of positive reinforcement (attention contingent on self-injury), negative reinforcement (escape from demands contingent on self-injury), automatic reinforcement (placement in a barren environment with no access to either attention or toys), and a control (no attention for self-injury, no demands, play materials available and attention contingent on the absence of self-injury). Their results showed specific functions of these various sources of reinforcement for 145 of the cases. Two hypothetical examples of analogue analyses for self-injury using multielement designs are presented in Figure 7.6. For Graph 1 in the figure there are higher levels of self-injury in the attention condition relative to the other conditions. These results imply that self-injury for this individual is maintained by social positive reinforcement. In Graph 2 we see higher levels of self-injury in the demand condition relative to the other conditions. The results in graph 2 imply that self-injury for this person is maintained by escape from task demands or social negative reinforcement.
Among the strengths of an experimental functional analysis are its objectivity and quantitative precision and its ability to analyze the effects of several variables. It has also been noted that the control condition included in an experimental analysis may indicate some temporary intervention strategies that can be implemented until the treatment program is designed and put into effect. Procedures to reduce aberrant behaviors will be discussed in detail in Chapter 10.
Although providing more conclusive data, one potential disadvantage with conducting such an assessment is that it may be difficult or impractical
Figure 7.6 Functional analysis results of self-injury for two individuals.
for use in many applied settings due to the stringent control necessary and also the limitations of staff, time, and facilities (Lennox & Miltenburger, 1989). However, Iwata, et al., (1990) point out that "this criticism is unwarranted because precisely the same requirements must be met in order to implement most treatment programs with any degree of consistency" (p. 310). In recent years, Wacker and colleagues (see, for example, Northup, Wacker, Sasso, Cigrand, Cook, & DeRaad, 1991) have shown how such analyses of aberrant behavior may be successfully carried out during one 90-minute outpatient clinic session. O'Reilly, O'Kane, Byrne, and Lancioni (1996) used a mini-reversal design during a 60-rninute therapy session to establish the antecedent variables of challenging behavior for a person with severe brain injuries. This research overcomes the potential problem of an extended experimental analysis delaying implementation of an effective treatment.
Another potential disadvantage, which has been suggested by LaVigna and Donnellan (1986), is that the analogue analysis may not be ecologically valid (that is, it may not mirror exactly the variables operating in the natural environment). This is a potentially serious problem, but it has been overcome in studies where experimental analyses have been successfully conducted in natural settings such as classrooms (Sasso & Reimers, 1988) and outpatient clinics with parents present (Northup et al., 1991). Additionally, Iwata et al., (1990) point out that a functional analysis does not reveal the functional variables involved but merely tests those variables that have been proposed through prior functional assessments. Hypothesized contingencies must be systematically identified prior to an experimental analysis. Therefore, all functional analyses must encompass additional information from background sources to facilitate how best to construct the analysis (i.e., to identify which variables to manipulate).
Research on the functional assessment and analysis of aberrant behavior has made a unique contribution to the behavioral assessment literature. These assessment technologies emphasize an examination of the environmental determinants of target behaviors. Interventions can therefore be selected based on the function of the behavior. Traditionally, applied behavior analysis assessment has focused on a detailed analysis of the target behaviors at the expense of examining functional variables. This traditional assessment approach results in the selection of intervention techniques that must override current maintaining contingencies in order to produce successful behavior change (Mace, 1994b). Implications for using such assessment techniques on the selection of intervention strategies will be discussed in detail in Chapters 9 and 10.
Applied behavioral assessment involves selecting and defining the behavior to be changed. Additionally, it is important for behavior analysts to identify the environmental conditions that influence the probability of the behavior targeted for change. These environmental variables can occur prior to and following the target behavior. A detailed knowledge of these contextual influences on the target behavior will help the behavior analyst to select interventions that may increase adaptive behavior or decrease maladaptive behavior.
Behavioral assessment is a rigorous process involving a number of steps. First, a socially relevant target behavior must be identified. Behaviors selected for change must be of primary importance to the individual client. Changes in the selected behaviors should maximize positive outcomes and minimize negative outcomes for the client. Target behaviors must then be operationalized to reveal a clear, complete, and objective description of the behavior to be changed. This clarifies to all involved (for example, client, family, therapist) the nature of the behavior to be changed, it also allows for an ongoing evaluation of changes in the behavior throughout the treatment process. Complex target behaviors can be broken down into a series of discrete behaviors known as a task analysis.
Second, a measurement system is selected to provide a continuous assessment of the target behavior before, during, and after the intervention. The measurement system should be sensitive to the relevant dimensions of the target behavior. It must also be possible to use the selected measurement system in an unobtrusive manner within an applied context. Common measures used include frequency, duration, latency, and interval protocol.
Third, observations must be conducted in such a manner as to yield a representative sample of the target behavior. It is important to consider when to observe, where to observe, and the length of time for each observation. Additionally, interobserver reliability should be conducted throughout the behavioral program. Agreement estimates provide an assessment of the consistency and accuracy with which the target behavior is measured.
Recent developments in the assessment and analysis of aberrant behavior were discussed in the final sections of this chapter. Functional assessment and functional analysis protocol examine the relationship between behavior and its environmental determinants. Prior to using any of these assessment techniques the target behavior must be systematically identified and operationalized according to the criteria outlined earlier in the chapter. Functional assessment protocols describe a set of techniques which include interview and observational methods to identify environmental events that evoke and maintain the aberrant responding. These functional assessment techniques produce correlational information regarding the controlling variables. Functional analysis protocols involve the systematic manipulation of hypothesized causal variables and therefore make a causal analysis possible. Whenever feasible, functional analyzes should be conducted in natural settings, or using materials from natural settings, in order to produce results that are externally valid. Interventions can then be tailored to the function of the behavior identified through these assessments.