Implementing Positive Behavior Support Within Preschool Settings: Group Functional Assessment and CW-FIT

Functional assessment of problem behavior

Prior to baseline, a Functional Assessment Interview (FAI; O’Neill et al., 1997) was conducted with each teacher. This was to identify the most common problem behaviors the children within the group engaged in and to gather information about them within their respective classroom to inform Antecedent Behavior Consequence (ABC) analysis. The FAI is a structured interview designed to attain information about an individual’s behavior(s) and the conditions under which it is likely or not likely to occur. Teachers were asked to provide a precise description of the most commonly occurring problem behavior(s) across all the children in the preschool classroom, rather than for one specific child. Identified behaviors included rough play, shouting in the classroom, noncompliance, and running in the classroom (see Table 1 for operational definitions). The section asking for personal information was not included in the interview. Teachers then indicated both antecedent and consequence variables, and identified the communication skills within the group, previous teaching strategies (both successful and unsuccessful), and a reinforcer inventory.

Chandler et al.’s (1999) GFA method was used to conduct ABC (O’Neill et al., 1997) data collection on the identified problem behaviors of each individual student within each group during the identified problematic period. This involved dividing the children into observation groups of three to four children. The researcher observed each child in sequential order and recorded the occurrence of the target behavior(s). A total of five 3-min observations (15 min) were undertaken per child as per Chandler et al. (1999; modified from 10 observations to five observations due to time constraints), ABC data were analyzed, and the prevalence of antecedents and consequences was tallied. Outcomes of the analysis identified the most commonly observed antecedent preceding and consequences following the problem behaviors from this tally for each group, such that the functionally equivalent social skills to teach could be identified. The first author, who was a trainee behavior analyst, had received extensive training on ABC data collection recording and completed the ABC data collection.

Baseline

Data collection was conducted on all dependent variables for the first 20 min of the problematic periods identified by teachers, for each group. During observation periods, the children engaged in their typical routines as designed within each classroom. Based on the location in which the children were seated, they were placed into clusters of three to four children, although they were not informed of this, to facilitate momentary time-sampling data collection. Frequency counts of target social skills and problem behavior were taken across the entire group, during each observation session, the number per session an aggregation of each child’s behavior. Data were recorded on the on-task behavior, of clusters of children, using 30-s momentary time sampling, signaled to the researcher via a vibrating timer. The researcher had a 5-s window to scan each cluster. At each 30-s interval, the researcher would rotate from one cluster of children to another and recorded engaged behavior if each child in the cluster was on-task at that point or a minus for on-task behavior, if any child was not attending to the material. If, at the end of the 30-s interval, a child was inhibiting others from attending to their materials, a minus was recorded for on-task behavior, and the presence of problem behavior was recorded in addition. Average on-task behavior for the whole class was calculated by adding the number of on-task intervals for each cluster and dividing it by the total number of possible intervals.

Group preference assessment

A modified version of Layer et al.’s (2008) group preference assessment was utilized. Pictures of the reinforcers were presented rather than requiring the children to associate a particular reinforcer with a color card to simplify the assessment procedure. First, the 10 most highly rated items identified in the reinforcer inventory completed by each teacher were selected to use in the group preference assessment. The children were also asked if they liked the suggested items and activities. The group preference assessment was conducted to identify each child’s preference before every group contingency started. To inhibit children influencing each other’s selection, the children were instructed to keep their choice private. In groups of three, children were each asked separately to select their preferred item: “point to the picture that they would like to do/have the most” from an array of four cards (pictures of the identified reinforcing items/activities). If they did not respond within 5 s, the vocal instruction was repeated. If they pointed to more than one picture, it was stated that “I’m not sure which one you want the most, pick again.” This picture was then put in a box, labeled the “voting box.”

Staff training

Teachers all participated in a Behavioral Skills Training (BST) program, which consisted of instruction, modeling, rehearsal, and feedback (Sarokoff & Sturmey, 2004). The teachers were provided with scripted lessons of the social skills comprising four steps employing direct instruction and a scripted procedure for the group contingency (i.e., written protocol denoting the target responses and the schedule of reinforcement delivery). Training was delivered individually with the researcher in a 2-hr session post-baseline and consisted of the review of a written document with an explanation of the rationale for the intervention and detailed explanations of each component (i.e., skills teaching, group contingency, and reinforcement). The modeling component of this training involved the researcher sitting in front of the teacher and modeling teaching each of the social skills using direct instruction (using definitions, modeling, and role-play) three times. Rehearsal involved each teacher practicing teaching each of the social skills scripts to the researcher 3 consecutive times. The researcher modeled the group contingency for a 3-min period where the researcher role-played being the teacher and the teacher role-played being a child. The timer was set for 30-s intervals and upon the timer sounding, the researcher modeled scanning the class and both praising the child for displaying social skills throughout the time period as well as awarding the points to a group. Rehearsal for the group contingency involved the researcher acting as the child and each teacher rehearsing the intervention for a 3-min period, 3 consecutive times. Verbal feedback (both positive and constructive) was given to teachers immediately following each rehearsal of the group contingency practice with opportunities for questions.

Throughout training, the researcher reminded the teacher of the importance of using group-wide and behavior-specific praise (Kamps et al., 2011). Each teacher was then required to rehearse each component of the intervention, with the researcher role-playing a child until they reached mastery criteria of 100% as indicated by performance observations (i.e., social skill steps taught, rationale for skill explained, modeling, and role-plays) across two consecutive trials. For the first 4 to 5 days of intervention implementation, teachers were provided with in situ training by the researcher until they were able to independently implement teaching with 100% fidelity as indicated by startup fidelity observations (i.e., social skill steps taught, rationale for skill explained, modeling, and role-plays) in the natural setting. The first author attended all intervention sessions. If implementation fell below 80% fidelity, BST would have been conducted again; if implementation or quality of implementation fell below 100% fidelity, teachers received onsite verbal feedback. Verbal feedback was provided in four instances for Group 1, three instances for Group 2, and one instance for Group 3 around giving reinforcement immediately, providing corrective feedback, and setting the timer at appropriate intervals.

Class-wide teaching

CW-FIT Tier 1, which included social skills instruction and an IGC, was delivered by teachers within preschool classroom settings. Functionally equivalent social skills for each group were identified, from the GFA, by the first author and clinical supervisor. The social skills identified were based on information identified from the FAI, which highlighted social skills that peers were engaging in, and the steps of the social skills were based on previous CW-FIT social skills. Scripted lessons were developed and adapted from previous CW-FIT research (Jolstead et al., 2016; Can I have help please?). Teachers delivered social skills instruction to the children through 10-min scripted lessons using Direct Instruction. Lessons included definitions of each skill and accompanying cues/hand gestures for each skill, modeling of examples and nonexamples, role-plays, and feedback. Scenarios were set up, where children were required to display the gestures for each step of a skill and explain the rationale before moving on to the next skill. Examples of the gestures for asking the teacher for help included raising one hand and holding the index finger to lips (to show waiting quietly), while holding the second hand in air before asking the question. The scripted lessons were repeated until all children had displayed the gestures for each skill and explained the rationale at least once. Visual aids, which outlined the specific steps of each of the social skills, were 24 in. × 36 in. posters in clear view of the children. The order in which the social skills were taught for each group was based on the results of the functional assessment. The social skills taught in Group 1 in the following order were: Can the class have 2 more minutes? (primary function: escape); (secondary function: teacher attention); Can I have a turn? (tertiary function: tangible). The skills taught in Group 2 and 3 in the following order were: Can I play a game with a friend? (peer attention); Can the class have 2 more minutes? (escape); and, within circle time, Can I do a Jumping Jack? (escape). An additional skill of “how to ignore other’s inappropriate behavior” was also taught in each group. This was the final skill taught across all groups, aiming to decrease bootleg reinforcement. The social skills were outlined before each IGC began. This process of testing and teaching ensured that children who were not maintaining the skills were identified (through not being able to complete the steps in the skill) and accessed the booster teaching sessions. Booster sessions consisted of a 10-min scripted lesson, delivered to the child by the teacher before the next intervention session.

IGC

The children were broken into groups of three to four depending on where the children were sitting. As described, prior to commencing the 20-min social skills session, each child indicated their preferred reinforcer by casting their anonymous vote. The group contingency was delivered by the teacher for 20 min of the designated problematic time period for each group. Prior to the intervention, the teacher outlined the rules of the group contingency, which was referred to as a “game,” and the target social skills (provided an explanation of each skill). A game chart (a poster 11 × 17 in.), which was used to keep track of team points, was placed on the wall. This chart listed the social skills in rows on the left and listed colors that represented each team along the top columns, providing cells to mark team points and the social skills that had occurred.

The teachers set a digital audible timer at intervals typically between 1.5 and 3 min. Each day, a point goal (i.e., how many points each team was required to reach to access reinforcement) was set by the teachers, and the team was required to reach this goal to receive a reinforcer. For each session, the point goal was determined by calculating 75% to 85% of possible timer beeps (e.g., if the timer was set for 3 min, the observation session 20 min was divided by 3, which equals 6.6, 75% of 6.6 is 4.9, so 5 points would be required that day). When the timer sounded, the teacher scanned each team (within a 5-s window) and awarded points and praise to a team contingent on all the children demonstrating engagement with the designated task at that time. Class-wide and individual behavior–specific praise was provided, by the teacher, to children for being on-task and for displaying the social skills throughout the game (i.e., “well done to all the teams, everybody is doing great listening today”; “fantastic job at raising your hand ____”). Throughout the group contingency, not dependent on the timer, the teachers awarded a group a point if all children, in a team, were displaying one of the taught target social skills. Minor behaviors that were not disruptive were ignored. On completion of the IGC, the teacher tallied team points and selected the “mystery motivator” available to the teams that met the predetermined criteria. The children were also praised again for using the social skills and being on-task. Data collection was identical to baseline and took place for the duration of the IGC. In addition, generalization data were also gathered for a 20-min observation during the problematic period, when the IGC was not in effect. These data were collected by the researcher.

Interobserver agreement (IOA) and treatment fidelity

Interobserver agreement was assessed by the first author. Video recordings were gathered for 20% of all intervention sessions to calculate IOA. The video recordings were independently reviewed by the first author using the same data collection methods outlined in baseline. This was carried out with no access to the data or knowledge of the number of behaviors that occurred in each session. IOA was calculated by summing up the total number of agreements between original data collection and the data collection from the video recording and dividing it by the complete number of timings. To specify the exact degree of agreement, the ratio was multiplied by 100 (Kazdin, 1982). IOA average across sessions ranged from 94.1% to 97.22%. Treatment fidelity data were collected during 90% of intervention sessions using a 13-item CW-FIT treatment fidelity checklist, on which the researcher scored each teacher. Treatment fidelity was calculated by dividing number of correct responses by the number of possible correct responses and gaining percentage values by multiplying by 100 and was found to be 91.3% (SD = 7.57) fidelity, on average, by the preschool teachers.

Social validity

The Intervention Rating Profile–15 (IRP-15; Witt & Martens, 1985) was completed by teacher participants post-intervention. This is a questionnaire that employs a 6-point Likert-type scale with 15 items in total to assess the social validity of an intervention. The questionnaire assesses variables such as if teachers thought there were any risks involved for participants as a result of the intervention, if it was appropriate for a given participant or situation, if it negatively impacted other student participants, or whether implementation was too time-consuming (Witt & Martens, 1985). Higher total scores reflect higher rates of social validity of an intervention, with total scores ranging from 15 to 90.

Generalization and follow-up probes

During baseline, intervention and post-intervention generalization probes were undertaken. Generalization probes consisted of a 20-min observation period, during the problematic period (on the hour period identified by the teacher), when the group contingency was not in effect. Data were recorded in an identical way to baseline for all target behaviors.

Follow-up probes were conducted for Group 1, during the 20 min in the identified problematic hour the group contingency was in effect, to examine if the intervention effects on the dependent variables were maintained 1 day, 1 week, and 2 weeks. Groups 2 and 3 were not available for follow-up probes.

Effect size estimate

Tau-U, which is a nonparametric, nonoverlap effect size method, was utilized, allowing examination of intervention effects between phases (percentage of data demonstrating improvement; Parker et al., 2011). Tau-U was calculated using an online calculator (Vannest et al., 2011) and interpreted as per Vannest and Ninci’s (2015) recommendations as dependent on context, a small change (0.20 improvement), a moderate change (0.20–0.60), a large change (0.60–0.80), and a very large change (above .80).