Functional Analysis in Public School Settings: A Systematic Review of the Literature

Abstract

Previous studies have shown that variations of experimental functional analysis (FA) can be used effectively as part of the functional behavioral assessment process in educational settings. The purpose of the current study was to review the literature pertaining to FA of problem behavior in schools focusing on method, topographies, and results. Results of this review showed that analog functional analysis (AFA) are conducted most often in school settings; but recent trends show that use of adapted formats (e.g., trial-based FA and brief FA) are increasing. Analysis of FA results identified most frequent topographies and maintaining variables of problem behavior in the reviewed studies. Trends in teacher-implemented FA are discussed, as well as discrepancies surrounding the terminology of FA in the literature base. Implications for further research, such as social validity of FA in public schools and use of multiple targeted topographies within school-based FA are presented.

Introduction

Functional behavior assessments (FBA) are commonly used as a means of identifying maintaining variables and designing effective treatment of problem behavior. This process often includes descriptive assessments, formal interviews with teacher and caregivers (O'Neill et al., 1997), direct observations (Haynes et al., 2011a), or in certain cases a more rigorous evaluation commonly referred to as functional analyses (FA). Experimental functional analysis (FA; i.e., analog FA, brief FA, trial-based FA, latency FA, precursor FA) is prevalent in behavior analytic research as a component of FBA (Hanley, 2012). The direct observation of problem behavior and corresponding experimental manipulation of conditions surrounding behavior used in FA has been applied to a variety of topographies of problem behavior such as self-injurious behavior (SIB; e.g., Iwata 1982/1994), elopement (e.g., Lang et al., 2010), aggression (e.g., Davis et al., 2014), and pica (e.g., Falcomata et al., 2007); and across settings such as clinics (e.g., Kurtz et al., 2003), homes (e.g., Wacker et al., 2013), and schools (e.g., Bloom et al., 2011). Although FA is considered by some to be the “gold standard” of behavior assessment in research and some clinical settings (Lloyd et al., 2016), the application of analog FA to settings such as public schools is often considered not practical (Solnick & Ardoin, 2010).

Criteria for an FBA to be conducted varies by state (Zirkel, 2011), but is suggested under the Individuals with Disabilities Education Act (IDEA) for students who frequently engage in problem behavior in a way that affects learning (IDEA, 2004). Not only do criteria for implementation vary, but the content that is included in an FBA varies as well. Recent reviews have found that only 56.7% of publications on FBA in schools included more than one of the previously listed components (Anderson et al., 2015). Research has also suggested that FBA interventions based on the results of FA show greater intervention effects than nonexperimental methods (Hurl et al., 2016). Although multicomponent FBAs utilizing FA procedures are often beneficial in school settings, 75.7% of FA conditions in school FBAs were conducted by researchers rather than school staff (Anderson et al., 2015).

Social Validity of FA in Schools

Despite evidence that FA can yield more effective intervention outcomes, it is still used infrequently by school staff as a tool in FBAs in public schools (Anderson et al., 2015). This gap in implementation by teachers and administrators can be explained by lack of training on FA procedures as part of the FBA process by special education administration, possibly due to logistical barriers of implementing FA in school settings (Lloyd et al., 2016).

Recent trends have shown researchers training classroom and school personnel in FA procedures (Rispoli et al., 2016) or conducting FA alongside classroom personnel (Rispoli et al., 2015). Results of previous work have indicated that special education teachers can effectively implement FA models following training (Berg et al., 2007; Sasso et al., 1992), in particular models that are less time intensive such as trial-based FA (LaRue et al., 2010; Rispoli et al., 2016; Bloom et al., 2011) or brief FA (Flanagan et al., 2019; LeGray et al., 2010; Northup et al., 1991). However, to date, no previous reviews of the literature have synthesized social validity data or information specific to school-based FA.

Barriers to FA in Schools

Although criticism of the research-to-implementation gap is warranted, school settings present multiple barriers to effective implementation of FA. Analog FA (AFA; Iwata, 1982/1994) involves multiple experimental manipulations with conditions of 10 min or more and typically a minimum of two staff members. During these conditions, establishing operations are presented to evaluate their potential effects on problem behavior that can in turn cause greater frequency and magnitude of the problem behavior being evaluated. Depending on the topography of the behaviors, this can lead to damage to staff, property, or the individual being assessed.

The large amounts of time and resources required to conduct such an assessment are often scarce in school settings. These procedures typically involve two trained individuals for implementation and data collection (teachers or assistants), a controlled environment (empty classroom, isolated area), and may involve high rates and intensity of the targeted behavior. In addition, it can take up to 3 hr to implement the procedures during the instructional day to effectively test all conditions. This requirement of time and space can at times be of greater detriment to the climate of the school than the problem behavior of the individual in the FA (Hurl et al., 2016). Furthermore, multiple implementers are typically needed for data collection purposes as well as for implementing the procedures and must also be trained in procedural fidelity. A strain is often imposed on educational personnel as well as students.

Current practice in FBAs include indirect (e.g., rating scales, interviews, questionnaires) and nonexperimental direct assessment (e.g., descriptive assessments, naturalistic observation) procedures. Indirect and nonexperimental direct assessment procedures typically require less preparation and assessment from school staff than experimental FA. These assessments can lead to appropriate identification of maintaining variables of problem behavior, which can be used to design effective interventions (Blood & Neel, 2007). Although interventions designed from results of an FA have been shown to be more effective (Hurl et al., 2016), the greater response effort of conducting an FA over nonexperimental FBA-based assessment procedures can create another barrier because both may yield similar intervention strategies or results.

However, despite the greater response effort, the inclusion of FA procedures may yield more consistent results regarding contingencies that maintain problem behavior. For example, recent findings have shown that nonexperimental FBA procedures conducted by different staff or caregivers with histories of intervening on the problem behavior of an individual often yield different functional outcomes (Nicholson et al., 2006; Shogren & Rojahn, 2003).

Variations of FA in Schools

In an effort to ease the difficulty of conducting FA in school settings, variations of traditional FA procedures have been shown to yield similar results and decrease the toll on school staff. One adaptation is the brief FA (BFA), which requires less time in each condition (i.e., 5 min) and “mini-reversals” (Wacker et al., 2004) with regard to test conditions and can also include contingency-reversals (i.e., approximations of function-based interventions) to identify the variables maintaining problem behavior (Northup et al., 1991). Trial-based FA (TBFA; LaRue et al., 2010; Bloom et al., 2011) is another adapted FA format and has been a common FA procedure used in schools (Solnick & Ardoin, 2010). TBFA procedures disperse conditions intermittently in short trials across the typical learning day and the frequency or occurrence of behavior is analyzed across conditions (Bloom et al., 2011). Latency-based FA (LFA; LaRue et al., 2010; Thomason-Sassi et al., 2011) is another adapted FA format that has been used in school settings in which problem behavior may be dangerous to students or staff. With the LFA, sessions are terminated at the onset of problem behavior and the latency between onset of the condition and occurrence of problem behavior is analyzed (LaRue et al., 2010). Finally, precursor FA (PFA) involves the antecedent manipulations of experimental conditions rather than manipulations of consequences (e.g., Smith & Churchill, 2002), which can be advantageous as it can mitigate risk by applying contingencies to precursor behavior (Smith & Churchill, 2002).

Research on FA methodologies have been increasing in the literature base possibly due to their effectiveness and the ability of school staff to be trained to implement them with fidelity (Rispoli et al., 2015). Although FA’s are traditionally conducted by researchers and school staff on students with problem behavior, application of these FA variations in general education classrooms and on students without disabilities is gradually increasing (McCahill et al., 2014).

Previous Literature Reviews

Two previous meta-analyses have evaluated the implementation of FA in public schools. The first was a quantitative review of consequence-based FAs conducted by teachers or experimenters in any public-school setting (Solnick & Ardoin, 2010). Solnick and Ardoin (2010) grouped FAs for analysis into “analog” FA and “classroom” FA according to session length and setting. Results of Solnick and Ardoin (2010) detailed that the majority of FA in public school settings were conducted by either teachers in classroom settings or therapists outside of the classroom. FA procedures typically involved four conditions with 10-min sessions across multiple replications. Despite the length of these analyses, only 58.4% of articles detailed intervention following assessment and only 22.92% led to individualized interventions based on reinforcement (Solnick & Ardoin, 2010).

The second meta-analysis assessed procedures implemented by school staff members but excluded studies in which FAs were conducted “outside their usual classroom setting” (Lloyd et al., 2016). Lloyd et al. (2016) found a large variety in FA procedures, session length, and number of sessions as well as an average of 4.7 conditions per FA. Participants in these FA were primarily diagnosed with emotional disturbance, intellectual disability, or autism spectrum disorder. Identification of maintaining variables as a result of the reviewed FAs was not reported (Lloyd et al., 2016).

Purpose

The application of FA technology in schools regardless of setting or implementer warrants the need for further evaluation of efficacy. The exclusion of private school settings from this review is an attempt to acknowledge the potential differences between resources at public schools as opposed to private special education schools and inpatient facilities. Synthesis of FA research in public schools may have implications for the widespread application of the technology, and may ultimately assist in the equipping of teachers with a valuable tool for the effective educational programming for their students. However, no review to date has analyzed variations of FA conducted in schools settings individually (i.e., TBFA, LFA, BFA). Further, the cumulative number of school-based FA publications utilizing these variations has increased considerably since previous reviews. Thus, an extension and update to previous reviews and syntheses of the literature are warranted. The present review sought to consolidate research on FA methodology in public schools in an effort to guide best practice in selection of procedures and implementation. Therefore, following research questions were addressed: (1) What FA procedures have been conducted in public school settings? (2) Who was the primary assessor during the FA? (3) In what specific settings in public schools have FAs been conducted? (4) What were the characteristics of the participants in the FA? (5) Which FA approaches led to identification of maintaining variables of behavior and what were the results?

Method

Studies were identified in July 2020 by a search of Academic Search Complete, ERIC, MEDLINE, PsycArticles, and PsychINFO using the search terms “functional analysis” and “public schools” with exclusionary term “gene*” due to fact that the term “functional analysis” is commonly used in genetic research. An initial search identified 11,219 records. Following deduplication, a total of 9,112 records entered preliminary title and abstract screening. Titles and abstracts were primarily screened for the inclusion of a functional analysis of behavior, as the terminology “functional analysis” is used in other fields for procedures outside of the scope of this review. Title and abstracts of included studies were again screened for school settings. If studies did not mention setting in title or abstract, they were included in the full text screen to ensure fidelity. A table of contents search was also conducted on additional journals, including Journal of Applied Behavior Analysis, Behavior Modification, School Psychology, Journal of Behavioral Education, Behavior Analysis in Practice, and Behavioral Interventions. Finally, a forward and backward ancestral search of all included studies was conducted to identify articles not identified through previous methods. Of the results, 173 studies were included for full-text screening. Following the full-text screen, 42 articles were included. A diagram of the search and screening process is represented in Fig. 1.

Fig. 1.

Study inclusion

Inclusion Criteria

Inclusion criteria were based off the framework of Hanley (2012), a seminal meta-analytic work on FA. Studies identified for inclusion in this review were those which showed evidence of (1) experimental manipulations of environmental variables (i.e., FA) by (2) using direct measurement of a problem behavior across (3) at least two conditions involving manipulation to determine relation (Hanley, 2012). With regard to setting a study had to be conducted in a public-school setting (excluding private schools, Head Start programs, in-patient therapy programs, etc.). Therapy centers and private schools were excluded from the current study due to the extreme variation in resources available to private institutions as opposed to public schools (e.g., higher number of trained behavior analysts and other support staff, higher staff to student ratios). Studies were excluded if (1) only descriptive methods of analysis were used (i.e., not visually represented with graphed data), (2) FA procedures were conducted on skill acquisition rather than problem behavior, or (3) FA procedures were solely focused on teacher training with no represented data on participant behavior during FA.

Analysis of Functional Analyses

Included studies in the present review that involved FA conducted under the setting requirements varied greatly in their procedures, placement, and characteristics of individuals involved. Therefore, we analyzed how these variations were conducted along the following criteria.

Participant and Assessment Characteristics

We identified participant characteristics including age, gender, ethnicity (if specified), diagnosis, and reported method of cognitive evaluation. Data not reported across any of these categories were noted as “not reported.”

We documented assessment methods across several categories including the primary assessor (e.g., researcher, administrator, teacher), setting within the school, and FA procedure used (e.g., analog, brief, trial-based, latency-based). We also identified and analyzed topographies targeted during the FA and potential correlations between method chosen and topography.

FA methods of included studies were categorized by FA procedure according to cited literature within the methods sections. These citations included seminal papers from each method of FA, including AFA (Iwata et al., 1984/1992), BFA (Northup et al., 1991), TBFA (Bloom et al., 2011), and LFA (LaRue et al., 2010; Thomason-Sassi et al., 2011). If these direct citations (or similar) were not listed, FA methods were evaluated and categorized by two independent observers (see Interrater Agreement).

FA Results

We analyzed all FAs in terms of maintaining variables of problem behavior. Results were analyzed across identified function and topographies. We documented all FAs that determined maintaining variables of behavior as “inconclusive” or “multiply-maintained” based on author interpretations. Inconclusive or multiply maintained FA that met criteria for inclusion also analyzed along using the same criteria.

Fidelity and Social Validity

Procedural fidelity, interobserver agreement (IOA), and social validity were reviewed and evaluated for all studies that reported data form these categories. Social validity measures varied and were therefore analyzed by publication (i.e., IRP-15, TARF) for accurate comparison. Studies that did not report data for a given category were not included in the analysis.

IOA

Thirty-four percent of included studies were coded by an additional reader to assess the reliability of the results. IOA was calculated by dividing agreements within code sheets divided by the total number of coded categories (20) and multiplied by 100%. The mean interrater agreement score was 97.3% (range = 75%–100%).

Results

A total of 42 studies met inclusion requirements for the review. Seventy-six different participants of varying disability categories were included in the reviewed studies, including some students with no diagnosed disability. The included studies were published in 17 different peer-reviewed journals from 1992 to 2019. Table 1 presents a summary of the included studies.

Table 1.

Included article characteristics

Citation	Model	Assessor	Setting	Diagnosis	Procedural fidelity	IOA	Social validity
Austin et al. (2015)	TBFA	Teacher, TA	Gen Ed	None	91%	100%	TQ
Berg et al., 2007)	AFA	Teacher	Special Ed	ID	N/A	96%	N/A
Bessette and Wills (2007)	AFA	TA	Special Ed	None	91%	94%	TQ
Broussard and Northup (1997)	BFA	Teacher & Researcher	Empty Classroom	ADHD	87%	90-95%	IRP-15
Casey and Merical (2006)	BFA	TA	Gen Ed	ASD	N/A	96%	N/A
Davis et al. (2014)	AFA	Researcher	Special Ed & Empty Classroom	ASD, SI, ID, VI, OHI, Arcadi Syndrome	N/A	95-99%	N/A
Dolezal and Kurtz (2010)	PFA	Staff Member	Special Ed	TBI, ID	N/A	94%	N/A
Dunlap et al. (1993)	BFA	Teacher	Special Ed	EBD	85%	N/A	N/A
Ellis and Magee (1999)	AFA	Teacher & Researcher	Empty Classroom	ADHD, ASD, Bipolar	N/A	82-98%	N/A
Ewing et al. (2001)	AFA	Researcher	Empty Classroom	ASD, ID	N/A	97%	N/A
Flanagan and DeBar (2018)	TBFA	Teacher	Special Ed	EBD	96%	95%	TARF
Flanagan et al. (2019)	TBFA & BFA	Teacher	Special Ed	EBD/SLD	90%	90-100%	TARF
Fragale et al. (2016)	AFA	Researcher	Special Ed	ASD	N/A	93-97%	N/A
Frea and Hughes (1997)	AFA	Researcher	Special Ed	ID	N/A	97-100%	N/A
Frieder et al. (2009)	AFA	Teacher	Empty Classroom	DD	100%	99%	N/A
Hagan-Burke et al. (2015)	PFA	Researcher	Gen Ed & Special Ed	EBD, ADHD	100%	93%	N/A
Hansen et al. (2019)	LFA	Researcher	Empty Classroom	ID, FAS, Epilepsy	N/A	88-96%	N/A
Kwak et al. (2004)	BFA	Researcher	Empty Classroom	None	N/A	86-98%	N/A
LaBrot et al. (2014)	BFA	Researcher	Special Ed	ADHD	100%	95-98%	N/A
Lambert et al. (2012)	TBFA	Teacher, TA	Gen Ed	DD	92%	93%	N/A
Lang et al. (2010)	TBFA	Researcher	Gen Ed & Special Ed	ASD	N/A	100%	N/A
LeGray et al. (2010)	BFA	Teacher	Gen Ed	None	100%	98%	ARP-R; IRP-15
LeJeune et al. (2019)	TBFA	Teacher	Special Ed	Down’s	99%	86%	Likert Scale
Lloyd et al. (2017)	TBFA	Teacher	Gen Ed &	ID, ASD	93%	96%	IRP-15
Lloyd et al. (2015)	TBFA	TA	Gen Ed & Special Ed	Down’s, ASD	93%	97%	IRP-15
McComas et al. (2003)	AFA	Researcher	Special Ed	Down’s, ID, VI, AI, EBD, SI, ASD	N/A	95-100%	N/A
Moore et al. (2009)	AFA	Teacher & TA	Special Ed	CP, DD, Quadriplegia	N/A	93-99%	N/A
Mueller and Kafka (2006)	AFA	Researcher	Empty Classroom	DD	100%	91-99%	N/A
Mueller et al. (2005)	BFA	TA	Special Ed	ASD	N/A	N/A	N/A
Nolan & Filter, 2012)	AFA	TA	Empty Classroom	OHI	97%	98%	N/A
Northup et al. (1994)	BFA	Teacher	Special Ed	ID, Goltz Syndrome	50-100%	85%	N/A
O’Connor and Daly (2018)	AFA	Researcher	Hallway	none	100%	100%	N/A
Rispoli et al. (2013)	AFA & TBFA	Teacher	Special Ed	Down’s/ID, ASD	100%	100%	N/A
Roantree and Kennedy (2012)	AFA	Researcher	Empty Classroom	ASD	N/A	88-95%	N/A
Sarno et al. (2011)	AFA	Teacher	Special Ed	ADHD	86-98%	92%	IRP-15
Sasso et al. (1992)	AFA	Teacher	Special Ed	ASD	89%	82%	TARF
Shumate and Wills (2009)	AFA	Reading Specialist	Gen Ed	None	N/A	N/A	N/A
Umbreit (1995a, 1995b)	BFA	Teacher	Special Ed	ID	N/A	100%	TARF
Umbreit (1995a, 1995b)	BFA	Researcher	Gen Ed	ADHD	N/A	97%	N/A
Vollmer et al. (1994)	AFA	Researcher	Empty Classroom	VI, SI	N/A	93-96%	N/A
Wilder et al. (2007)	TBFA	Researcher	Gen Ed	None	N/A	91-100%	N/A
Wright-Gallo et al. (2006)	AFA	Teacher	Special Ed	EBD	N/A	89%	N/A

Note. AFA analogue functional analysis, BFA brief functional analysis, LFA latency functional analysis, PFA precursor functional analysis, TBFA trial-based functional analysis, TA teacher’s assistant, Gen Ed general education classroom, Special Ed special education classroom, ADHD attention deficit/hyperactivity disorder, AI auditory impairment, ASD autism spectrum disorders, ID intellectual disability, EBD emotional/behavioral disorder, FAS fetal alcohol syndrome, OHI other health impairment, VI visual impairment, TARF Treatment Acceptability Rating Form (Reimers et al., 1991), IRP-15 Intervention Rating Profile-15 (Martens et al., 1985), TQ Teacher Questionnair

Functional Analysis Models

Analog FAs were conducted most often and was represented in 20 studies (47.6%). Brief FA were conducted in 10 studies (23.8%), trial-based FA was conducted in 10 of the studies (23.8%), with 1 included study using a latency FA (2.4%), and 1 using a precursor FA (2.4%).

Participant and Setting Characteristics

Participants across all studies totaled 76, with 21 having no disability or a disability that was not specified by the authors (27.6%). Of the remaining participants, 19 participants were diagnosed with ASD (26%), and 13 were diagnosed with an emotional behavioral disorder (17.1%). Other disabilities included ID (12 participants; 15.8%), attention deficit hyperactivity disorder (8 participants; 10.5%), developmental delay (6 participants; 7.9%), Down’s syndrome (3 participants; 3.9%), other health impairment (1 participant; 1.3%), and various physical disabilities (6 participants; 7.9%).

Settings in which the FA were conducted varied across studies, with 21 FAs conducted in special education classrooms (47.7%), 9 in general education classrooms (20.5%), 11 in empty classrooms (25%), and 3 in a combination of two classrooms (6.8%). FA models varied across settings as well, with five trial-based FAs and four brief FAs conducted in general education classrooms (11.1% and 8.9%, respectively) along with one analog FA (2.2%). In special education classrooms, 10 analog FAs (22.2%), 6 brief FAs (13.3%), 6 trial-based FAs (13.3%), 1 precursor FA (2.2%), and 1 latency FA (2.2%) were also conducted (note: percentages totaled >100% due to some studies conducting multiple models of FA in the same setting). Seven analog FAs were conducted in empty classrooms (15.6%), as well as two trial-based FA and brief FA (4.4%, respectively) and one latency FA (2.2%). Finally, one analog FA, trial-based FA, brief FA, and precursor FA were conducted in a combination of two placements (2.2% respectively) and one AFA was conducted in a hallway of a school building (2.2%).

Implementer

A large portion of the studies involved researchers as the main implementer of the FA procedures (16 studies, 36.3%). Sixteen studies were implemented by teachers while researchers collected data or coached teachers during the FA or through preimplementation training (36.3%). A combination of researchers, teachers, and teaching assistants (TA) were identified as implementers in five studies (11.4%). A teaching assistant was the sole implementor in five studies (11.3%). One FA (2.3%) was implemented by a reading specialist (Shumate & Wills, 2009) and a “school staff member” (Dolezal & Kurtz, 2010), respectively.

Topographies of Problem Behavior

Table 2 shows the topographies of problem behavior addressed and their maintaining variables as determined by the FA. In cases where multiple topographies were grouped under “problem behavior” as the dependent variable measured in a given study, all topographies listed were documented and indicated in the table (e.g., if one student engaged in elopement, disruption, and SIB, and the FA indicated escape as the maintaining variable, each topography was indicated individually under “escape”).

Table 2.

Behavior topographies and maintaining variables of problem behavior

Topography	Escape	Attention	Tangible	Automatic	Multiple	Inconclusive
Off Task	7	5	0	0	2	0
Disruption	5	6	1	1	5	0
Aggression	11	10	3	0	11	1
Non-compliance	4	3	3	0	2	0
Self-Injurious behavior	3	2	1	3	1	3
Property destruction	5	5	1	0	4	0
Vocal outburst	6	12	4	1	7	1
Elopement	2	1	3	0	4	0
Tantrum	3	2	0	0	3	0
Other	3	4	0	1	0	0

Across studies, the 76 participants engaged in 161 individual topographies. The 161 topographies were organized into 10 categories based on the terminology used in the studies. Across these 161 topographies, the following functions were identified: (1) escape (30.4% of topographies), (2) attention (30.4% of topographies), (3) multiple functions (23.6% of topographies), (4) tangible (9.9% of topographies), and (5) automatic (3.7% of topographies), and 3.1% were found to be inconclusive. The topography with the highest frequency of occurrence was aggression (22.3%), followed by vocal outburst (19.2%), disruption (11.2%) and property destruction (9.3%). Other topographies included off-task (8.7%), noncompliance (7.5%), SIB (8%), elopement (6.2%), and tantrum (5%). Eight occurrences were grouped as “other” (5%), which were defined as “undesirable behavior” without further description of topography (Dunlap et al., 1993), “inappropriate social behavior” (Frea & Hughes, 1997), and object mouthing (Mueller & Kafka, 2006).

Functions of Problem Behavior

Table 3 shows the identified maintaining variables of problem behavior across FA formats. Totals were calculated by analyzing all FAs conducted in a study and documenting each occurrence in which a function was identified with a given participant. For example, in a given study in which trial-based FA was implemented and four participants were shown to have an attention function and one was shown to have an escape function, four attention functions and one escape function was indicated for TFA for that study. If one participant was identified with multiple functions, the FA was credited as “multiple.”

Table 3.

Maintaining variables of problem behavior by FA method

	Attention	Escape	Tangible	Automatic	Multiple	Inconclusive
AFA	13	11	5	2	8	5
TBFA	2	1	2	0	7	0
BFA	4	6	0	2	2	0
PFA	1	1	0	0	0	0
LFA	2	0	0	0	2	0

Note. AFA analogue functional analysis, BFA brief functional analysis, LFA latency functional analysis, PFA precursor functional analysis, TBFA trial-based functional analysis

Analog FA was conducted most often in school settings (n = 20) and identified attention-maintained behavior as occurring most frequently. Brief FA was conducted in 11 studies and identified escape-maintained behavior most frequently. Trial-based FA was conducted in 10 studies and identified multiply maintained behavior most often. Precursor FA was conducted one time and was found to identify both attention and escape (n = 1). Finally, latency FA was conducted in two studies and identified attention-maintained behavior in two participants and multiply maintained behavior two participants.

Fidelity and Social Validity

Percentage of fidelity and inter-observer agreement were summarized and presented in Table 1. For the included studies, 21 of 42 studies (50%) reported procedural fidelity data on FA implementation with an average of 93.8% (range: 50%–100%). IOA data were included for 39 of 42 studies (92.9%) for an average of 97.1% (range: 82%–100%).

Social validity data were reported for 12 of 42 studies (28.6%). Of the 12 studies, 5 used the Intervention Rating Profile-15 (IRP-15; Martens et al., 1985) and 4 evaluated data using the Treatment Acceptability Rating Form (TARF; Reimers et al., 1991). Two studies used a researcher-created questionnaire, and two studies implemented a researcher-created Likert scale and the Assessment Rating Profile-Revised (APR-R; Eckert et al., 1999), respectively.

Validity scores for IRP-15 range from 15 to 90 using a 15-question Likert scale from 1 to 6. Scores for two studies using the IRP-15 were 84.34 for an average of 5.62 per scale, indicating high social validity (three studies were excluded because IRP-15 was only used to evaluate intervention, not FA). Validity scores for the TARF range from 17 to 119 using a 17-question Likert scale from 1 to 7. Scores for three studies using the TARF were 99.3 for an average of 5.8 per scale, indicating high social validity (one study was excluded because the TARF only evaluated intervention, not FA). For the remaining studies that utilized unique questionnaires or Likert scales, all studies showed moderate to high acceptability of FA implementation following training.

Discussion

The present review summarized the current literature on FA in public school settings while updating and expanding on previous reviews by analyzing the individual variations of FA within public schools, as well as inclusion of social validity data. The widening of inclusion criteria to any FA conducted within a public school as well as requiring a minimum number of experimental manipulations yielded a variety of results that further illuminates how FA procedures are used in these settings.

Results of this review showed that teacher-led FAs in public schools tend to be aided by the researcher as a coach or data collector, but that a variety of FA procedures can be implemented effectively by teachers in their classrooms. Shorter formats, such as trial-based or brief FA (i.e., 49% of FA procedures identified), tended to be chosen within studies implemented by teachers (20 studies; 71.4%). This shift in the literature toward shorter duration methods are likely due to the previously addressed barriers to implementing FA in school settings. The current results suggest that adaptations to the FA process allow teachers to effectively conduct FAs using shorter formats and with fewer resources while still maintaining a high level of fidelity.

Results pertaining to functions of behavior showed that participants were frequently identified as having multiply maintained problem behavior regardless of the method of FA chosen by the researchers. This result may be due to the difficulty of producing a completely controlled environment for FAs to be conducted. It is also possible that participants’ learning history or previous exposure to similar reinforcement contingencies may affect the high percentage of cases in which multiple functions were identified (Jessel et al., 2014). The majority of the studies (i.e., 87.5%) took place within classroom settings with other peers or adults present. This finding is notable because the presence of other peers or adults could function as potential confounding variables during a FA. Although these variables were often controlled by the researchers through positioning, classroom procedures, or marking trials as “failed,” the possibility of interference still remains. Furthermore, many of the included studies did not report the number of failed trials during the FA procedure. Although this may seem insignificant, a large number of failed trials while conducting an FA could invalidate results and/or confound results. Invalidated and/or confounded results might in turn result in Type I and/or Type II errors, which severely limit the extent to which the results might effectively inform treatments that successfully address problem behavior (i.e., treatments may be developed that are not matched to the true functions of problem behavior because they are based on confounded assessment results). Future studies should evaluate the extent to which such results might affect the effectiveness of subsequent treatments.

The current results also suggest that the frequent use of broad topographies within the included studies may be another potentially confounding element in the application of FA procedures in school settings. Studies frequently use a general term for the dependent variable such as “problem behavior” (e.g., Frieder et al., 2009) or “disruptive behavior” (e.g., Umbreit, 1995a, 1995b), which often include multiple topographies. Although multiple topographies may all be maintained by the same function in the same setting, it is possible that one or more topographies are maintained by a different function, which has the potential to confound the results (e.g., Beavers & Iwata, 2011; Derby et al., 1994; Derby et al., 2000). In experimental designs that use reversal elements to demonstrate functional relations (e.g., BFA; TBFA), the isolation of topographies may be possible; however, none of the included studies that utilized a general dependent variable (e.g., “problem behavior,” “disruptive behavior”) separated topographies within their analyses. Similar to the issue pertaining to invalid (i.e. Type I or Type II errors) and/or confounded results, inconsistencies with the treatment of general or topography-specific problem behavior may also have implications regarding the effectiveness of subsequent treatments informed by FA (i.e., Type I and/or Type II errors may result from the way in which aspects of problem behavior [general approach vs. topography-based] are targeted during the FA). Thus, it is possible that the treatments that are developed as a result of FAs will not be appropriately matched to the contingencies that are operating in the natural environment to produce problem behavior because of Type I and/or Type II errors (e.g., inappropriate application of reinforcement to respective topographies of problem behavior with different maintaining variables). Future studies should evaluate the extent to which these types of interactive effects might “misinform” subsequent treatments and affect their effectiveness.

Terminology of Functional Analysis

Previous literature reviews (e.g., Lloyd et al., 2016) utilized definitions of FA that resulted in the inclusion of analyses that did not include basic FA procedures (i.e., AFA, TBFA, BFA, LFA) to identify functions of problem behavior (e.g., demand assessment only) or used nonexperimental methodologies (e.g., structural descriptive assessment, high-preference vs. low-preference assessment, curriculum analysis). We chose to focus only on basic FA procedures (i.e., AFA, TFA, BFA, LFA) that (1) were experimental, (2) had the purpose of identifying the function of problem behavior, and (3) have effects that have been replicated across numerous studies and a variety of participants, settings, and behaviors, such as analog FA (Iwata et al., 1992), trial-based FA (Sigafoos & Saggers, 1995), brief FA (Northup et al., 1991), and latency FA (Thomason-Sassi et al., 2011). Therefore, the current review extends previous literature reviews on FA in public schools by incorporating stricter definitions of FA and broadening the study inclusion criteria in terms of who implemented the FA. These modifications allowed for a broader view of the experimental manipulations that have been conducted in public schools and how trends in school-based FA may have changed over time. However, as the amount of research on FA has increased along with the breadth of variations, future literature reviews may need a revised or more systematic set of criteria (e.g., a minimum requirement of experimental manipulations, more strict and standardized procedures for control conditions).

Limitations

The current results should be interpreted in light of several potential limitations. First, although a systematic online search was conducted as meticulously and systematically as possible, some results may have been inadvertently excluded from the current review. Second, outside of the criterion that there be a minimum number of manipulations (Hanley, 2012), no indicators of study quality were considered for inclusion. Third, interpretation of the maintaining variables of problem behavior in the reviewed studies was determined via the study authors’ interpretation of the results in lieu of independent analysis of the included results. Although inconsistent interpretations of FA results are likely mitigated by the peer-review process, the possibility of inconsistencies in interpretation remains; thus, it is a potential limitation of the present review.

Implications for Future Research

The findings of the current review suggest at least three potential avenues for future research. First, analysis of time requirements and social validity between methods of FA should be conducted in school settings to determine whether one method may be more efficient while maintaining effectiveness (i.e., reliable identification of functions, effective generation of treatments based on results). In particular, studies aimed at validating effective treatments across FA formats in school settings is warranted. Research in this area could lead to more widespread use of FA in public school settings, as well as more effective teacher training.

Second, FA-based research is needed in school settings regarding the implications of using general dependent variables/target behaviors with multiple topographies relative to the separating of individual topographies. Such studies may be useful to determine if the frequency of multiply maintained behavior is due to potentially confounding variables or if it is an accurate representation of the response–reinforcer relations demonstrated by individuals in school settings. Although more time and resources would be required to conduct FAs across multiple topographies, it is possible that the effectiveness of the treatments that are informed by and subsequently generated may be differentially affected (e.g., more or less effective) based on the configuration of the FA in this way (i.e., FA across multiple topographies vs. FA of combined topographies). Further the results of such studies could reveal information about the feasibility of FA procedures being used in environments such as public schools where experimental control is difficult.

Third, controlled large-scale or replication studies comparing intervention effectiveness and outcomes following FA compared to nonexperimental FBA methods in school settings would be of value. Although general research has been conducted in terms of experimental versus nonexperimental FBA (Hurl et al., 2016) as well as large-scale record reviews (Mueller et al., 2011), the overall small sample size of previous studies may limit the extent to which they may be confidently generalized to all school settings. Large-scale replication studies comparing the effects of interventions based on FBAs that do not include experimental methods (i.e., FA) versus interventions based on FBAs that do include experimental methods) could reveal important results pertaining to the way in which problem behavior is assessed in public schools.

Last, although a moderate percentage of the reviewed studies (51.2%) evaluated the effects of treatments based on the results of their assessment and demonstrated positive results, future studies should continue to systematically evaluate treatments based on the different FA formats. In particular, more studies are needed to confirm the treatment utility of the various FA formats. Future studies might also compare the relative effects of treatments that are informed by respective FA formats when the assessment results of FAs do not align.

Conclusion

Widespread inclusion of FAs into school-based FBAs may still be far off. Although the standard of inclusion of FA procedures is held and met in many ABA-based clinics and therapy centers, the amount of time, resources, and training needed to conduct FAs in school settings present barriers for widespread use. In particular, many constraints and variables exist in public schools that are distinct from other settings that make the public school context unique. Therefore, a focus on the literature pertaining exclusively to FA methods in public school settings is warranted. As the literature on the implementation of FAs in schools has grown, adaptations of AFA such as BFA and TBFA have shown increasing promise in terms of their effective inclusion in the FBA process. As teachers and researchers continually attempt to move forward in designing effective interventions for problem behavior in school settings, FA methodology may be a vital tool. Although more research on its application and effectiveness relative to current practice is needed, the use of FA as part of the FBA process may provide school staff members and parents a clearer path to their desired outcomes.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declarations

Conflict of Interest

The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.