Teachers Training Paraeducators to Implement Systematic Prompting Practices for Students with Significant Disabilities

Abstract

Teachers are often tasked with training paraeducators to deliver instruction to students with significant disabilities, but there is little research that involves teachers training paraeducators. In this single-case design study, we supported teachers to deliver tiered training that first involved group training for teams of paraeducators, and then follow-up coaching for the subset who did not meet criterion-level implementation fidelity. Three teachers at three different schools delivered tiered training on systematic prompting to nine paraeducators who taught students with significant disabilities. After tiered training, all nine implemented simultaneous prompting in both targeted (d = 0.73) and generalization situations (d = 0.76), as well as least-to-most prompting in both targeted (d = 0.65) and generalization situations (d = 0.37). Five met criterion-level adherence with group training alone, and four required coaching. Eight of the nine paraeducators also improved their implementation quality. Students made progress on individualized goals aligned with paraeducator-delivered instruction both in targeted (d = 0.99) and generalization situations (d = 0.14). These findings suggest that teachers can effectively implement tiered training that enables paraeducators to deliver evidence-based instruction to students with significant disabilities.

1. Introduction

In the United States, it is common for special education paraeducators to provide support and instruction to students with disabilities under the supervision of a licensed special education teacher (Carter et al., 2009). Paraeducators are not required to hold a teaching license and are often hired without any prior experience or training in education (Suter & Giangreco, 2009). Education requirements for paraeducators vary across states and schools, with some schools requiring only a high school diploma, and others requiring two years of study in higher education in any field, or a minimum level of performance on an assessment in reading, writing, and mathematics (Bisht et al., 2021). Paraeducators are an integral part of special education teams that serve students with significant disabilities. We define significant disabilities as students with a disability label of intellectual disability, autism, or multiple disabilities who have significant enough cognitive disabilities that it is not appropriate for them to participate in general standardized assessments even with accommodations. In the United States, these students are exempted from the general state-mandated standardized assessment and are instead assessed using an alternate assessment based on alternate achievement standards.

Paraeducators typically outnumber licensed special education teachers by more than 3:1 in programs for students with significant disabilities (Suter & Giangreco, 2009). One of the primary responsibilities of these paraeducators is to provide one-to-one instruction to students (Carter et al., 2009). In fact, paraeducators report spending substantially more time delivering instruction than special education teachers (Suter & Giangreco, 2009). Given that paraeducators greatly outnumber licensed teachers and provide more instruction than licensed teachers, the education of students with significant disabilities is profoundly shaped by paraeducator-delivered instruction.

The Individuals Education Improvement Act (IDEIA, 2004) allows for paraeducators to deliver instruction with the caveat that they be appropriately trained and supervised. Specifically, the law states that “paraeducators and assistants who are appropriately trained and supervised, in accordance with State law, regulation, or written policy…be used to assist in the provision of special education and related services” (20 U.S.C. 1412(a)(14)). The challenge lies in how to appropriately train and supervise paraeducators who deliver instruction to students with significant disabilities. In practice, efforts to train and supervise paraeducators are often lacking. Supervising teachers report that paraeducators are often inadequately trained to provide instruction, and that instructional strategies remain a significant training need (Walker et al., 2017). Most paraeducators report that their only training on instructional practices is informal, on-the-job training that is idiosyncratic and rarely competency-based (Carter et al., 2009).

A rapidly growing research literature provides guidance for how to train paraeducators more effectively. According to two systematic reviews of the literature, there are at least 49 experimental studies that test the efficacy of paraeducator training on implementation of practices with students with significant disabilities (Brock & Anderson, 2021; Brock & Carter, 2013), with nearly half being published in the past decade. Across this emerging literature, a number of themes have emerged. First, paraeducators roles should be clearly defined, and they should receive competency-based training that is aligned with their roles. Researchers have demonstrated that general, undefined support does not result in optimal outcomes (Brock & Anderson, 2021), and in some cases, can lead to adverse outcomes (Chung et al., 2019). However, when paraeducators are trained to implement an evidence-based practice with fidelity, they can contribute to improving student outcomes (Brock & Anderson, 2021; Brock & Carter, 2013). Second, efforts to train paraeducators should include a key combination of strategies: a checklist that details implementation steps, modeling of implementation steps, and performance feedback on implementation of the steps. These recommendations align with meta-analyses of the broader staff training literature in special education (Brock & Carter, 2017; Brock et al., 2017). Third, training should be sustained over time. Researchers have repeatedly demonstrated that one-shot workshops have very little impact on paraeducator teaching (e.g., Brock & Carter, 2015). Instead, paraeducator training should be sustained over time to allow for paraeducators to receive performance feedback multiple times (Brock & Anderson, 2021; Brock & Carter, 2013).

Although we know more today than ever before about how to best train paraeducators, there are still critical gaps in the research literature. Specifically, we have identified two critical limitations to the existing literature. First, very few studies involve implementation by teachers—the individuals who are typically responsible for paraeducator training. In their recent review of the literature, Borosh et al. (2023) only identified 11 moderate- or high-quality experimental studies that tested teacher-delivered training for paraeducators. Even fewer studies focus specifically on paraeducators who teach students with intellectual and developmental disabilities. These included teachers training paraeducators to promote alternative and augmentative communication (AAC; Andzik & Cannella-Malone, 2019; Wermer et al., 2017), systematic prompting (Britton et al., 2017; Heinrich et al., 2016; Mason et al., 2017), behavior-specific praise (Scheeler et al., 2018), and peer support arrangements (Brock & Carter, 2016). While these studies demonstrate initial promise for teacher-delivered paraeducator training, further replication is needed. Second, all seven of the aforementioned studies exclusively involved one-to-one coaching models. Therefore, there is no research-validated model for how teachers might train more than one paraeducator at the same time. This is critical, as teachers report that that it is difficult for them to find time to train paraeducators when balancing their other responsibilities (Sobeck et al., 2021).

One alternative to an exclusive reliance on coaching is a tiered model where coaching is only implemented when group training does not enable paraeducators to implement a practice with criterion-level fidelity. Brock et al. (2021a) demonstrated that tiered training could efficiently enable paraeducators to implement systematic prompting strategies with fidelity and generalize their implementation to new teaching situations. They trained 13 paraeducators to implement simultaneous prompting and least-to-most prompting. All 13 paraeducators implemented and generalized both procedures after training was complete, and 9 paraeducators only required group training to do so. While this model is promising and offers a possible pathway for teachers to more efficiently train groups of paraeducators, it has not yet been implemented by teachers.

We address these limitations in the present study by testing the efficacy of teacher-implemented tiered training on systematic prompting for paraeducators who teach students with significant disabilities. Specifically, we address the following research questions:

• What is the efficacy of teacher-implemented tiered training on paraeducator adherence to systematic prompting strategies (i.e., simultaneous prompting and least-to-most prompting) with students with significant disabilities? What are the contributions of the individual tiers of training, including group training (Tier 1) and one-to-one coaching (Tier 2)?
• When directed to use systematic prompting strategies in a new teaching situation for which they have received no coaching or feedback, to what degree do paraeducators generalize their adherence?
• How does paraeducator implementation quality change after receiving tiered training?
• What progress do students make on individualized goals after receiving instruction from paraeducators who have participated in tiered training?

2. Materials and Methods

2.1. Participants and Settings

All procedures were approved by a university Institutional Review Board (IRB). All teachers and paraeducators provided their consent, all students’ parents provided their permission, and all students provided assent prior to participating in this study. We recruited three teachers from three separate school districts who agreed to provide tiered training to paraeducators whom they supervised. The school districts served students from suburban and rural communities in a Midwest state. We recruited teachers by first asking administrators in the three districts to identify teams of special education teachers and paraeducators who served students with significant disabilities, and then meeting with teachers who were referred to us. In order for a school to participate in the study, we had to obtain consent from both a teacher and at least two paraeducators who provided instruction to students with significant disabilities (i.e., students who were eligible for their state’s alternate assessment for students with significant cognitive disabilities). We recruited four paraeducators in School 1, three in School 2, and two in School 3. For each paraeducator, the teacher identified a student with a significant disability to whom the paraeducator regularly delivered instruction who would benefit from systematic prompting, and we obtained permission from the student’s family and assent from the student. In one case, two paraeducators taught the same student (i.e., School 2 Paraeducators 2 and 3). Participants included 3 White female teachers, 9 White paraeducators (i.e., 2 male and 7 female), and 8 students. Additional characteristics of the teachers, paraeducators, and students are summarized in

Table 1. Characteristics of teachers, paraprofessionals, and students.

School #	Teachers			Paraeducators			Students
	Gender	Years Exper.	Highest Educat.	#	Gender	Years Exper.	Age	Grade	Disability	IQ Score	Adaptive Behavior Score
1	Female	4	Master’s	1	Male	3	13	7	MD	NV a	64 b
				2	Female	4	12	6	ASD	NR	<70 c,e
				3	Female	12	14	8	MD	NV	NR
				4	Female	3	15	8	ASD	NR	49
2	Female	1	Under-graduate	1	Female	NR	9	3	ID	52 d	NR
				2	Female	NR	9	3	MD	NV a	NR
				3	Male	<1	9 *	3 *	MD *	NV a,*	NR *
3	Female	13	Master’s	1	Female	2	13	7	ASD	NR	54 c
				2	Female	4	14	8	ASD	NR	48 c

Note: Schools and paraeducators are numbered consistently in this fashion for ease of cross-referencing throughout the manuscript. Only primary educational labels are reported. ASD = Autism Spectrum Disorder; ID = Intellectual Disability; MD = Multiple Disability; NR = Not Reported; NV = No Valid Score. ^a Naglieri Nonverbal Ability Test; ^b Vineland Adaptive Behavior Scales, 3rd Edition; ^c Adaptive Behavior Assessment System, 3rd Edition; ^d Wescheler Intelligence Scale for Children, 5th Edition; ^e A numerical score was not included in the report; rather, the narrative stated that the score fell in the extremely low range (score < 70). * Paraeducator 7 taught the same student as paraeducator 6.

. All students spent time in both special education and general education classrooms, and our observations of paraeducator-implemented instruction took place in special education classrooms.

2.2. Experimental Design and Procedures

We implemented a concurrent multiple probe across participants using a single-case design (Ledford & Gast, 2024). The introduction of the intervention was staggered across the three schools; however, data were collected at the level of individual paraeducators and students. We randomized the order in which the intervention was introduced across schools using a computerized random number generator, which enhances the strength of the design (Kratochwill & Levin, 2014).

2.2.1. Student Goal Selection, Teacher Training, and Task Analyses of Chained Skills (Pre-Baseline)

Student Goal Selection. We met with each special education teacher to identify individual goals for each student participant. We required that the learning targets (a) be included as a goal in their Individualized Education Program (IEP) or part of the general education curriculum, (b) include behaviors, either discrete or chained, that could be objectively measured, (c) be appropriate to target with systematic prompting, (d) be possible to target with materials already available in the classroom or that could be easily produced, and (e) have not already been mastered by the target student. Each teacher selected two goals per student. The goals were either assigned to be a target goal or a generalization goal using a computerized random number generator. See

Table 2. Focus of student goals targeted by paraeducators.

School Number	Para Number	Primary Goal	Generalization Goal
1	1	Telling time to the half hour	Identifying simple fractions
	2	Identifying first name in print	Stamping name onto written work
	3	Typing first name	Identifying community signs
	4	Identifying coins by name and value	Identifying community signs
2	1	One-digit addition	Identifying coins by name and value
	2	Identifying letters by their sound	Identifying numbers 1–10
	3	Identifying letters by their name	Identifying coins by name and value
3	1	Counting combinations of 1, 5, and 10-dollar bills	Two-digit addition
	2	Counting manipulatives up to 10	Reading high-frequency sight words

for a list of outcomes targeted by goals in target and generalization conditions for each student. For target goals, paraeducators received instruction and performance feedback on their implementation of systematic prompting to teach this goal during group training. For generalization goals, paraeducators did not receive instruction or feedback related to teaching the goals until the one-on-one coaching portion of this study, if needed (see Section 2.2.5).

Teacher Training. The first author met with each teacher to train them on the group training procedures. During these meetings, which lasted approximately 45 min, the first author oriented them to a 19-page teacher manual. The teacher manual included identical content as the written materials that paraeducators received, with the addition of a training checklist on the left margin of each page. The training checklist included a step-by-step description of how to conduct the training. Information contained in the margins included (a) prompts to refer to key information, (b) prompts to provide examples from the teacher’s own classroom, and (c) steps for modeling procedures and conducting role play of the systematic prompting procedures. The first author provided an overview of the teacher manual and highlighted how it paralleled the written materials that paraeducators received. They then walked the teacher through the entire coaching manual, explaining every item in the training checklist and modeling key training behaviors. Teachers were invited to ask questions throughout the training, and the first author invited questions again after they had explained and modeled all training behaviors.

Task Analyses of Chained Skills. The week following the teacher training, the teacher held one training session with the paraeducators, which focused on (a) identifying whether the skills they were teaching were either discrete or chained, and (b) conducting task analyses for any chained skills. The training lasted approximately 30 min. Training strategies included (a) sharing written descriptions and examples of discrete skills, chained skills, and task analyses; (b) providing didactic instruction; and (c) supporting paraeducators to record task analyses of chained skills they were targeting and provide feedback. At the conclusion of the pre-baseline training session, task analyses were completed for all chained skills (both target skills and generalizations skills).

2.2.2. Written Directions (Baseline)

We distributed written directions to all paraeducators for how to implement both simultaneous prompting and least-to-most prompting during the baseline condition. The written directions included a definition of each prompting procedure (i.e., simultaneous prompting and least-to-most-prompting), an explanation of when and why the prompting procedure should be used, and an implementation checklist for each procedure. We instructed the paraeducators to read the handouts, but did not provide any additional training or support during the baseline condition. We distributed these checklists so that the paraeducators would have a conceptual understanding of the procedures we were asking them to implement, and so we could isolate the effects of the tiered training above and beyond the effects of providing written directions. Based on findings in previous studies, written directions have limited effects and are unlikely to result in optimal paraeducator implementation of systematic prompting strategies (Brock & Anderson, 2021).

2.2.3. Group Training for Paraeducators

Each of the three special education teachers provided training across three sessions over three weeks for paraeducators at their respective school. The first author attended each training session to support the teacher and ensure adherence to the training protocol. The sequence of topics was designed to build on each other and incorporate feedback on previously taught strategies in each subsequent lesson.

The focus of the first training session was simultaneous prompting. At the start of the session, the teacher checked that the paraeducators had a copy of the trainee manual, including didactic materials, implementation checklists, and planning sheets. Then, the teacher conducted the training by (a) providing a rationale for simultaneous prompting, (b) explaining the steps of simultaneous prompting and providing examples, (c) modeling the implementation steps, (d) directing paraeducators to practice through role play, and (e) providing feedback to paraeducators following the role play by praising steps that were followed with adherence and offering suggestions for fixing mistakes. The role play and feedback continued until the paraeducators practiced implementing the strategy successfully. At the end of the session, the first author introduced the paraeducators to the video-feedback portion of the training. Paraeducators were given a tablet computer (i.e., Amazon Fire) and mini-tripod and directed to video record their implementation of simultaneous prompting with the target students so that they could share their performance with the group at the next training session. The paraeducators were told that they only needed to video record their implementation of the target goal (i.e., the goal for which they received instruction during training). Instruction that targeted the generalization goal was not recorded during the group training condition and no feedback was provided for that skill during group training.

The second training focused on feedback on the video recording of simultaneous prompting and an introduction to least-to-most prompting. First, the teacher briefly reviewed the implementation steps for simultaneous prompting. Each paraeducator then shared their video of their implementation of simultaneous prompting with the group. After each video, the teacher provided feedback on the paraeducator’s performance and also provided the other paraeducators an opportunity to provide feedback. The teacher would repeat the same process with each video. After reviewing the videos, the teacher introduced least-to-most prompting in the same fashion as described for simultaneous prompting. Additionally, the teacher discussed and modeled how to record data on the student performance during least-to-most prompting sessions, either using data sheets provided by the teacher or data sheets included as part of the trainee manual. At the end of the session, the teacher directed paraeducators to video record their implementation of both simultaneous prompting and least-to-most prompting before the next training session.

The final training session focused on providing feedback for both simultaneous and least-to-most prompting. Video feedback was provided in the same fashion as described in training session two for simultaneous prompting.

2.2.4. Maintenance

No additional training was provided in the maintenance condition. Procedures were identical to the baseline condition in which paraeducators had access to written directions.

2.2.5. Individual Coaching

If a paraeducator failed to demonstrate criterion-level adherence (i.e., ≥85% of steps implemented correctly) following training for two consecutive sessions for either procedure (i.e., simultaneous prompting or least-to-most prompting) for either student goal (i.e., target goal or generalization goal) during the maintenance condition, the special education teacher delivered a 15 min one-to-one coaching session for that paraeducator. Coaching focused on the goal and procedure for which the paraeducator did not demonstrate criterion-level adherence. Coaching sessions were conducted in the same manner as the feedback portion of training sessions two and three, but in a one-to-one format. Unlike the group training, however, coaching could include feedback on the generalization skill if adherence while targeting the generalization skill fell below the criterion. A member of the research team (i.e., either the first or second author) was present at all coaching sessions.

2.3. Dependent Measures and Recording

2.3.1. Classroom Observations

We observed paraeducators in their classrooms as they were delivering one-on-one instruction to the target students. The time of observations was chosen in consultation with the classroom teachers and was designed to occur at the time of day during which the paraeducators would naturally be working with the target student. All observations involved the paraeducator providing instruction on both goals (i.e., target and generalization) chosen for the students prior to the start of intervention, and then the student being probed on each goal. We observed paraeducators delivering instruction using both simultaneous prompting and least-to-most prompting on both goals during each observation. The paraeducator delivered instruction using simultaneous prompting and least-to-most prompting and completed the student probe for one goal before moving on to the second goal. Directions and measurement were identical for both the target goal and generalization goal. First, we directed the paraeducator to implement simultaneous prompting as if they were teaching the skill for the first time. Next, we directed the paraeducator to implement least-to-most prompting as if they had already taught the skill for several days and the student had demonstrated initial progress. The directions were phrased in that manner (i.e., “teaching the skill for the first time” and “had already taught the skill”) in order to teach the paraeducators to pair the procedures with the situations in which they were most appropriate. Lastly, we directed the paraeducator to deliver a student probe to gauge student progress. During the student probe, the paraeducators were directed to withhold any prompts so that we could gauge independent student performance. We measured implementation fidelity in two ways: adherence to steps of the prompting protocol, and quality of implementation of systematic prompting. Adherence and quality represent two distinct dimensions of implementation fidelity (O’Donnell, 2008).

Adherence to Steps. The primary dependent variables were the adherence to the implementation steps for simultaneous prompting and least-to-most prompting, measured using implementation checklists. Steps for simultaneous prompting included (a) delivering a controlling prompt within 1 s of providing a cue or task direction, (b) delivering the same topography of controlling prompt for each trial, and (c) responding to the student response by either providing specific praise for a correct response or repeating the trial after an incorrect response. Steps for least-to-most prompting included (a) waiting 3–5 s following the cue or task direction for the student to provide a response, (b) providing specific praise for a correct independent response, (c) delivering prompts of increasing intensity after incorrect responses within the same trial, and (d) delivering the same hierarchy of prompts across multiple trials. The implementation steps were based on implementation checklists developed by the National Professional Development Center on Autism Spectrum Disorders (Neitzel & Wolery, 2009a, 2009b). We recorded the data as the percentage of steps implemented correctly, which we calculated by dividing the number of steps implemented correctly by the number of total steps attempted multiplied by 100.

Quality of Implementation. We measured implementation quality as a second, descriptive variable. Following each session, we scored the implementation quality using a six-item rating scale. Each item was rated using a 4-point Likert-type scale with three being the highest quality and zero being the poorest quality. Items focused on the pace of instruction, consistency of delivering the same cue or task direction, immediacy of prompting after an error, quality of verbal praise, immediacy of reinforcement after correct responses, and overall quality of implementation. We computed average scores across items as a measure of overall quality, with a possible range of 0–3.0. The complete quality instrument is published in Brock et al. (2021b).

Student Performance on Individualized Goals. As part of each observation, we administered probes to the students for both the training goal and the generalization goal. During the probe, the student had the opportunity to provide 10 discrete response (or 8 for students working on identifying coins and their values) or 10 chained responses associated with the target and generalization goals described in Table 2. We calculated the percentage of correct responses by dividing the number of correct responses by the number of opportunities to respond and multiplying by 100.

2.3.2. Observer Training and Reliability

The second author, a doctoral candidate in special education, was the primary data collector for this study. She was already trained by the first author on the observation protocol and had conducted a number of studies using the observation protocol prior to the current study. The second author trained one additional graduate student (the third author) and three undergraduate students as additional data collectors, with the undergraduate students primarily collecting data for reliability purposes. Before collecting live data in the classroom, each observer was required to (a) review the coding manual with the second author, (b) score 100% on a written test of coding definitions, (c) achieve at least 95% agreement with the second author when coding training videos, and (d) achieve at least 95% agreement with the author in a live observation.

Two observers collected data during 27.9% of all observations across participants and conditions. We computed point-by-point agreement for each behavior. The average overall agreement for adherence to simultaneous prompting was 91.1% (range = 66.7–100% across observations), and adherence to least-to-most prompting was 83.7% (range = 33.3–100%). Exact agreement for implementation quality ratings was 80.2% (range = 0–100%), while agreement within one point of the primary observer was 94.5% (range = 66.7–100%).

2.4. Procedural Fidelity

We measured procedural fidelity for the teacher training, group training, and coaching conditions. For the teacher training, the first author completed a checklist at each meeting documenting that they trained the teacher by (a) orienting them to the teacher manual and written materials for paraeducators; (b) covering every item in the training checklist; and (c) inviting and answering all questions. The second author attended one of the three teacher training sessions and confirmed that 100% of steps were followed with fidelity. During the group training and coaching sessions, the first author attended all sessions and confirmed that 100% of the steps were followed with fidelity using the training checklist (for training sessions) or a coaching checklist (for coaching sessions).

2.5. Data Analysis

The data were analyzed in two ways. First, we conducted visual analysis to determine whether we could be confident that experimental effects were demonstrated for each dependent variable. Specifically, we inspected the data patterns within and across conditions in terms of level, trend, variability, and immediacy of effect (Ledford & Gast, 2024). Second, we summarized the magnitude of effects using a between-case standardized mean effect size (BC-SMD; Pustejovsky et al., 2014). The BC-SMD is computed using a multi-level model to account for the nesting of time within participants, and it estimates the mean difference between the baseline and intervention conditions across participants while adjusting for variation within and between cases (Pustejovsky et al., 2014). We used the calculator developed by Pustejovsky et al. (2023) to calculate a BC-SMD for each dependent variable using Restricted Maximum Likelihood (RML) estimation and allowing for case-level random effects in both baseline and treatment conditions. The BC-SMD is a design-comparable estimate, meaning that it produces an effect size directly comparable to Cohen’s d in group design studies (Pustejovsky et al., 2014).

3. Results

We report on the results of (a) paraeducator adherence to systematic instructional strategies, (b) paraeducator implementation quality, and (c) student performance on individualized goals.

3.1. Paraeducator Implementation of Systematic Instructional Strategies

We measured two dimensions of implementation fidelity: (a) adherence to steps, and (b) implementation quality. The results from each are reported below, including a visual analysis of the data in terms of level, trend, variability, and immediacy of effects. The adherence to the steps for simultaneous prompting and least-to-most prompting across training and generalization setting is graphed in Figure 1. Implementation quality is described in

Table 3. Implementation quality by paraeducator and condition.

School Number	Paraeducator Number	Average Implementation Quality
		Baseline	Intervention
1	1	2.24 (1.83–2.50)	2.71 (2.60–3.00)
	2	2.20 (1.67–2.67)	2.86 (2.50–3.00)
	3	1.88 (1.00–2.50)	2.41 (2.00–2.83)
	4	2.31 (1.67–2.83)	2.69 (2.50–3.00)
2	1	2.28 (2.00–2.50)	2.00 (1.67–2.67)
	2	1.94 (1.67–2.33)	2.50 (2.50–2.50)
	3	1.89 (1.67–2.33)	2.50 (2.50–2.50)
3	1	2.31 (2.00–2.50)	2.80 (2.50–3.00)
	2	1.68 (1.00–2.60)	2.20 (2.00–3.00)

.

3.1.1. Adherence to Steps for Simultaneous Prompting

We measured adherence to steps for simultaneous prompting for the trained skill (i.e., the teacher provided feedback on performance during group training and coaching) and the generalization skill (i.e., the teacher did not provide feedback on performance during group training and coaching).

Targeted Training Situations. The percentage of correctly implemented steps for simultaneous prompting for the training goal is represented by filled circles in Figure 1. During the written instructions condition, all paraeducators implemented simultaneous prompting below the criterion. Eight of the nine paraeducators had at least one session with 0% of the steps implemented with fidelity; the lowest performance, for Paraeducator 1 in School 3, was 50%. Although the baseline performance was variable, in only two cases was performance consistently trending toward improved implementation fidelity. After the introduction of the group training condition, fidelity improved for all paraeducators to the criterion level (i.e., >85%). For seven of the nine paraeducators, this level was maintained during the maintenance phase. For the two paraeducators that did not maintain criterion levels (i.e., Paraeducator 3 in School 2 and Paraeducator 2 in School 3), coaching was provided. After coaching, both paraeducators implemented simultaneous prompting at the criterion level. The magnitude of the effect, based on the BC-SMD (Pustejovsky et al., 2023), was medium (d = 0.73; SE = 0.32; 95% CI = 0.06–1.39).

Generalization Situations. The percentage of correctly implemented simultaneous prompting steps for the generalization goal is represented with open circles in Figure 1. During the written instruction condition, eight of the nine paraeducators implemented simultaneous prompting below the criterion. Eight of the nine had at least one session with 0% fidelity. The lowest performance for Paraeducator 1 in School 2 was 20% of the steps implemented correctly, and it was 50% for Paraeducator 2 in School 2. After the introduction of the group training condition, fidelity improved to criterion-level performance (i.e., 85%) for six of the nine participants. For Paraeducator 3 in School 1, performance was either 100% or 0%, with most days at 0%. For Paraeducators 1 and 2 in School 3, performance did not meet the criterion, but improved after coaching was provided for the target skill. The magnitude of the effect, based on the BC-SMD (Pustejovsky et al., 2023), was medium (d = 0.76; SE = 0.24; 95% CI = 0.25–1.27).

3.1.2. Adherence to Steps for Least-to-Most Prompting

We measured adherence to the steps for least-to-most prompting on the targeted goal and the generalization goal.

Targeted Training Situations. The percentage of correctly implemented steps for least-to-most prompting for the training goal is represented by filled squares in Figure 1. During the written instruction condition, all paraeducators implemented simultaneous prompting below the criterion. Five of the nine paraeducators had at least one session with 0% of the steps implemented with fidelity; the lowest performance for Paraeducators 2 and 4 in School 1 and Paraeducator 1 in School 3 was 50%. The lowest performance for Paraeducator 3 in School 2 was 25%. Although baseline performance was variable, in only one case was performance consistently trending toward improved implementation fidelity. After the introduction of the group training condition, fidelity improved for two of the nine paraeducators to the criterion level (i.e., >85%). Six paraeducators continued to perform at the criterion level during the maintenance phase. Coaching was provided to the paraeducators who did not meet criterion-level performance. After coaching, these paraeducators demonstrated performance at the criterion level. The magnitude of the effect, based on the BC-SMD (Pustejovsky et al., 2023), was medium (d = 0.65; SE = 0.25; 95% CI = 0.15–1.15).

Generalization Situations. The percentage of correctly implemented steps for least-to-most prompting for the generalization goal is represented by open squares in Figure 1. During the written instruction condition, all paraeducators implemented simultaneous prompting below the criterion (i.e., 85%). Three of the nine had at least one session with 0% fidelity. The lowest performance for the six paraeducators who scored above 0% fidelity was between 25 and 50%. After the introduction of the group training condition, fidelity improved to criterion-level performance (i.e., 85%) for six of the nine participants. Five of the nine paraeducators maintained their performance at the criterion or above. In cases where paraeducators did not meet the criterion during maintenance (e.g., Paraeducator 2 in School 2, Paraeducator 2 in School 3), coaching was provided for the targeted skill. After coaching, performance was variable but increased to the criterion after subsequent coaching sessions. The magnitude of the effect, based on the BC-SMD (Pustejovsky et al., 2023), was small (d = 0.37; SE = 0.31; 95% CI = −0.26–1.00).

3.1.3. Implementation Quality

Implementation quality was rated on a four-point scale ranging from 0 (poor quality) to 3 (excellent quality). Across paraprofessionals, the mean level of quality in the baseline condition was 2.08 (range = 1.68–2.31) and in the intervention condition was 2.52 (range = 2.00–2.86). Eight of the nine paraeducators improved their implementation quality between these phases. The only paraeducator who did not improve was Paraeducator 1 in School 2. For means and ranges by condition for each paraeducator, see Table 3.

3.2. Student Progress on Individualized Goals

Student performance was measured as a secondary descriptive variable. Student performance was measured as the percent of correct responses during a probe for both the training goal (filled triangles) and generalization goal (open triangles) and is graphed in Figure 2. Because it was a secondary descriptive variable, it is not appropriate to draw causal conclusions for this variable. Furthermore, a lack of stability in baseline performance for many of the students does not make it possible to confidently conclude that improvement in student performance can solely be attributed to post-training paraeducator-implemented instruction. In general, students did make progress over time, although not all students met the performance criterion (i.e., >80%) for all goals. Specifically, five out of the nine students met the performance criterion for the training goal, and three met the criterion for the generalization goal. For training goals, the magnitude of the effect, based on the BC-SMD (Pustejovsky et al., 2023), was large (d = 0.99; SE = 0.31; 95% CI = 0.36–1.62). For generalization goals, the magnitude of the effect was small (d = 0.14; SE = 0.16; 95% CI = −21–0.49).

4. Discussion

Many students with significant disabilities receive a substantial portion of their instruction from paraeducators who have not been well-trained to implement evidence-based practices (Carter et al., 2009). Existing research provides guidance for how to better prepare paraeducators, but important gaps remain in the literature. Specifically, very few studies involve teacher-implemented training for paraeducators, and none of them involve teachers training more than one paraeducator at a time (Brock & Anderson, 2021). More feasible solutions are needed in which teachers can train multiple paraeducators simultaneously. In this study, three teachers delivered tiered training to nine paraeducators. First, all paraeducators received group training, and then one-to-one coaching was only delivered when paraeducators did not achieve a training criterion. After tiered training, all nine paraeducators implemented simultaneous prompting and least-to-most prompting with criterion-level fidelity in both target and generalization situations. Five met criterion-level fidelity with group training alone, and four required coaching to achieve the criterion in one of the targeted or generalization situations. The findings from this study contribute to the literature in a number of ways.

First, teachers can feasibly deliver tiered training to paraeducators. In this study, three teachers and three different schools demonstrated the ability to feasibly deliver group training to all paraeducators on their team, and then coaching to those who required further support in order to implement with fidelity across situations. While previous studies have demonstrated the efficacy of tiered training for paraeducators (e.g., Brock et al., 2021a) and there are published guides for how teachers can implement such training (i.e., Brock, 2022), this is the first experimental study in which teachers have implemented tiered training for paraeducators. It is important to note that across all three schools, arranging a time for group training required support from administrators who either facilitated staff coverage to enable paraeducators to be released from other responsibilities in order to attend the training during the school day, or allowed paraeducators to flex their time when students were not present (e.g., staying after student dismissal for later than the contract time in order to complete group training, and then being permitted to leave earlier than the contract time on other days to offset the extra time). Therefore, teacher facilitation of tiered training may only be feasible given this type of support from administrators.

Second, teacher-delivered tiered training enables paraeducators to implement systematic prompting strategies with fidelity. At the conclusion of tiered training, all nine paraeducators implemented both simultaneous prompting and least-to-most prompting with fidelity across two different situations. Five paraeducators acquired and generalized fidelity with group training alone, and four required one-to-one coaching to reach criterion fidelity in one or more situations. Teachers were able to train paraeducators to the criterion more efficiently using a tiered approach than if they had provided one-to-one coaching to every paraeducator. Furthermore, this is one of only a handful of studies on staff training that included training features intentionally designed to promote generalization, and to measure the degree to which generalization occurred (Shawbitz & Brock, 2023). Based on the results from this study, modeling multiple exemplars is a strategy than enables the majority of paraeducators to generalize fidelity to new situations, but in some cases, performance feedback is still required in the generalization context (i.e., sequential modification; Stokes & Baer, 1977).

Third, all students made progress on at least one individualized goal, but a lack of progress on some goals points to a need for teachers to direct paraeducators to make adjustments to their instruction. Student progress was measured as a secondary descriptive variable, and variability and trends during the baseline make it difficult to draw strong conclusions about the impact of paraeducator implementation of evidence-based prompting strategies on student outcomes. What is clear is that for a subset of students, paraeducator implementation of these strategies alone did not result in acquisition of IEP goals before the end of this study. Because students with significant disabilities sometimes make slow and steady progress over time, it is possible that some of these students might have mastered their goals if data collection were continued across a longer period of time. Another explanation is that, in some cases, paraeducators did not acquire criterion-level adherence until relatively late in the study, so some students only experienced the prompting strategies as designed for a small number of sessions. In addition, it is likely that some students would have benefited from adjustments to the instruction, such as changing reinforcement approaches, using antecedent-based interventions to increase engagement, or using a different instructional approach. By measuring paraeducator fidelity, it is possible for teachers to eliminate poor paraeducator fidelity as a reason for lack of student progress and then proceed to systematically make adjustments to the instructional plan.

4.1. Implications for Practice

The findings from this study have important implications for teachers and administrators. First, the findings from this study are consistent with a growing body of literature that suggests paraeducator roles should be clearly defined, and that they should receive competency-based training that is sustained over time and includes an implementation checklist, modeling, and performance feedback. Second, administrators should ensure that teachers have sufficient time to plan and deliver paraeducator training, and that there are opportunities to deliver the training during the hours that paraeducators are paid to work. Based both on the literature (e.g., Sobeck et al., 2021) and our work in the three schools in this study, it seems this time is typically not available in the absence of intentional efforts by administrators. Third, this study provides initial evidence that teachers should consider using a tiered training model when a team of paraeducators will be using the same practice. Given a large team of paraeducators, some paraeducators may be able to acquire and generalize implementation fidelity with group training alone, while others may require one-to-one coaching to achieve or generalize fidelity. A critical component of this style of training is measuring paraeducator implementation fidelity and making data-based decisions about what subsequent training or coaching might be required. Live observations may not always be feasible, and this study provides an illustration of how video-based observations can be sufficient for assessing fidelity and delivering feedback.

4.2. Limitations and Future Directions for Research

Limitations in this study point to future directions for research. First, two of the nine paraeducators demonstrated progress toward implementation fidelity during the baseline condition for one or more of the prompting procedures, making it impossible to make claims about the efficacy of the group training in those specific cases. It is possible that these paraeducators were able to make some progress from the written directions that were made available in the baseline condition. In future studies, researchers might seek to replicate this by testing if a subset of paraeducators demonstrate implementation fidelity with written directions alone. Second, student data were variable during the baseline condition. Measured as a secondary descriptive variable, in some cases, student progress trended upward before paraeducators had received training. We suspect that in some cases, the expectation of being observed resulted in paraeducators more consistently and systematically targeting student goals, and that these changes may have sometimes led to student progress even when prompting practices were not implemented with fidelity. In future studies, researchers might seek to measure if simply observing paraeducators deliver instruction focused on a specific goal—in the absence of any training—might lead to student progress on that goal. Third, we collected less maintenance data than we originally planned because of student and staff absences that could not be made up when the school year ended. In future studies, researchers might begin their studies earlier in the school year to mitigate this problem. Fourth, we did not collect social validity data in this study. In future studies, researchers might gauge the perceptions of both teachers and paraeducators about the training procedures.

4.3. Conclusions

Coaching is a powerful tool for supporting paraeducators to implement evidence-based practices, but it is not feasible to ask teachers to provide separate one-to-one coaching to every paraeducator on their team. This study demonstrated that teachers can feasibly implement tiered training in which coaching is only provided when paraeducators do not acquire or generalize implementation fidelity with group training alone. Given administrator support, this more efficient approach may enable teachers to more feasibly and effectively support their paraeducators.