Clay et al., 2021

Overview of the Article

  • Focus: Feasibility and preliminary efficacy of using Virtual Reality–based Behavioral Skills Training (VR-BST) to teach preservice clinicians to implement Functional Communication Training (FCT)
  • Authors & Affiliations: Casey J. Clay, Brittany A. Schmitz, Bimal Balakrishnan, James P. Hopfenblatt, Ashley Evans, SungWoo Kahng; Universities of Missouri & Rutgers
  • Publication: Journal of Applied Behavior Analysis, 2021, Vol. 54, pp. 547-565
  • Core research questions
    • Can VR-delivered BST (instructions ➔ modeling ➔ rehearsal ➔ feedback) bring novice students to mastery when teaching FCT for escape- and attention-maintained problem behavior?
    • What are the associated time, cost, and social-validity outcomes?

Key Concepts & Definitions

  • Behavioral Skills Training (BST)
    • Components: written/verbal instructions, modeling, rehearsal, feedback
    • Commonly used to teach functional behavior assessment/intervention skills to staff, teachers, and parents
  • Functional Communication Training (FCT)
    • Differential-reinforcement procedure teaching a socially acceptable communicative response (FCR) that accesses the reinforcer previously maintaining problem behavior
    • Evidence-based for reducing severe challenging behavior in Autism Spectrum Disorder (ASD)
  • Virtual Reality (VR)
    • Immersive computer-generated environment experienced through head-mounted displays (HTC Vive® in this study)
    • Presence: psychological state in which virtual objects/actors are experienced as real; requires accurate self-location & possibilities for action
  • AutSim©
    • Custom VR simulation of an autism clinic treatment room, including a motion-captured child avatar capable of scripted problem behavior and FCRs

Rationale for VR-BST

  • Traditional in-vivo BST exposes trainees & clients to risks
    • Aggression, self-injurious behavior (SIB) ➔ staff injuries (Hostetler 2019, Whitman 1976)
    • Infectious-disease transmission (e.g., SARS, tuberculosis; Schablon 2009)
    • Logistical costs of human confederates (training, scheduling, social distancing)
  • VR advantages
    • Eliminates direct contact; repeatable, safe rehearsal; automated or remote delivery; potentially inexpensive with modern consumer hardware

Development of AutSim© VR Simulation

  • Design goal: high experiential congruence with a real autism clinic room
  • Technical elements enhancing presence
    • High-detail architectural modeling (Autodesk 3DS Max)
    • Motion-captured child avatar
    • 6-degree-of-freedom navigation & hand interaction via HTC Vive controllers
    • One-way mirror, clock, FCR card, table/chairs accurately scaled to real room (10 m × 7 m)
  • Software stack: Unity 3D engine + C# scripting; flexible to migrate to standalone headsets (e.g., Oculus Quest)

Method

Participants & Setting

  • N = 13 female undergraduate Health-Science majors (no prior ABA/FCT training)
  • Freshmen = 2, Sophomores = 6, Juniors = 2, Seniors = 3
  • Sessions in a cleared conference room; consent obtained; optional compensation \le 30\text{ USD}

Experimental Design

  • Non-concurrent multiple-baseline across participants
  • Component analysis of BST
    1. Instructions (baseline)
    2. Modeling
    3. Rehearsal + feedback
  • Mastery criterion: \ge 90\% correct implementation across 3 consecutive sessions

Session Types

  1. Attention FCT (60 scripted actions/session)
    • EO: therapist withdraws attention
    • FCR prompt on FT 30\text{ s}; deliver 30\text{ s} of high-quality attention for FCR
    • Ignore screaming, hitting, property destruction, threatening approach
  2. Escape FCT (30 scripted actions/session)
    • EO: present academic/functional task
    • 3-step guided compliance hierarchy (instruction ➔ instruction + model ➔ physical guidance)
    • FCR prompt on FT 30\text{ s}; deliver 30\text{ s} break; maintain demands for problem behavior

Response Measurement

  • Data collectors observed physical trainee behavior + live overhead VR feed
  • Each child action = a scored opportunity; coded correct vs error using operational definitions (see Tables 1 & 2)
  • Percentage correct = \frac{#\;\text{Correct}}{#\;\text{Correct}+#\;\text{Errors}}\times100

Interobserver Agreement (IOA)

  • Scored in 10\text{-s} intervals
  • Attention sessions: IOA mean =85\% (range 76\text{–}97\%; 46\% of sessions)
  • Escape sessions: IOA mean =94\% (range 77\text{–}100\%; 95\% of sessions)

Social-Validity Assessment

  • 11-item Likert 1\text{–}7 survey (adapted from Higgins 2017) via Qualtrics; anonymous responses from 12/13 participants

Time & Cost Tracking

  • Personnel time derived from grant budgets; direct monetary costs capped by small internal grant
  • Key 1-year expenditures (Table 4):
    • Programmer GA: 10\text{ h/wk} \times 40\text{ wk} = 400\text{ h} ➔ 10{,}015\text{ USD}
    • Lead researcher oversight: 10\text{ h/wk} \times 52\text{ wk} = 520\text{ h} ➔ 19{,}500\text{ USD}
    • Research assistants for sessions: \approx 180\text{ h} ➔ 3{,}780\text{ USD}
    • Hardware/software: gaming laptop 1{,}649\text{ USD} + Unity/other licenses 1{,}500\text{ USD}
    • Participant payments: 200\text{ USD}
    • Total direct cost: 26{,}629\text{ USD} (indirect costs not included)

Results

Attention-Maintained FCT (Figure 2)

  • Instructions phase mean =73\% (range 57\text{–}88\%)
    • Participant “April” reached mastery during instructions alone
  • Modeling phase mean =93\% (range 80\text{–}100\%); six more participants mastered
  • Feedback phase mean =99\% (range 99\text{–}100\%); remaining three reached mastery
  • Most frequent error: incorrect EO arrangement
  • Least frequent error: provision of attention for problem behavior
  • Time to mastery: mean 716\text{ min} post-Instructions/Modeling; mean 85.6\text{ min} post-Feedback

Escape-Maintained FCT (Figure 3)

  • Instructions phase mean =25\% (range 3\text{–}51\%); no mastery
  • Modeling phase mean =72\% (range 41\text{–}91\%); “Jennifer” mastered
  • Feedback phase mean =95\% (range 90\text{–}98\%); remaining four mastered
  • Most frequent error: failure to differentially reinforce compliance vs guided compliance
  • Least frequent error: failure to deliver break after FCR
  • Time to mastery: mean 66.8\text{ min} post-Instructions/Modeling; mean 78.6\text{ min} post-Feedback

Social-Validity Findings (Table 3 highlights)

  • Highest satisfaction: modeling component (mean 6.7/7)
  • Overall technology/setup satisfaction: 5.8/7
  • Perceived utility for other clinical skills: 5.8/7
  • Lowest mean ratings (≈ 4.7\text{–}4.9/7) tied to comfort performing while observed, especially when questions were prohibited (Instructions/Modeling phases)
  • Open-ended feedback stressed enjoyment of VR, but discomfort without immediate Q&A and difficulty tracking timing/wording rules without feedback

Cost & Efficiency Insights

  • Average individual training time < 100 minutes to mastery—comparable or faster than published live BST durations
  • VR removes cost of confederates, reduces trainer staffing; one-time programming cost amortizable over large cohorts

Discussion & Implications

  • Efficacy: VR-BST reliably produced mastery; 7/13 participants succeeded without feedback, suggesting virtual contingencies may provide automatic “programmed feedback” (natural consequences of correct/incorrect actions)
  • Safety: Eliminates risk of trainee injury, client injury, and disease transmission during rehearsal of severe behavior intervention
  • Scalability & Access:
    • Compatible with consumer headsets; future autonomous modules (gamification, automated feedback) could remove live trainers entirely
    • Potential for remote/telehealth deployment, benefiting rural or pandemic-restricted training contexts
  • Educational Technology Perspective: Illustrates fusion of “technology of tools” and “technology of process” (Twyman 2014); calls for more behavior-analytic R&D in automated, competency-based VR learning

Limitations

  • No assessment of generalization to live clients
  • Only two behavioral functions (attention, escape) targeted
  • Correct-response topographies not individually coded; limits fine-grained error analysis
  • Indirect costs and potential graphics/feature upgrades not analyzed

Future Research Directions

  • Direct comparison of VR-BST vs traditional BST on efficacy, safety, cost (e.g., randomized group designs)
  • Test maintenance & generalization to in-vivo contexts and additional functions (e.g., tangible, automatic reinforcement)
  • Expand AutSim© with variable client avatars, automated feedback, data-analytics dashboards
  • Investigate training effect on trainee response variability and generalization through programmed scenario diversity

Ethical, Practical & Philosophical Considerations

  • Reduces ethical concerns of exposing clients/trainees to risk prior to mastery
  • Raises questions about balancing technological immersion with humanistic elements in clinician preparation
  • Supports broader move toward competency-based, data-rich education in health sciences and applied behavior analysis

Quick-Reference Numerical Summary

  • Participants: 13 undergraduates
  • Scripts per session: 60 (attention) / 30 (escape)
  • IOA: 85\% (attention), 94\% (escape)
  • Mastery criterion: \ge 90\% correct × 3 consecutive sessions
  • Time to mastery (post-full BST): Attention \approx 86\text{ min}; Escape \approx 79\text{ min}
  • Direct project cost (1 yr): 26{,}629\text{ USD}