Lee Mason

Purpose: For Mexican families of autistic children, finding access to specialized services such as applied behavior analysis can be challenging. Yet behavior-analytic intervention is critical for remediating stimulus overselectivity within the verbal repertoire. Here the authors present a case study on the use of telehealth to deliver behavior-analytic intervention to improve the verbal behavior of an autistic child who lived with her mother in northeastern Mexico. Design/methodology/approach: Weekly sessions consisted of natural environment training that was interrupted every 10 min to conduct a 1-min fluency probe. After establishing a stable baseline, a clinician-led intervention was implemented, in which the behavior analysts manipulated the environment to condition different verbal operants. This was followed by a parent-led intervention, in which the participant’s mom manipulated the environment and reinforced the verbal operants. Findings: Throughout the clinician-led intervention phase, there was broad overlap among the number of correct and incorrect verbal responses per minute. Similar levels of variability could be seen at the onset of parent-led intervention, but the two trendlines soon diverged, showing a clear separation between the increasing number of correct responses per minute and the decreasing number of incorrect responses per minute. Originality/value: The authors extend prior research to show that function-based approaches to verbal behavior intervention can be effective when delivered via telehealth. While clinician-led intervention showed limited efficacy, it likely served as a critical component of training. Although limited by a quasi-experimental design, the results suggest telehealth-based verbal behavior intervention was effective for remediating functional language skills.

We briefly introduce the science of behavior analysis and the extent to which it overlaps with the model of ecological affordances described by Ko and Neuberg. Although we largely agree with the description of human-environment interactions offered by the affordance-management framework, we aim to refine the author’s discussion of goal seeking to avoid circular reasoning and afford validation through experimentation.

In this chapter, we provide a conceptual overview of a deep neural network (DNN) algorithm as applied to the prediction of human behavior. DNNs are one among many powerful and quickly evolving machine learning (ML) algorithms capable of performing tasks previously attributable only to humans. The objective in training a DNN is to develop a standalone predictive model that can perform independently when presented with entirely new/unseen input values. DNNs and a wide range of related ML algorithms are becoming increasingly prevalent across disciplines as researchers design neural networks to answer novel questions and forecast nuanced outcomes regarding human behaviors. Our chapter focuses on this quickly evolving technology as applied to behavior analytic investigations. With a special emphasis on generalization and validation strategies applied to small-n clinical datasets, we provide a conceptual framework for one type of deep neural network. However, the principles we discuss are applicable to a much broader range of neural network algorithms.

Background & Aims: Access to specialized healthcare for children with autism spectrum disorder has long been hindered by limited availability of qualified clinicians, particularly in rural and economically disadvantaged areas. The COVID-19 pandemic exacerbated these disparities, overwhelming traditional healthcare models and highlighting the need for innovative solutions. Telehealth emerged as a critical tool during the pandemic, facilitating continued access to care, particularly for underserved populations. We aimed to investigate the extent to which functional language deficits can be identified and assessed in children with autism, and ruled out in neurotypical children, via telehealth. Methods: This study investigated the feasibility and effectiveness of using telehealth for assessing verbal behavior deficits in children diagnosed with autism, comparing them to neurotypical peers, and exploring the potential for telehealth to expand access to specialized intervention. Twelve children, seven of whom had an autism diagnosis, participated in telehealth-based functional analyses of verbal behavior. Results: Results indicated that children with autism exhibited statistically significant differences in across functional language domains, whereas no such differences were identified in neurotypical peers. No false positives or false negatives were detected. At follow-up, participants with autism showed statistically significant improvements in verbal language skills. Conclusions: Our findings suggest that telehealth is a reliable and valid method for assessing the verbal behavior of children with autism. This research supports the use of telehealth to bridge healthcare gaps for families in remote or underserved areas. By providing an accessible platform for functional language assessment, telehealth offers a promising avenue for enhancing early intervention and improving long-term outcomes. Implications: Our findings highlight the importance of individualized interventions to address verbal behavior deficits of children with autism. Future research is necessary to explore the scalability of these findings and their application in broader, more diverse populations.

The most effective treatments for autism spectrum disorder have generally focused on single-subject data to personalize behavior-analytic intervention at the individual level. Though interventions based on applied behavior analysis have been effective for treating the core deficits of autism, the lack of large-scale data analysis has prohibited the use of precision medicine for identifying subtypes of responders within this population. Language deficits are characteristic of an autism diagnosis, particularly for those more profoundly affected by the disorder, who may not develop functional communication despite early intervention. Yet language development is often described as a progression through stages that may not accurately describe the manner in which individuals with autism acquire language skills. In the current study, we use binomial logistic regression to examine the statistically significant predictors of exteroceptive control over verbal behavior. Using a common curriculum-based assessment of verbal behavior, we examined the intake scores of 118 children with autism who received early intensive behavioral intervention. Results show that an individual’s ability to mand reliability predicts their ability to tact, but no statistically significant variables were related to mand control. We discuss our findings within the context of intervention for individuals with autism whose verbal behavior is under differentiated sources of control.

This study explores the use of diversity indices, typically employed in ecological studies, to measure verbal behavior in children with autism spectrum disorder. By drawing comparisons between species richness and abundance, we applied Shannon entropy and beta diversity measures to assess the diversity and distribution of a verbal repertoire over time. This analysis utilized the outcomes of three verbal operant experimental analyses that use a multi-element design to test for the occurrence of four elementary verbal operants. Using archival data, diversity measures were then used to conduct a detailed analysis of the richness and evenness of verbal responses. The results demonstrate that beta diversity and Shannon entropy effectively capture the richness and distribution of the verbal repertoire over time. These findings suggest that diversity indices can provide a robust framework for assessing language development and the effectiveness of interventions. This research underscores the importance of integrating ecological diversity models into behavioral science to better understand complex human behaviors like language. Future applications of this method may allow for more complex analysis of verbal behavior within and across individuals and interventions.

A growing body of literature supports the use of echoic prompts toward conditioning other functional language skills. However, many individuals with autism spectrum disorder do not emit echoic behavior. Identifying the prerequisite skills of an echoic repertoire may be beneficial for intervention planning and clinical decision making. A chart review was conducted for 118 patients with autism who received early intensive behavioral intervention. We examined the Level 1 scores on the Verbal Behavior Milestones Assessment and Placement Program (Sundberg, 2014) for all patients. Using their scores on the echoic skills subtest as a dependent variable, we ran a binomial logistic regression in which the remaining eight domains, along with age and sex, served as independent variables. Our logistic regression model was statistically significant: X ² (10) = 109.61, p < .001. The model explained 86.0% of the variance in echoic verbal behavior and correctly classified 95.8% of cases. Patients who demonstrated a mand repertoire were greater than 3.5 times more likely to also exhibit an echoic repertoire. Additionally, vocal play and motor imitation were associated with an increased ability to echo. Using binomial logistic regression, we identified three predictors influencing the echoic behavior of children with autism. Patients who demonstrated these three skills were statistically significantly more likely to echo the vocal verbal behavior of others. Additional research is now needed to confirm a functional relationship between each of these predictors and echoic control.

As an exploration of machine learning strategies applied to behavior analytic investigations, we sought to predict the likelihood that particular children with Autism Spectrum Disorder (ASD) would develop echoic verbal behavior. After retrieving 143 case records of children assessed with the VB-MAPP from a data repository, we reserved 22 case records for holdout analysis. Through iterative development, the model achieved 100% accuracy in predicting the presence or absence of an echoic repertoire. While the functional value of predicting a known outcome might be questioned, it is only possible to gauge the accuracy of any prediction technique when known outcomes are available. Thus, rather than trying to predict a completely unknown outcome, our experimental preparations were directed at gauging the multilayer perceptron’s (MLP) ability to predict at high levels of precision, sensitivity, specificity, and overall prediction accuracy while accessing a limited dataset from one children’s healthcare setting. We discuss procedures for employing this machine learning strategy in conjunction with a receiver operator characteristics analysis of our findings. We developed this study as a general guide for using one of several quickly evolving neural network methodologies for behavior analytic researchers, and we examine specifics regarding MLP functions and operations as they relate to providing more efficacious treatment for children with ASD.

Chapter 10 focused on the distinct functional properties of verbal behavior, such as mand control (i.e., requesting). Mand training is a technique for strengthening a child’s ability to request access to specific reinforcers, like toys, snacks, activities, and interactions with other people. Research has shown that verbal behavior may need to be explicitly reinforced as a request for it to be useful as such. Mand training may lead to an increase in spontaneous language, an increase in social initiations, and a reduction of problem behavior. However, sometimes challenging behavior interferes with skill acquisition. When challenging behavior occurs at high rates or presents a potential for harm, mand training may be combined with extinction through a chaining procedure known as functional communication training.

[This corrects the article DOI: 10.1007/s40614-024-00404-6.].