Over the course of a year, we examined the verbal behavior of Ryan, a 7-year-old boy whose parents had emigrated from India. Ryan was exposed to multiple languages at home, including English, Telugu, and Tamil. At the start of the present study, Ryan’s verbal behavior consisted solely of single-word responses in English. Table 1 shows the languages used as a speaker and listener by members of Ryan’s verbal community.

Several of the neighboring families were also from India, and Ryan’s parents primarily communicated with them in either Telugu or Tamil. Ryan’s multilingual verbal community presented a unique opportunity to assess his functional speech across non-English languages. This study took place in the participant’s home where he received early intensive behavioral intervention (EIBI) from a Registered Behavior Technician (RBT). A supervising Board Certified Behavior Analyst independently recorded the frequency and topography of Ryan’s tacts, mands, echoics, and intraverbals for all trials of every assessment. Trial-by-trial interobserver agreement measured 100%.

A verbal operant experimental (VOX) analysis (Enriquez et al., 2023) was conducted in English to assess Ryan’s present levels of functional speech. Since Skinner’s (1957) introduction of the elementary verbal operants, researchers studying language remediation in children with ASD have primarily focused on four sources of environmental control: mand, tact, intraverbal, and echoic (DeSouza et al., 2017). Given that pure sources of control rarely occur in the natural environment (Michael et al., 2011), we acknowledge the potential for supplementary variables.

A VOX analysis extends prior research on functional analyses of verbal behavior (see Plavnick & Normand, 2013) to examine the functional interdependence of verbal operants by assessing the same topographies under different sources of control. We plotted the frequency of verbal responses observed across each operant class on a radar chart to create a polygonal profile of Ryan’s verbal behavior. Polygonal language profiles allow for visual and quantitative analyses using a normalized first moment of area (Q’; Porter & Niksiar, 2018); a measure analogous to area under the curve, ranging from 0.00 to 2.00.

The VOX analysis began by allowing Ryan to select from a variety of preferred items to assess his labeling (i.e., tact control). Upon picking up an item, the RBT inquired, “What is it?” To control for potential tact confounds, the RBT ensured that: (a) access to the item was not restricted (i.e., mand control), (b) the name of the item was not provided (i.e., echoic control), nor (c) was a verbal description of the item presented (i.e., intraverbal control). If he labeled the item, the RBT provided praise. For example, while holding a gel-filled, thermoplastic rubber ball Ryan said, “Squishy,” and the RBT said, “Yes, it is a squishy!” After 20 s, the target item was removed, and Ryan was encouraged to select another response target. We repeated this process until we assessed labeling with three different items: “Pig,” “Squishy,” and “Turtle.”

Next, we used a multiple-stimulus without replacement preference assessment to assess Ryan’s requesting (i.e., mand control) for each of the three items identified in the labeling condition. We placed all three items on the floor in front of Ryan and asked him to choose one. Upon selecting one, the RBT removed the other two items. After 20 s, the RBT removed the target item and asked, “What do you want?” To control for potential mand confounds, the RBT ensured that: (a) the item was not physically present (i.e., tact control), (b) the name of the item was not provided (i.e., echoic control), nor (c) was a verbal description of the item presented (i.e., intraverbal control). If he requested the item, the RBT gave it to him. For example, having selected the squishy ball, the RBT hid it behind her back to evoke Ryan’s request, “Squishy.” When he said, “Squishy,” she gave it to him for another 20 s. The first item was removed, and Ryan was asked to select one of the two remaining items. We repeated this process repeated until we assessed requesting with all three items: “Squishy,” “Pig,” and “Turtle.”

Then we assessed Ryan’s echoing (i.e., echoic control), using the name for each item emitted by Ryan in the prior conditions as antecedent verbal stimuli. Throughout these next two conditions, Ryan was engaged in an activity unrelated to the items being assessed to abolish their reinforcing value. To control for potential echoic confounds, the RBT ensured that: (a) the item was not physically present (i.e., tact control), (b) access to the item was not restricted (i.e., mand control), nor (c) was a verbal description of the item presented (i.e., intraverbal control). The three antecedent verbal stimuli were presented semi-randomly to systematically vary the order of presentation from the previous two conditions, thereby controlling for potential sequencing effects. For each word he echoed, the RBT provided praise. For example, the RBT said, “Say, squishy.” If Ryan said “Squishy,” the RBT replied, “Good saying squishy!” We continued this process until we assessed echoing the names of all three items: “Turtle,” “Pig,” and “Squishy.”

Finally, we assessed Ryan’s conversing (i.e., intraverbal control) about each of the three items. Ryan was engaged in an activity unrelated to any of the response targets, and at 20-s intervals, he was asked to complete a fill-in-the-blank statement specific to Ryan’s interaction with each item. These intraverbal frames were structured to occasion Ryan’s response to complete the sentence by saying the name of a corresponding response target. To control for potential intraverbal confounds, the RBT ensured that: (a) the item was not physically present (i.e., tact control), (b) access to the item was not restricted (i.e., mand control), nor (c) was the name of the item provided (i.e., echoic control). Additionally, the order in which the RBT presented the three fill-ins was varied intentionally from the previous three conditions to control for potential sequencing effects. If Ryan answered the fill-ins, the RBT provided praise. For example, while Ryan was playing with a musical toy, the RBT said, “You squeeze the ….” If Ryan’s said, “Squishy,” the RBT enthusiastically repeated, “Squishy!” We repeated this process until we assessed conversing about all three items: “Pig,” “Turtle,” and “Squishy.”

After Ryan’s functional language had been assessed for each of the three items, a second round of the assessment was conducted. The Labeling condition came first to allow Ryan to select three novel verbal response targets, but we varied the sequence of the other conditions from the first round: Conversing, Echoing, and Requesting. Two rounds of each condition were necessary to achieve an adequate sample size to analyze Ryan’s verbal behavior.Footnote 1 All conditions were completed in one session, with the total assessment lasting approximately 30 min.

Another VOX analysis was conducted to assess progress and identify present levels of functional language. Since the initial VOX analysis, 30-h per week of EIBI had been conducted in English. Mom continued to speak to Ryan in Telugu, and Dad spoke to him in English. Ryan – now 7.5 years old – continued to be exposed to Tamil intermittently, but we did not observe his verbal community directly reinforcing Tamil as either speaker or listener. We reconducted the VOX analysis as described above.

To comprehensively assess Ryan’s verbal behavior, we conducted a VOX in Telugu and another in Tamil. The procedures in both assessments were identical to those used in the English assessment except that the RBT delivered antecedent verbal stimuli in Telugu or Tamil and consequences followed vocalizations in Telugu or Tamil.Footnote 2 That is, for the Labeling conditions, the RBT called for Ryan to respond in English (e.g., “What is that?”), Telugu (e.g., “Adi ēmiṭi?”), or Tamil (e.g., “Atu eṉṉa?”). For the Requesting conditions, the RBT called for Ryan to respond in English (e.g., “What do you want?”), Telugu (e.g., “Nīku ēmi kāvāli?”), or Tamil (e.g., “Uṅakku eṉṉa vēṇnum?”). For the Echoing conditions, the RBT called for Ryan to respond in English (e.g., “Say …”), Telugu (e.g., “Ceppu …”), or Tamil (e.g., “Sollu …”). For the Conversing conditions, the RBT called for Ryan to respond in English (e.g., “Choo, choo goes the …”), Telugu (e.g., “Cū, cū veḷtāḍu …”), or Tamil (e.g., “Cū, ccū celkiṟatu …”).

The intensity of Ryan’s behavior-analytic intervention was titrated down for the subsequent 6-month treatment authorization. Ryan received only six hours per week of focused behavior-analytic intervention, and he attended school two days per week. Furthermore, his services were interrupted for consecutive weeks when COVID-19 spread through his family, and again when his family traveled to a Tamil-speaking region of India for a month. At the end of the second 6-month authorization period, Ryan was 8 years old. We conducted three VOX analyses in English, Telugu, and Tamil to assess Ryan’s functional verbal behavior across languages.