Participants

Participants included three children from a center providing applied behavior analytic services on the East Coast of the United States. Idris, a 4-year-old African-American boy, completed the study in 58 days. Elisia and her monozygotic twin, Delicia, both 4-year-old White girls, completed the study in 96 and 98 days, respectively. Each child came from a monolingual, English-speaking household. The children’s parents expressed interest in their children learning a vocabulary in a second language. Spanish was selected as the foreign language in consultation with caregivers.

Inclusion criteria required a formal diagnosis of ASD from an external professional and limited or no challenging behaviors, as reported by the Board Certified Behavior Analyst (BCBA) and based on data collected daily in the classroom. Participants also needed to score either level 2 or 3 on the Echoic, Listener Responding, Mand, Tact, and Intraverbal modules of the Verbal Behavior Milestones Assessment and Language Programs (VB-MAPP; Sundberg, 2014). The Childhood Autism Rating Scale-2 (CARS-2; Schopler et al., 2010) was completed by each child’s BCBA and caregivers at the end of the study to provide additional information on the participant’s characteristics. The experimenter administered the Preschool Language Scale-Fifth Edition (PLS-5; Zimmerman et al., 2011) to evaluate each participant’s language proficiency using the auditory comprehension and expressive communication subscales. Table 1 shows assessment results.

Table 1 Participant demographics and language assessment scoresSetting and Materials

Sessions occurred in a preschool room with a table, chairs, picture cards (9×9 cm), timers, a camera, toys, and edible reinforcers. Each meeting lasted 30–45 min, averaging 146.7 s per session, consisting of six trials, with six to 12 sessions per day. The experimenter had 6 years of applied behavior analytic experience and intermediate Spanish proficiency, ensuring accurate pronunciation. Stimuli were common objects (see Supporting Information A).

Dependent Variables and Response Measurement

Trained observers collected data using paper and pencil data sheets to score the participant’s response in each trial throughout all conditions. The experimenter (first author) collected data in vivo during study sessions. Secondary observers, who were undergraduate psychology students at the time of the study, collected data from video recordings of the sessions. The experimenter used behavioral skills training to train secondary observers to a predetermined mastery criterion. Once observers met this criterion, they collected data independently. The dependent variables included echoic, listener responding, tact, and intraverbal responses. A correct echoic response required the participant to repeat the word spoken by the experimenter within 10 s of the vocal discriminative stimulus (SD): “Say (name of the word).” A correct echoic response required demonstrating formal similarity and point-to-point correspondence or close resemblance with the actual word in either English or Spanish. As an example of formal similarity and point-to-point correspondence, if the experimenter asked the participant to say “grater,” a correct echoic response was the participant repeating “grater.” As an example of formal similarity and close approximation, if the experimenter said “dinero,” a correct approximation might be “dimero,” in which the child substitutes the /m/ for the /n/ sound but maintains the overall sound pattern of the modeled word. Substitutions of native for non-native phonemes were scored as correct (e.g., an English speaker learning Spanish pronounces “pero” with an English-like /ɹ/ instead of the Spanish trill /r/). An incorrect response occurred if the participant responded after 10 s, failed to respond, or provided an unrelated response, such as saying “cutting.” Echoic responses were assessed for all targets in both English and Spanish.

For listener responding, a correct response was defined as pointing to the corresponding picture in the corresponding language within 10 s of the instruction. The experimenter presented an array of three pictures and instructed the participant to “Point to (name of the picture in the corresponding English or Spanish).” An incorrect response included any of the following: responding after 10 s, not responding at all, or failing to point to the corresponding picture. Listener responses were assessed for all potential targets in both languages.

To ensure that participants learned tacts in the target language without influence from instructions, vocal instructions were delivered only in English (i.e., the participants’ native language). The specific SDs (i.e., “What is this in English/Spanish?”) were selected to maintain consistency across languages and to promote intraverbal emergence during subsequent intraverbal probes by establishing clear, language-specific verbal antecedents (i.e., the words English and Spanish, respectively). During the English tact baseline, intervention, pretest, and posttest, the vocal SD was “What is this in English?” During the Spanish tact baseline, intervention, pretest, and posttest, the vocal SD was “What is this in Spanish?” A tact was considered correct when the participant emitted a vocal response in the corresponding language that accurately matched the visual stimulus in the presence of the instruction “What is this in English/Spanish?” An independent correct response was recorded if the participant’s response occurred within 10 s of the SDs specified above. An incorrect response included failing to respond within 10 s, providing a response that did not correspond to the visual stimulus (e.g., saying “yellow” when shown a picture of a sponge), or responding after the 10-s interval.

During intraverbal trials, a correct response was recorded when participants vocalized the Spanish translation of the English word or vice versa within 10 s of the vocal SD “Rallador in English is…” or “Grater in Spanish is…” An incorrect response included any of the following: responding after 10 s of the SD, not responding, or providing a response that did not correspond to the vocal SD. The experimenters conducted probes for both English-to-Spanish and Spanish-to-English intraverbals to determine if bidirectional relations between stimuli in the two languages developed.

For all measures, the percentage of independent correct responses for each session was determined by dividing the number of independent correct responses by six (representing the six trials per session) and then multiplying the result by 100%.

Error Analysis

The experimenters examined the types of errors made by participants during each baseline and intervention session. Participants’ error responses were transcribed and classified by data collectors following post-session review of video recordings. Analyzing these errors enabled the researchers to form hypotheses about the sources of stimulus control for incorrect responses and assess whether the types of errors varied across different conditions and comparisons. The errors were classified into four categories: (a) errors associated with the same or different language (category 1), (b) errors originating from the same target set or a different target set (category 2), (c) no response (category 3), and (d) other (category 4). Categories 1 and 2 pertain to responses involving targets presented within the experimental conditions. A detailed description of each error category is provided in Table 2.

Table 2 Error categories, descriptions, and examplesInterobserver Agreement and Procedural Fidelity

Observers were trained via behavioral skills training (Sarokoff & Sturmey, 2004) to ≥90% IOA across three consecutive sessions. An agreement was determined when both observers either recorded a correct response or scored an incorrect response. IOA was calculated by dividing the number of agreements by the total of agreements plus disagreements and multiplying by 100%. The mean IOA was 99.8% (range: 94.5–100%) across all preassessments, pre- and posttests, and instructional conditions for all participants. For baseline and intervention sessions, 100% of sessions in both the monolingual and bilingual conditions were scored for IOA for correct and incorrect responses, with mean agreement of 100% for Idris, 100% for Elisia, and 100% for Delicia. For error analysis, an agreement was defined as both observers assigning the same error category to a given response. IOA for error analysis was also scored in 100% of baseline and intervention sessions in both monolingual and bilingual conditions, using the same method mentioned above, with a mean IOA of 100% for all participants.

The same independent observers also recorded data on procedural fidelity for 100% of sessions during preassessments, pre- and posttests, and intervention sessions. Independent observers used task analyses to document procedural fidelity for each session. A separate task analysis was developed for each procedure (e.g., preference assessment, baseline, echoic probes); the number of steps in each task analysis ranged from 5 to 14. For example, for tact instruction at 10-s, the task analysis included the following steps: (a) seating the participant at the table or on the floor across from or next to the experimenter, (b) at the start of each session, experimenter present an array of three to four tangible/edible reinforcers and instruct the participant to “Pick one” (or similar); proceed only after a selection is made, (c) experimenter present the target card within the participant’s view and deliver the target question (e.g., “What is it in English?”), (d) if the participant emits an independent correct response within 10 s, experimenter deliver praise and the selected reinforcer (one edible or 30-s access to the tangible), (e) if no response occurs within 10 s, experimenter provide a prompt, (f) if the participant emits a correct prompted response within 10 s of the prompt, experimenter deliver praise, (g) if the participant emits an incorrect response or does not respond following the prompt, experimenter implement error correction, repeating up to three times, and (h) if the participant does not respond correctly after three error-correction attempts, terminate the trial and proceed to the next trial. Task analyses are available upon request. Refer to Supporting Information B for the task analysis of the 10-s instructional procedures. Procedural fidelity was calculated by dividing the number of correct opportunities by the total number of opportunities and multiplying by 100. The mean procedural fidelity was 100% across all preassessments, pre- and posttests, and instructional conditions for all participants.

A second observer independently collected procedural fidelity data for 38.5% of monolingual sessions and 35.0% of bilingual sessions. IOA for procedural fidelity was calculated as described previously. The mean procedural fidelity agreement for all preassessments, pre- and posttests, and instruction was 99.3% (range: 96–100%) for Idris, 99.3% (range: 94.7–100%) for Elisia, and 99% (range: 97–100%) for Delicia.

Experimental Design

An adapted alternating treatments design, with within- (N = 4) and between-subject replications (N = 3; Rodriguez et al., 2025; Kazdin, 2019), compared monolingual and bilingual tact instruction. Stimuli and conditions were randomized to prevent more than three consecutive sessions of the same condition. Once the mastery criterion was reached in one condition, sessions proceeded exclusively in the remaining condition. Stimulus generalization, listener responses, and bidirectional intraverbals were assessed with pretest–posttest probes.

Procedure

Please see Fig. 1 for a flowchart depicting the procedures employed in this study.

Fig. 1

Flowchart illustrating the breakdown of study procedures

Stimulus Selection and Assignment

In the bilingual condition, we taught three tacts in English and the corresponding tacts in Spanish concurrently (simultaneously). In contrast, we taught a separate set of six stimuli in English only for the monolingual condition. All instructions were provided in English (the participants’ native language), such as asking, “What is this in Spanish?” Translations for the target words were verified using Merriam-Webster’s Spanish-English dictionary and online resources (e.g., Google Translate) to ensure accuracy and correct pronunciation. To control for prior learning history, the experimenters included only words that were not present in participants’ repertoires, as indicated by incorrect responding across languages during the pre-assessment (described below). The selected words represented common items that participants might encounter in their environment (e.g., mosquito, knob, plunger).

Stimuli were assigned on the basis of several criteria meant to ensure consistency across conditions (Cariveau et al., 2021) and increase visual disparity within each set (Halbur et al., 2021). The criteria included: (a) no more than three syllables per word; (b) each word in a set began with a different initial letter; (c) words within a set were distinct in spelling and sound, with only minimal overlap in endings; (d) visual stimuli differed in shape and color to enhance visual distinction; (e) participants performed similarly across target stimuli during the tact, listener responding, and intraverbal pretests; and (f) participants could pronounce each word in both English and Spanish during the echoic assessment (described below). The first four criteria were established by the experimenters prior to data collection, while the remaining two were assessed during pre-intervention data collection.

A target was replaced if participants (a) achieved an average response rate above chance level (i.e., 33.3%) during the listener responding sessions, and (b) correctly identified the target during the tact assessment and pretest probes (described below). The stimuli consisted of nine color pictures (six for the monolingual condition and three for the bilingual condition), displayed on picture cards with a white background. During tact instruction, a corresponding color card was presented for each condition, determined by the color preference assessment (described below). Throughout the four within-subject replications of this study, a total of 36 stimuli were taught (24 stimuli for the monolingual condition and 12 for the bilingual condition). Although only 12 stimuli were used in the bilingual condition, each was taught in both English and Spanish, resulting in 24 total targets. The list of stimuli taught to each participant across comparisons and conditions is available as Supporting Information A. Across assessments and conditions, all sessions had an equal number of trials.

Preference Assessments

The experimenter conducted multiple stimulus without replacement (MSWO) preference assessments (DeLeon & Iwata, 1996) to identify participants’ preferred stimuli, including toys, edibles, and colors, and establish a preference ranking. Separate assessments were conducted for tangible and edible items due to differing motivating operations, with edibles being more influenced by pre-session access (Miller et al., 2025). Preference was determined on the basis of the third MSWO session (Conine et al., 2021), with the top three preferred items from edibles and tangibles selected. For colors, the third and fourth-ranked colors were chosen to avoid biases (Luczynski & Hanley, 2009) and assigned to experimental conditions.

Pre-assessment

We conducted two types of pre-assessments—tact and echoic—as described below. During these assessments, we did not offer any programmed consequences, feedback, or error correction for either correct or incorrect responses; the experimenter simply moved on to the next trial or concluded the assessment after the final trial. The experimenter incorporated mastered trials after every two target response trials (Pizarro et al., 2021); for example, motor imitations and actions (e.g., “Show me clapping!”) or fill-in-the-blank questions (e.g., “The sky is __”) that participants had previously mastered in the classroom. When participants responded correctly in mastered trials, we provided praise (e.g., “Good job!”) and other preferred items (such as toys or edibles). Each session comprised six trials of target stimuli (i.e., stimuli subsequently targeted for tact instruction) and three mastered trials.

Tact Assessment

The experimenter conducted a tact assessment to determine whether participants could accurately tact visual stimuli in English and Spanish. Words were excluded from the study if participants demonstrated independent correct responses. During the tact assessment, the experimenter presented a visual stimulus (i.e., a picture) and asked the question, “What is this in English/Spanish?” and allowed the participants 10 s to provide their responses.

Echoic Assessment

The experimenter performed an echoic assessment to evaluate if the participant could pronounce the words in both English and Spanish and to ascertain whether echoic prompts could be used during tact instruction. In this assessment, the experimenter instructed the participant by saying, “Say [name of item]” and then allowed 10 s for a response. No visual stimuli were presented during the echoic assessment.

Pretest and Posttest Probes

We administered three types of probes: listener responding, intraverbal, and generalization (described below). The experimenter conducted four sessions for each probe—two sessions prior to the beginning of the study (pretest) and two sessions after mastery of each condition (posttest). The procedures for pretest and posttest were identical, with sessions structured similarly to the tact assessment; each session included six trials with teaching targets and three trials with mastered targets. For each probe, the experimenter instructed the participant to identify the targets in both English and Spanish within the same session.

Listener Responding Probe

The experimenter administered listener-responding probes to evaluate whether participants could respond as listeners to the stimuli included in the study. During these probes, the experimenter presented three visual stimuli (i.e., three pictures from the same set; called comparison stimuli) on picture cards arranged horizontally in front of the participant, with each spaced approximately 10 cm apart. The participant was instructed to “Point to [name of item]” in either English or Spanish. The positions of the comparison stimuli were randomly altered (i.e., left, middle, right) across trials.

Intraverbal Probe

The experimenter assessed intraverbal responses through bidirectional relations between English and Spanish. In both the pretest and posttest phases, the experimenter conducted two sessions of six trials for each stimulus set across both conditions. In the English-to-Spanish trials, the experimenter stated, “[English name of item] in Spanish is __,” for example, “Grater in Spanish is __?” In the Spanish-to-English trials, the instructor stated, “[Spanish name of item] in English is…,” such as, “Rallador in English is __?” Visual stimuli were not presented during these sessions.

Generalization Tact Probe

The experimenter conducted generalization probes to assess whether the participant exhibited stimulus generalization by tacting visually similar stimuli. Similar stimuli were selected by maintaining the critical features of the items while varying nonessential attributes (Layng, 2018), such as identifying a “stapler” despite variations in size, shape, or color. The procedure was similar to the tact assessment. The experimenter included two distinct exemplars of each target and presented each exemplar twice within the sessions (i.e., once in English and once in Spanish).

Baseline

Following the assessments and probe sessions, the experimenter implemented baseline conditions to evaluate the participant’s level of responding across experimental conditions before beginning tact instruction. The procedures for the baseline sessions were identical to those used in the tact assessment described above. A total of four baseline sessions were conducted for each condition. During the baseline sessions, the experimenter presented a visual stimulus (i.e., a picture card) and asked, “What is this in English/Spanish?” allowing participants 10 s to provide their responses. Each session consisted of six trials, with mastered trials interspersed after every two target trials.

Tact Instruction

During tact instruction sessions, the experimenter presented a visual stimulus to the participants and asked, “What is this in English/Spanish?” The experimenter then delivered echoic prompts (i.e., vocalizations of the correct responses) using a constant prompt-delay procedure (Walker, 2008). Initially, the echoic prompts were provided immediately (0-s delay) after the presentation of the visual stimulus and the question. The mastery criterion for the 0-s delay was two consecutive sessions with 83.3% or more correct responses, and the mastery criterion for the terminal delay was two consecutive sessions with 83.3% correct independent responses. Once the participant’s responding met the mastery criterion, the delay between the visual stimulus presentation and the echoic prompt increased to 10 s. When the participant responded incorrectly, we implemented an error correction procedure that consisted of the experimenter delivering the prompt 0 s after an incorrect response from the participant, followed by the next trial. The error correction procedure was implemented up to three times. If the participant did not respond correctly on the third attempt, the experimenter moved on to the next trial. Independent correct responses (i.e., vocalizations of the corresponding name in English or Spanish) resulted in the cancellation of the scheduled prompt presentation, such that no prompt was delivered following a correct response. If the participant made an error, the experimenter provided the prompt and recorded the response as incorrect. The experimenter delivered social and tangible/edible consequences contingent on participants’ responses. Praise followed both independent correct and prompted responses, whereas tangible/edible reinforcers were provided only for independent correct responses. The magnitude of the social consequences was greater for independent correct responses compared to correct prompted ones (Karsten & Carr, 2009). For prompted correct responses, the experimenter offered a single statement of praise (e.g., “Good job!”). In contrast, for independent responses, the experimenter provided enthusiastic praise (“You are doing a great job!”) and granted access to tangible or edible reinforcers for 30 s.

If a participant provided an incorrect response three times following the error correction procedure, the experimenter proceeded to the next trial without offering social consequences. Each session consisted of six trials, and no mastered trials were presented during the tact instruction sessions. The participants had the same exposure to targets across conditions (i.e., with an equal number of targets and trials in each condition).

Experimental Conditions Condition 1 (Monolingual)

To emulate monolingual instruction, participants were taught to tact items in a single language (i.e., English). We chose English for its social relevance, considering that it was the primary language used by all participants in both school and home settings. In these sessions, all six targets were presented in English, with each target representing a different item. The experimenter taught all six targets simultaneously (i.e., within the same session, concurrently), using the question, “What is this in English?” for each target. Each session consisted of six trials, with each target presented once per session in a variable order predetermined by the researcher.

Condition 2 (Bilingual Simultaneous)

We selected simultaneous (i.e., concurrent) teaching procedures because previous studies have demonstrated that they lead to higher levels of emergent relations as compared to sequential (i.e., serial) teaching procedures (Thordardottir, 2019; Wunderlich & Vollmer, 2017; Wunderlich et al., 2014). That is, we taught participants to tact items in both languages concurrently (i.e., within the same sessions). In these sessions, the experimenter taught three targets in English and three targets in Spanish. Corresponding vocal SDs indicated which language was expected for the correct response; for example, the experimenter asked, “What is this in English?” for the English response and “What is this in Spanish?” for the Spanish response. To encourage the formation of conditional discrimination between the two languages, we alternated between tact instruction in English and Spanish for each target (i.e., back-to-back, with close temporal proximity; Cariveau et al., 2016). For instance, if a trial began in English, the next trial presented the same target in Spanish. The order of the three targets was block-randomized for each session. Each session consisted of six trials (i.e., three trials prompted in English and three trials prompted in Spanish), with each target in the corresponding language presented only once per session (see Supporting Information C for a sample intervention datasheet).

Modification to Protocols

Elisia and Delicia required modifications to their procedures to meet the mastery criterion due to consistently low levels of independent correct responses for the foreign language equivalents across conditions. A modified error correction procedure was introduced for both participants (Carroll et al., 2015). This procedure provided Elisia and Delicia with an additional opportunity to respond independently following the initial error correction when they answered incorrectly. Specifically, after they correctly echoed the experimenter’s prompted response, the experimenter asked them to identify the same target independently in the corresponding language within the same trial. For Elisia, the modified error correction was implemented in the bilingual condition only during the first comparison, as she had already mastered the targets in the monolingual condition when the modification was implemented. However, the modification was applied to both conditions for all subsequent comparisons. In contrast, for Delicia, the modified error correction was introduced simultaneously for both monolingual and bilingual conditions during the first comparison.

Later on, owing to continuous low levels of independent correct responding and to enhance the differentiation of language responses and promote conditional discrimination, different colored cards were used to represent each language. For example, a rectangular paper-sized red card represented Spanish, while a green card of the same size represented English. Before each session, the experimenter explained to Delicia, “We have red for Spanish and green for English,” and placed both color cards in front of her. During the session, the experimenter positioned the target card on the corresponding red or green card, depending on the expected response, and instructed, “What is this in English/Spanish?” For instance, if the trial required a Spanish response for “grater,” the experimenter said, “We have red for Spanish; what is this in Spanish?” while placing the target card for “grater” on the red card background.

Maintenance Probe

The experimenter evaluated the retention of tacts in both languages for up to 6 months. Maintenance probes were administered at 1, 2, and 4 weeks after the participants met the mastery of each set of tact instruction. For Idris, a 5-month maintenance probe was conducted for all previously mastered sets across both conditions. Additionally, 4-month maintenance probes were conducted for all previously learned sets across conditions for Elisia and Delicia. During these sessions, the procedures were similar to the tact instruction sessions, except that no error correction procedures were implemented after an incorrect response; the experimenter simply moved on to the next trial.