Comparing Manipulations to Enhance Stimulus Salience during Intraverbal Training

Abstract

Children with autism spectrum disorder (ASD) may have difficulty acquiring intraverbal behavior. The present study compared manipulations of stimulus salience (i.e., volume increase, elongation) to teach intraverbals (e.g., “You drink [juice]” and “You drink from [cup]”) to three participants diagnosed with ASD whose pre-treatment responding suggested restricted stimulus control. We used an adapted alternating treatments design to compare the efficacy and efficiency of increased volume, elongated, and unmodified antecedent verbal stimuli on correct intraverbal responses. Results suggested the volume increase condition was the most efficacious and efficient for two participants, whereas all conditions were similarly efficacious and efficient for one participant. High levels of responding maintained as the stimulus salience manipulations were removed.

Intraverbal behavior is a critical component of human social interactions and is commonly taught in educational settings. An intraverbal is a verbal operant under the control of verbal stimuli, and it does not have point-to-point correspondence with the stimulus (Skinner, 1957). Mastery of intraverbals establishes the foundation for advanced repertoires such as verbal conditional discriminations (e.g., tell me a red fruit, tell me a green fruit, tell me a red vegetable; Sundberg & Sundberg, 2011). Typically developing children often engage in verbal conditional discriminations under multiple sources of control between the ages of 3 and 5 years old (Sundberg & Sundberg, 2011). However, verbal conditional discriminations and more complex intraverbal behavior is often not as readily acquired by children with ASD (Sundberg & Sundberg, 2011). Thus, intraverbal training frequently occurs within comprehensive behavioral intervention, and the identification of efficient and efficacious ways to teach complex intraverbals is of crucial importance (Sundberg & Sundberg, 2011).

Nevertheless, previous research shows many learners with ASD engage in responding under faulty or restricted stimulus control (e.g., Kisamore et al., 2016; Sundberg & Sundberg, 2011). Restricted stimulus control occurs when a learner’s verbal behavior comes under the control of only a portion or subset of the antecedents that should come to control a response. For example, when asked, “what do you eat” and “what do you eat with,” a learner may engage in the same response (e.g., “tacos”) in the presence of both antecedent verbal stimuli. Emitting the same response following both antecedent verbal stimuli suggests that the preposition “with” is not sufficiently salient. Therefore, the learner’s behavior is not under multiple sources of control and differential responding is not observed (Axe, 2008). Although previous studies have evaluated methods for reducing restricted stimulus control, such as use of differential observing responses (DOR; e.g., Kisamore et al., 2016) and training with blocked trials (e.g., Ingvarsson et al., 2016), these procedures required an extended duration of training and several procedural modifications to produce mastery of the targeted verbal conditional discriminations for some participants.

A method to potentially prevent restricted stimulus control during verbal behavior instruction is to modify aspects of the antecedent stimulus to increase the salience of stimuli for learners. Summers et al. (1993) compared extra-stimulus prompts (e.g., models, position prompts) and within-stimulus prompts (manipulation of stimulus salience of antecedents) to teach children with developmental disabilities to engage in auditory-visual conditional discriminations that included “in” and “on” with blocks and a container. The procedure consisted of repeating the auditory stimulus (i.e., “in” or “on”) three times loudly. Following correct responding, the volume (i.e., loudness) and number of repetitions of the auditory stimulus were faded successfully across trials. Results suggested the stimulus-salience modifications were more efficacious than extra-stimulus prompts. For the remainder of the present paper, we will refer to these procedures as manipulations of stimulus salience (rather than within stimulus prompts) as they were designed to facilitate acquisition by enhancing stimulus salience instead of prompting a specific response.

Striefel et al. (1978) also investigated potential ways to increase the salience of antecedent stimuli while teaching auditory-visual conditional discriminations in a matrix training evaluation to adolescents with intellectual impairment. One procedure emphasized one word in the antecedent stimulus by presenting that stimulus at a louder volume. Another procedure included elongating part of the antecedent stimulus for 3 s. The third procedure consisted of repeating a portion of the antecedent stimulus twice. Each procedure was successfully faded in intensity following correct responding.

Although manipulations of stimulus salience have successfully reduced or prevented restricted stimulus control during teaching of various skills (Striefel et al., 1978; Summers et al., 1993), we could not find any previous studies that applied these methods to intraverbal training. Furthermore, a previous review on intraverbal relations (e.g., Axe, 2008) and stimulus salience (Halbur et al., 2021) recommended investigating manipulations of stimulus salience in antecedent verbal stimuli to teach intraverbal behavior. Thus, the purpose of the present investigation was to compare the efficacy and efficiency of two manipulations of stimulus salience to teach complex intraverbals to learners with ASD. We also evaluated performance following the removal of any added manipulations to stimulus salience. Said another way, the present investigation had four primary research questions, (a) What are the effects of volume increase versus elongate on intraverbal responding while implementing those procedures?, (b) What are the effects of volume increase versus elongate on subsequent intraverbal responding once those procedures are removed?, (c) How does the more efficacious and efficient procedure (volume increase or elongate) compare to training with neither stimulus salience procedure (i.e., typical training)?, and (d) How does the more efficacious and efficient procedure compare to no training (i.e., control).

Method

Participants

Three six-year-old boys diagnosed with ASD participated in the study. All participants had intervention goals related to teaching complex intraverbals and displayed restricted stimulus control during intraverbal probes. Mohammed was Middle Eastern and communicated using short sentences. Lester was European American and communicated using complete sentences. Teddy was European American and emitted 3- to 6-word mands, tacts, and intraverbals. Teddy and Mohammed’s intraverbal and listener responding by feature, function, and class (LRFFC) scores on the Verbal Behavior Milestones Assessment and Placement Program (VB-MAPP; Sundberg, 2008) were Levels 2/3, and Lester’s intraverbal and LRFFC scores on the VB-MAPP were Level 3.

Setting and Materials

Sessions took place in a quiet area in a clinic (Mohammed and Lester) or in the participant’s home (Teddy). Each area included a table, chairs, data collection materials (e.g., pens, clipboards, timers), preferred items to use as reinforcers (identified via brief multiple stimulus without replacement preference assessments and participant’s vocal mands), and video recording equipment. A Kindle™ was used to play recorded audio files for Mohammed and Lester. Table 1 shows the intraverbal targets and responses for each participant. We used components of a logical analysis to equate targets across conditions (Cariveau et al., 2021). We equated conditions so that the action in each antecedent verbal stimulus (e.g., drink, smell) was one syllable and started with a different letter. Responses were all 1–2 syllables. Targets across conditions were of similar familiarity to the participants. If the target questions were used for more than one participant, they were assigned to different conditions to counterbalance targets when possible.

Table 1.

Targets across participants and conditions

P	Condition
	Control	No modification	Volume increase	Elongate
Mohammed	What do you drink? Water What do you drink from? Cup	What do you smell? Flower What do you smell with? Nose	What do you wash? Hands What do you wash with? Soap	What do you count? Numbers What do you count with? Fingers
Teddy	What do you brush? Teeth What do you brush with? Toothbrush	What do you wash? Hands What do you wash with? Soap	What do you cut? Paper What do you cut with? Scissors	What do you drink? Milk What do you drink from? Bottle
Lester	What do you count? Numbers What do you count with? Fingers	What do you fly? Kite What do you fly with? Wind	What do you smell? Food What do you smell with? Nose	What do you watch? TV What do you watch with? Eyes

P = Participant

Experimental Design

An adapted alternating treatments design (Sindelar et al., 1985) was used to compare participant responding to volume increase and elongate conditions for all participants. A multiple baseline design across behaviors was also used to analyze the effects of volume increase on correct intraverbal responses for Mohammed across three tiers and for Teddy across two tiers. The conditions included two manipulations of stimulus salience (i.e., increased volume and elongate), treatment as usual (i.e., no-salience modification) condition, and a control condition. Sessions were conducted in blocks of four (i.e., one of each condition) until the initial phase change criteria of two consecutive sessions with at least 83% (5/6) was met. Following meeting the phase change criteria in the increased volume or elongate condition, the manipulation was removed (i.e., changed to 0% increased volume or 0% elongate) and intervention continued until responding met the mastery criterion of two consecutive sessions at 83% without the modification present. This was done to evaluate performance in the absence of the modifications. A discontinuation criterion was used to prevent extended exposure to ineffective procedures. The discontinuation criterion was defined as the absence of an increasing trend in correct responses after a minimum of seven sessions of each condition and a minimum of three sessions beyond mastery of another condition. Sessions continued if increasing trends were observed. If a condition was discontinued, then the targets assigned to that condition were taught with the most efficacious condition.

Response Measurement

Our primary dependent variable was the percentage of correct responses and meeting the mastery criterion when the manipulation (i.e., volume increase or elongation) was removed. Correct responses were defined as the participant engaging in the response that corresponded to the antecedent verbal stimulus within 5 s. An error was defined as the participant engaging in any other vocal response that did not correspond to the antecedent verbal stimulus. Errors also included vocal stereotypy and whispers. No responses were defined as the participant not engaging in a vocal response during the response interval, and included bodily noises (e.g., grunts).

A secondary dependent variable was the efficiency of procedures, defined as the number of treatment sessions required to reach the mastery criterion.

Interobserver Agreement and Procedural Integrity

A secondary observer collected data live or via video independently on each dependent variable for a minimum of 32% of sessions for each participant. Trial-by-trial interobserver agreement (IOA) was calculated. Trials were only scored as an agreement if the data matched across all components (i.e., correct responses, prompted responses, and errors). Interobserver agreement was calculated by dividing the number of trials with an agreement by the total number of trials per session, multiplied by 100. Refer to Table 2 for mean IOA for each participant.

Table 2.

Interobserver agreement and procedural integrity for each participant

P	Participant responding			Duration		Procedural integrity
	% w/IOA	IOA	Range	IOA	Range	% w/PI	PI	Range
Mohammed	32	100	–	84	0-100	32	98	83.3–100
Teddy	43	96	83.3–100	100	–	43	99	83–100
Lester	47	91	16.6–100	100	–	47	97	83.3–100

The percentage column shows the mean percentage of sessions with IOA and PI. P = participant, IOA = interobserver agreement, PI = procedural integrity

Procedural integrity was recorded for a minimum of 32% of sessions across participants. Correct integrity was recorded for a trial if the experimenter implemented every component correctly (i.e., presented the correct S^D according to the data sheet, waited the allotted response interval, implemented the correct modifications if applicable, and provided the correct consequences). The volume increase and elongation were considered correct if it was presented as planned by the experimenter (i.e., Teddy) or the correct Kindle™ file was played (i.e., Mohammed and Lester). Trials were scored as a 1 (all components correct) or a 0 (any component was incorrect). The percentage of procedural integrity was calculated by dividing the total number of trials scored as a 1 by the total number of trials per session, multiplied by 100. Refer to Table 2 for mean procedural integrity data for each participant.

Procedures

Each session consisted of six trials with two targets presented in a pseudo-random order three times each. The presentation order of targets was pseudo-random because the same target was not presented more than twice in a row. During each session, the experimenter established ready behavior (i.e., the child sitting at the table oriented toward the instructor), presented the target antecedent verbal stimulus vocally (Teddy) or via an audio file on a Kindle™ (Mohammed and Lester), and provided the allotted response interval. A typical presentation for each of the sessions that had a sound file (Lester and Mohammed) was that the whole file was presented between 70–75 decibels, which is the normal conversational level. Each word was presented for approximately .3 s.

Baseline

The antecedent verbal stimulus was presented without modification (e.g., typical conversational volume and no elongated words) across conditions in baseline. Correct responses resulted in general praise (e.g., “Great!”). No prompts or other feedback were provided for errors or trials with no responses. Mastered tasks (e.g., tacts, intraverbals, and listener responses unrelated to the targets) were presented during the inter-trial interval approximately every two trials, and praise and access to a preferred tangible item were provided for 20 s following correct responses to the interspersed mastered tasks. Least-to-most prompting was used for mastered tasks, as needed.

Treatment Comparison

Treatment sessions in all conditions except control began with two 0-s prompt delay (PD) sessions, during which an immediate vocal model of the correct response was provided. The experimenter allowed 5 s for a correct prompted response to the vocal model. If the participant echoed the vocal model within 5 s, praise and access to a preferred item for 20 s was provided. If the participant did not respond correctly within 5 s of the vocal model, the experimenter moved on to the next trial. Data for the 0-s PD session are omitted from the figures (available upon request from the second author). Thereafter, the experimenter implemented a 5-s PD for the remainder of the treatment sessions. During the 5-s PD, the experimenter waited 5 s for a response following the presentation of the antecedent verbal stimulus. Correct responses resulted in praise and a preferred item for 20 s. An error or no response resulted in the delivery of a vocal model of the correct response and a 5-s response interval for the participant to engage in a correct prompted response. A correct prompted response resulted in the delivery of praise and a preferred item for 20 s. Following mastery, the experimenters conducted sessions without the modifications (i.e., increase volume or elongation) to identify if participants would continue to respond correctly when the manipulation of the antecedent verbal stimulus was removed.

Control

The control condition was identical to baseline. Approximately one probe of the control condition was conducted for every three sessions of each intervention condition, with slight variation across participants.

Treatment as Usual (no Modification, PD)

The antecedent verbal stimulus was not modified, and the PD procedure was implemented in each trial.

Volume Increase

The antecedent verbal stimulus was modified to increase the volume of the preposition (e.g., with, for). The experimenter approximately doubled the volume of her vocalization of the preposition for Teddy, and all other words in the antecedent verbal stimulus were stated at conversational level (i.e., as done in baseline). To help with implementation of the increased volume, only one person (a graduate student) ran Teddy’s treatment sessions across all conditions, and the experimenter received training and ongoing monitoring to produce the vocalizations at the correct volume, speed, and consistency.

For Mohammed and Lester, the experimenter played an audio file on a Kindle™ that had a pre-recorded antecedent verbal stimulus that included the same volume manipulation for the preposition. In the recordings, the preposition was emphasized between 95–100 decibels (i.e., considered to be as loud as a hairdryer or hand drill), whereas the rest of the antecedent verbal stimulus was recorded between 70–75 decibels (i.e., volume that falls within the normal conversation range). All other procedures were the same as the no-modification condition.

Elongate

The antecedent verbal stimulus was modified to elongate the preposition. The experimenter vocally elongated the preposition by approximately 1.5 s (e.g., “wwwiiiiiiith”; Teddy) or played an audio file on a Kindle™ that had a pre-recorded antecedent verbal stimulus that included the same elongated preposition for 1.5 s (Mohammed and Lester) presented at a conversational volume. All other words in the antecedent verbal stimulus were stated at conversational level and were not elongated (i.e., as done in baseline). Just like with the increased volume condition, to further help with implementation of the vocal elongate (i.e., Teddy), only one person ran Teddy’s treatment sessions and received training and ongoing monitoring to produce the vocalizations at the right volume, speed, and consistency.

Results

Table 3 shows sessions to mastery across conditions for each participant. Figures 1 through 3 show the individual results for each participant. Each tier represents a different condition, and sessions are alternated across the tiers as they were presented in the randomized order. Low to moderate levels of correct responding were observed in baselines across participants. Figure 1 depicts Mohammed’s results. During baseline in all conditions, Mohammed emitted responses associated with the action (e.g., wash; he said, “hands”) rather than the action plus preposition (e.g., wash with; he also said, “hands”). During treatment, Mohammed’s responding met the phase change criterion in the increased volume condition (top tier) in four treatment sessions (including the 2 0-s sessions not shown in the figure), and mastery in 11 sessions (i.e., modifications were successfully removed from 100% to 0%). In the elongate condition (bottom tier), Mohammed met the phase change criterion in 11 treatment sessions, and mastery in 19 sessions. The increased volume condition was added to the treatment as usual and control conditions (second and third tiers, respectively) resulting in mastery level performance.

Table 3.

Sessions to mastery across conditions for each participant

P	Increased volume	Elongate	Treatment as usual	Increased volume (subsequent implementation)
Mohammed	11	19	N/A	7, 9
Teddy	7	9	11	8
Lester	6	6	6, 4	N/A

0-s PD sessions are included in these totals for each condition

Fig. 1.

Mohammed’s percentage of correct responses across conditions in an adapted alternating treatments design separated into different panels. Note. BL = Baseline; VI = volume increase; PD = prompt delay

Fig. 3.

Lester’s percentage of correct responses across conditions in an adapted alternating treatments design separated into different panels. Note. BL = Baseline; VI = volume increase; PD = prompt delay

Figure 2 depicts Teddy’s results. During baseline in all conditions, Teddy emitted responses associated with the action rather than the action plus preposition. During treatment, Teddy’s responding met phase change criterion in the increased volume condition (second tier) in four treatment sessions, followed by mastery in seven sessions. His responding met the phase change criterion in the elongate condition (bottom tier) in seven treatment sessions, and mastery in nine sessions. In the treatment as usual condition (second tier), he met mastery in 13 sessions. The increased volume condition was added to the control condition (third tier) resulting in mastery level performance (i.e., with 0% modification) in five treatment sessions.

Fig. 2.

Teddy percentage of correct responses across conditions in an adapted alternating treatments design separated into different panels. Note. BL = Baseline; VI = volume increase; PD = prompt delay

Figure 3 depicts Lester’s results. During baseline in all conditions, Lester emitted responses associated with the action rather than the action plus preposition. During treatment, Lester’s responding met mastery in all conditions in six training sessions. Due to similar efficacy of conditions, the treatment as usual condition was added to the control condition (i.e., being the least intrusive procedure), which produced mastery level performance in two training sessions.

Discussion

The present study is the first of our knowledge to investigate the efficacy and efficiency of manipulations of stimulus salience of antecedent verbal stimuli to teach intraverbal behavior to children with ASD. Results suggested the volume increase condition was the most efficient procedure for two of three participants (Mohammed and Teddy), and all conditions were similarly efficient for one participant (Lester). For Mohammed and Teddy, both stimulus salience manipulations (increased volume and elongate) led to increased responding when compared to control and treatment as usual. The results replicate previous research on the efficacy of manipulations of stimulus salience for listener training (e.g., Striefel et al., 1978; Summers et al., 1993) and suggest manipulations can be used to enhance the salience of antecedent verbal stimuli during intraverbal training for children with ASD.

For all participants, the preposition (i.e., with, from) was the portion of the antecedent verbal stimulus that was modified. This modification matched the error pattern displayed by participants in baseline; participants typically emitted the response associated with the action (e.g., brush; they said, “teeth”) rather than the action and preposition (e.g., brush with; they also said, “teeth). Therefore, their baseline responding suggested restricted control by the action in the antecedent verbal stimulus. However, if some other part of the antecedent verbal stimulus controls responding (e.g., the action plus preposition), future research could evaluate making stimulus salience modifications to different portions of the antecedent verbal stimulus, and these modifications should be participant-specific and based on error patterns in baseline. In addition, future research should take additional steps to consider the conditional stimulus control of the intraverbal. For example, the stimulus sets should be arranged such that all relevant components overlap and vary within a set (e.g., nouns and prepositions presented in a variety of combinations) (e.g., Aguirre et al., 2019).

Consistent with previous research (e.g., Striefel et al., 1978; Summers et al., 1993), the manipulations were able to be removed for all participants in the present investigation. As such, we did not need to implement gradual fading (Schreibman, 1975). Nevertheless, the removal of the manipulation or successfully fading manipulations to antecedent verbal stimuli remains an area for future investigation, as this may be necessary for some learners.

Stimulus assignment into sets was considered in the present investigation (Cariveau et al., 2021), such that baseline levels were relatively similar across conditions for each participant. Despite this consideration, a potential limitation is that our stimulus sets for each condition were relatively small (i.e., two targets per condition each presented three times). These set sizes were identified based on available intraverbal targets under convergent control for which error patterns were observed during probes conducted during clinical service delivery. Future research may choose to consider larger set sizes, replicate methods with additional participants, and consider the use of alternative experimental designs, as two targets per set may not be sufficient (Sidman, 1987).

In the present investigation, we exposed any discontinued conditions to the more efficacious procedure across participants. For example, increased volume was added to control and treatment as usual conditions for Mohammed, and increased volume was added to the control condition for Teddy. In each of these cases, once the modification was added, mastery-level performance was attained. However, the study was limited in that we did not have multiple sets assigned to each condition across participants, which would have permitted within-subject replication of outcomes. Due to the limited within-subject replication, it is possible that extraneous or motivational variables may have impacted the acquisition of target sets. Additionally, it is possible that some participants may have had idiosyncratic results across replications or may acquire targets across sets more efficiently. However, we observed the most consistent replication with the volume increase condition (i.e., replicated the effects of VI on correct intraverbal responses both within and across participants). Further replication of the procedures remains an important area of investigation.

Future research should also consider the efficacy of other procedures that modify the putative salience of antecedent verbal stimuli. For example, repeating a word to increase the salience of portions of the antecedent verbal stimulus (i.e., double word; Striefel et al., 1978) was not evaluated within the present investigation. Future research could also compare the efficacy and efficiency of manipulations of stimulus salience on antecedent verbal stimuli (i.e., increased volume, elongate, double word) to other procedures that have been used to remediate restricted stimulus control such as specific orienting cues (e.g., Koegel et al., 1981), DOR (e.g., Kisamore et al., 2013), and blocked trials (Ingvarsson et al., 2016) for intraverbal behavior. It is possible that certain manipulations may be more socially valid and feasible to implement during practice; thus, research could also evaluate the social validity of methods to increase the salience of stimuli.

Data Availability

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethical Approval

All procedures were performed in accordance with the ethical standards of the institutional review committee and with the 1964 Helsinki declaration and its later amendments.

Conflict of Interest

The authors of this manuscript declare no conflict of interest regarding this manuscript.