Abstract
In this study, we extended the literature on the generalization of negatively-reinforced mands in three young children with autism spectrum disorder (ASD). First, we used example and nonexample stimuli embedded in mand training to teach a new, socially appropriate, negatively-reinforced mand to reject unpreferred food items while continuously assessing mand discrimination. Second, we evaluated the discriminated generalization of the newly acquired mand by using untrained example and nonexample stimuli. Finally, we conducted maintenance probes to examine if the new, discriminated mand occurred over time in the absence of training. Results suggest that our mand training produced acquisition of a discriminated negatively-reinforced mand in all three children. Data indicate that the newly acquired, discriminated mand generalized to untrained food items and was maintained after training was discontinued. We discuss the conceptual significance and clinical implications of using example and nonexample stimuli to produce acquisition, generalization, and maintenance of negatively-reinforced mands in young children with ASD and language delays.
Young children with autism spectrum disorder (ASD) may be exposed to a variety of aversive stimuli during typical routines in their natural environment. Examples of aversive stimuli include, but are not limited to, noisy or crowded environments, social interactions, academic tasks, and unpreferred items or activities. An aversive stimulus is a stimulus for which termination functions as a reinforcer (Skinner, 1953). Children with ASD who have communication deficits may not emit socially appropriate and effective communicative responses to escape or avoid aversive stimuli and, thus, may resort to nonvocal behaviors (e.g., pushing away, dropping, covering eyes) or to problem behavior (e.g., elopement, screaming, aggression; Drasgow & Halle, 1995; Groskreutz, Groskreutz, Bloom, & Slocum, 2014; Yi, Christian, Vittimberga, & Lowenkron, 2006). These communication deficits may limit a child’s participation in his or her natural environment by increasing the likelihood of social isolation and by restricting friendships (Knott, Dunlop, & Mackay, 2006).
Teaching children with ASD and communication deficits to use socially appropriate and effective negatively-reinforced mands to escape or avoid aversive stimuli may empower them to navigate complex environments by increasing the probability of successful communication and by preventing problem behavior (Choi, Reilly, Sigafoos, & Lancioni, 2010; Falcomata, Wacker, Ringdahl, Vinquist, & Dutt, 2013; Fritz, Iwata, Hammond, & Bloom, 2013; Kelley, Shillingsburg, Castro, Addison, & LaRue, 2007; Kreibich, Chen, & Reichle, 2015; Reeve & Carr, 2000; Sigafoos et al., 2004). A mand is “a verbal operant in which the response is reinforced by its characteristic consequence and is under the functional control of relevant conditions of deprivation or aversive stimulation” (Skinner, 1957, pp. 35-36). Mands that occur under a state of aversive stimulation or satiation are referred to as negatively-reinforced mands (Drasgow, Sigafoos, Halle, & Martin, 2009). For example, a child who just had several drinks (i.e., aversive stimulation or liquid satiation) may sign “no, thank you” (i.e., a negatively-reinforced mand) when offered a drink by a social partner.
A unique characteristic of negatively-reinforced mands is that some variables other than the state of aversive stimulation or satiation may exert control over their occurrence (Winokur, 1976). In the example presented previously, the drink offered by a social partner functioned both as a visual discriminative stimulus and as an aversive stimulus that exerted control over the occurrence of the negatively-reinforced mand “no, thank you.” Thus, a negatively-reinforced mand may occur for different reasons. One reason is the presence of the discriminative stimulus that evokes the mand due to the history of reinforcement in its presence. Another reason is the presence of a relevant motivating operation (i.e., an internal state of aversive stimulation or satiation) that decreases the reinforcing value of the consequence and, consequently, increases the probability of the mand that resulted in its termination. Nearly all negatively-reinforced mands are controlled by a negative reinforcer that also functions as a discriminative stimulus (Winokur, 1976).
One strategy that has been used to teach socially appropriate and effective negatively-reinforced mands to children with ASD is mand training (Sundberg & Partington, 1998). Mand training typically involves the (a) identification and assessment of the child’s existing mand repertoire and the environmental variables that serve as aversive stimuli; (b) selection of a novel, socially appropriate, and effective mand that serves the same function or provides access to the same reinforcer as the existing, inappropriate mands; and (c) differential reinforcement to teach the novel mand. Differential reinforcement is a strategy that involves withholding access to the reinforcer (i.e., extinction) following the emission of an existing, inappropriate mand or problem behavior, while reinforcing the novel, socially appropriate mand (Sigafoos, Arthur-Kelly, & Butterfield, 2006).
One of the most significant components of mand training is the selection of the novel mand to be taught during training. Several researchers have emphasized the importance of teaching specific mands to increase the probability that the response emitted by a child will contact the specific reinforcer relevant to the motivating operation (i.e., state of satiation or deprivation) in effect at that time and enhance the clarity of communication (Sundberg, 2001; Sundberg & Partington, 1998). An example of a specific negatively-reinforced mand would be “stop the music” emitted by a child in the presence of loud music played on the radio.
Although selecting and teaching specific mands is an effective approach of mand training for most children, it may not be very effective for nonvocal children with severe communication deficits and slow acquisition histories. For these children, a more feasible alternative may be to teach a generalized mand (e.g., “no, thank you,” “stop”). A “generalized mand” is defined as a mand that allows a child to access a variety of reinforcers or avoid and escape multiple aversive stimuli encountered in his or her environment without being taught first a discrimination among specific reinforcers (Keogh & Reichle, 1985).
Teaching a generalized mand for these children may be a more effective approach than teaching a specific mand for several reasons. First, a generalized mand such as “no” allows a child to reject a multitude of unpreferred items or activities presented by social partners in the natural environment. Second, a generalized mand is more functional than a specific mand because numerous opportunities for emitting the mand occur very often during typical activities in the natural environment (Sigafoos et al., 2004). Finally, previous research on mand acquisition for children with ASD and communication delays suggests that these children require extensive training to acquire multiple mands (Chezan, Drasgow, Martin, & Halle, 2016; Drasgow, Martin, Chezan, Wolfe, & Halle, 2016). Consequently, teaching specific mands may not be effective in those situations in which the child needs to acquire a socially appropriate mand to prevent the likely acquisition of problem behavior and to function effectively in his or her environment (Reeve & Carr, 2000).
One critical component of mand training that has not been extensively researched in the literature is facilitating the acquisition of discriminated mands (Drasgow, Halle, Ostrosky, & Harbers, 1996; Martin, Drasgow, Halle, & Brucker, 2005; Reichle, Byiers, & Reeve, 2018; Reichle & McComas, 2004). A small body of literature provides empirical evidence on the effectiveness of mand training in teaching discriminated, positively-reinforced mands to young children with ASD and language delays (e.g., Drasgow, Halle, & Phillips, 2001; Gutierrez et al., 2007; Gutierrez, Vollmer, & Samaha, 2010). A discriminated mand is a mand that occurs in the presence of a motivating operation (e.g., satiation) and does not occur in the absence of the motivating operation. Findings of these studies have highlighted several aspects that facilitate the acquisition of discriminated mands. One aspect relates to the conditions under which a mand occurs. Specifically, a discriminated mand should occur under certain stimulus conditions and not under other stimulus conditions. For example, a nonvocal child who engages in physical aggression when asked to complete work is taught to sign “break” to escape or avoid a task demand presented by the teacher. Signing “break” is a discriminated mand if the child emits the sign when the teacher presents the task demand (i.e., an aversive stimulus) and refrains from using the sign when there is no task demand (i.e., no aversive stimulus present).
Another aspect is the emission of a mand that the verbal community reinforces with a stimulus that corresponds to the relevant state of aversive stimulation or satiation. For example, the sign “no, thank you” emitted by a nonvocal child in the presence of an unpreferred item may or may not be reinforced (i.e., removal of the aversive stimulus or unpreferred item if the social partner does not understand sign language). However, if the child uses a picture symbol or the vocal response “no, thank you” in the presence of the same adult, the probability that the adult will reinforce the response by removing the aversive stimulus is greatly enhanced. In other words, the verbal community may gain a discriminative stimulus function and exert some control over the response that is likely to be reinforced under specific conditions (Drasgow et al., 2001; Winokur, 1976).
The acquisition of discriminated mands is critical for promoting generalization while avoiding generalization errors (Horner, Bellamy, & Colvin, 1984). A generalization error is the emission of a newly acquired mand in an inappropriate situation in the presence of untrained stimuli. If a child first signs “no, thank you” when offered a cookie, an untrained preferred food item, and then immediately grabs the cookie and consumes it, a generalization error has occurred. One strategy that has the potential to promote generalization while reducing the likelihood of generalization errors consists of using examples and nonexamples when teaching negatively-reinforced mands. Examples consist of stimuli that are likely to evoke a specific mand in the presence of an aversive stimulus or a state of satiation. For example, a child who just finished playing on an iPad for 2 hr is more likely to say “no, thank you” when the teacher offers him the iPad during a break than when he has not played on the iPad for the last 2 hr. Nonexamples consist of stimuli that are less likely to evoke a specific mand when an aversive stimulus or a state of satiation is not in effect. For example, a child is less likely to sign “no, thank you” when he has not had a snack for the last 3 hr and he is offered a cookie (a preferred food item) by the teacher.
A commonly used generalization-promoting procedure consists of using sufficient examples during training to promote mand acquisition and evaluating generalization in the presence of untrained stimuli (Falcomata et al., 2013; O’Neill & Sweetland-Baker, 2001; Stokes & Baer, 1977; Yi et al., 2006). Assessing generalization of mands only in the presence of untrained stimuli does not provide any evidence of whether generalization errors occur and, thus, has the potential to inflate data and lead to inaccurate conclusions. For example, a child who uses a card to indicate “not now” when presented with a puzzle (untrained stimulus) should not use the card when no puzzle is presented. If the emission of the mand is not assessed in those situations when the puzzle is not offered and other conditions remain the same, the researcher or the therapist does not have data to document the presence or absence of generalization errors.
The use of nonexamples to promote generalization of negatively-reinforced mands has received very limited attention in the literature on mand training and generalization (Chezan et al., 2016; Groskreutz et al., 2014). The findings of the previous literature on negatively-reinforced mands are important for several reasons. One reason is the emerging empirical evidence demonstrating that mand training promotes the acquisition of discriminated, negatively-reinforced mands in young children with ASD. Furthermore, the results of previous studies reveal that young children with ASD can learn not only multiple discriminated mands, but they are able to alternate mands when social partners do not respond immediately to their first communicative attempt (Chezan et al., 2016). Another reason relates to the importance of using multiple example and nonexample stimuli to promote generalization without errors of the newly acquired mand to novel social partners and untrained environments (Groskreutz et al., 2014).
Although the findings of the studies listed previously contribute to our understanding of discriminated mands generalization, additional research is warranted. Thus, in this study, we endeavored to extend the existing literature on discriminated, negatively-reinforced mands by examining several aspects that promote acquisition, discrimination, and generalization of negatively-reinforced mands. First, we examined the effects of mand training consisting of example and nonexample stimuli on promoting the acquisition of discriminated, negatively-reinforced mands in young children with ASD. In prior studies, researchers assessed the acquisition of a discriminated mand within the context of probes interspersed throughout the baseline and mand training rather than continuously examining its emission during each session (Groskreutz et al., 2014). Using nonexample stimuli during each session represents not only an opportunity to assess mand discrimination, but it may also provide additional opportunities for teaching the mand. For example, a food item identified as preferred at the beginning of a study may change its reinforcing value over time and become an aversive stimulus, a situation that requires the emission of a negatively-reinforced mand and constitutes a teaching opportunity. Therefore, embedding both example stimuli and nonexample stimuli during training provides multiple learning opportunities for mand acquisition, which are extremely important for young children with ASD and limited communication who may require extensive training ranging from hundreds to thousands of trials to acquire a tact or a mand (Carr, Binkoff, Kologinsky, & Eddy, 1978).
Second, we assessed whether teaching the new mand using multiple example and nonexample stimuli produces discriminated generalization of the newly acquired mand to untrained food items. Groskreutz et al. (2014) used a generalization-promoting strategy (i.e., multiple exemplars) implemented sequentially across four aversive stimuli until generalization occurred to untrained stimuli and in novel settings. Although training implemented sequentially across stimuli is an effective procedure in promoting mand acquisition and generalization, teaching the mand concurrently across multiple stimuli may offer additional details related to the effectiveness of using examples and nonexamples in promoting mand discrimination and generalization. Finally, we evaluated the discriminated maintenance of the newly acquired negatively-reinforced mand in the presence of both untrained examples and untrained nonexample. Maintenance of discriminated, negatively-reinforced mands has not been evaluated in previous studies (Chezan et al., 2016; Groskreutz et al., 2014). Promoting maintenance is a critical aspect of any effective communication intervention program intended to promote mand acquisition that lasts over time in the absence of training.
Our research questions were as follows:
Method
Participants
Four young children with a diagnosis of ASD participated in the study, but only three children completed the study. We report background information and data only for the three children who completed the study. Children were enrolled in the study if they (a) had a formal diagnosis of ASD; (b) were between 2 and 6 years old; (c) had limited or no functional speech; (d) used nonvocal behaviors, such as gestures or body movements, to reject unpreferred food items; and (e) had parental consent. Teachers, therapists, or parents nominated the children for participation in the study because they did not emit socially acceptable negatively-reinforced mands when offered an unpreferred food item by a social partner.
Adrian was a 3-year 8-month-old boy diagnosed with ASD at the age of 23 months based on an evaluation conducted by a local physician. At the time of the study, Adrian was attending a half-day preschool program for children with ASD and other developmental disabilities at a public school. He was receiving 11 hr of applied behavior analysis (ABA) services at home. Adrian was ambulatory, but he needed assistance with toileting, dressing, and using utensils when eating. His communication objectives during the ABA therapy consisted of vocally imitating sounds (e.g., “m,” “ba,” “ah,” “uh,” and “pa”) and using a Picture Exchange Communication System (PECS; Bondy & Frost, 2001). He had approximately 25 pictures in his PECS repertoire representing foods, toys, and electronics. Adrian was able to scan multiple pictures contained in his PECS binder, select a picture of a preferred item, create a sentence using the selected picture (e.g., “I want Goldfish”), and point to the sentence strip while looking at a social partner. However, he used PECS inconsistently throughout the day and refrained to gestures to access preferred items and activities. Adrian did not have any pictures in his PECS binder related to a negatively-reinforced manding function and used gestures (e.g., pushing away, dropping to floor) to escape or avoid unpreferred items or activities.
Julian was a 3-year 4-month-old boy who received a diagnosis of ASD at the age of 29 months based on an evaluation conducted by a local physician. Julian was attending a half-day preschool program for children with ASD at a local public school. He was receiving ABA services for approximately 6 hr per week at a private center-based program for children with ASD. Julian was ambulatory and needed minimal assistance with dressing and eating skills. He had emerging toileting skills and was receiving toileting training as part of his ABA therapy. Julian emitted approximately five mands (e.g., “hey,” “look,” “car”); however, his speech was very often unintelligible unless the social partner was familiar with Julian. He did not use sign language, PECS, or other communication aids to access items and activities in his environment. His communication objectives addressed during the ABA therapy consisted of establishing a mand repertoire by using PECS and improving his listener repertoire. Julian used gestures, such as running away, pushing away, and covering his eyes when offered an unpreferred item or activity by a social partner.
Justin was a 5-year 3-month-old boy diagnosed with ASD at the age of 19 months based on an evaluation conducted by a physician. At the time of the study, Justin was attending a program for children with ASD and other developmental disabilities at a local public school. He was not currently receiving ABA services. Justin was ambulatory and needed minimal assistance with dressing, toileting, and using utensils while eating. He was nonvocal and had a limited sign language repertoire consisting of approximately three signs (i.e., “more,” “candy,” “juice”); however, he used these signs only when prompted by parents. Justin had good motor imitation skills and was able to follow one-step directions. He used gestures, including reaching and leading, to access preferred items and activities. He also used gestures, such as turning away, running away, and pushing away, when presented with an unpreferred item or activity by a social partner.
Settings
For Adrian, all assessment and training sessions were conducted at a small table in the living room (5 × 5 m) located in his house. The room contained a dining table with chairs, two couches, a plastic box with toys, and a small table with one chair. The room was the location in which Adrian received one-on-one ABA therapy 4 days per week. All sessions occurred in the afternoon at the beginning of his scheduled ABA therapy session. For Julian, all assessment and training sessions occurred at a table in a classroom (8 × 8 m) located in the private center for children with ASD. The classroom contained three large tables with chairs and three shelves with books and toys stored in plastic containers and represented the setting in which typical activities occurred throughout the day. All sessions were conducted in the morning before his scheduled snack. Three or four peers were present in the room during most of the sessions. For Justin, all assessment and training sessions were conducted at a table in the dining room (5 × 5 m) located in his house. The room contained a dining table with chairs and was the location in which Justin served his snacks throughout the day. All sessions were conducted in the afternoon before his scheduled snack. His younger sibling and one of his parents were present in the room during each session. For all children, each assessment session lasted approximately 15 min. For Adrian and Julian, each training session lasted approximately 10 min, whereas the training session for Justin lasted approximately 15 min. The trainer implemented all assessment and training sessions in a one-on-one format. We conducted one session per day, 3 to 4 days per week.
Target Behaviors and Recording System
We recorded three target behaviors: (a) existing positively-reinforced mands (i.e., communicative responses used to access preferred food items), (b) existing negatively-reinforced mands, and (c) new, alternative, negatively-reinforced mands. Existing positively-reinforced mands were reaching for Adrian, reaching and babbling for Julian, and reaching for Justin. Existing negatively-reinforced mands were pushing away, dropping, giving back, and squeezing for Adrian; dropping, whining, covering eyes, running away, and turning away for Julian; and giving back and pushing away for Justin. Table 1 presents each mand topography and its corresponding definition.
Mand Topographies and Definitions for Existing Mands.
We selected the new, alternative, negatively-reinforced mands through consultation with each child’s therapist, teacher, or parents. For Adrian, the target mand was defined as picking up a rejection card and extending his arm with the card facing toward the trainer. The rejection card consisted of a laminated square card (3 in. × 3 in.) marked with an X in the middle. For Julian, the target mand was defined as picking up a rejection card and extending his arm with the card facing toward the trainer paired with vocally stating “no, thank you” within 3 s of being offered a food item by the trainer. However, the vocal topography was not a required part of the mand because of Julian’s unintelligible speech and the difficulties associated with prompting a vocal mand alone. For Justin, the target mand was signing “no” defined as extending his index and middle finger and tapping his thumb with one or both fingers.
We used a response-per-opportunity recording method to collect data on the three target behaviors. An opportunity consisted of a trial that began with the trainer presenting a preferred or an unpreferred food item that required a response from the child. For each trial, we recorded the type of the trial (i.e., preferred or nonexample; and unpreferred or example), the food item presented to the child, the child’s target behaviors, including existing positively- and negatively-reinforced mands, and the prompted or independent new, negatively-reinforced mand, and whether the child consumed the item requested.
To determine the acquisition of discriminated mands, we further classified these target behaviors as correct or incorrect based on the correspondence between the child’s communicative response and his subsequent behavior. We recorded a correct response if the child emitted an existing positively-reinforced mand to request a preferred or unpreferred food item and then immediately consumed it. We also recorded a correct response if the child emitted a negatively-reinforced mand to avoid a preferred or unpreferred food item offered by a social partner and did not emit additional mands (i.e., existing positively-reinforced mands) to request the item and then consume it. During baseline, we recorded the occurrences of existing negatively-reinforced mands to demonstrate that motivating operations were in effect. During mand training, generalization, and maintenance conditions, the new, negatively-reinforced mand was considered correct and we continued to record the occurrences of existing negatively-reinforced mands to demonstrate that differential reinforcement resulted in the acquisition of the new, alternative mand while decreasing the level of existing negatively-reinforced mands.
We recorded an incorrect response when the child emitted (a) an existing positively-reinforced mand to request a preferred or unpreferred food item that he did not consume, (b) an existing or a new negatively-reinforced mand to avoid an item offered by a social partner followed by an existing positively-reinforced mand to request the item and then consume it, or (c) a new, negatively-reinforced mand simultaneously with or immediately preceding or following an existing negatively-reinforced mand (e.g., use the rejecting card with one hand while pushing the item away with the other hand, giving back and then using the rejecting card, or signing “no” and then pushing away).
Experimental Design and Trainers
We used a partially nonconcurrent multiple-baseline design across participants to evaluate the effects of the mand training in producing acquisition, generalization, and maintenance of discriminated, negatively-reinforced mands (Watson & Workman, 1981). The partially nonconcurrent multiple-baseline design consisted of the concurrent implementation of baseline sessions for the first two children and the nonconcurrent baseline implementation for the third child. We used this design because it allowed us to assess the effects of mand training on discriminated mands while demonstrating a functional relation between these two variables under those circumstances in which one child withdrew from the study and was replaced by another child who started in baseline. In a nonconcurrent multiple-baseline design, the replication of effects of an intervention on a target behavior at different points in time allows a researcher to control for common threats to internal validity, such as maturation, history, and instrumentation and, thus, demonstrate experimental control (Campbell & Stanley, 1963; Harvey, May, & Kennedy, 2004). The multiple-baseline design consisted of the same sequence of baseline, mand training, generalization, and maintenance conditions across the three children.
One trainer implemented the sessions for Adrian. The trainer was a doctoral student in the Department of Special Education. She was also a Registered Behavioral Technician™ (RBT™) who has been providing Adrian ABA services 4 days per week in the home setting for approximately 1 year prior to the beginning of the study. Two trainers implemented the sessions for Julian. The first trainer was a RBT™ who has been working with Julian for approximately 9 months prior to his enrollment in the study. She had a BS in psychology and was working toward her master’s degree in ABA. The second trainer was a certified early childhood special education teacher who was enrolled in an ABA Certificate Program and had more than 16 years of experience working with children with disabilities. She has been working with Julian for approximately 8 months prior to the beginning of the study. Both trainers were full-time employees at the private center-based program attended by Julian and were working with him on a daily basis during typical activities. Two trainers conducted the sessions for Justin. The first trainer was the same trainer who implemented the sessions for Julian. The second trainer was the first author who was a certified behavior analyst at the doctoral level with more than 14 years of experience working with individuals with disabilities in school, residential, and home settings. Both trainers worked with Justin only for the duration of the study and they did not provide services outside the context of the study.
Procedures
Identification of food preference and existing mands
We first conducted interviews with therapists, teachers, and parents to identify preferred and unpreferred food items and the topographies corresponding to positive and negative reinforcement manding functions for each of the three children. We then observed each child in his natural environment while being offered preferred and unpreferred food items to validate the information obtained from therapists, teachers, and parents. Finally, we used the information gathered through interviews and direct observation to conduct a structured assessment.
Structured assessment
We used a structured assessment (Halle & Meadan, 2007) to systematically assess the hypothesized function of existing mands. Structured assessment sessions consisted of a predetermined number of example and nonexample stimuli based on the number of preferred and unpreferred food items identified for each child. We conducted four assessment sessions over 4 days. Each assessment session contained half of the food items identified for a child and each food item was presented twice throughout the four assessment sessions. For Adrian, we identified 12 food items (i.e., six preferred and six unpreferred). Each assessment session consisted of six trials for a total of 24 trials. For Julian, we identified 20 food items (i.e., 10 preferred and 10 unpreferred). Each assessment session consisted of 10 trials for a total of 40 trials. For Justin, we identified 22 food items (i.e., 10 preferred and 12 unpreferred). Each assessment session consisted of 11 trials for a total of 44 trials. The full list of preferred and unpreferred food items included in the structured assessment is available from the corresponding author upon request. Example and nonexample stimuli were randomly presented throughout the session with no more than two consecutive same-type stimuli. Prior to the implementation of the structured assessment, the trainers (i.e., therapists and teacher) received training consisting of written instructions, explanation, and demonstration on how to implement the procedures.
Assessment sessions consisted of the trainer first gathering materials (e.g., plates, spoons, cups, and food items) and placing them in the proximity of where the child would sit, but out of his sight and reach. Then, the trainer brought the child to the table and had him sit. For nonexample (i.e., preferred) stimuli, the trainer presented a small portion of a food item by placing it on a plate (solid item) or in a cup (liquid items) on the table in front of the child, but out of his reach. Next, the trainer waited 3 to 5 s for the child to respond. If the child emitted a mand to request the food item (i.e., positive reinforcement function), the trainer allowed the child to consume the item contingent on his response. After the child consumed the item, the trainer presented the next stimulus. If the child emitted a mand to reject the food item (i.e., negative reinforcement function), the trainer immediately removed the item contingent on the child’s response and waited 3 to 5 s before presenting the next stimulus. If the child did not respond within 3 to 5 s, the adult moved the item closer to the child and allowed him an additional 5 s to respond. If the child did not respond within 5 s, the trainer removed the item. After 3 to 5 s, the trainer presented the next stimulus with a different food item. For example (i.e., unpreferred) stimuli, the procedures were the same with one exception. The trainer offered the food item by holding it within 10 inches in front of the child’s chest.
Baseline
Baseline procedures were the same as the structured assessment procedures except the number of stimuli and the presence of the rejecting card for Adrian and Julian. Baseline sessions consisted of five nonexample stimuli and five example stimuli for a total of 10 trials. We used six of the 10 stimuli (i.e., three example and three nonexample) for mand training and four of the 10 stimuli (i.e., two example and two nonexample) for generalization probes. We selected five preferred and five unpreferred food items by considering two factors. First, based on the structured assessment data, we selected the food items that consistently evoked a mand with a clear function pertaining to positive and negative reinforcement. Specifically, we selected those items that the child either consumed or rejected during each trial when the item was offered. Second, we asked parents, teachers, or therapists to rank the perceived reinforcing value of the preferred and unpreferred items evaluated during the structured assessment based on the child’s behavior in the presence of those items outside the context of the study. We included nonexample stimuli for two reasons: to reduce the overall aversiveness of the session and to establish a control condition that allowed us to assess the discriminated use of the newly taught negatively-reinforced mands. For Adrian and Julian, the rejection card was available during baseline sessions.
Mand training
Mand training sessions consisted of three nonexample stimuli and three example stimuli for a total of six trials. For Justin, we increased the number of trials to 18 after 15 training sessions (Session 38). We made this change because his summer schedule was unpredictable (i.e., he was not available to attend three to four training sessions every week) and his parents notified us that they would be relocating to a different area soon. We used prompts, constant time delay, differential reinforcement, and error correction procedures to teach the new mands. The trainer reinforced all existing positively-reinforced mands and all prompted and independent new, negatively-reinforced mands (i.e., rejection card for Adrian and Julian; signing “no” for Justin) and placed on extinction all existing negatively-reinforced mands (e.g., pushing away, covering eyes, giving back). Similar to baseline, each trial began with the trainer gathering materials, bringing the child to the table, and having him sit.
The protocol for conducting a nonexample (i.e., preferred item) trial was the same as that employed during baseline sessions except the number of trials. For example, for unpreferred-item trials, during the first three sessions, the trainer used a 0-s delay to prompt the new mand and immediately removed the unpreferred food item contingent on the prompted mand. We used full-physical prompts for Adrian, full physical and an echoic prompt for Julian, and full physical and a model prompt for Justin. Beginning with the fourth session, the delay was increased to 3 to 5 s to provide the child an opportunity to emit the mand independently. If the child emitted a mand independently, the trainer removed the food item contingent on the child’s response. If the child engaged in an existing negatively-reinforced mand, the trainer immediately interrupted the response and prompted the child to ensure the correct response. If the child emitted a positively-reinforced mand, the trainer allowed the child to access the item and recorded whether the child consumed the item. The acquisition intervention criterion was 100% correct negatively-reinforced mands for three consecutive training sessions.
Discriminated generalization
After a child met the acquisition intervention criterion, we conducted generalization probes. The purpose of generalization probes was to determine if children used the newly acquired, discriminated negatively-reinforced mand when offered untrained food items. The protocol for conducting a generalization probe was the same as the protocol implemented during baseline sessions except the number of trials and the food items. A generalization probe consisted of two nonexample stimuli and two example stimuli for a total of four trials implemented in a random sequence. We used four of the food items included in the baseline sessions, but not used during training. For Julian and Justin, we conducted only one generalization probe. For Adrian, we decided to conduct a second generalization probe with four different food items based on his performance on the first generalization probe.
Discriminated maintenance
We conducted two maintenance probes for Adrian (i.e., three and six weeks posttraining; Sessions 29 and 31) and Julian (i.e., three and four weeks posttraining; Sessions 31 and 33) and one maintenance probe for Justin (i.e., one-week posttraining; Session 46). The purpose of the maintenance probes was to determine if children continued to emit the newly acquired, discriminated negatively-reinforced mand after the training was discontinued. The protocol for conducting a maintenance probe was the same as the protocol implemented during baseline sessions. Each maintenance probe consisted of five nonexample stimuli and five example stimuli for a total of 10 trials implemented randomly with no more than two consecutive same-type stimuli and contained all the food items used during baseline sessions.
Interobserver Agreement (IOA)
The trainer served as the primary recorder and the first author served as the reliability observer on all assessment and training sessions. The primary recorder and the reliability observer independently collected data during each phase of the study. We collected IOA data on at least 28% of the sessions across all phases of the study. We used point-by-point agreement ratio (Kazdin, 2011) to calculate agreement on the occurrence of positively- and negatively-reinforced mands. An agreement was scored if both the primary recorder and the reliability observer recorded the same topography of the mand emitted by a child during each trial. A disagreement was scored if there was a discrepancy between the primary recorder and the reliability observer. We divided the total number of agreements by the total number of agreements plus disagreements and multiplied the quotient by 100 to calculate the percentage agreement.
For Adrian, we calculated IOA for 50% (10 of 20) of the nonexample and example trials during structured assessment sessions, for 40% (10 of 25) of the trials during baseline sessions, for 28.5% (18 of 63) of the trials during the mand training sessions, and for 50% (4 of 8) of the trials during generalization and maintenance sessions. Mean agreement was 100% for both nonexample and example trials during structured assessment sessions and baseline sessions and 94.4% (range, 67%-100%) for nonexample trials and 100% for example trials during mand training sessions. Agreement was 100% for nonexample and example trials during generalization and maintenance sessions.
For Julian, we calculated IOA for 50% (6 of 12) of the nonexample and example trials during structured assessment sessions, for 50% (15 of 30) of the trials during baseline sessions, for 38.8% (21 of 54) of the trials during mand training sessions, for 100% (4 of 4) of the trials during generalization sessions, and for 50% (4 of 8) of the trials during maintenance sessions. Mean agreement was 100% for both the nonexample trials and the example trials during the structured assessment and baseline sessions and 100% for the nonexample trials and 95.2% (range, 67%-100%) for the example trials during mand training sessions. Agreement was 100% for nonexample and example trials during generalization and maintenance sessions. For both Adrian and Julian, the low end of the IOA on nonexample trials represents a disagreement in which children handed the trainer the rejection card together with the unpreferred food item on one of the three trials during one session. For this trial, the primary recorder coded the response as correct, whereas the reliability observer scored an incorrect response. At the end of the session, the primary recorder and the reliability observer discussed the disagreement by reviewing the operational definition and clarifying the difference between a correct and incorrect response to prevent further disagreements.
For Justin, we calculated IOA for 45.4% (10 of 22) of the nonexample trials and for 54.5% (12 of 22) of the example trials during structured assessment sessions, for 37.5% (15 of 40) of the trials during baseline sessions, for 28.2 % (66 of 234) of trials during mand training sessions, for 100% (1 of 1) of the generalization sessions, and for 50% (1 of 2) of the maintenance sessions. Mean agreement was 100% for both nonexample trials and example trials during structured assessment, baseline sessions, and mand training sessions. Agreement was 100% for nonexample and example trials during generalization and maintenance sessions.
Procedural Fidelity
The reliability observer collected procedural fidelity data for at least 28% of the sessions across all structured assessment, baseline, mand training, generalization, and maintenance sessions. Procedural fidelity data were recorded by using a checklist consisting of a task analysis of the steps required to implement a session. The reliability observer coded each step as completed or not completed. The procedural fidelity was calculated by diving the number of steps completed by the total number of steps completed plus not completed and multiplying the quotient by 100.
For Adrian, mean procedural fidelity was 97.5% (range, 95%-100%) for the nonexample trials and 100% for the example trials during structured assessment sessions, 100% for baseline sessions, and 100% for the nonexample trials and 99% (range, 94.3%-100%) for the example trials during mand training sessions. Procedural fidelity was 100% for both the nonexample trials and the example trials for generalization and maintenance sessions. For Julian, mean procedural fidelity was 100% for the nonexample trials and 95.8% (range, 91.6%-100%) for the example trials during structured assessment sessions, 100% for baseline sessions, and 100% for the nonexample trials and 99.3% (range, 95.3%-100%) for the example trials during mand training sessions. Procedural fidelity was 100% for both the nonexample trials and the example trials for generalization and maintenance sessions. For Justin, mean procedural fidelity was 97.5% (range, 95.8%-100%) for nonexample and example trials during structured assessment sessions, 100% for the nonexample trials and 97.5% (range, 75%-100%) for the example trials during baseline sessions,100% for the nonexample trials and 98.6% (range, 87.5%-100%) for the example trials for mand training sessions. Procedural fidelity was 100 % for both the nonexample trials and the example trials for generalization and maintenance sessions.
Social Validity
We used a questionnaire to collect social validity data on the acceptability of the intervention and the children’s outcomes at the end of the study. The questionnaire consisted of seven closed-ended items on a 5-point Likert-type scale and one open-ended question. The Likert-type scale ranged from 1 (strongly disagree) to 5 (strongly agree). The items required the trainer to rate several aspects related to intervention (e.g., ease of implementation, interference with daily responsibilities, and the use of the procedures in the future) and to the target behavior (e.g., usefulness and relevance for the child, applicability to other activities and settings, and impact on the child’s quality of life and independent functioning). The open-ended question provided trainers the opportunity to make comments related to other aspects of the study that were not captured in the closed-ended items. For Justin, we collected social validity data by interviewing parents during a posttraining meeting.
Results
Structured Assessment
Table 2 shows the results of the structured assessment. All children emitted multiple topographies of positively- and negatively-reinforced mands in the presence of preferred and unpreferred food items. Adrian used mostly reaching to access preferred food items and pushing away to reject unpreferred food items. He emitted negatively-reinforced mands (e.g., pushing away, dropping) to reject preferred food items on several occasions and used a positively-reinforced mand (i.e., reaching) on one occasion to request an unpreferred food item which he consumed. Julian emitted predominantly reaching to request preferred food items and dropping or running away to reject unpreferred food items. He used a negatively-reinforced mand (i.e., dropping) to reject a preferred food item on one occasion. Justin emitted usually reaching to access preferred food items and giving back or pushing away to reject unpreferred food items. He did not emit a negatively-reinforced mand when offered a preferred food item during the structured assessment sessions.
Topographies of Positively- and Negatively-Reinforced Mands During Structured Assessment.
Acquisition of Discriminated Mands
Figure 1 displays the percentage of correct, discriminated mands across baseline, mand training, generalization, and maintenance for the three children. Data suggest a functional relation between mand training consisting of prompting and differential reinforcement and negatively-reinforced mands in young children with ASD as illustrated by the change in the level and trend of negatively-reinforced mands from baseline to intervention at three different points in time.
Percentage of existing positively-reinforced mands (squares), existing negatively-reinforced mands (triangles), and new negatively-reinforced mands (circles).
For Adrian, baseline data indicate that he emitted positively-reinforced mands when offered a preferred food item and existing negatively-reinforced mands when presented with an unpreferred food item on 100% of the trials prior to intervention. He emitted the new negatively-reinforced mands at 0 levels during baseline. There was no variability or trend in Adrian’s baseline data. After the first three 0-s delay mand training sessions (Sessions 6, 7, and 8), data indicate an upward trend and an increase in his level of performance on the new, discriminated negatively-reinforced mand to 100%. Adrian’s level of performance on the negatively-reinforced mand was variable over the next 11 training sessions and then stabilized at 100% beginning with Session 24. Adrian needed 21 sessions and 126 trials to reach the acquisition training criterion (i.e., 100% correct mands over three consecutive training sessions).
For Julian, baseline data show that his performance on the positively-reinforced and existing negatively-reinforced mands was 100%, whereas this performance on the new negatively-reinforced mand was at 0 levels. There was no variability or trend in Julian’s performance during baseline. Acquisition data reveal an upward trend and an increase in his level of performance on the new mand after the first six training sessions. After two consecutive sessions (i.e., Sessions 23 and 24) at 100%, data indicate minimal variability illustrated by a slight decrease in Julian’s level of performance following a 2-week break due to medical reasons (Session 26). Beginning with Session 27, Julian’s performance began to increase and stabilized at 100% during Sessions 28, 29, and 30. Julian needed 18 sessions and 108 trials to reach the acquisition training criterion.
For Justin, data reveal that he emitted positively- and existing negatively-reinforced mands at 100% prior to training and his level of performance on the new negatively-reinforced mand was at 0 levels. Baseline data show no variability or trend in Justin’s performance. After training began, Justin’s level of performance on the new negatively-reinforced mand remained at 0 levels for the first eight training sessions (Sessions 24-31) and then varied over the next five sessions (Sessions 32-36). Beginning with Session 37, data indicate an upward trend and a gradual increase in Justin’s level of performance to 100% over the next six training sessions (Session 37-42). Although a slight increase in Justin’s level of performance has been documented beginning with Session 37, a marked increase in performance was noted beginning with Session 38 which corresponds to an increased number of trials (i.e., 18) presented during training. Justin needed 21 sessions and 234 trials to reach the acquisition training criterion.
Generalization and Maintenance of Discriminated Mands
Generalization data reveal that two of the three children (i.e., Julian and Justin) emitted the newly taught, discriminated negatively-reinforced mand when offered untrained food items. For one child (i.e., Adrian), data show moderate generalization levels in the presence of untrained items. Specifically, he emitted the new mand in the presence of two of the four untrained food items presented during the two generalization probe sessions. Maintenance data show that all three children maintained and emitted the newly acquired, discriminated mand after the intervention ended.
Social Validity
Social validity data reveal that the trainers rated the intervention as easy to implement (M = 5), applicable to other items or activities encountered by children in their environment (M = 5), and having the potential to improve children’s effective functioning in their environments (M = 5). Trainers did not rate the intervention as being time-consuming (M = 1) or as interfering with their daily responsibilities (M = 1.5). They were also willing to continue the intervention in the future (M = 4.5) and considered that the child learned a useful and relevant skill (M = 4.5). Justin’s parents reported during the posttraining meeting that their child learned an important skill that could be used across environments. Furthermore, his parents stated that they noticed an increase in the use of existing signs in his repertoire and a combination of the newly acquired negatively-reinforced mand with an existing mand. This happened also twice during mand training when Justin signed “no chip.” His parents also expressed interest in learning more about different strategies to promote generalization of the new mand to untrained environments.
Discussion
Our first purpose in this study was to evaluate the effects of example and nonexample stimuli embedded in mand training on producing acquisition of discriminated negatively-reinforced mands in young children with ASD. Our second purpose was to assess whether the mand training promoted discriminated generalization of the newly acquired mand to untrained food items. Our third purpose was to determine whether children continued to emit the new, discriminated mand after training was discontinued. Before we introduced mand training, all children used existing, inappropriate negatively-reinforced mands to reject unpreferred food items and positively-reinforced mands to access preferred food items. Results suggest that all children acquired a socially appropriate and discriminated negatively-reinforced mand. Data indicate high levels of generalization to untrained food items when the trainer conducted example and nonexample trials for two of the three children. Furthermore, no generalization errors were documented during probes to untrained items. Maintenance data reveal that all three children continued to emit the newly acquired, discriminated mand after training was discontinued.
The findings of the study extend the limited body of literature on the effectiveness of mand training in producing acquisition and generalization of negatively-reinforced mands to avoid aversive stimuli other than task demands, such as academic school work (e.g., Chezan et al., 2016; Drasgow et al., 2016; Groskreutz et al., 2014; Yi et al., 2006). These findings are relevant for practitioners working with young children with ASD and language delays in applied settings. Specifically, children with ASD and language delays may resort to problem behavior when exposed to a variety of unpreferred items and activities in their natural environment, unless they acquire socially appropriate and effective mands to communicate with social partners (Reeve & Carr, 2000). Thus, using mand training to teach children negatively-reinforced mands to escape or avoid a multitude of items that are not reinforcing may empower them to exert control over their environment and function effectively when navigating complex situations.
Several aspects of the study have conceptual significance and thus merit discussion. The first aspect is related to the role of example and nonexample stimuli embedded in mand training to promote acquisition of discriminated, negatively-reinforced mands. Our data provide empirical evidence that the example and nonexample trials used during mand training facilitated the acquisition of discriminated mands as illustrated by the emission of a positively- or negatively-reinforced mand based on the reinforcing value of the food item rather than on the type of the food item or trial (preferred or unpreferred) presented by the therapist. For example, we documented multiple instances during mand training when children emitted the new negatively-reinforced mand to reject an item identified as preferred during the structured assessment and, thus, providing evidence that children may attend to the motivating operation (i.e., satiation or aversive stimulation) rather than to an external discriminative stimulus (i.e., the food item presented by the therapist). Similarly, children emitted positively-reinforced mands to mand for an unpreferred food item which they then consumed.
This finding makes a significant contribution to the literature on generalization of negatively-reinforced mands because it represents one of the first studies to examine the role of nonexample stimuli in promoting discriminated manding and, therefore, it sets the foundation for future studies examining the acquisition and generalization of mands to include assessing where the mand should occur and where it should not occur. Although our study focused on examining the generalization and maintenance of discriminated mands, this line of research could be extended beyond mand training to examining the effects of example and nonexample stimuli on producing discriminated skills other than mands, such as vocational, independent living, or social skills. When designing future studies, researchers should consider the findings of past research assessing the role of example and nonexample stimuli within the context of general-case programming, a generalization-promoting strategy, used to teach a multitude of skills to students with moderate and significant cognitive disabilities, including dressing (Day & Horner, 1986), street crossing (Horner, Jones, & Williams, 1985), telephone use (Horner, Williams, & Steverly, 1986), food and drink preparation (Tekin-Iftar & Birkan, 2010 ), table setting (Lehman, Neill, & Proctor, 2009), and requests for assistance (Chadsey-Rusch, Drasgow, Reinoehl, & Halle, 1993).
A novel aspect of our study consists of training and assessing the acquisition of the discriminated mand across multiple example and nonexample stimuli within each session rather than by administering probes that allowed us a more nuanced understanding of how discrimination is achieved. Specifically, the results of the study reveal that the acquisition of discriminated mands may vary within and across children. Specifically, Adrian’s level of performance on discriminated mands during mand training was more variable compared with the performance of the other two children. It is important to note that the variability in Adrian’s performance was not due to discrimination errors or chance responding, but rather to using successive approximations of the new negatively-reinforced mand. For example, during nonexample trials when presented with a preferred food item that was not reinforcing at that time, he would use a combination of an existing mand (e.g., giving back) and the new mand (i.e., using a rejection card) to reject the item; this pattern of responding did not occur during example trials when the trainer offered an unpreferred food item. Thus, these findings suggest that although Adrian emitted the correct topography of the mand during example (or unpreferred) trials, he was not yet able to emit the correct topography when the trainer conducted a nonexample (or preferred) trial and instead he emitted simultaneously the existing and the new mands or used a combination of the two topographies.
Adrian’s behavior does not support the hypothesis of chance responding as illustrated by his consistent pattern of responding across all nonexample trials in which this error occurred. The analysis of Adrian’s responding during the nonexample trials indicated that the mand he emitted, although incorrect topographically, was discriminated and corresponded to the motivating operation in effect at that time. For example, as soon as a preferred food item was available but not reinforcing (i.e., a state of satiation present), Adrian immediately rejected the item by using a combination of an existing and the new, negatively-reinforced mand. He never emitted a positively-reinforced mand immediately after rejecting the preferred food item that would indicate that his responding was due to chance during these trials. Furthermore, the same preferred food item that was rejected during one session was consumed during a different session, which indicates that the reinforcing value of the same item fluctuated over time. Future research should examine the environmental variables, such as the type of trial presented, that may exert control on the acquisition of a targeted, discriminated mand during mand training.
A second novel aspect is related to the discriminated generalization and maintenance of the newly acquired, negatively-reinforced mand to untrained food items. Generalization data reveal that all three children emitted the new, discriminated mand when presented with untrained food items. Although the finding related to generalized mands adds to the existing empirical evidence on generalization of negatively-reinforced mands in children with ASD suggesting that when multiple exemplars are used during mand training generalization to untrained items is likely to occur (Chezan et al., 2016; Groskreutz et al., 2014; Yi et al., 2006), the maintenance of discriminated mand has not been previously examined. Data reveal that all children in this study continued to emit the new, discriminated mand after the intervention ended.
Adrian’s performance (i.e., 100% correct responding) on the two maintenance probes merits further explanation in relation to his performance (i.e., 50% correct responding) on the generalization probes. Although each maintenance probe included all untrained example and nonexample stimuli assessed during generalization and Adrian did not receive additional training, one would expect that his level of correct responding would be similar to his responding on the generalization probes. The difference in the level of correct responding between generalization and maintenance probes can be explained by the presence of the motivating operations. Motivating operations influence the reinforcing value of a food item and, thus, produce variability in responding. For example, Adrian emitted simultaneously both an existing mand and the novel mand (i.e., incorrect response) when presented with a preferred food item during the generalization probes, whereas he emitted a positively-reinforced mand and consumed the same food item (i.e., correct response) when it was offered during the maintenance probes. This pattern of responding suggests that the same item has different reinforcing values at different times and evoked different responses relevant to the motivating operation in effect.
A third novel aspect of this study consists of embedding nonexample stimuli during generalization and maintenance probes. Using nonexample stimuli to evaluate generalization provides evidence related to the occurrence of the newly acquired, discriminated mand in the presence of untrained items and allows researchers to identify generalization errors, such as overgeneralization or undergeneralization, which would not be otherwise identified. Data indicate the mand training was effective in producing generalization while preventing errors as demonstrated by the emission of the newly acquired mand in a discriminated way in the presence of untrained nonexamples. Thus, the findings of the study add additional empirical evidence to the limited body of literature on the role of nonexample stimuli in assessing the generalization of negatively-reinforced mands to untrained items (Chezan et al., 2016; Groskreutz et al., 2014).
One interesting finding related to generalization to untrained nonexamples consists of Adrian’s level of performance during generalization probes after mand acquisition. When the trainer presented Adrian with an untrained nonexample (or preferred) food item during the first generalization probe, he used a combination consisting of an existing mand (i.e., giving back) and the new mand (using the card) to reject the item. Because of the limited number of opportunities (i.e., two) to emit a mand to untrained nonexamples, we wanted to obtain more information to better understand the error in Adrian’s response. Therefore, we decided to use two additional untrained nonexamples identified during the initial structured assessment to examine this error within the context of a broader assessment with items whose value may have changed from the time when we conducted the assessment to the time when we assessed generalization.
Data on the second generalization probe revealed that the same error occurred when the trainer conducted trials with additional untrained nonexample items. One potential explanation for Adrian’s error may be related to response efficiency (Drasgow et al, 2001). Response efficiency involves first the comparison of the reinforcement probability of each response available and then selecting the one that has been associated with a history of reinforcement and requires less effort (Horner & Day, 1991). The fact that Adrian just acquired a new mand in his repertoire that had a shorter history of reinforcement compared with the existing mands combined with the possibility that the existing mands were reinforced outside the context of the study likely influenced the simultaneous emission of both mands. It is possible that it was easier and more adaptive for Adrian to emit both topographies simultaneously in an attempt to access the reinforcer. Therefore, further examination of the influence and contribution of each variable related to response efficiency in promoting generalization of discriminated negatively-reinforced mands is warranted.
Another important finding of our study consists of social validity. Trainers rated the intervention as easy to implement, applicable to other items or activities, not time-consuming, and relevant for increasing children’s functioning in their environment. They reported that the child learned a relevant and useful skill and were willing to continue the procedures in the future. The social significance of the new mand was also reported by Justin’s parents who also noticed an increase in the use of existing signs in his repertoire in combination with the newly acquired negatively-reinforced mand. These findings have clinical significance. Specifically, our data documented that practitioners working with children with ASD in applied settings can not only implement mand training with fidelity but also rate the procedures and outcomes as being socially valid. The social validity of procedures is essential to their adoption and successful implementation by practitioners providing services to children with ASD (Callahan, Henson, & Cowan, 2008; Kazdin, 1977). It is important to note that the therapists who implemented the mand training in this study had training in ABA. Future studies should determine whether these procedures can be replicated with teachers or parents who do not have training in behavior analysis. The need for studies examining the social validity of interventions with children with ASD and their implementation by teachers in classroom settings has been emphasized in the literature on evidence-practices in ASD in an attempt to identify various factors that impede the use of evidence-based interventions for children with ASD in schools (Callahan et al., 2017).
Readers should consider several limitations when interpreting the findings of this study. First, we taught a general mand (i.e., “no”) to children enrolled in this study. Although in conventional spoken language, a communicative response such as “no” may not be responded to as a mand, young children with ASD who have limited or no negatively-reinforced mands in their repertoire may benefit from learning a general mand as a first step before acquiring more complex verbal operants (Groskreutz et al., 2014). Second, we determined the negative reinforcement function of existing mands within the context of a structured assessment, but we did not conduct a functional analysis of existing mands present in the children’s repertoires prior to implementing mand training. Researchers conducting future studies on this topic could use other types of systematic stimulus preference and avoidance assessments (Groskreutz et al., 2014) or a functional assessment of verbal behavior to determine the function of existing verbal responses (Lerman et al., 2005). Third, we did not use a specific assessment to determine the shift in children’s preference across time in relation to preferred and unpreferred food items used throughout the study. However, we considered children’s behavior as an indicator of their preference and analyzed the pattern of their behavior across sessions to identify the shift in preference. For example, if the trainer offered a child an item categorized as unpreferred and the child emitted a positively-reinforced mand and consumed the item on repeated trials with the same item, we inferred that there was a shift in the child’s preference over time. Fourth, the limited number of generalization probes may limit the conclusions related to the discriminated generalization of newly acquired negatively-reinforced mands. Although we embedded the example and nonexample probes to untrained items in the baseline and maintenance sessions, we conducted only one or two generalization probes after children met the acquisition training criterion. We decided not to conduct continuous generalization probes throughout the study because previous empirical evidence suggests that mands do not generalize to untrained items until the targeted mand is trained in the presence of multiple exemplars and the training criterion is achieved (Groskreutz et al., 2014; Stokes & Baer, 1977; Yi et al., 2006). Fifth, although data demonstrate a functional relation between mand training and negatively-reinforced mands, the strength of the experimental control may have been weakened for two reasons. One reason is the delayed effect of the mand training for two children. For example, data displayed in Figure 1 suggest an effect of mand training as illustrated by an increase in the emission of the novel mand following four sessions at the 3-s delay for Julian and following five sessions at the 3-s delay for Justin. The second reason is the increased number of trials (i.e., 18) conducted with the third child. We do not know whether an increased number of trials would have produced a faster acquisition of the new mand for the first two children. Finally, we implemented the intervention in the settings (i.e., home and center-based program) in which children received their daily ABA therapy in a one-on-one format with the trainer. Although, these settings represented the natural environment in which instructional activities occurred for children included in this study, they may not be considered a typical environment in which daily activities occur for some young children with ASD. Therefore, future research should replicate the findings of this study in more natural settings (e.g., inclusive classrooms in public schools) while training typical communication partners, such as teachers or parents to implement the intervention.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the Office of Research at Old Dominion University.