We compared three methods for teaching beginning conversation skills (i.e., intraverbals in the form of answering questions about object features or functions) to individuals with language delays. We examined transfer-of-control, interspersal, and general instruction plus instructive feedback. Adding requests into instructional sessions did not result in faster acquisition of conversation skills when compared to general instruction plus instructive feedback. However, differences were observed in how long each type of session took to conduct with participants. In addition, two out of three participants correctly answered questions related to the instructive feedback stimuli even though specific instruction was never introduced. Finally, no single training condition was associated with improved maintenance relative to the other conditions. Results are discussed in terms of the efficiency of instruction with consideration given to the potential benefits of including requests into instructional sessions even when the amount of time needed to conduct sessions is increased.
Keywords: intraverbal, transfer-of-control, interspersal, instructive feedback
Evaluation of Three Methods for Teaching Beginning Conversation Skills to Individuals with Language Delays
Skinner (1957) developed a system by which verbal behavior, or language, could be conceptualized within the framework of the principles of behavior analysis. This conceptualization of verbal behavior differed from traditional conceptualizations of language by focusing on the environmental events occurring prior to and following the verbal response of the individual as the speaker. Skinner suggested that verbal behavior was maintained by consequences delivered by another person (i.e., the listener). The unit of analysis in Skinner's theory of verbal behavior was the verbal operant. In general, this unit of analysis focuses on the functional relationship between verbal behavior and the independent variables responsible for evoking and maintaining it (Cooper, Heron, & Heward, 2007). Skinner identified a number of different verbal operants, with echoics (i.e., verbal imitations), mands (i.e., requests), tacts (i.e., object or activity labels), and intraverbals representing the elementary verbal relations of which more difficult types of verbal behavior are shaped (Sundberg & Michael, 2001).
As described by Skinner (1957), the intraverbal is under the control of a verbal cue (e.g., question) that is different than the expected response (e.g., answer) and is maintained by generalized reinforcement such as reciprocal conversation or praise. It is considered the building block of conversation skills because it describes most types of verbal interactions. A functional repertoire of conversation skills is important for successful integration into home, school, and work settings. However, research focusing on Skinner's account of verbal behavior has largely overlooked the intraverbal (Sautter & LeBlanc, 2006).
Most studies related to teaching intraverbals have evaluated transfer-of-stimulus-control procedures by introducing and fading prompts that evoke intraverbal responding (e.g., Finkel & Williams, 2001; Partington & Bailey, 1993; Sundberg, San Juan, & Dawdy, 1990). Transfer of stimulus control refers to the process of fading out contrived controlling variables (e.g., response prompts, motivating operations) from the instructional arrangement so that responding is eventually under the control of naturally occurring antecedents (Cooper et al., 2007). Finkel and Williams, for example, compared the effects of using textual prompts (i.e., presenting the correct response in written form) and echoic prompts (i.e., presenting a spoken model of the correct response) while teaching answers to social questions to an individual diagnosed with autism. Both prompting procedures were effective in establishing intraverbal behavior; however, the written response prompts resulted in faster acquisition of intraverbals and greater use of full sentence responses.
Further studies aimed at identifying more efficient methods for teaching intraverbals are needed due to the time constraints often placed on individuals who implement verbal behavior programs. In addition, more research is needed in this area given the importance of building verbal repertoires or communication skills for individuals with language delays. Research on other verbal operants, such as echoics (i.e., verbal imitations) and tacts (i.e., object or activity labels), suggests that intraverbal instruction may be enhanced by capitalizing on the unique controlling features of the mand (e.g., Arntzen & Almås, 2002; Carroll & Hesse, 1987; Drash, High, & Tudor, 1999). Mands, or requests, occur when an individual is motivated to change a state of deprivation, satiation, or aversive stimulation. Mands are maintained by whatever is being requested (i.e., specific reinforcement). Because the mand results in specific reinforcement and is the only verbal operant that allows the speaker to control his or her environment (Sundberg & Michael, 2001), motivation to engage in mand responses may be higher than for other types of verbal operants. Thus, incorporating mands into the instruction of other verbal operants may increase the individual's motivation to learn.
Carroll and Hesse (1987), for example, compared the effects of teaching tacts alone to a condition during which the experimenters alternated between mand and tact learning trials with six preschool-aged children. The participants acquired the tacts more rapidly during the combined training. Arntzen and Almås (2002) replicated these findings with individuals with developmental disabilities. However, in both investigations, the experimenter evaluated acquisition under the training conditions only. Thus, two variables -- seeing the object and desiring access to the object -- may have controlled accurate responding under the mand-tact condition. It remained unclear whether the learners would label objects if they did not also desire access to those objects.
Despite the potential benefits of this approach, interspersing trials that target mands with trials that target another verbal operant (e.g., intraverbals) might compromise efforts to teach more efficiently. As an alternative, initial intraverbal learning trials could incorporate the controlling features of the mand and then fade them gradually. As described by Sundberg and Partington (1998), initial intraverbal instruction could focus on things that interest the individual (e.g., conversation related to preferred items). The instructor could withhold access to these highly preferred items while delivering the cue for the intraverbal. Contingent on a correct response, the individual would gain access to the item along with some type of nonspecific reinforcement (e.g., praise). The instructor then could gradually remove the controlling stimuli for the mand (i.e., restriction of preferred items and contingent access to them) until the response only occurred following the intraverbal cue. This transfer-of-stimulus control procedure has not yet been evaluated during intraverbal instruction.
Another body of research suggests that a procedure called "instructive feedback" also might enhance intraverbal instruction. Teachers using instructive feedback present new information following the learner's response to a different instruction (i.e., during the consequence of a learning trial; Werts, Wolery, & Holcombe, 1995). For example, when teaching an individual to identify a lion from a picture card, a teacher might show the learner a picture card of a lion and ask, "What is this?" If the individual emits the correct response (i.e., "lion"), the teacher might respond by saying, "Great job! That is a lion. You spell lion 'L-I-O-N.'" In this example, the teacher delivered praise as reinforcement for the targeted response (i.e., vocalizing the name of the pictured stimulus), as well as extra information (i.e., the correct way to spell lion). The extra information (i.e., how to spell lion) is the instructive feedback stimulus. A response to the instructive feedback stimulus is not required and no contingencies are in place for responding or not responding to the stimulus. Generally, the main purpose of including instructive feedback stimuli is to increase the efficiency of instruction by allowing students the opportunity to acquire new information without significantly increasing instructional time.
Research findings on instructive feedback suggest that learners acquire much of the information presented as instructive feedback, although results have been inconsistent (Werts et al., 1995). Perhaps more importantly, benefits are generally seen when the information previously presented as instructive feedback is subsequently targeted in training (e.g., Holcombe, Wolery, Werts, & Hrenkevich, 1993; Wolery, Doyle, Ault, & Gast, 1991). This method could provide clinicians with a highly efficient mechanism for priming new skills that will be later introduced during instruction. That is, verbal behavior programs could incorporate intraverbal stimuli as instructive feedback during direct intraverbal instruction or during instruction targeting other verbal operants (e.g., tacts or mands).
Further research is needed to directly evaluate efficient methods for teaching intraverbals. The purpose of the current investigation was to evaluate the efficiency of three methods for teaching intraverbals to individuals with language delays. We examined the advantages of incorporating mands into training via transfer-of-stimulus-control or task interspersal procedures by comparing performance under those conditions to a condition under which instruction did not include mands. We also assessed whether learners acquired intraverbals presented as instructive feedback during learning trials for other intraverbals. Finally, we evaluated potential differences in the generalization and maintenance of responding across training conditions.
Participants and Settings
Three children with language delays participated in the study. They were selected from children enrolled in a clinic providing early intervention and behavioral services and from an early intervention program for children with autism and other language delays. Participants were screened for inclusion using the Assessment of Basic Language Learning Skills, Revised (ABLLS-R; Partington, 2006). In general, inclusion criteria were selected to ensure that participants verbally imitated words, requested items when prompted and unprompted in the presence of the item, labeled preferred items, and did not reliably answer questions related to item functions, classes, or features (more detailed information about specific inclusion criteria can be obtained from the first author). In addition, inclusion criteria required that praise functioned as a reinforcer for participants' behavior (additional details provided below).
Greg was a 12-year-old boy diagnosed with autism, attention deficit hyperactivity disorder (ADHD), and asthma. He primarily spoke using four- to five-word utterances and was enrolled in a behavior analytic early intervention program for 3 months prior to this investigation. Greg used asthma medication on an as-needed basis during the investigation. Kyle was a 4-year-old boy diagnosed with autism who spoke in three- to four-word utterances and used full sentences for some requests. He was enrolled in a behavior analytic intervention program for approximately 4 months prior to entering the current investigation. He was not taking medication during the course of the investigation but began to receive a gluten-free diet between the completion of training for the first stimulus set and the initiation of training for the second stimulus set. Mary was a 3-year-old girl with no formal diagnosis but was being treated for language delays. She communicated using two- to three-word utterances along with complete sentences on occasion (e.g., to request for items). She was enrolled in a behavior analytic early intervention program for 10 months prior to the beginning of the investigation. She resided in a bilingual household where her mother exclusively spoke in Spanish and her father spoke in English and Spanish. She was not taking any medications at the time of the investigation. None of the participants were diagnosed with any sensory or physical impairments. Two other individuals were enrolled in the study but withdrew participation prematurely for reasons unrelated to their participation in the study (e.g., withdrew from the early intervention program).
Trained graduate students, undergraduate students, or clinic staff conducted sessions during various phases of the investigation. Sessions were conducted in areas designated for individual instruction in all settings. Each area was equipped with materials necessary to conduct the sessions, including a table, chairs, area partitions, preferred stimuli, a stopwatch, and data sheets. Participants were blocked from engaging with other objects present in the area (e.g., trash can, books, extra desks) when necessary.
Target Behaviors, Measurement System, and Procedural Integrity
Observers collected data on the number of independent responses, modeled responses, error responses, and no responses. Independent responses were defined as correct responses emitted following an intraverbal cue and prior to a model of the correct response. Modeled responses were defined as correct responses emitted following a model of the correct response and prior to the end of the trial; modeled responses could only occur during training sessions. Error responses were defined as incorrect responses emitted at any point during the learning trial. No response was defined as no response emitted within 10 s of an intraverbal cue. Experimenters and observers used pencil and paper to collect frequency data on the participant's responses in each session. Observers recorded responses during each trial as soon as the response was emitted or once the trial duration ended (10 s elapsed, as signaled by a timer). One trial consisted of the presentation of the appropriate cue, a response emitted by the participant or the end of the trial duration, and presentation of the programmed consequence. Data on each response type were converted to percentage of trials by dividing the total number of each response type by the total number of trials and multiplying that number by 100%. During training sessions, data also were collected on session duration by starting a stopwatch when the first cue was delivered and stopping the stopwatch at the end of the consequence of the last learning trial. The main dependent variable was the total number of training sessions to meet the mastery criteria. Fewer training sessions under a particular condition would suggest a more efficient training method. Secondary analyses of efficiency were conducted by comparing the average session length for each training condition, with shorter session durations indicating a more efficient training method.
A second observer independently collected data during at least 48% of baseline, acquisition probe, training, generalization, and maintenance sessions. Agreement between observers for responding was calculated by comparing the recorded responses of each observer on a trial-by-trial basis. Agreement was defined as observers scoring the same type of response emitted during each trial. Interobserver agreement coefficients for responding were calculated by dividing the total number of agreements by the total number of agreements plus disagreements and multiplying that number by 100%. Interobserver agreement coefficients for session duration were calculated by dividing the smaller of the recorded session durations by the larger and multiplying that number by 100%. Agreement coefficients were collapsed across baseline, acquisition probe, training, generalization, and maintenance sessions for responding and session duration. Mean agreement for Greg was 97.7% (range, 75% to 100%) for responses and 99.2% (range, 81.5% to 100%) for session duration. Mean agreement for Kyle was 98.3% (range, 90% to 100%) for responses and 99.3% (range, 82.3% to 100%) for session duration. For Mary, mean agreement was 97.2% (range, 50% to 100%) and 99.8% (range, 98.2% to 100%) for responses and session duration, respectively.
Baseline, acquisition probe, training, generalization, and maintenance sessions were also scored for procedural integrity related to the delay to the prompt (defined as the total number of seconds between the delivery of the intraverbal cue and delivery or no delivery of a prompt of the correct response) and consequence delivery (defined as the type of stimulus change initiated by the experimenter following a response from the participant). Sessions were videotaped and data collectors scored each videotaped session for integrity as described below. Integrity was defined as the experimenter waiting the correct number of seconds between the presentation of the intraverbal cue and the prompt (if necessary; i.e., delay to the prompt) based on the condition in place and delivery of the programmed consequence based on the participant's response and condition in place. In other words, during baseline, acquisition probe, generalization, and maintenance sessions, a correct delay was scored if the therapist did not deliver a model of the correct response during the 10-s trial interval. During training sessions, a correct delay was scored if the therapist waited the number of seconds indicated on the data sheet based on the increasing time delay before presenting a model of the correct response. A correct consequence delivery was scored during baseline, acquisition probe, generalization, and maintenance sessions if the experimenter delivered praise following a correct response or ignored incorrect and no responses. During training, a correct consequence delivery was scored if the experimenter only delivered an error correction for incorrect responses during both mand conditions and an error correction plus instructive feedback during the instructive feedback condition, delivered no consequence for no response during all conditions, delivered praise plus instructive feedback for correct or modeled responses during instructive feedback sessions, delivered praise plus 20-s access to the item during mand transfer-of-control sessions, delivered 20-s access to the item during mand trials for correct responses or no consequence for incorrect mand responses, and delivered praise only during intraverbal trials for correct or modeled responses during mand interspersal sessions. Final integrity scores were converted to the percentage of trials with the correct delay used and consequence delivered by dividing the number of trials with integrity by the total number of trials and multiplying by 100%. Integrity was scored for at least 19% of all sessions for each participant. Mean integrity scores for participants were collapsed across sessions and stimulus sets and were as follows: Greg 98% (delay; range, 80% to 100%) and 96.7% (consequence delivery; range, 80% to 100%), Kyle 98.7% (delay; range, 85% to 100%) and 97.3% (consequence delivery; range 85% to 100%), and Mary 93.8% (delay; range, 50% to 100%) and 92.9% (consequence delivery; range, 50% to 100%).
The effects of the training on intraverbal acquisition were examined via a nonconcurrent multiple baseline across stimulus sets design and a multiple probe design. The three conditions were compared twice for each participant by alternating between the different training conditions. Each stimulus set consisted of training three different intraverbals (one under each condition) and presenting one intraverbal as instructive feedback. Thus, a total of eight intraverbals were taught to each participant in this study (i.e., four items per stimulus set).
Each participant was exposed to three pre-training assessments (i.e., preference assessment, reinforcer assessment, pre-training probes); intraverbal training; and acquisition, generalization, and maintenance probes. Results from the preference assessment (Fisher et al., 1992) were used to identify items to use in each stimulus set. The items included were those that participants requested in the natural environment. Assignment of stimuli to each condition was counterbalanced within and across participants based on preference rankings to control for differences in preference between the stimuli. For example, if the highest preferred item was assigned to the interspersal condition for participant 1 in the first evaluation, then the second preferred item might be assigned to that condition in the second evaluation. Then, for participant 2, the third preferred item might be assigned to this condition during the first evaluation and the fourth preferred item in the second evaluation. This was done for each condition of the investigation across all participants. Table 1 lists the intraverbal stimuli used during training for each participant across both stimulus sets. The participant did not have access to the selected items outside of the training sessions for the duration of the evaluation to ensure that the motivation to request items was present during training. The reinforcer assessment was conducted to evaluate whether praise functioned as a reinforcer for participants. In addition, the pre-training probes evaluated whether participants engaged in mands and intraverbals related to the items used during training. Items were included if the participant requested the item but did not answer targeted questions related to it. More details about all pre-training assessments are available from the first author upon request.
Intraverbal baseline. During baseline, five trials were conducted with each of the four intraverbal stimuli used during training, resulting in a total of 20 intraverbal trials per session. That is, five trials were conducted for the intraverbal stimuli assigned to the transfer-of-control condition, the instructive feedback condition (training stimulus and instructive feedback stimulus), and the interspersal condition, resulting in all four intraverbals being probed in one session. Order of presentation was randomized prior to each session. During each trial, the referent item was out of the participant's view. The experimenter presented the intraverbal cue (e.g., "What is something that is a vehicle and is yellow?"). Participants had 10 s to respond following the presentation of the intraverbal cue. Praise was delivered contingent on the targeted response. There were no consequences in place for nontargeted responses or no response. A new trial was initiated once 10 s elapsed from the presentation of the intraverbal cue, regardless of when a response was observed during that time. For example, if a correct response was emitted at 1 s following the presentation of the intraverbal cue, the experimenter waited 9 s prior to initiating a new trial. This was done to equate the length of a trial within each condition. Baseline consisted of at least three sessions and continued until stable levels of responding were observed for each of the intraverbal stimuli evaluated.
Acquisition probes (transfer of control, instructive feedback training, instructive feedback stimulus, and interspersal). Acquisition probes were conducted for each intraverbal used during training and presented as instructive feedback prior to each series of training sessions (described below). In other words, acquisition probes always preceded training sessions (with the exception of the first series of training sessions). Acquisition probes were identical to baseline sessions and were conducted as a test of intraverbal acquisition in the absence of the different training conditions (e.g., mand interspersal, specific reinforcement, instructive feedback presentation). The training condition was discontinued contingent on three consecutive acquisition probes with 80% correct independent intraverbal responses related to the stimulus being trained during that condition. Stimuli being trained under other conditions remained in training until the criteria were met for those stimuli during acquisition probe sessions. All stimuli continued to be presented during the acquisition probes regardless of whether mastery criteria were met. Generalization probes were initiated once the participant met mastery criteria for all intraverbals (training and instructive feedback stimuli).
General training procedures. A progressive time delay prompting procedure was used in all training conditions (Charlop, Schreibman, & Thibodeau, 1985). That is, the initial delay to the prompt was 0 s. The delay was increased by 1 s within and across sessions following three consecutive trials with correct independent or modeled responses until mastery criteria were met (see Acquisition Probes above) or a 10-s delay was reached, whichever came first. If a 10-s delay was reached, prompts were discontinued following two consecutive sessions with 80% correct independent or modeled responses at the 10-s delay. Training was conducted in 30-min blocks, which consisted of one acquisition probe and at least one session for each training condition. Participants were provided with at least a 1-min break between each session, which was necessary for the experimenter to reorganize materials.
Mand to intraverbal transfer-of-control training (Mand TC). The referent item was placed out of the participant's view. The experimenter delivered the intraverbal cue (e.g., "What is something that is a vehicle and is yellow?") and waited for the corresponding duration of time based on the increasing time delay procedure described above. The experimenter immediately delivered praise and 20-s access to the referent item if the participant engaged in the targeted response (e.g., "bus") within the delay following the prompt. The experimenter verbally modeled the correct response (e.g., "bus") if the participant did not respond within the delay period. The experimenter immediately delivered praise and 20-s access to the item if the participant engaged in the targeted response following a model and prior to the end of the trial. No consequence was delivered and the next trial was initiated if the participant did not respond after the model and prior to the end of the trial. The experimenter immediately delivered an error correction (i.e., "No." followed by a model of the targeted response) if the participant responded incorrectly at any point during the trial (i.e., engaged in a nontargeted response). No reinforcement was delivered for a targeted response following an error correction. A new trial was initiated once 10-s elapsed from the initial presentation of the intraverbal cue, regardless of when a response was observed during that time. However, following a correct independent or modeled response, the trial ended upon the termination of the reinforcement interval since it exceeded the 10-s trial limit. The delay to the model prompt was increased as described above. Reinforcement fading would have been initiated in order to transfer control from the combined intraverbal and mand to intraverbal only following five consecutive sessions with 100% independent responding. However, reinforcement fading was never initiated because all participants met mastery criteria during acquisition probe sessions prior to meeting the reinforcement fading criteria during the training sessions. Transfer-of-control sessions were discontinued as previously described.
Instructive feedback training (IF TRN). This condition was similar to the Mand TC condition with the following exceptions. First, reinforcement for independent or modeled correct responses consisted of praise only. Second, instructive feedback related to novel intraverbal information was presented during the consequence of the learning trial. In other words, instructive feedback was delivered immediately following praise, error corrections, or the end of the 10-s trial interval with no response. For example, following the intraverbal cue "What's something that is green and squishy?" and the independent response "water snake," the experimenter said, "Good job." followed by, "A dog is furry and barks [instructive feedback stimulus]." The instructive feedback stimulus (IF STIM) was a different intraverbal from that taught during training. A response to the instructive feedback stimulus was not required and there were no consequences for responses related to it. The delay was increased and sessions were discontinued as previously described. Direct intraverbal training was initiated for the IF STIM as described for the training stimulus in this condition if the mastery criteria for the intraverbal IF STIM (e.g., "What is furry and barks?") were not reached prior to training termination. However, if two consecutive probe sessions were obtained at or above 80% for the IF STIM when training was terminated for the IF TRN, one more IF STIM probe session was conducted prior to initiating direct instruction to determine if mastery would be observed (Greg and Kyle). Once the mastery criteria for all intraverbal stimuli were reached, generalization probes were initiated (with the exception of Mary).
Mand interspersal training (Mand INT). Sessions were similar to the above conditions with the following exceptions. The experimenter alternated between known mand and unknown intraverbal trials within each session. Mand trials always preceded intraverbal trials. An identical response was required for the known mand and unknown intraverbal trials. Intraverbal trials were conducted as described above for the training stimulus in the instructive feedback condition, except no instructive feedback was delivered. Mand trials began by the experimenter briefly showing the relevant item to the participant and then removing the item from the participant's view. The experimenter delivered the prompt, "What do you want?" once the item was removed from view. Participants had 10 s to request the item briefly shown to them. Contingent on correct mand responses, the experimenter delivered 20-s access to the item. Incorrect responses (i.e., requests for items not used in the training situation) did not result in access to the item and the intraverbal trial was initiated. The participant did not access the item and an intraverbal trial was initiated if the participant did not respond within 10 s of the mand prompt. Ten trials each of mands and intraverbals were conducted, for a total of 20 trials. The delay was increased and sessions terminated as previously described.
Generalization probes (GEN). Generalization probes related to the intraverbals taught during training or presented as instructive feedback were conducted following acquisition of all intraverbals. The purpose of these probes was to evaluate whether participants acquired symmetrical intraverbal responses related to the training stimuli and instructive feedback stimulus. Generalization probes were similar to baseline and acquisition probe sessions with five trials conducted for each trained intraverbal stimulus (training and instructive feedback), resulting in a total of 20 intraverbal trials per session. However, for each intraverbal taught during training, the experimenter delivered a novel but related intraverbal cue. For example, if during training the individual was taught to respond to "What is something you sing into that makes you louder?" with the correct response being "microphone," the generalization probe consisted of "Tell me something about a microphone" with the correct response(s) being "You sing into it," "It makes you louder," or a combination of the two. At least five generalization probe sessions were conducted (Mary's first stimulus set was the only exception).
Maintenance probes (MTN). Maintenance probe sessions were conducted for all participants (except for Mary's first stimulus set) following the generalization probes for each condition. Maintenance probes were identical to baseline sessions and were conducted for the stimuli taught in each condition. Maintenance probes were conducted one week following the last generalization probe with the exception of the first trained set for Kyle. Kyle's maintenance probe for the first trained set was conducted 6 weeks following the generalization probes due to a scheduled 2-week school closure that was immediately followed by temporary closure due to weather-related problems. One maintenance probe was conducted for each stimulus set.
Results of the intraverbal acquisition, generalization, and maintenance probes are presented in Figure 1 for all participants. Data from the mand trials are available from the first author upon request. Recall that acquisition probes were implemented immediately prior to the training sessions (with the exception of the first training day) to assess acquisition of the intraverbals outside of the training conditions. Before training, Greg (top panel) never engaged in correct intraverbal responses with either stimulus set (M = 0%). Throughout training, levels of Greg's correct intraverbal responses during the acquisition probes were similar across conditions for both stimulus sets (Set 1 - Mand TC and Mand INT, M = 42.9%; IF TRN, M = 37.1%; IF STIM M = 40%; Set 2 - Mand TC M = 93.3%; Mand INT, M = 80%; IF TRN stimulus, M = 70%; IF STIM, M = 100%). Similarly, no consistent differences in generalized responding were observed across conditions for either stimulus set (Set 1 - Mand TC, M = 100%; Mand INT, M = 92%; IF TRN, M = 84%; IF STIM, M = 96%; Set 2 - Mand TC, M = 77.5%; Mand INT, M = 90%; IF TRN and IF STIM, M = 75%). In terms of maintenance, responding was nearly identical across conditions for both trained stimulus sets (Mand TC, 100%; Mand INT, 100%; IF TRN, 80%; IF STIM, 100%).
Kyle's intraverbal acquisition, generalization, and maintenance probes are presented in the middle panel of Figure 1. Mean levels of responding during baseline were 0% across both stimulus sets and all conditions. Like Greg, levels of correct intraverbal responses were similar across conditions and stimulus sets (Set 1 - Mand TC, M = 35%; Mand INT, M = 48.9%; IF TRN, M = 47.5%; IF STIM, M = 31.1%; Set 2 - Mand TC, M = 80%; Mand INT, M = 72%; IF TRN, M = 72%; IF STIM, M = 52%). Generalized responding was not observed for any stimulus for the first stimulus set, regardless of the training condition (M = 0%). However, generalized responding was observed for the second stimulus set, with similar levels of responding across training conditions (Mand TC, M = 88%; Mand INT, M = 88%; IF TRN, M = 84%; and IF STIM, M = 92%). Maintenance of responding was observed for the Mand INT (80%) and IF STIM (100%) conditions, but not for either the Mand TC (0%) or IF TRN (0%) conditions for the first stimulus set. Similar levels of maintenance were observed for the second stimulus set across all training conditions (Mand TC, Mand INT, and IF TRN = 100%; IF STIM = 80%).
Results of Mary's intraverbal acquisition, generalization, and maintenance probes are presented in the bottom panel of Figure 1. For the first stimulus set, similar levels of responding were observed across all training conditions (Mand TC, M = 38%; Mand INT, M = 36%; IF TRN, M = 34%). Mastery criteria were not met for the IF STIM when presented as instructive feedback during the training sessions (M = 0%) or following direct instruction (M = 4%). For the second stimulus set, similar levels of correct intraverbal responses were observed across the Mand INT (M = 62%), Mand TC (M = 56%) and IF TRN (M = 41.8%) conditions. Again, Mary never met the mastery criteria for the IF STIM for the second stimulus set prior to and after direct instruction (M = 1.5% and M = 0%, respectively). Unlike Greg and Kyle, she did not demonstrate generalized responding for any intraverbal stimulus for either stimulus set (M = 0%). Responding during maintenance probes was slightly higher for stimuli trained under Mand INT and IF TRN conditions (100%) relative to the Mand TC condition (80%).
The number of training sessions needed to meet the mastery criteria is presented in Figure 2 for all participants. Data for Greg are presented in the top panel, Kyle in the middle panel, and Mary in the bottom panel. For all participants, fewer training sessions were needed to master the intraverbals for the second stimulus set relative to the first stimulus set for all training conditions. The same or a similar number of sessions was needed for stimuli trained under all conditions across stimulus sets for Greg (Set 1 - all conditions, n = 7; Set 2 - Mand TC, n = 3; Mand INT, n = 5; IF TRN, n = 4) and Kyle (Set 1 - Mand TC and IF TRN, n = 8; Mand INT, n = 9; Set 2 - Mand TC, n = 3; Mand INT, n = 5; IF TRN, n = 4). Direct instruction was never initiated for the IF STIM for Greg and Kyle because mastery was obtained during probes. For Mary, the same number of sessions was needed across conditions for the first stimulus set (n = 10); however, for the second stimulus set, fewer sessions were required for stimuli trained under the Mand TC and Mand INT conditions (n = 5) compared to the IF TRN condition (n = 11). Mary never mastered stimuli presented as instructive feedback (IF STIM), even when direct instruction was initiated.
Figure 3 displays data from all participants' training sessions. For Greg (top panel), higher, more stable levels of independent responding were observed under the Mand TC condition for both stimulus sets (M = 70% and M = 83.3% for Sets 1 and 2, respectively) than under the Mand INT condition (M = 30% and M = 80%) and the IF TRN condition (M = 24.3% and M = 65%). For Kyle (middle panel), the highest levels of independent responding for the first stimulus set were observed under the Mand INT (M = 68.9%) condition, followed by the IF TRN condition (M = 56.3%), with the Mand TC condition associated with the lowest levels of responding (M = 46.3%). Different patterns of responding were observed for the second stimulus set, with the highest levels of responding associated with the Mand TC condition (M = 90%). The IF TRN condition was associated with the second highest levels of responding (M = 72.5%). Lowest levels of responding were observed under the Mand INT condition (M = 32%), with an overall decreasing trend in responding observed across sessions. For Mary (bottom panel), correct responding increased more rapidly under the Mand INT (Set 1, M = 91%; Set 2, M = 68%) and Mand TC (Set 1, M = 77%; Set 2, M = 68%) conditions. The IF TRN condition produced the lowest levels of responding (Set 1, M = 48%; Set 2, M = 41.8%) across both stimulus sets. When training was initiated for the IF STIM, responding remained low, and Mary never met the mastery criteria for either stimulus set (Set 1, M = 20%; Set 2, M = 0%).
The duration of each participant's training sessions are presented in Figure 4. Data represent the average session duration for all training sessions in each condition across stimulus sets for each participant. Overall, IF TRN training sessions took less time to implement (Greg, M = 3.1 min; Kyle, M = 2.9 min; Mary, M = 3.2 min) relative to both types of mand sessions. In addition, Mand TC sessions took less time to implement (Greg, M = 5.3 min; Kyle, M = 6.5 min; Mary, M = 6.2 min) than Mand INT sessions (Greg, M = 7.4 min; Kyle, M = 8.2 min; Mary, M = 8.0 min).
Conclusions and Guidelines for Practitioners
The current investigation evaluated the efficacy of three methods for teaching intraverbals, or beginning conversation skills, to children with language delays. Participants generally required a similar number of training sessions to meet the mastery criteria for the intraverbals taught under each training condition. However, the instructive feedback sessions took less time to complete than both mand conditions, and the mand transfer-of-control sessions took less time than the mand interspersal sessions. Results also indicated that two participants (Greg and Kyle) acquired intraverbals presented as instructive feedback without specific training. Finally, all participants demonstrated faster acquisition for the second stimulus set relative to the first stimulus set.
Results suggest some important considerations when selecting training methods to use during verbal behavior or language instruction. Although all conditions required a similar number of training sessions, the duration of these sessions differed significantly. The instructive feedback sessions were the fastest to conduct followed by the mand transfer-of-control sessions. Clinicians and educators might consider choosing the most efficient strategy based on the time needed to complete the session to maximize instructional time. On the other hand, Sundberg and Michael (2001) indicated that parents and therapists often report mand training to be more enjoyable than instruction involving other verbal operants. Although data were not formally collected on each participants' preference for the different training conditions, they appeared to enjoy the mand sessions more than the instructive feedback sessions in that they expressed more enthusiasm by clapping and smiling when they received access to preferred items. Thus, incorporating mands (i.e., requests) into training might be beneficial even if instructional sessions take a little longer to complete because they may be more enjoyable for the individual. Conversely, the mand interspersal condition did not enhance training outcomes and was associated with the longest session durations.
An unexpected outcome related to the mand interspersal training sessions was also observed for Kyle's second stimulus set. Correct intraverbal responding was observed at generally high levels (i.e., 70% in the first training session) when training was initiated. Mand responses were also high (i.e., 80%; data not presented). However, both mand and intraverbal responding began to decrease following the first training session. In fact, Kyle met the mastery criteria for this target during the acquisition probes even though intraverbal responding continued to decrease across the training sessions. Interaction effects across conditions may provide one possible explanation for the decrease in correct responding during these training sessions. The incorrect responses emitted during these sessions were related to the stimuli used in other training conditions. For example, Kyle would often respond with the word "music" during these sessions, which was the item used in the Mand TC training condition. Thus, it appears that he was motivated to access both the microphone and music during the interspersal sessions, which may have influenced which response was emitted during each trial. It is important to consider how fluctuations in preference and motivation can influence responding during instructional sessions. A learner's behavior may be motivated by items that are not being used in a given instructional situation. To remedy this problem, practitioners should frequently re-assess preferences to ensure that highly preferred items are being used during instruction. Individuals tend to lose interest in things that are use too frequently or repeatedly as reinforcers.
Although the overall findings are inconsistent with those of previous research utilizing combined training methods (e.g., Arntzen & Almås, 2002; Carroll & Hesse, 1987), results extend previous research in several ways. First, this investigation extends the results of Arntzen and Almås and Carroll and Hesse by evaluating a similar procedure (mand interspersal) while teaching intraverbal behavior. In addition, the current investigation improved on these previous studies by utilizing a more stringent method for assessing mastery. That is, the experimenter probed for mastery outside of training conditions in the current study whereas performance was evaluated only during training sessions in previous studies. It is possible that the different findings obtained were a result of how mastery was probed. Subsequently, it might be important to distinguish true mastery of each individual skill from perceived mastery as a result of the combined training techniques. Practitioners should consider the ultimate goal of instruction. If the targeted skill will not be used in isolation, probing for mastery during training could be warranted. However, mastery should be probed outside of training conditions if the ultimate goal is for isolated skill use. This could help determine whether the individual is able to use the skill under more natural situations and in contexts other than the instructional sessions.
Despite the potential benefits of the mand transfer-of-control procedure, we could only teach a limited number of intraverbals because the responses had to be related to highly preferred items that could be delivered during training. Practitioners will need to consider alternative teaching strategies once basic intraverbal skills are mastered. If used, specific reinforcement should always be paired with nonspecific reinforcement (i.e., praise) when teaching these types of skills. In addition, it may be beneficial to intersperse more complex intraverbals (e.g., answering personal questions) with previously mastered concrete intraverbals. Programmed reinforcement, whether specific or nonspecific, should be faded out to ensure that the behavior eventually comes under the control of the naturally occurring consequences.
Results of the generalization probes suggested that while participants learned to answer questions about items based on two descriptors, they were not likely to be able to use those descriptors later to describe the item. To this end, practitioners should recognize that specific intervention may be needed to teach both types of skills, even though they are seemingly related. Intervention sessions could include multiple opportunities for the individual describe things not present when asked with reinforcement provided for correct descriptions. Other types of generalization should be programmed as well (Stokes & Baer, 1977). Instructional programming for any behavior must include ways to ensure that the skill being taught is used outside of instructional time. Thus, practitioners should consider teaching using multiple examples of the skill, in many different settings, and with many different people cuing the behavior. Naturalistic training (e.g., embedded instruction) might also increase the likelihood that generalized use of the skill will occur.
In sum, results suggest a number of guidelines for practitioners. First, consider the amount of time available for instruction. If time is limited, beginning conversation skills can be taught by presenting the question, waiting for a response, verbally prompting the correct response when needed, and delivering an appropriate consequence for the response. Also, practitioners might consider providing the learner with extra information that will be targeted later (i.e., using instructive feedback). Doing so will not significantly increase instructional time, and the learner may acquire the information without having to be specifically taught. Higher levels of motivation during instruction might also be achieved by combining praise with access to a preferred item that is related to the questions being used during instruction. Although this procedure will only be useful for certain items, early instruction related to conversation skills should focus on things that interest the individual. By including preferred items during instruction, it is also possible that instructional sessions will be more enjoyable for learners and educators.