Learning Channels: The Role of Compound Stimuli in the Emergence of Intraverbal Relations in Children on the Autism Spectrum

Abstract

Teaching tact and intraverbal responses based on function–feature–class to children with language delays can result in the emergence of untrained relational responses. The purpose of this study was to compare the effects of compound stimuli in discriminated operants (i.e., different combinations of hear, see, touch, and taste) on the acquisition of object–attribute relations, on the emergence of untrained attribute–object relations, and on the acquisition and emergence of same–different relations between objects and their attributes. All the participants were on the autism spectrum and between 4 and 12 years old. Participants who did not meet the mastery criterion or show emergent intraverbal responses during initial training trials completed a fluency-based practice phase. Overall results showed that all six participants required fewer trials to meet the criterion in the condition involving compound stimuli (e.g., HearSeeSay plus Touch, Taste, or Sniff) as compared to the HearSeeSay-alone condition. In addition, participants required fewer fluency practice timings in the condition involving compound stimuli to meet fluency aim.

Language and communication skills are crucial in enabling children (including those on the autism spectrum) to navigate their environment more independently (Goldsmith et al., 2007). Sundberg and Sundberg (2011) demonstrated that, for some individuals with autism, the acquisition of basic verbal operants such as manding, tacting, and receptive language may not result in the acquisition of other more complex verbal operants or intraverbal responses. Thus, before programming to teach advanced tact and intraverbal responses to children, it is crucial to teach basic listener and tact vocabulary (e.g., a variety of nouns, verbs, adjectives, prepositions, and pronouns). Once the listener-responding skill of selecting items based on their function–feature–class has been established, a transfer of stimulus control procedure may be utilized to transfer the listener responses to tact and then tact to intraverbal responses. Further, the same function–feature–class relations can be taught as intraverbal reversals (e.g., ask “What do you do with a pillow?” and the child says the function “sleep on a pillow”). Because previous research in the field of applied behavior analysis has shown that verbal operants are functionally independent of each other, multiple researchers have developed and recommended specific training protocols to transfer stimulus control from one response class to another (e.g., Braam & Poling, 1983; Ingvarsson, Tiger, Hanley, & Stephenson, 2007; Luciano, 1986; Partington & Bailey, 1993; Sundberg & Partington, 1988).

Skinner’s (Skinner, 1957) analysis of verbal behavior conceptualized how early language skills can be taught using principles of reinforcement, and how language can further develop from single words to phrases. However, it may not be feasible to teach every exemplar through direct reinforcement contingencies (Sundberg & Michael, 2001). Consequently, there is a need for teaching procedures wherein multiple-exemplar training can be combined with other procedures to teach conditional discrimination that results in the emergence of various untrained verbal responses. Although previous research has shown that training in one verbal operant can sometimes result in the emergence of another verbal operant (e.g., Petursdottir, Carr, Lechago, & Almason, 2008a; Petursdottir, Olafsdottir, & Aradottir, 2008b), few procedures have been tested in teaching and the generalization of advanced intraverbal skills based on categorization. Thus, it is important to synthesize various behavior-analytic theories and applications (e.g., transfer of stimulus control, contextual cues, learning channels, fluency, and rules that allow for concept formation) to select and implement an effective and efficient instructional procedure to build advanced intraverbal repertoires in children diagnosed with autism and other developmental delays.

According to Zentall, Galizio, and Critchfield (2002), “when the stimuli within and between categories vary along relatively simple dimensions (e.g., wavelength, size, brightness), categorization is readily conceived in the same terms as stimulus discrimination and generalization” (p. 238). A category is a class of stimuli that evokes common responses in certain contexts; thus category and stimulus class are often used interchangeably by many practitioners and authors (Zentall et al., 2002). Advanced verbal behavior and intraverbal training can promote concept building and categorization, but responding may fail to generalize across untrained stimuli (Sundberg & Sundberg, 2011). According to Sundberg and Sundberg (2011), the challenges associated with learning intraverbal behavior are partly due to the highly complex nature of the antecedent verbal stimuli that control the intraverbal response. These antecedent stimuli often involve discriminated operants, conditional stimulus control, and compound stimuli (Eikeseth & Smith, 2013). Eikeseth and Smith (2013) explained how compound stimuli, or additional sensory prompts, could play a crucial role in increasing stimulus salience in various ways, especially while teaching advanced intraverbal relations to children diagnosed with autism.

The use of sensory prompts can be discussed within the context of learning channels, a concept introduced by Haughton (Haughton, 1980), who described behavior that forms a sense–action pair (e.g., TouchSay, HearWrite). Multiple sense modalities can be used to present a stimulus (i.e., visual [see], auditory [hear], tactile [touch], kinesthetic [motion], olfactory [sniff], and gustatory [taste]), and responses following in relation to the stimulus can also occur through a wide variety of different action topographies (e.g., pointing, writing, speaking). According to Lindsley (1992), an antecedent stimulus can be presented through various sense modalities, and responses may also occur via multiple sense–action modalities. He proposed the term “learning stream” as a more expansive and inclusive term to explain sense–action pairs that have more than one sense aspect or action, operating either concurrently, sequentially, or both (e.g., HearSeeTasteSay, as in the current study). Likewise, rather than only a single response being evoked, multiple responses may occur, again concurrently, sequentially, or both. Although there is some research on the use of learning channels with children with autism (e.g., Lin & Kubina, 2004; Lindsley, 1992), there has not been a direct analysis of the role of stimuli involving different sensory features (e.g., Hear vs. HearSniff) in teaching categorization and advanced intraverbal skills. Wolf (Wolf, 1963) showed enhanced stimulus control in lab rats when the combination of stimulus components across same (i.e., combinations of different types of light stimuli) and different (i.e., presentation of light and tone stimuli) sense modalities was involved. Thus, learning channels can be explained as compound stimuli in discriminated operants, where multiple sensory components of a stimulus can increase the stimulus salience when teaching the feature–item tact and intraverbal relations. In addition, the use of learning channels involving compound stimuli may prevent the perennial problem of rote intraverbal responding, where a learner emits only one response regardless of differing contextual variables. The current study compared the effects of two different conditions: one in which a HearSeeSay learning channel was used, and one in which an additional learning channel “input” was involved (Hear, Touch, or Taste). The concept of learning channels extends the understanding of Skinner’s (Skinner, 1957) analysis of verbal behavior to include the role of sensory modalities as compound stimuli to teach various attributes.

The current study also included a fluency component to examine the effects of fluency practice on response acquisition. Multiple studies have demonstrated the use of fluency-based training to teach a variety of skills (e.g., Heinicke, Carr, LeBlanc, & Severtson, 2010; King, Moors, & Fabrizio, 2003; Kubina & Yurich, 2012). Typically, fluency practice timings are designed to target accurate responding that will maintain at an optimal rate and promote generative learning (Lindsley, 1995). Practice timings (Binder, 1996), used as a teaching tactic for fluency building, generally result in the critical learning outcomes of performance retention or maintenance, endurance, stability, application, and generativity (Binder, 1996).

The purpose of the current study was to examine whether the addition of sensory prompts of Touch, Taste, and Sniff to instructional trials involving a HearSeeSay learning channel to teach objects and their attributes would (a) impact trials to criterion during training and (b) lead to the emergence of untrained relational responses. Additionally, the study included a fluency-timing component for participants who failed the initial tests of untrained responses in order to assess whether fluency-based practice would strengthen the stimulus control of directly trained relations and facilitate the emergence of untrained responses.

Method

Participants, Setting, and Materials

Six participants were recruited to participate in this study: Adam, a 9-year-old boy; Nela, an 8-year-old girl; Emmanuel, an 11-year-old boy; Larry, a 9-year-old boy; Trey, a 6-year-old boy; and Ahaan, a 5-year-old boy. All participants had a diagnosis of autism, except for Trey, who had received a diagnosis of pervasive developmental disorder–not otherwise specified. Participants were selected based on their scores on the Verbal Behavior Milestones Assessment and Placement Program (VB-MAPP, second edition; Sundberg, 2008). Inclusion criteria were as follows: Participants demonstrated the skill of matching to sample across identical and nonidentical items, demonstrated listener responding by function–feature–class but not through tact and intraverbal responses, had all the basic listener responding and tact responses across 200 nouns, and did not demonstrate any high-frequency or high-intensity problem behaviors (e.g., aggression), as measured by the VB-MAPP barrier assessment (p. 25-29). Informed consent was obtained from all individual participants’ parents/guardians; assent was obtained from all participants.

Sessions were conducted in a small therapy room approximately 9 ft × 11 ft (2.7 m x 3.4 m) in size. The room contained a table and two chairs, where the participant sat across from the experimenter (i.e., Aarti Thakore). Materials included data sheets, a timer, clicker counters, treatment integrity forms, social validity forms, a bucket full of toys, and a box of 32 common objects for testing and training (see Appendix Table 3 for a list of objects).

Table 3.

Four Sets of Objects

Phase I
Set A	Set B	Set C	Set D
(HearSeeSay)	(HearSeeTouchSay)	(HearSeeSay)	(HearSeeTouchSay)
Hard	Hard	Hot	Hot
Rock	Rock	Coffee	Coffee
Block	Block	Soup	Soup
Soft	Soft	Cold	Cold
Cotton	Cotton	Ice	Ice
Pillow	Pillow	Popsicle	Popsicle
(HearSeeSay)	(HearSeeTasteSay)	(HearSeeSay)	(HearSeeTasteSay)
Sweet	Sweet	Clean	Clean
Syrup	Syrup	Teddy bear	Teddy bear
Marshmallow	Marshmallow	Towel	Towel
Sour	Sour	Dirty	Dirty
Pickle	Pickle	Trash/Garbage	Trash/Garbage
Cranberries	Cranberries	Sock	Sock
Phase II
(HearSeeSay)	(HearSeeTouchSay)	(HearSeeSay)	(HearSeeTouchSay)
Sharp	Sharp	Rough	Rough
Pencil	Pencil	Basketball	Basketball
Scissors	Scissors	Pieapple	Pieapple
Blunt	Blunt	Smooth	Smooth
Eraser	Eraser	Cup	Cup
Spoon	Spoon	Apple	Apple
(HearSeeSay)	(HearSeeTasteSay)	(HearSeeSay)	(HearSeeTasteSay)
Crunchy	Crunchy	Dry	Dry
Chips	Chips	Peanut	Peanut
Pretzels	Pretzels	Crackers	Crackers
Chewy	Chewy	Juicy	Juicy
Bagel	Bagel	Grapes	Grapes
Gummies	Gummies	Watermelon	Watermelon

Note. Every participant completed one HearSeeSay condition and one HearSeeTouchSay condition, as well as one HearSeeTasteSay or HearSeeSniffSay condition.

Adam: HearSeeSay (Set A); HearSeeTouchSay and HearSeeSniffSay (Set D)

Nela: HearSeeSay (Set C); HearSeeTouchSay and HearSeeTasteSay (Set B)

Emmanuel: HearSeeSay (Set C); HearSeeTouchSay and HearSeeTasteSay (Set B)

Larry: HearSeeSay (Set A); HearSeeTouchSay and HearSeeSniffSay (Set D)

Trey: HearSeeSay (Set C); HearSeeTouchSay and HearSeeTasteSay (Set B)

Ahaan: HearSeeSay (Set A); HearSeeTouchSay and HearSeeSniffSay (Set D)

Response Measurement

The dependent variables in the study were the participants’ number of correct and incorrect responses in all experimental phases and conditions. A correct response was defined as making a vocal-verbal response that specified the trained feature when given an item, such as saying “soft” when asked “How does a pillow feel?” and saying “pillow” when asked to “tell me something that feels soft.” A response was scored as correct if it matched the responses listed on the data sheet, corresponding to the question asked. For example, during the tacting pretest, the correct response was vocalizing the name of the item when it was held up (e.g., saying “pillow” when the researcher held up a pillow). In the training and testing conditions, when the researcher asked “How does a pillow feel?” and the participant said “soft,” this was counted as a correct response. During the fluency timings, a correct response was counted when the participant vocalized the attribute of the corresponding object (e.g., when the experimenter asked “How does a pillow feel?” the participant said “soft”).

An incorrect response was defined as a response that did not match the response listed on the data sheet, including any vocal-verbal response that did not specify the attribute of the presented item. For example, when the researcher asked “How does a pillow feel?” responses such as “hard,” “pillow,” and “I don’t know” were scored as incorrect. If the participant did not vocalize any response within 3 s of the end of the question, this was scored as incorrect. If the participant’s response was technically correct but did not match the one listed on the data sheet (e.g., saying “smooth,” rather than “soft,” when asked “How does a pillow feel?”), then the experimenter recorded that response and said “Yes, a pillow feels smooth, and ..?” then waited an additional 3 s for the participant to respond.

All correct vocal-verbal responses across all training and testing conditions were scored as 1, and incorrect vocal-verbal responses were scored as 0. During the fluency timings, the experimenter used two different hand counters to track the number of occurrences of correct and incorrect responses, and values were recorded on a data sheet, as well as on a successive timings standard celeration chart (Timings Chart, Behavior Research Company, Laurence, KS 66103-3351).

Interobserver Agreement and Treatment Integrity

Interobserver agreement for dependent variables was calculated for between 47.6% and 91.1% of sessions for all participants across all conditions using a trial-by-trial formula, with a resulting score of 100% agreement across all sessions. An observer filled out a treatment integrity checklist during 92% of sessions, and a second observer also scored treatment integrity for 33% of sessions. Both treatment integrity and treatment integrity interobserver agreement scores were 100% for all sessions.

Experimental Design

A single-subject alternating-treatments experimental design was used, with a separate set of instructional stimuli in each condition. Two instructional conditions were alternated within each experimental session, and the procedure was replicated with an additional stimulus set for each condition. The order of conditions within each session alternated each day; for example, on one day a participant may have completed trials specific to the HearSeeSay condition first, followed by the HearSeeTouchTasteSay condition, and on the next day, the participant completed the conditions in the reverse order, HearSeeTouchTasteSay followed by HearSeeSay. Stimulus sets were counterbalanced across participants. Once the participant met the criterion with the first stimulus set in a given condition, the same procedures were repeated with a second stimulus set for that same condition. Within each condition, participants completed pretest, training, and posttest phases assessing both directly trained and derived relational responses specific to the objects involved in that condition. Participants also completed fluency timings if they did not meet the criterion during testing conditions, as described in the procedure section. Every participant completed one HearSeeSay condition and one HearSeeTouchSay condition, and one HearSeeTasteSay or HearSeeSniffSay condition (see Appendix Table 3).

Procedures

Training and Testing Relations of Sameness Across Arbitrary Stimuli

This screening procedure assessed whether participants could demonstrate the component skills of arbitrary applicable relational responding based on relations of sameness, which were prerequisite skills for the subsequent instructional conditions. Participants were directly trained A-to-B and A-to-C stimulus relations, and then probes were conducted for B-to-A, C-to-A, C-to-B, and B-to-C relations. During training, pictures of abstract shapes were used for each class; the experimenter provided verbal praise contingent on correct responses and followed an error correction procedure contingent on incorrect responses. The experimenter trained the relations of sameness across B1 and B2, then A2 and B2, and finally A3 and B3. Then, the experimenter repeated the same procedure to train a relation of sameness between A1 and C1, A2 and C2, and A3 and C3. Finally, the experimenter tested for equivalence relations by probing for the following responses using Sets 1, 2, and 3: presenting Item C1 as the sample and Item B1 as the comparison, along with two unrelated pictures (e.g., B2 and B3). During testing, the experimenter did not provide feedback for correct or incorrect responding. If the participants responded correctly to at least three of four trials on the aforementioned derived relations, they continued on to the subsequent phase. If the participant did not respond correctly to at least three of four trials for each probe type, then the experimenter retrained the base relations, and then retested. All six participants demonstrated equivalence, the necessary prerequisite skill to continue participating in the study.

Object Tacting Pretest

The purpose of the object-tacting pretest was to determine which objects the participant could correctly tact the name of, but would not say features of those objects. There were a total of 32 objects across the four sets of object–attributes (see Appendix Table 3), with eight objects in each trial block. That is, the experimenter conducted a block of eight testing trials across four testing trial blocks. Next, the experimenter held up one object at a time and asked, “What is it?” (an arbitrary relation: object ⟶ name). During each trial, the experimenter waited at least 3 s for the participant to name each object, one object at a time. There were no programmed consequences for correct or incorrect responses. After every four trials, the experimenter provided verbal praise for participating.

If the participant incorrectly named an object, the experimenter then trained that spoken response before beginning the next block of tact pretest trials. Training procedures were identical to testing trials, except that now the experimenter provided vocal feedback (e.g., “That’s correct, it’s a pillow,” or “Try again.”) and then repeated the question and stated the correct response. Training continued until the participant responded with 100% accuracy across three consecutive trial blocks. The experimenter then immediately presented four testing blocks (in the same manner as the object-tacting pretest described previously) with no feedback contingent on responding. All participants responded with 100% accuracy across all four testing trial blocks.

Object–Attribute Relation Pretest

Figure 1 displays all relations between stimuli that were trained and tested over the course of the study. During the object–attribute relation pretest, the experimenter tested the object–attribute relations for all 32 objects across the four sets of objects to be used in training. For example, the experimenter held up a pillow and asked the participant, “How does a pillow feel?” (arbitrary relation: see an object and hear the name of the object ⟶ attribute). The experimenter then waited 3 s for the participant to respond, and there were no programmed consequences for correct or incorrect responses. The experimenter conducted a total of three testing trials for each of the 32 objects, with all trials delivered in an interspersed order. If the participant responded incorrectly during at least two of the three testing trials for all objects, then the experimenter began the object–attribute training condition using those same objects.

Fig. 1.

Trained and Untrained Relations During HearSeeTouchTasteSniffSay (Multiply Controlled Tact) Conditions. Note. Solid lines indicate directly trained relations, and dotted lines indicate untrained relations

If the participant correctly said the attribute of any objects in the set on at least two of three trials, then those objects were removed and another set of objects was presented, to ensure that there were 32 objects that the participant could correctly say the name of, but could not otherwise say the object’s attributes. Then the participants completed both tacting and object–attribute pretesting trials for any additional objects as needed. For example, if a participant correctly named the attribute of two objects, then those two objects were excluded from the study and the experimenter completed tacting and object–attribute pretesting trials for two new objects. If the participant responded incorrectly when asked to name each object’s attribute, these objects were included in the group of 32 total objects used in subsequent phases. Each participant completed procedures involving eight objects in the HearSeeSay training condition and eight objects in the HearSeeTouchTasteSniffSay training conditions, and then both conditions were replicated with the remaining eight objects used for each.

HearSeeSay Object–Attribute Relation Training Condition

During this condition, the experimenter randomly selected one of the four sets of objects (either Set A, B, C, or D) for each participant; each included 8 of the 16 objects selected from the pretests. The experimenter then trained each exemplar using a progressive prompt delay across three steps. The progressive prompt delay was similar to the delayed prompting procedure described by Braam and Poling (1983). In the current study, the experimenter gradually increased the time delay across 15 training trials, where, during the first 5 trials, the experimenter delivered an immediate prompt with a 0-s prompt delay. This was followed by a 3-s prompt delay for the next five training trials. Once the participant emitted correct responses, the experimenter then increased the prompt delay to 5 s across the last five training trials. All the targets during the training trials were presented in a random order to avoid sequencing effects.

Once the participant met the criterion for each object, the experimenter interspersed trials with all objects. Each session, the experimenter presented a total of 15 training trials per object, for a total of 120 trials during one condition, in a given session (an arbitrary relation: object + name of object ⟶ attribute, as in the previous phase). The criterion to complete the aforementioned training condition was 90% accuracy across the interspersed trials of all participants, and then the participant moved on to the testing phase for that set of objects.

HearSeeTouchSay Object–Attribute Relation Training

This training condition was conducted in the same manner as the HearSeeSay training condition described previously, except that the experimenter handed the object to the participant during each trial. For example, the researcher said, “Here, touch this pillow and feel it. How does a pillow feel?” so the participant could touch and feel the object before saying its attribute, such as “soft” (an arbitrary relation: object + texture of the object + name of the object ⟶ attribute). The participants were asked to touch and feel the objects for 3 s, and then the experimenter would start a new trial. The participant was asked to touch and feel the objects across all the training trials under this condition.

HearSeeTasteSay Object–Attribute Relation Training

The experimenter used the same training steps as in the HearSeeSay training condition described previously, except that in this phase the experimenter gave a small piece of the object and then said, for example, “Here, taste this. How does a lemon taste?” so that the participant tasted it before responding (an arbitrary relation: object + taste of the object + name of the object ⟶ attribute). If the participant refused to taste the food, then the experimenter explained, “You can spit it out in a paper towel or a trash if you don’t like it.” All participants were willing to taste the food and spit out the food they did not like. For example, during the HearSeeTouchTasteSniffSay condition, Trey said that he did not like to taste cranberries and pickles, and he spat them out after his initial tasting. Also, Adam did not like the taste of cranberries and asked if he could spit them out.

HearSeeSniffSay Object–Attribute Relation Training

The experimenter used the same training steps as in the HearSeeSay training condition described previously, except that in this phase the researcher held the object close to the participant’s nose and said something like “Here, smell this. How does the sock smell?” and prompted the participant to sniff or model how to sniff, so that the participant could sniff the object before responding (an arbitrary relation: object + smell of the object + name of the object ⟶ attribute). If the participant tried to touch the object, then the experimenter physically blocked the participant or removed the item and said, “Please, do not touch the item. Just sniff.”

Testing for Attribute–Object Relations

Once a participant met the criterion in a training condition described previously, the experimenter then tested the eight attribute–object relations across the two sets of objects that were directly trained (see Fig. 1 and Table 1). This phase was conducted in the same way as the HearSay object–attribute pretest (described previously), except that the participants were tested on mutually entailed intraverbal relations (arbitrary relation: attribute ⟶ name of the object). In this condition, there were no programmed consequences given for correct or incorrect responses. The experimenter conducted a total of three testing trials for each attribute–object relation, and all trials were interspersed. If the participant scored below 90% accuracy on the test of attribute–object relations, then they began the object–attribute fluency-timing phase. If the participant responded with 90% accuracy or above, then they completed the attribute relation of sameness and opposition pretest.

Table 1.

Each Experimental Phase, the Corresponding Verbal Operant(s), and Sample Trials

Fluency Timings Across Object - Attribute Relations

The fluency aim for the HearSeeSay (part tact and part intraverbal) fluency timings in the current study was set at 15 correct responses per minute (see the fluency aims in Kubina & Yurich, 2012). The experimenter put all the objects from a stimulus set out on the table and conducted three practice timings per session. During the practice timing, the experimenter pointed to each object in an array and said “[Name of object] feels,” one object at a time for each object, and if the participant failed to emit the correct vocal response within 2 s, the experimenter pointed and said the next object on the list (arbitrary relation: hear the name of the object and see the object ⟶ say the related attribute). The order of trials was interspersed, and the experimenter provided vocal feedback (e.g., “yes” or “correct” for correct response and “next” for the incorrect response) following each response. The experimenter also gave feedback at the end of each practice timing (e.g., “Great, you got 15 correct responses,” or “You got 5 correct and 5 incorrect responses in a minute. Let’s try again.”). The experimenter then reviewed the incorrect responses with the participant and practiced each response using an additional echoic prompt (i.e., vocal imitation, where the experimenter said the correct responses and waited for the participant to imitate the correct responses vocally) once before beginning the next practice timing.

If the participant failed to meet the aim during any of the three timings during Day 1, then the experimenter conducted fluency timings again on the following day and continued until the participant met the fluency aim. The number of timings completed for a given stimulus set ranged from 1 to 15 (see Table 2 for the total number of fluency practice timings each participant completed during each condition). After the participants met the aim, they again completed the test for attribute–object intraverbal relations (an arbitrary relation). After the practice timings, if the participants responded correctly on less than 90% of testing trials for attribute–object relations, then the experimenter directly trained the attribute–object relations using the same prompt-delay procedure that was used to train object–attribute relations.

Table 2.

The number of correct responses per minute during 1-min fluency practice timings for all participants. Timings had a fluency aim of at least 15 correct responses per minute

	Object – Attribute Set 1		Same – Different Attribute Set 1		Object – Attribute Set 2		Same – Different Attribute Set 2
	HearSeeSay	HearSeeTouch TasteSniffSay	HearSeeSay	HearSeeTouch TasteSniffSay	HearSeeSay	HearSeeTouch TasteSniffSay	HearSeeSay	HearSeeTouch TasteSniffSay
Adam	10, 14, 16	14, 14, 15	13, 14, 11, 14, 16	11, 14, 16	12, 14, 16	--	11, 13, 16	--
Nela	8, 10, 10, 14, 13, 13, 14, 12, 15, 16	13, 14, 16	10, 14, 12, 14, 14, 16	14, 14, 16	12, 12, 13, 14, 14, 16	12, 16	14, 16	--
Emmanuel	10, 13, 12, 13, 14, 14, 14, 14, 16	12, 14, 16	12, 13, 12, 14, 16	13, 14, 16	12, 11, 10, 14, 14, 13, 13, 12, 13, 14, 14, 16	N/A	N/A	--
Larry	14, 14, 16	--	10, 12, 12, 12, 14, 16	14, 13, 16	12, 10, 12, 13, 16	13, 14, 16	13, 14, 14, 16	--
Trey	13, 12, 12, 14, 14, 16	14, 16	11, 12, 13, 12, 13, 14, 14, 14, 13, 14, 14, 13, 12,14,16	13, 14, 16	10, 13, 14, 14, 15	14, 13, 16	13, 13, 16	--
Ahaan	13, 13, 16	14, 14, 16	11, 12, 12, 10, 13, 16	14, 14, 16	12, 13,16	16	10, 11, 11, 13, 14, 16	16

Note. Timings had a fluency aim of at least 15 correct responses per minute, and dashes represent that fluency timings were not required for that participant in that condition.

Same–Different Relation Pretest

This testing condition was targeted arbitrary relations, where the experimenter conducted a pretest using the HearSay learning channel (i.e., pure intraverbal fill-ins without any visual prompts) across all the attributes from the object–attribute sets that were directly trained. This pretest was the same as the object–attribute pretest, except that the experimenter tested for the relation of sameness and difference across all the attributes in each set of objects (an arbitrary relation: hear the attribute ⟶ say the attribute that is the same or not the same). For example, the experimenter said, “Soft is not same as” and waited 3 s for the participant to say “hard,” or the experimenter said, “Soft is same as” and waited 3 s for the participant to say “soft.” (To keep the language simple and concise for the participants, the experimenter used “not same” instead of “different.”) During the pretest, if the participants responded correctly on at least 90% of trials, they advanced to the testing for the relation of sameness and distinction across objects. If the participant responded with below 90% accuracy, then they completed the training the same and different relations across the attributes.

Training the Same and Different Relations Across the Attributes

During this phase, the experimenter trained relations of sameness and difference between the two attributes across the trained set of objects (hear the attribute ⟶ say the attribute, an arbitrary relation). The experimenter used an echoic prompt to establish the rule of “same” and “not same” across each attribute. The participants completed training trials on the relation of sameness and difference (i.e., “not same”) across attributes (e.g., soft is not same as hard, and soft is the same as soft) using a HearSay learning channel, across all four attributes that were directly trained in the previous training phases. The experimenter conducted a total of 10 trials for each attribute term, and these trials were interspersed. During a trial, the experimenter said, for example, “Soft is the same as ..?” and trained the participant to say “soft.” The experimenter used the same three-step prompt-delay training procedure used to train object–attribute relations, but with modified instructions. For example, the experimenter stated, “Soft is not same as ..?” and trained the response “hard.” The participant continued the training until they responded with at least 90% accuracy on each relation.

Testing for the Emergence of Same and Different Relations Across the Related Objects

This testing phase was conducted in the same manner as the attribute–object testing phase described previously, except that the participants were tested on the emergence of untrained relations of same and different across all objects within the specific set of trained objects, through a HearSay learning channel (hear the name of the object ⟶ say the name of the object, an arbitrary relation). For example, when the experimenter said, “Tell me something that feels same as a pillow” and the participant said “cotton,” “pillow,” or “cotton and pillow,” this was counted as a correct response; if the participant said “pillow,” then the experimenter said, “Yes, pillow and what else?” Similarly, the relations across objects were tested (e.g., soft and hard, crunchy and chewy, dry and juicy). There were no programmed consequences for correct and incorrect responding. If the participants responded correctly on at least 90% of the trials, then their participation in the current condition for the targeted stimulus set was complete. If the participant responded correctly to less than 90% of trials, then the experimenter conducted fluency timings for the same and different attribute relations.

Fluency Timings Across the Same and Different Attribute Relations

This procedure was the same as the object–attribute fluency timings described previously, except that the targets were sets of the related same and different attributes (arbitrary relation: see the object and hear the attribute ⟶ say the attribute that is the same or not the same). For example, the experimenter said, “Soft is the same as,” and the correct response was “soft,” and the experimenter said, “Soft is not same as,” and the correct response was “hard.” The aim was to respond with at least 15 correct responses and no more than 1 incorrect response in 1 min. The experimenter conducted a maximum of six fluency timings per day until the participant met the aim. Finally, the participant completed the testing for attribute relations of sameness and difference (described previously). When the experimenter said, “A pillow feels the same as” and the participant said “blanket” instead of “cotton” (as per the data sheet), then the experimenter said “Yes, blanket and..?” and recorded all the object names that shared the relation of sameness with the pillow (i.e., feels soft as a pillow). However, for consistency in data collection, the experimenter counted the response as a correct response and assigned a score of 1 when it included the target response (e.g., “pillow”) as well.

Results

Figures 2 and 3 display data for all six participants across all experimental conditions for stimulus Sets 1 and 2. Because responding was similar across participants, what follows is a detailed summary of one participant’s responding, along with an overall summary of responding for all participants. During the HearSeeSay condition, Adam required a total of 44 trial blocks to meet the mastery criterion across all the objects in Set 1. The first five training trial blocks involved a 0-s prompt-delay procedure, where the echoic prompt was delivered immediately. These trial blocks are not displayed on the graph, as there was no opportunity for independent correct responses to occur. During the initial pretest of HearSeeSay object–attribute relations, Adam showed 0% accuracy. Thus, he completed training for those eight object–attribute relations. Once Adam met the mastery criterion of 90% correct responding and his performance remained stable across three consecutive sessions, he completed the attribute–object relations testing phase. The data in the attribute–object testing phase show performance accuracies of 63%, 50%, and 50%, respectively. As the responding during attribute–object testing was below 90%, the experimenter introduced the fluency practice timings phase. Adam completed a total of three fluency practice timings across all trained object–attribute relations. The one data point on the graph depicts his accuracy on the final fluency practice timing (see Table 2 for all the other fluency practice timings). Once Adam met the fluency criterion of 15 correct responses per minute, he completed the attribute–object testing phase again. Following the fluency practice timings, he showed 100% correct responses on three consecutive testing trial blocks. Next, Adam completed a pretest phase for same–different relations for the same attributes targeted in the previous training. Adam’s performance showed 0% accuracy, so he began training on same–different attribute relations in the HearSeeSay condition. Once Adam met the criterion of 90% across three consecutive trial blocks, the experimenter then presented testing trials for same–different relations across objects. Adam scored 50%, 63%, and 50% across three consecutive testing trial blocks, below the 90% criterion, so he completed fluency practice timings for same–different relations across attributes. Following the fluency practice timings, he responded with accuracies of 100%, 88%, and 100%, respectively, on the attribute–object relations test.

Fig. 2.

Adam’s, Nela’s, and Emmanuel’s Percentage of Correct Responses across All Training and Testing Phases for Both HearSeeSay and HearSeeTouchTasteSniffSay Conditions. Note. The dotted lines represent pretest and posttest trial blocks within each training phase. Closed filled circles indicate HearSeeTouchTasteSniffSay condition, open squares indicate HearSeeSay condition; Each phase is abbreviated as following: O = object; A = attribute; FPT = fluency practice timing; S-D O= same and different relations across objects; S-D A = same and different relations across attributtes

Fig. 3.

Larry’s, Trey’s, and Ahaan’s Percentage of Correct Responses across All Training and Testing Phases for Both HearSeeSay and HearSeeTouchTasteSniffSay Conditions Note. The dotted lines in the graphs represent pretest and posttest trial blocks within each training phase. Closed filled circles indicates HearSeeTouchTasteSniffSay condition, open squares indicate HearSeeSay condition; O = object; A = attribute; FPT = fluency practice timing; S - D O = same and different relations across objects; S - D A = same and different relations across attributes

During the HearSeeTouchSniffSay condition, Adam reached the mastery criterion of 90% correct responding after completing a total of 35 training trial blocks. Adam did not show derived attribute–object reversal responses and required fluency practice timings of object–attribute relations, after which he responded with 100% accuracy for attribute–object reversal responses. Next, the experimenter trained Adam on same–different relations across the eight attributes in the HearSeeTouchSniffSay condition. The data show that Adam required 10 trial blocks to reach the criterion of 90% across three consecutive trial blocks. Again, he did not show the emergent same –different responses across objects and was given the fluency practice timings for same–different relations across attributes. Following fluency practice, Adam scored 100% across three consecutive sessions. He required fewer training trial blocks to reach the mastery criterion for object–attribute relations and same–different relations in the HearSeeTouchTasteSay condition compared to the HearSeeSay condition (for fluency-timing data, see Table 2). However, he did require the fluency practice timings for both conditions to meet the accuracy criterion of 90% or above across untrained relations. During the HearSeeSay condition for Set 2, Adam required a total of 33 trial blocks to reach the mastery criterion of 90% correct responding during the object–attribute training phase. Even for Set 2, Adam required fluency practice timings to show the derived attribute–object relations with 100% accuracies during the testing trial blocks. The experimenter then presented the same–different pretest across attributes from Set 2. Adam responded with 100% accuracy during the pretest; thus, the experimenter conducted the same–different test across related objects, on which Adam scored 50%, 38%, and 63% accuracy, respectively. As a result, he completed fluency practice timings for same–different attributes, after which he demonstrated 100% accuracy across three testing trial blocks for same–different relations across objects related to trained attributes. During Set 2, Adam required fewer training and testing trials for the HearSeeTouchTasteSniffSay condition compared to the HearSeeSay condition. However, he showed emergent responses during HearSeeTouchSniffSay without fluency practice timings.

There were several differences in responding across participants. Nela did not show the emergence of derived attribute–object relations and same–different relations across the related objects, for either Set 1 or Set 2. Thus, she received fluency practice timings for both the sets (see Figure 4). However, she required fewer fluency practice timings during the HearSeeTouchTasteSay condition compared to the HearSeeSay condition for both Sets 1 and 2. Emmanuel also did not show emergent attribute–object relations and same–different relations across the objects during Set 1 and, thus, required fluency practice timings in both conditions. However, he only received fluency practice timings for attribute–object intraverbal relations, as he showed emergent same–different relations across objects following the same–different training across attributes for Set 2. Again, he required fewer fluency practice timings during the HearSeeTouchTasteSay condition compared to the HearSeeSay condition Figs. 5 and 6.

Fig. 4.

Successive Timings Standard Celeration Chart Displays Number of Correct Responses per Minute During Nela’s Fluency Practice Timings for Both Attribute–Object Relations and Same–Different Relations Note. Participants only completed fluency timings after failing tests of derived relational responding. The fluency criterion was 15 correct responses per minute and is indicated by the aimstar A. Numbers indicate celeration values (i.e., change in frequency per week) for each set of timings data

Fig. 5.

Total Number of Correct Responses Completed in Each Condition Note. The top panel displays correct responses for the attribute–object intraverbal relations test and the bottom panel displays correct responses for the same–different relations across objects test across all participants for stimulus Sets 1 and 2. The grey bars indicates HearSeeSay condition, and the black bars indicate HearSeeTouchTasteSniffSay condition

Fig. 6.

Total Number of Fluency Practice Timings Completed in Each Condition Note. The left panel displays attribute–object relations, and the right panel displays same–different relations across objects

Larry also completed the HearSeeTouchSniffSay condition with fewer training and testing trials compared to the HearSeeSay condition for Sets 1 and 2. Similar to Set 1, he required fluency practice timings both for attribute–object reversal responses and same–different relations across objects during the HearSeeSay condition for Sets 1 and 2. However, during the HearSeeTouchSniffSay condition for Set 2, he showed emergence of same–different relations across the related objects without fluency practice timings across the attributes. Similar results were observed for Trey and Ahaan as well.

Participants’ parents, teachers, and paraprofessionals filled out a social validity questionnaire. All 12 individuals reported that they noticed an increase in the participants’ overall descriptive language, in that participants were using the attributes targeted in the study to describe what they were eating or why they did not like certain items. One therapist reported that “Yes, when asked ‘How does it taste or feel?’ he responds using adjectives appropriately.” Another teacher reported that “he enjoyed working in this study and has benefited.” One parent reported that “I have definitely noticed his use of adjectives now at home.” One therapist noted that Nela, Emmanuel, Larry, and Trey now “sometimes use adjectives to describe something they smell or touch during activities.” At the end of the study, the experimenter also asked each participant if they liked or did not like tasting, touching, and/or smelling things they learned. Five out of six participants said they liked to taste, touch, and/or sniff items.

Discussion

The purpose of the current study was to compare the effects of training phases with different combinations of learning channels that involved additional discriminative control on participants’ acquisition of intraverbal relations and emergence of novel intraverbal relations related to objects and their attributes. Overall, results showed that all six participants required fewer training trials to meet the criterion in the condition involving an additional learning channel input (HearSeeTouchTasteSay) as compared to the HearSeeSay-alone condition. In addition, when participants received additional practice in establishing the trained object–attribute relation using fluency practice timings, they showed an increase in the number of untrained relations. Thus, the results of the current study illustrate the importance of including contact with the environment through an additional learning channel input when building tacting and intraverbal repertoires related to objects and their attributes.

The current study may provide some insight into the role of multiple learning channels as compound stimuli. Results support the assertion that combinations of sensory prompts, along with verbal discriminative stimuli, might have increased stimulus salience and thus enhanced the verbal stimulus control across complex tact and intraverbal object–attribute relations. These findings are consistent with those demonstrated by Wolf (1963), Wolf (1963) in his study discussed how the summative effect of combined CSs (Conditioned Stimuli) reported by Hull (1943), Pavlov (1960), and (Kimble, 1961). The similar effect was found in the experiment done by Wolf (1963), where the result showed “combine S^DS” (p. 346). This enhanced stimulus control occurred during the combination of stimulus components rather than during the simple unitary stimulus condition. Similar to the findings in previous studies on the transfer of stimulus control (e.g., Braam & Poling, 1983; Ingvarsson et al., 2007; Luciano, 1986; Perez Gonzalez, Garcia-Asenjo, Williams, & Carnerero, 2007;Partington & Bailey, 1993; Petursdottir et al., 2008a; Petursdottir, Olafsdottir, & Aradottir, 2008b; Sundberg & Partington, 1988) and reinforced multiple-exemplar training (e.g., Berens & Hayes, 2007), the current study showed how transfer of stimulus control and functions across multiple exemplars involving reinforcement of relational responses can facilitate the acquisition of untrained tact and intraverbal responses. In the current study, because the objects were displayed on the table during object–attribute relational training, same–different training across attributes, and fluency practice timings, responding during these conditions was likely multiply controlled (i.e., both tact and intraverbal), allowing for the emergence of untrained intraverbal responses during testing for the emergence of attribute–object relations. Additionally, adding multiple discriminative stimulus control using learning channels potentially demonstrated convergent control over a multiply controlled tact, and then extended that convergent control to the intraverbal, thus highlighting the role of multiply controlled tacts on the acquisition of advanced intraverbal responses and reversals. Although results of the current study are consistent with previous research on concept formation (e.g., Zentall et al., 2002) in terms of combining various stimuli in teaching some concepts via categorization, there is a need for future research to assess these components separately to teach various tact and intraverbal relations.

Two participants, Adam and Ahaan, showed generalized responding across objects. For example, they correctly vocalized the names of other objects in the environment when asked “Tell me something that feels hard,” in addition to vocalizing the names of objects used during training phases. Their responding also showed generalization across stimulus Sets 1 and 2. For example, when asked “How do scissors feel?” in Set 2, both participants responded “hard,” which was correct, but not the targeted response “sharp.” Ahaan, in particular, showed multiple occurrences of generalized responding across both stimulus sets. For example, when asked “How do chips taste?” he would say “crunchy” and also “dry,” and when asked “How does watermelon taste?” he responded “sweet and juicy.” He had been trained on the object–attribute relations of marshmallows and syrup as being sweet during Set 1, and his responding showed generalization of the attribute “sweet” across the objects in Set 2 (e.g., watermelon) along with the trained object–attribute relation (e.g., watermelon–juicy). Similarly, he completed training on object –attribute relations for chips and pretzels as crunchy, and peanuts and crackers as dry, and he responded with the attribute “dry” when presented with chips and pretzels, in addition to the targeted response “crunchy.” This study did not directly assess or train multiple attributes of the instructional stimuli; rather, only one attribute was related to each stimulus during training in terms of sameness. In a more naturalistic setting, one would probably train multiple attributes at once. Therefore, future studies could assess the effects of instructional procedures involved in teaching more than one attribute using learning channel inputs corresponding to each attribute. Parent and teacher responses on the social validity form also provided some evidence of generalization beyond the experimental context. For example, in answers given on the social validity form, Larry’s teacher said that he began to use the words “soft” and “hard” to describe his toys. However, the current study did not formally assess the impact of the intervention on skill generalization and maintenance, an area for future research.

The current study did allow for examining the role of echoic prompts in future responding in upcoming phases (i.e., the echoic-to-tact transfer or echoic-to-intraverbal transfer) as other similar equivalence studies have shown. Echoic prompts were delivered at a similar rate within both the HearSeeSay and HearSeeTouchTasteSniffSay conditions. Overall, participants in the current study required less echoic prompting during the HearSeeTouchSniffSay condition as compared to the HearSeeSay condition. However, the current study did not directly compare the number of echoic prompts required across the two conditions. Thus, future studies could assess the number of echoic prompts required in training trials across different learning channel conditions, as well as the role of echoic prompts used during fluency practice timings, in light of the fact that fluency timings often provide increased opportunities to practice, and hear, correct responses as compared to other instructional activities. Although participants did not receive formal listener-responding training as part of the current study, as it was a prerequisite, it likely occurred in some form when they were learning the verbal responses associated with the more abstract (or less salient) conditions, such as adjectives (soft, hard, etc.), which requires training various stimuli that may be composed of very different tacts versus a tact response class in which stimuli are more similar in some way. This may have played a role in participant responding in subsequent phases (i.e., the echoic-to-tact transfer or echoic-to-intraverbal transfer).

The experimenter also observed that three out of the six participants engaged in low-magnitude verbal behavior strategies such as self-echoics, which Lowenkron (Lowenkron, 1984; Lowenkron, 2006) explained as the role of joint control in stimulus selection, conditional discrimination, word–object bidirectionality (Petursdottir, Olafsdottir, & Aradottir, 2008b), selection-based autoclitics, and generalized responding. For example, during the training trial blocks for object–attribute relations, the researcher observed both Adam and Ahaan vocalizing the first letters of the objects during trials; the experimenter asked, “How does a pillow feel?” and the participant said, “P . . . Pillow feels soft,” and, similarly for cotton, “C . . . Cotton feels soft.” Then, during attribute–object testing trials, when the experimenter stated, “Tell me something that feels soft,” the response would be “P, C . . . Pillow cotton.” This joint control of vocalizing letters with the object names could have facilitated the emergence of untrained relations during initial testing trial blocks for these two participants, and may be similar to the common practice of using acronyms as mnemonics in everyday life to remember lists of things or related concepts. Thus, future studies could evaluate the role of this type of joint control strategy and additional participant vocalizations in the acquisition of object–attribute intraverbal relational responding.

In the current study, four out of six participants asked to touch or taste foods or objects during the HearSeeSay condition training trials following exposure to HearSeeTouchTasteSniffSay training trials. Thus, the use of multiple learning channels may be the preferred way for some children on the autism spectrum to learn object features, as appeared to be the case in the current study, and an instructional history with additional learning channels (whether formal or more naturalistic) may lead to increased contact with environmental stimuli using those same modalities. Although further research is needed on the topic, we assert that one important consideration in carrying out this type of instructional procedure is the degree to which a child may demonstrate over- or undersensitivity, or approach or avoidance related to the stimulus attributes, including sound, taste, texture, and so on, and how this may relate to what learning channels are preferred by the learner and promote engagement and positive rapport between the therapist and child. Thus, the use of different learning channel inputs is part of a natural learning process, it may make learning more enjoyable, and contact with the relevant taste, smell, and texture of an instructional stimulus may promote acquisition and generalization of instructional targets involving that stimulus. Although the use of additional learning channels during instruction could potentially be distracting for some learners, it may also facilitate on-task behavior for other learners and allow them to contact multiple stimulus properties that they can “relate to” the arbitrary stimulus (i.e., the name of the item). Moreover, the addition of sound, taste, texture, and other stimulus attributes is crucial to make sure the learner can respond both verbally and via the targeted sense rather than through rote learning.

In the current study, the experimenter observed that, for five of six participants, similar lengths of time were required to present a trial block in the HearSeeSay condition as compared to the HearSeeTouchSniffTasteSay condition. Trey was the exception, as he was sensitive to tasting anything that had a strong flavor, and thus he was resistant to try some edible items in the study (e.g., pickles, cranberries, syrup). However, because the researcher offered him the option to spit out the food after taking a bite, he agreed to participate and tasted the food during HearSeeTasteSay trials. Thus, adding the additional step of allowing the participants to spit out the food that is less preferred after taking a bite could be a strength of the procedure in that it allows for learners to engage in the task even with nonpreferred food. However, it may also present a potential limitation, if allowing a learner to spit out food during instructional trials leads to feeding issues outside the instructional context, such as during mealtimes at school and at home. Although no behavioral issues emerged with participants in the current study, this is a consideration for future studies.

Additionally, it is worth noting that, at least for Trey, some of the foods may have had an aversive function, the impact of which was not directly assessed in the current study. During the HearSeeTasteSay condition, in Phase 1 Trey requested to spit out sour stimuli (e.g., cranberries and pickle), and he required a total of 54 trials to meet criteria compared to 25 trials during Phase 2, where he did not request to spit out the offered stimuli. However, this is a very small sample to be able to derive any conclusion. Previous research has shown aversive stimuli and negative reinforcement contingencies can have a rapid and strong effect on future responding (see Magoon & Critchfield, 2008). Therefore, future studies could examine any difference in how quickly a participant learns relations involving more pleasant or neutral stimuli as compared to those that function as aversive stimuli. As a general practice, and in situations involving aversive stimuli in particular, it is critical that experimenters and clinicians attend to participant assent and assent withdrawal.

Fluency practice in the current study allowed for repeated practice, and as a result, the trained responses became more efficient over time. As previously discussed, fluency-based practice can lead to key learning outcomes, including application and generativity (Binder, 1996), both of which were demonstrated by participants’ performances on the tests of emergent intraverbal relations and generalization of responding to novel stimulus sets in later phases of training. However, the current study did not directly assess the relationship between fluency and the emergence of untrained relations, it was used just to provide additional practice across the trained relations when participants had failed tests of emergent intraverbal relations. Therefore, future research should explore (a) the relation between meeting specific fluency aims and the emergence of untrained relational responding and (b) the resulting celeration values when comparing different learning channel instructional approaches. It is important to note that the current study’s fluency-timing procedures placed a procedural ceiling on responding in that the participants could not make a correct response until the researcher had presented an instruction. Therefore, future studies should modify the fluency timings to reflect a free-operant format (e.g., a SeePointSay or SeeSay worksheet that displays the objects and contextual cues of “same” and “different” or “not same”).

It is also important to mention that during instructional trials, some of the attributes used were true opposites of one another (e.g., soft and hard) but some of them were not (e.g., crunchy and chewy, dry and juicy), so the contextual cue “not same” was used. The cue “not same” was used to keep the instructional sequence consistent with “same” and “not same” and simpler for the participants. Future research can include training and testing using cues of “same” and “opposite” involving a continuum that includes other possible “difference” relations. Finally, future research can extend this type of instructional procedure to establish other types of relations (e.g., hierarchical, comparison). Research in the area of equivalence-based training and derived relational responding has demonstrated that multiple-exemplar training allows for concept formation and categorization, and that contextual cues can promote the emergence of untrained (derived) responses in relation to the stimulus classes without directly training the relations between every single stimulus (Hayes, Fox, Gifford, Wilson, Barnes-Holmes,& Healy, 2001; Berens & Hayes, 2007). This has been demonstrated with mand, tact, and intraverbal instruction with young children (e.g., Hayes, Barnes-Holmes, Roche, & Eds, 2001a; May, Hawkins, & Dymond, 2013; Murphy, Barnes-Holmes, & Barnes-Holmes, 2005; Petursdottir, Olafsdottir, & Aradottir, 2008b).

The current study added to the body of research on the effectiveness of different learning channel combinations (i.e., adding Touch, Taste, and Sniff inputs to a HearSeeSay instructional procedure) and fluency practice timings in promoting the acquisition of object–attribute relations for children on the autism spectrum. The results indicate that all participants mastered intraverbal relational responses with fewer trials in conditions involving additional learning inputs (Touch, Taste, and Sniff) as compared to HearSeeSay alone. Adding a fluency practice timing component contingent on failed tests of derived relational responding resulted in an increase in the emergence of untrained relations for all participants during retesting.

Finally, the present study was the first to examine the impact of different learning channels on intraverbal relations, but it does not provide information or comparison about how participants of other ages and diagnoses (or lack thereof) would respond in training and testing phases involving different learning channels, and whether fluency timings would be necessary for those participants to reach mastery. Participants in this study did not demonstrate object–attribute relational responding when initially tested during the intraverbal pretest, and the current study’s procedure allowed for tact (TouchTasteSniffSay) to intraverbal (HearSay) transfer only across two exemplars. Thus, future studies should include more than two objects to train each attribute and include the objects in the pretest phase to ensure that tact responses are under the control of relevant physical features and are not merely intraverbal responses. Future studies in this area should also include pretests targeting various learning channels (i.e., TasteSay vs. TouchSay or SniffSay). Last, future research should evaluate the efficiency of specific learning channel formats across different participant populations, and different types of fluency practice timings (HearSay vs. HearSeeSay), to continue to explore the critical attributes of an effective protocol to teach intraverbal responses based on attributes and categorization.

Compliance with Ethical Standards

Conflict of Interest

The authors declare that they have no conflict of interest.

Ethical Approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee (the Chicago School of Professional Psychology Institutional Review Board, expedited review under Category 7) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.