Correspondence Between Vocal-Verbal Behavior and Go/No-Go Responses During the Successive Matching-to-Sample Procedure

Abstract

In the current study, eight college students were exposed to a successive matching-to-sample (S-MTS) procedure utilizing non-verbal auditory stimuli consisting of common sounds. During emergent relations tests, participants were asked to talk aloud, and their vocal-verbal statements were transcribed and categorized as class-consistent, class-inconsistent, or irrelevant. All participants met emergence criterion for symmetry and four did so for transitivity/equivalence. Analysis of vocal-verbal statements showed a positive correlation between class-consistent statements emitted by participants and correct selection responses during S-MTS tasks. Such results suggest possible verbal mediation during emergent relations tests.

Conditional discrimination training via matching-to-sample (MTS) can establish unrelated stimuli as substitutable for one another (i.e., equivalence classes; Sidman & Tailby, 1982). During a typical MTS trial, an observing response toward a sample stimulus produces a minimum of three comparison stimuli, one of which is assigned as positive. One possible limitation of this procedure is that auditory comparisons can never be presented simultaneously, limiting its use for the establishment of auditory–auditory conditional relations (Dube et al., 1993). On the other hand, the successive matching-to-sample (S-MTS) procedure permits the use of auditory stimuli as samples and comparisons, as only one stimulus is presented at a time. During a S-MTS trial, a response to the sample (e.g., selecting it) causes it to disappear and a comparison to appear in its place. If the comparison belongs to the same class as the sample (i.e., positive), the learner is taught to select it (a go trial), and if sample and comparison are unrelated (i.e., negative), the learner is taught to refrain from selecting it (a no-go trial). This procedure has been successful in establishing visual–visual (Howland et al. 2021; Lantaya et al., 2018), auditory–visual (Zhelezoglo et al., 2021), and auditory–auditory conditional relations (Hanson & Miguel, 2021; Hanson et al., 2022).

Hanson and Miguel (2021) showed that following S-MTS training, 25 of 32 (78%) college students formed equivalence classes across familiar and unfamiliar verbal stimuli. Moreover, some participants could also relate these verbal stimuli intraverbally, likely by having echoed samples and comparisons during training (e.g., “Boj goes with soch and megh”). Thus, it is possible that correct go/no-go responses were also influenced by supplementary stimuli produced by participants’ verbal behavior (Miguel, 2016; 2018).

In a follow-up study, Hanson et al. (2022) replicated this procedure using nonverbal auditory stimuli (i.e., everyday sounds such as chopping vegetables) that could not be echoed. Six of eight participants passed equivalence tests. Even though participants could not echo, they tacted nonverbal stimuli and some related these names intraverbally (e.g., “The toaster sound goes with the beeping and the writing sounds”) during respective posttests. These results suggested that although echoic mediation may not be necessary to establish auditory classes, participants could have still related samples and comparisons verbally during S-MTS posttests, which could have aided in correct selections1 (Miguel, 2018).

Although participants in Hanson et al. (2022) engaged in tacts and intraverbals during respective posttests, there is no evidence they did so while completing S-MTS posttests. Talk aloud procedures, such as the protocol analysis (e.g., Austin & Delaney, 1998) and the silent dog method (e.g., Cabello et al., 2004), in which participants are instructed to talk aloud while solving tasks, have been used in previous research to assess the possible role of covert mediation (Vie & Arntzen, 2017). Thus, the purpose of the current study was to replicate and extend Hanson et al. by including a protocol analysis during which participants were required to talk aloud while completing S-MTS posttests.

Method

Participants, Setting, and Materials

Eight undergraduate students were recruited from a large public university and received course credit regardless of performance. Sessions were conducted remotely (see Hanson & Miguel, 2021) on a laptop using Microsoft Visual Basic^®. Stimuli were auditory clips of everyday sounds (see Table 1), which were grouped across three 3-member classes (i.e., A1, B1, C1; A2, B2, C2; A3, B3, C3). Written instructions (see Table 2) were emailed to participants prior to their first session. Participants were required to read and vocally summarize instructions before each condition.

Table 1.

Experimental stimuli

	A	B	C
Class 1	(ice in cup)	(chopping)	(sliding door)
Class 2	(toaster)	(scanner)	(signature)
Class 3	(rocket)	(razor)	(film rewind)

Table 2.

Written experimental instructions for participants

Condition	Instruction
BA/CA and BC/CB Pretest	Once the task begins, you will hear a sound from the computer. After you hear the sound, a green box will appear on the screen. Click the green box to hear a second sound. After you hear the second sound, a white box will appear on the screen. If you think the first and second sound go together, click the white box and say “click.” During this phase, you get no points as feedback. Remember, you can read these instructions at any time. Place the instructions in the designated location.
Baseline Training (Consequences)	During this phase you will learn how to group sounds together. Once the task begins, you will hear a sound from the computer. After you hear the sound, a green box will appear on the screen. Click the green box to hear a second sound. After you hear the second sound, a white box will appear on the screen. If you think the first and second sounds go together, click the white box and say “click.” If the sounds do not go together, then do not click the box and wait for the box to disappear. You will get points when sounds go together, and you will not get points when they don't. For the first few trials I will help you with the answer. After that you will have 4 seconds to respond on your own. If you do not respond, I will help you. You can read these instructions at any time. The harder you try, the faster this will go. Place the instructions in the designated location.
Baseline Posttest (No Consequences)	Continue clicking the white box and saying “click” for sounds that go together as before. During this time no points will be presented. Remember, you can read these instructions at any time. Place the instructions in the designated location.
Protocol Analysis	We are interested in how people solve problems. We want you to think out loud for the duration of the experiment. So that you understand what we mean by thinking out loud, an example would be solving a math problem while vocally describing each step that you take while solving it. For example, if I were to solve the problem 2 x 20, I would say aloud, “First I would put the 2 under the 0. Then I would multiply the 0 by the 2, making 0, and I would write the 0 under the problem. Next, I would multiply the bottom 2 by the top 2, getting 4. Finally, I would write the 4 under the problem, next to the 0, making a total of 40.” Try this now with problem number 1 (below). We are not concerned with whether you get the problem correct or how quickly you solve it. You will continue to speak aloud for the remainder of the study. Problem 1: 12 x 5 Problem 2: 11 x 4 Problem 3: 10 x 6 Problem 4: 5 x 10
BA/CA and BC/CB Posttests	This is a new phase. Use what you have learned so far to figure out what sounds go together. Once the task begins, you will hear a sound from the computer. After you hear the sound, a green box will appear on the screen. Click the green box to hear a second sound. After you hear the second sound, a white box will appear on the screen. If you think the first and second sounds go together, click the white box, and say “click.” If the sounds do not go together, then do not click the box and wait for the box to disappear. During this phase, you will get no points as feedback. Remember, you can read these instructions at any time. Place the instructions in the designated location. Continue to speak aloud as you complete the tasks.

Dependent Variables

The main dependent variable was the percentage of correct go/no-go responses. Correct go/no-go responses were defined as independently clicking on related sample and comparison combinations during go trials and refraining from clicking on unrelated combinations during no-go trials. Additional measures included reaction times to comparisons, trials to criterion in baseline training, intraverbals, and tacts made by the participants during the protocol analysis (i.e., talk aloud) while completing S-MTS posttests. Intraverbals included class-consistent, class-inconsistent, and irrelevant statements. Class-consistent intraverbals were defined as any vocal-verbal statement(s) that correctly identified two or more stimuli as being related or unrelated, according to experimenter-identified classes (e.g., “The sound of the ice [A1] goes with the sound of the door closing [C1]”). Class-inconsistent intraverbals were defined as any vocal-verbal statement(s) that incorrectly identified two or more stimuli as being related or unrelated (e.g., “The door closing [C1] does not go with the chopping vegetables [B1], so I’m not going to click on the white square”). Irrelevant statements were defined as any vocal-verbal statement(s) that did not refer to stimuli used in the study and/or their relation (e.g., neutral statements such as “Yeah” or “Okay”). Tacts consisted of experimenter-defined tacts (i.e., tacts identical to the name of the audio clip used) or unique tacts (i.e., tacts uniquely emitted by the participant such as calling ice in glass [A1] “coins in jar”) within intraverbal statements during S-MTS posttests.

Procedure

The experiment employed a one-to-many training structure. Conditions were presented in the following order: pretraining, BA/CA (symmetry) pretest, BC/CB (equivalence) pretest, AB/AC baseline training, baseline posttest, protocol analysis practice, BA/CA posttest with protocol analysis, BC/CB posttest with protocol analysis, and post-experimental interview (see Table 3). Baseline training, and BA/CA pre and posttests, included 24 trials per block, whereas BC/CB pre- and posttests included 36 trials per block.

Table 3.

Procedural differences

Conditions	Hanson et al. (2022)	Current study
Pretraining	X	X
BA/CA pretest	X	X
BC/CB pretest	X	X
Baseline training	X	X
Baseline posttest	X	X
PA practice		X
BA/CA posttest	X
BC/CB posttest	X
BA/CA posttest w/PA		X
BC/CB posttest w/PA		X
Tact test	X
Intraverbal test	X
Post-experimental interview	X	X

PA = protocol analysis

Each S-MTS trial included an auditory sample played by the computer, with each stimulus lasting approximately 4 s (Green, 2001). After the presentation of the sample, a green box appeared in the center of the screen and participants clicked on it, causing it to disappear. Next, an auditory comparison played by the computer was followed by the appearance of a white box in the center of the screen. If the two stimuli were related, the participant clicked on the white box (go trial) and if unrelated, they refrained from clicking on it (no-go trial). The white box remained on the screen for 8 s regardless of participants’ responses. All trials were separated by a 2 s intertrial interval.

Pretraining

This condition served to familiarize participants with the computer software. All trials followed the format described above with recognizable sounds (e.g., “green,” “red”). To proceed to the experimental conditions, participants needed to respond correctly across four consecutive trials.

BA/CA (Symmetry) and BC/CB (Transitivity/Equivalence) Tests

Participants were required to score 67% or lower across one to two blocks of both BA/CA and BC/CB pretests that were conducted identical to the S-MTS procedure described previously. No prompts or consequences were provided during tests. Emergence criterion for BA/CA posttests was 92% correct across two consecutive blocks and was 94% correct across two consecutive blocks for BC/CB posttests. Participants moved to BC/CB posttests regardless of performance on BA/CA posttests. BC/CB posttests were terminated if a decreasing or stable trend of incorrect responding below emergence criterion was displayed across three consecutive blocks.

AB/AC (Baseline) Training and Testing

During this condition, participants learned to relate AB and AC stimuli, and trials were conducted as described above. If a participant did not independently respond during go-trials within 4 s from the appearance of the white box, the experimenter verbally instructed them to click the box (i.e., “Click”). Both independent and prompted go responses were followed by the emission of a tone from the computer and 10 points at the top of the screen. Each additional correct response, prompted or independent, resulted in the addition of 10 points to the score. No programmed consequences followed a correct response on no-go trials or an incorrect response across both go and no-go trials. Mastery criterion was set at 100% independent (i.e., no prompting) correct responses across two consecutive blocks. Participants were then required to complete one subsequent block at 100% without the tone and points. If participants did not score 100% during this block, the emission of the tone and points were reinstated. Training was terminated if participants failed to meet criterion within 24 blocks.

Protocol Analysis

Before starting S-MTS posttests, participants were first provided with a math problem and were instructed to explain the steps they would take to solve the problem. If participants engaged in incorrect responses (e.g., stating only the solution) the experimenter modeled the correct response (e.g., “First I will multiply the 5 by the 2…”). Once participants correctly described one math problem, they were asked to talk aloud (e.g., “That sound goes with the last sound, so I am going to click the box”) during S-MTS posttests (see Table 2; Austin & Delaney, 1998). If at any point participants stopped talking aloud while completing S-MTS posttests, the participant was told “remember to talk out loud while you are completing the task” and given the opportunity to reread the instructions. If a participant continued to complete S-MTS posttests without talking aloud the procedure would have stopped and the participant would have been required to reread the instructions and complete a math problem while talking aloud before proceeding, but this never happened.

Post-Experimental Interview

Following posttests, the experimenter asked participants two questions to add additional information as to possible mediating behaviors: (a) “How did you decide when to click the white box?” and (b) “How did you decide during the last part, when you had to use what you learned?”

Experimental Design

The current study utilized a two-tier nonconcurrent multiple baseline design across dyads (Watson & Workman, 1981) in which participants in the second tier underwent double pretests. They were assigned to each tier based on the order in which they were recruited. This design controlled for the possibility that repeated exposure would be sufficient for the establishment of equivalence classes while preventing that too many pretest exposures would establish spurious stimulus control (Lantaya et al., 2018).

Interobserver Agreement and Procedural Fidelity

Interobserver agreement (IOA) was calculated for all sessions and trials (Kazdin, 2011). A secondary observer collected data on the percentage of correct go and no-go trials across S-MTS conditions and the transcribed participants’ vocal-verbal statements recorded during the protocol analysis. IOA was calculated by dividing the number of agreements by the number of agreements plus disagreements and multiplying by 100. IOA for S-MTS tasks averaged 99.27% (range, 98.69–100%), and IOA for vocal-verbal statements recorded during the protocol analysis averaged 95.86% (range, 94–97%).

Procedural fidelity data were recorded by the secondary observer, which measured the correct implementation of prompts by the experimenter during baseline training. Fidelity was computed by dividing the number of correct trials by the number of correct plus incorrect trials and multiplying by 100. Procedural fidelity averaged 98.7% (range, 93–100%).

Results

Figure 1 illustrates the results of training and testing conditions. Participants completed baseline training at an average of 246 trials (range, 96–504 trials). Emergence criterion for BA/CA posttests was met within two blocks across participants, and emergence criterion for BC/CB posttests was met within two–three blocks by four participants (P1, P4, P6, and P7). The remaining four participants (P2, P3, P5, and P8) did not meet emergence criterion for BC/CB posttests. Average reaction times for correct go trials across participants during pretests was 2.84 s (range, 1.28–7.70 s), during baseline training was 1.59 s (range, 1.18–2.20 s), during BA/CA posttests was 2.60 s (range, 1.01–4.91 s), and during BC/CB posttests was 3.00 s (range, 1.11–4.59 s).

Fig. 1.

Percentage of correct go and no-go responses, percentage of experimenter-defined and unique tacts, and percentage of class-consistent intraverbals across participants. Note. Exp Tact = experimenter-defined tact; Sym = symmetry; Trans = transitivity; BL train = baseline training; BL Test = baseline training when programmed consequences were removed; % CC IV = percentage of class-consistent intraverbals

All participants engaged in a combination of experimenter-defined or unique tacts within intraverbal statements during the protocol analysis (e.g., “That’s the pen [C2] sound and I know that goes with the button sound [B2] and the toaster sound [A2]”). Participants who engaged in a higher number of class-consistent intraverbals while completing BC/CB posttests made more correct selection responses, whereas participants with a lower percentage of class-consistent intraverbals made fewer correct selections. Participants made very few irrelevant statements which did not appear to have a detectable effect on performance. Figure 2 depicts the relation between the percentage of class-consistent intraverbals and the percentage of correct selection responses across participants. The experimenters calculated a Pearson r, and a strong positive relation between changes in these two variables was found (r = .9953). Figure 3 depicts the relation between the percentage of class-inconsistent intraverbals and percentage of incorrect selections across participants, and a Pearson r suggests a strong positive correlation (r = .9979).

Fig. 2.

Percentage of class-consistent intraverbals and percentage of correct selection responses across participants

Fig. 3.

Percentage of class-inconsistent intraverbals and percentage of incorrect selection responses across participants

The results of the post experimental interview differed across participants. Three of four participants who mentioned attending to the feedback provided by the computer required fewer training trials to reach criterion than those participants who did not. Interestingly, participants who reported categorizing or reciting the names of the stimuli were also the ones who met emergence criterion.

Discussion

The purpose of the current study was to extend Hanson et al. (2022) to assess whether participants were tacting or intraverbally relating stimuli while completing S-MTS tasks. Four of eight participants met emergence criterion during transitivity/equivalence posttests, providing additional support for the S-MTS procedure to establish equivalence classes with non-verbal stimuli. The protocol analysis showed that participants who successfully passed equivalence tests also correctly tacted and related stimuli intraverbally while completing S-MTS tasks.

It is possible that while speaking aloud during S-MTS tasks, participants vocalizations generated auditory stimuli that served to evoke the correct selection. For example, on a correct go trial, P6 frequently tacted either all three class members, including the sample, or the two class members that matched the sample prior to touching the comparison. Chastain et al. (2022) showed that participants who learned to emit class-consistent statements required fewer trials to criterion and made fewer errors during training, whereas those who were taught to engage in class-inconsistent statements made most of the recorded errors and produced low equivalence yields. Although our data seem consistent with a mediational account of stimulus equivalence (Horne & Lowe, 1996; Miguel, 2018) future research should explore the specific type of mediation (if any) that is at play during S-MTS tasks (e.g., Lowenkron, 2006).

The current study is not without limitations. First, participants may only have vocalized because they were asked to do so. Second, the protocol analysis could have contributed to the lower equivalence yields. Future research could require participants to engage in incompatible behavior (e.g., singing a song; Clough et al., 2016) to determine if it would disrupt S-MTS performance. Moreover, asking participants to talk aloud only after they first meet emergence criterion during S-MTS posttests may serve to assess whether the protocol analysis itself affects test performance. Finally, although there was a lack of replication on the emergence of equivalence classes within dyads, there was replication across dyads. This (along with other possible limitations of the non-concurrent multiple-baseline design) was mitigated by the fact that stimulus relations were arbitrary, and verbal reports were correlated with S-MTS performance. Thus, we can be confident that equivalence class performance was not a product of variables other than the ones manipulated by the experimenter.

The results of the current study suggest that the go/no-go S-MTS procedure can lead to both auditory equivalence classes and verbal categorization (i.e., intraverbals). Although correlational in nature, our data add to a relatively large body of research (e.g., Carp & Petursdottir, 2015; Chastain et al., 2022; Cordeiro et al., 2021; Diaz et al., 2020; Jennings & Miguel, 2017; Petursdottir et al., 2015) suggesting that when verbally sophisticated participants are presented with complex tasks, they may engage in mediating verbal behaviors whose response products may serve as supplemental stimuli that can affect their performance (Miguel, 2018).

Funding

No funds were received to conduct this study.

Data Availability

All data are available upon request.

Code Availability

Not applicable.

Declarations

Conflicts of Interest

The authors have no conflicts of interest.

Ethics Approval

All procedures and recruitment protocols were approved by the university’s Institutional Review Board (IRB).

Consent to Participation/Consent to Publish

Informed consent was provided by all participants in the study.

Author Note

This study is based on a thesis submitted by the first author under the supervision of the third author to the Department of Psychology at California State University, Sacramento in partial fulfillment of the requirements for a M.S. in Applied Behavior Analysis.