Do Targets with Persistent Responses Affect the Efficiency of Instruction?

Abstract

The efficacy and efficiency of instruction may be reduced as a result of persistent response patterns to targets. The current project exposed participants to tact training with one set of targets. Thereafter, the efficacy and efficiency of teaching different responses to the previously trained set of targets was compared to tact training with a novel set of targets. Results showed that targets with pre-established responses took longer to acquire than targets without pre-existing responses for both participants.

The efficacy and efficiency of training procedures is an important consideration when teaching verbal behavior to learners with autism spectrum disorder (ASD) and other developmental and intellectual disabilities. One barrier to efficacious and efficient instruction is persistent response patterns to targets. Learners may engage in the same incorrect response during each presentation of a target, such as saying, “fish” each time a picture of a penguin is presented, although the response “fish” does not occur to other targets. This type of consistent error may result in delayed acquisition (e.g., Kodak et al., 2019).

Persistent response patterns may be established during instructional programs via continuous or intermittent reinforcement. For example, Grow et al. (2011) reported persistent responding to a specific position in an array during conditional discrimination training, which likely continued due to intermittent reinforcement and either delayed or prevented acquisition. In comparison, Kodak et al. (2019) showed delayed acquisition for one participant with ASD due to a persistent error to one target in a stimulus set despite arranging extinction and prompts following errors. In addition, some error patterns may develop as a result of training responses to the same stimulus sequentially. For example, Hanney et al. (2019) taught tacts of auditory stimuli (e.g., tiger growl) to which two participants with ASD already had acquired tacts of the visual stimuli. Results showed previously acquired visual tacts interfered with acquisition of auditory tacts for one participant.

To our knowledge, no studies have directly manipulated persistent response patterns as an independent variable to examine the effects on skill acquisition. Rather, persistent responses are reported in the results section and included in hypotheses about delayed acquisition. Therefore, additional research is needed to directly evaluate the effects of persistent response patterns on acquisition. The current, translational study sought to evaluate the effects of persistent responding on acquisition by reinforcing responses to one set of stimuli and comparing the efficacy and efficiency of instruction to stimulus sets with and without previously reinforced responses.

Method

Participants, Setting, and Materials

Allen and Elliot were 6- and 7-year-old White and Middle Eastern males, respectively, with ASD who received comprehensive behavioral intervention at a university-based clinic for at least seven months. Participants were selected for inclusion because they had a history of engaging in persistent errors to a portion of stimuli in previous tact and intraverbal programs. Both participants had acquired at least 200 tacts prior to participation, engaged in no or low levels of problem behavior, and had mild cognitive delays. Participants completed sessions in individual cubicles in a classroom that contained tables, chairs, and work materials.

Unknown visual stimuli of historical figures (e.g., Bill Gates, Oprah Winfrey) were selected for inclusion due to novelty (no history of training) and no consistent responses. Each stimulus was printed in color and laminated on a 6.35 cm x 8.89 cm card. Targets were selected based on probes (described below). Ten stimuli were divided into two sets of five targets, which were equated across sets based on number of syllables and visual features of stimuli. Refer to Table 1 for a list of targets assigned to each stimulus set and condition. Set 1 was included in the initial comparison, and set 2 was included in the replication.

Table 1.

Targets across Participants, Sets, and Phases

Participant	Set	Pre-Training Responses Training Responses (First Names)	(Last Names)
	1	Ruth, Hilary, Jeff, Ronald, Bil	Bader Ginsburg, Clinton, Bezos, Reagan, Gates
		N/A	James, Winfrey, Jackson, Obama, Mandela
Allen	2	Michael, Rosa, Carl, Beyonce, Tina	Jordan, Parks, Lewis, Knowles, Turner
		N/A	Day O'Connor, Zuckerberg, White, Klein, Lennon
	1	Lebron, Oprah, Jesse, Michelle, Nelson	James, Winfrey, Jackson, Obama, Mandela
Elliot	2	N/A	Bader Ginsburg, Clinton, Bezos, Reagan, Gates
		Sandra, Mark, Betty, Calvin, John	Day O'Connor, Zuckerberg, White, Klein, Lennon
		N/A	Jordan, Parks, Lewis, Knowles, Jordan, Parks, Lewis, Knowles

Note. Cells marked N/A indicate that pre-training did not occur with the corresponding stimulus set prior to training last names.

Other session materials included a preferred item (e.g., iPad®), stopwatches to time the reinforcement intervals and record duration, video cameras to record sessions, previously mastered tacts for use during error correction (Allen only), and data-collection materials. Allen’s mastered tacts were previously taught tacts to which he engaged in correct independent responses during a probe of the visual stimulus. The preferred items included in reinforcement intervals during training were selected based on prior preference assessments (e.g., MSWO; Carr et al., 2000) and participant mands.

Response Measurement, Interobserver Agreement, and Treatment Integrity

The primary dependent variables were independent correct responses and session duration. Independent correct responses were defined as a vocalization that corresponded to the visual stimulus and occurred within 5 s of the participant looking at the stimulus. Data on independent correct responses were converted to a percentage by dividing the number of independent correct responses by the total number of trials per session, multiplied by 100. Session duration began right before the first trial and concluded after the last trial or the last reinforcement interval (if applicable). Total session duration was calculated by summing the duration of all training sessions in a condition.

A second observer collected data on independent correct responses during sessions for Allen and Elliot in the first comparison and replication. Trial-by-trial interobserver agreement (IOA) was calculated for independent correct responses. A trial was scored as a 0 if the primary and secondary data collectors scores did not match for that trial or a 1 if the scores matched. The trial scores were summed and divided by the total number of trials in the session, then multiplied by 100. In addition, treatment integrity data were collected for Allen’s and Elliot’s sessions in the first comparison and replication. Each trial was scored as a 1 (all components implemented with integrity) or a 0 (one or more components were not implemented with integrity). The percentage of treatment integrity was calculated by adding the scores for each trial and dividing by the number of trials per session, multiplied by 100. See Table 2 for IOA and treatment integrity data.

Table 2.

Interobserver Agreement and Treatment Integrity across Participants

P	Interobserver Agreement			Duration		Treatment Integrity
	%w/IOA	IOA%	Range	IOA%	Range	%w/TI	TI	Range
Allen Set 1	67	99	80-100	94	0-100	63	99	80-100
Elliot Set 1	59	99	80-100	97	0-100	56	93	60-100
Allen Set 2	36	100	N/A	100	N/A	36	93	60-100
Elliot Set 2	33	99	88-100	100	N/A	33	94	60-100

Note: P=participant, IOA= interobserver agreement, TI= treatment integrity, N/A= not applicable

Experimental Design

An adapted alternating treatments design embedded within a nonconcurrent multiple baseline design across stimulus sets was used in this study. The adapted alternating treatments design alternated between two conditions; one condition included pre-training of targets to establish a consistent response to the targets (e.g., first names of historical figures before training last names), and the other condition included targets not exposed to pre-training (e.g., only last names were trained; see Table 1). Sessions of the two conditions alternated until responding met the mastery criterion in at least one condition. Thereafter, sessions of the remaining condition continued until responding met the mastery criterion, which was two consecutive sessions with 100% independent correct responses.

Probes and Baseline

Probes were conducted to ensure that both participants did not respond correctly to any targets. Probe and baseline sessions consisted of five trials with each stimulus presented once per session. During each trial, the therapist held up the visual stimulus and waited up to 5 s for a response. No feedback was provided for correct or incorrect responses. Approximately every two trials, the therapist provided descriptive praise for appropriate session behavior (e.g., sitting still in the chair) and 20-s access to a preferred item.

Pre-Training

Only one set of target stimuli was exposed to pre-training (i.e., first names of the historical figures) to establish a recent history of reinforcement for a specific response to each stimulus. The trial arrangement for pre-training was identical to baseline, except for the inclusion of prompts and reinforcement. The therapist implemented a 5-s prompt delay in which independent correct responses resulted in praise and 20-s access to a preferred item. Incorrect responses resulted in a vocal model prompt by the therapist every 5 s until the participant engaged in a prompted correct response. Prompted correct responses resulted in praise. Error correction was implemented following prompted correct responses and involved re-presenting the target stimulus and waiting the 5-s response interval for Elliot. Error correction for Elliot continued until he engaged in an independent correct response or five error-correction trials occurred. For Allen, error correction involved interspersing a mastered tact between the prompted correct response and subsequent re-presentation of the target (Plaisance et al., 2016). During interspersal, the therapist presented a previously mastered visual stimulus, waited 5 s for a correct tact response, and provided praise following a correct tact response or a model prompt of the mastered tact. Thereafter, the trial was re-presented, and this sequence of error correction continued until Allen engaged in an independent correct response or he completed a total of five error-correction trials.

Training

Training occurred for stimulus sets in both conditions. Procedures were identical to pre-training except the targeted responses were only last names of the historical figures (e.g., “Gates” for Bill Gates) rather than first names. That is, only last name responses produced reinforcement in this phase.

Results and Discussion

Allen’s data are shown in Fig. 1. Responding in baseline was 0% across all stimulus sets and conditions. During pre-training of first names, the mastery criterion was met in 11 sessions for set 1 in the initial comparison and 7 sessions for set 2 in the replication. Responding to last names remained at 0% during the second exposure to baseline across all targets and conditions. During training of last names in the initial comparison (set 1), the mastery criterion was met in 7 versus 5 sessions for the targets with pre-established responses versus no pre-established responses, respectively. During training of last names in the replication (set 2), the mastery criterion was met in 10 versus 6 sessions for the targets with pre-established responses versus no pre-established responses, respectively. Therefore, Allen required more training sessions to reach the mastery criterion for both sets of stimuli with pre-established responses. Elliot’s data are shown in Fig. 2. Responding in baseline was 0% across all stimulus sets and conditions. During pre-training of first names, the mastery criterion was met in 18 sessions for set 1 in the initial comparison and 11 sessions for set 2 in the replication. Responding to last names remained at 0% during the second exposure to baseline across all sets and conditions. During training of last names in the initial comparison (set 1), the mastery criterion was met in 22 versus 13 sessions for the targets with pre-established responses versus no pre-established responses, respectively. During training of last names in the replication (set 2), the mastery criterion was met in and in 9 sessions for both stimulus sets. Therefore, Elliot also required more training sessions for stimuli with pre-established responses, although this outcome occurred in one of the two comparisons. Allen’s and Elliott’s training duration data across sets in each condition are reported in Table 3. Both participants required a longer training duration for the stimulus sets with pre-established responses in all four comparisons. Thus, pre-established responses to stimuli delayed acquisition of tacts for both participants in all four comparisons.

Fig. 1.

Percentage of Independent Correct Responses Across Sessions for Allen

Fig. 2.

Percentage of Independent Correct Responses Across Sessions for Elliot

Table 3.

Total Training Sessions and Duration Across Conditions

Participants	Sets	Targets with Pre-Training	Targets without Pre-Training
Allen	1	7 sessions (21 min)	5 sessions (14 min)
	2	10 sessions (24 min)	6 sessions (21 min)
Elliot	1	22 sessions (68 min)	13 sessions (33 min)
	2	9 sessions (28 min)	9 sessions (24 min)

During training, participants continued to engage in the pre-trained responses even though reinforcement was no longer provided for doing so during training of last names. For example, 88.5% and 91.3% of Allen’s errors and 66.7% and 83.2% of Elliot’s errors consisted of the pre-established responses (first-name tacts) during training in the first comparison and replication, respectively. However, over time, participants began responding correctly or engaged in more variable errors (i.e., responses not included in pre-training). Target stimuli with no pre-trained responses also resulted in errors or no responses during early training sessions; however, those errors were not consistent. Therefore, engaging in consistent responses to each stimulus, rather than varying incorrect responses, was the only identified response pattern that resulted in delayed acquisition.

The results of the present investigation and Hanney et al. (2019) suggest careful consideration is needed when selecting targets to include in tact training. First, researchers and practitioners should carefully consider whether to include stimuli to which learners engage in consistent responses (i.e., errors). The effects of consistent errors on acquisition in comparison studies remains unknown, although some researchers have reported persistent responding during training that delayed acquisition of one or more stimuli assigned to a specific condition (e.g., Kodak et al., 2019). Stimuli that occasion consistent errors could be excluded from stimulus sets in comparison research to prevent uncontrolled differences in stimulus sets across conditions that may influence the efficiency of instruction and bias the comparison. Second, some persistent responses are not established via trial-based instruction and may be adventitiously reinforced in the natural environment. For example, a child may tact a swimming penguin as a fish while at an aquarium, and her parents may provide attention and laughter (i.e., conditioned reinforcers for the tact; Skinner, 1957). If practitioners must subsequently teach tacts of stimuli with previously established errors, practitioners could conduct discrimination training with the target stimulus and a second stimulus associated with the error pattern, similar to the mixed trials described by Hanney et al. (2019). For example, if the client consistently says, “fish” in the presence of pictures of penguins, tact training with exemplars of both fish and penguins could be conducted to teach the relevant discriminations.

The results also have implications for the format of tact training. Sequential tact training with the same stimuli commonly occurs during skill acquisition programming, such as teaching object tacts prior to teaching categorical tacts of the same targets (Ribeiro and Miguel 2020). Delayed acquisition during sequential training of tacts could be prevented by teaching both tacts simultaneously (akin to mixed trials; Hanney et al., 2019). However, research is needed to evaluate the efficiency of this recommendation.

The current investigation was a translational study to investigate the effects of persistent responding on the efficacy and efficiency of skill acquisition. We did not select stimuli to which the participant already engaged in persistent errors because we could not control the history of reinforcement for those responses. Rather, we established a history of reinforcement for a set of responses and evaluated whether that history of reinforcement affected the efficacy and efficiency of teaching new responses to those same stimuli. In future studies, researchers could identify stimuli to which learners already engage in consistent errors and compare acquisition of those stimuli to stimuli that do not occasion consistent errors.

One limitation was the use of different error-correction procedures for both participants. For Elliot, we used the re-present until independent method. For Allen, we began by using the re-present until independent method but did not observe acquisition of targets during pre-training. Therefore, we introduced a task interspersal method of error correction (Plaisance et al., 2016) which required Allen to attend to the visual stimulus to engage in a response, rather than covertly or overtly echoing the correct prompted response in between trial re-presentations. Future researchers should consider using individualized error-correction strategies that promote the development of stimulus control during training and use the same error-correction procedure throughout training.

Declarations

Ethical approval

This study was approved by the appropriate institutional human subjects committee and was performed in accordance with the ethical standards as laid down in the 1964

Declaration of Helsinki and its later amendsments or comparable ethical standards. Informed consent was obtained from participants.

Conflicts of interest

The authors declared that there are no conflicts of interest related to the current study.