The Analysis of Verbal Behavior (2020) 36:1–20

https://doi.org/10.1007/s40616-019-00124-y

Comparing Error Correction to Errorless Learning: A

Randomized Clinical Trial

Justin B. Leaf 1,2 & Joseph H. Cihon 1,2 & Julia L. Ferguson 1 & Christine M. Milne 1,2 &
Ronald Leaf 1 & John McEachin 1

Published online: 19 February 2020

# Association for Behavior Analysis International 2020

Abstract
Errorless learning and error correction procedures are commonly used when teaching
tact relations to individuals diagnosed with autism spectrum disorder (ASD). Research
has demonstrated the effectiveness of both procedures, as well as compared them. The
majority of these studies have been completed through the use of single-subject
experimental designs. Evaluating both procedures using a group design may contribute
to the literature and help disseminate research related to the behavioral science of
language to a larger audience. The purpose of the present study was to compare an
errorless learning procedure to an error correction procedure to teach tact relations to 28
individuals diagnosed with ASD through a randomized clinical trial. Several variables
were assessed, including the number of stimulus sets with which participants reached
the mastery criterion, responding during pre- and postprobes, responding during
teaching, efficiency, and the presence of aberrant behavior. The results indicated that
both procedures were effective, efficient, and unlikely to evoke aberrant behavior,
despite participants in the error correction condition engaging in significantly more
independent correct responses and independent incorrect responses.

Keywords discrete-trial teaching . error correction . errorless learning . most-to-least

prompting . tacting

Intervention programs for children diagnosed with autism spectrum disorder (ASD)
commonly include teaching language skills (Barbera & Rasmussen, 2007; Greer &
Ross, 2008; Petursdottir & Carr, 2011; Sundberg & Partington, 1998). Within behav-
iorally based intervention programs, language skills are often classified and targeted
based on Skinner’s (1957) conceptualization of verbal operants. This often includes

* Justin B. Leaf
[email protected]

1
Autism Partnership Foundation, 200 Marina Drive, Seal Beach, CA 90808, USA
2
Endicott College, Seal Beach, CA, USA
2 The Analysis of Verbal Behavior (2020) 36:1–20

directly targeting a variety of tact relations through the use of discrete-trial teaching
(DTT; e.g., Grow, Kodak, & Clements, 2017; Majdalany, Wilder, Smeltz, & Lipschultz,
2016), which is a common approach used in behaviorally based intervention programs
(Green, 1996; Lovaas, 1987, 2003; Smith, 1999, 2001). A discrete trial involves three
core components: (a) a discriminative stimulus (e.g., a teacher-delivered instruction),
(b) the learner’s response, and (c) a teacher-delivered consequence based on the
learner’s response. These three core components of DTT do not, by themselves,
eliminate or prevent learner errors; therefore, approaches to prevent, reduce, or correct
errors have been developed (Mueller, Palkovic, & Maynard, 2007). Generally, these
approaches to address errors within DTT could be put into two broad categories: (a)
errorless learning procedures and (b) error correction procedures. However, it should be
noted that these categories are not mutually exclusive and can commonly occur in
combination.
Errorless, or near-errorless, learning procedures involve attempting to prevent errors
during all teaching sessions (Mueller et al., 2007). There are several techniques that are
used to achieve near-errorless learning that have been described in the literature,
including constant time delay (e.g., Ledford, Gast, Luscre, & Ayres, 2008), progressive
time delay (e.g., Walker, 2008), simultaneous prompting (e.g., Leaf, Sheldon, &
Sherman, 2010), stimulus shaping (e.g., Schilmoeller, Schilmoeller, Etzel, & LeBlanc,
1979), stimulus fading (e.g., Moore & Goldiamond, 1964), and most-to-least
prompting (e.g., Demchak, 1990). A common theme across each of these systems is
to begin teaching with the most assistive prompt in order to minimize the likelihood of
an error. Prompts are faded gradually in a manner that maximizes the probability of
continual correct responding while decreasing the probability of errors.
Errorless learning, more generally, has been endorsed by many professionals as the
desired approach for teaching individuals diagnosed with ASD (e.g., Green, 2001;
Mueller et al., 2007). For instance, Gast (2011b) stated,

In light of research supporting the effectiveness and efficiency of such “near-

errorless” response-prompting procedures as CTD [constant time delay], PTD
[progressive time delay], and SLP [system of least prompts], it is hard to justify
the use of any trial and error procedure, [or] error correction alone. (p. 237)

When reviewing errorless learning procedures, Mueller et al. (2007) came to the same
conclusion, stating, “Because there are many potential negative side effects of incorrect
responding when using trial and error procedures, teaching procedures can be arranged
to eliminate or reduce responding to incorrect choice stimuli” (p. 698). Furthermore,
when selecting approaches to train staff in the implementation of DTT, researchers have
opted for training staff to use DTT with errorless learning procedures such as most-to-
least prompting, which may be due the effectiveness of most-to-least prompting over
other prompting methods (e.g., Day, 1987; McDonnell & Ferguson, 1989; Miller &
Test, 1989) and recommendations for practitioners to use errorless learning procedures
(e.g., Kayser, Billingsley, & Neel, 1986; McDonnell & Ferguson, 1989).
Error correction, on the other hand, involves procedures that are used following an
incorrect response that will increase the probability of a correct response on subsequent
trials (Cariveau, La Cruz Montilla, Gonzalez, & Ball, 2018). Similar to errorless
learning procedures, correct responses by the learner set the occasion for reinforcement.
The Analysis of Verbal Behavior (2020) 36:1–20 3

There are many error correction procedures that often include vocal feedback (e.g.,
stating “no” following an incorrect response; Leaf et al., 2019), modeling the correct
response (Carroll, Joachim, St. Peter, & Robinson, 2015), a short time-out (e.g., turning
away from the learner; Carroll et al., 2015), multiple response repetition (Carroll,
Owsiany, & Cheatham, 2018), or re-presentation of the trial (Carroll et al., 2015).
Although these are common components of error correction procedures, research
evaluating them commonly includes slight procedural deviations (Cariveau et al.,
2018), and it is likely that variations and combinations of these procedures in clinical
practice will occur. Some common procedures that could be classified as error correc-
tion procedures that have been empirically evaluated include, but are not limited to, no-
no-prompt procedures (e.g., Leaf et al., 2010), single response repetition (e.g., Worsdell
et al., 2005), and multiple response repetition (e.g., Worsdell et al., 2005). A common
theme across each of these error correction procedures is providing the learner with the
opportunity to respond independently prior to the instructor taking action to influence
how the learner responds.
Although many studies have demonstrated the effectiveness of error correction
procedures in isolation or in combination with errorless learning procedures (e.g.,
Carroll et al., 2015; Kodak, Fuchtman, & Paden, 2012; McGhan & Lerman, 2013;
Plaisance, Lerman, Laudont, & Wu, 2016; Townley-Cochran, Leaf, Leaf, Taubman, &
McEachin, 2017), some have argued against the use of procedures that allow the
learner to err and are in favor of errorless learning strategies (e.g., Green, 2001; Mueller
et al., 2007). For example, Gast (2011b) asked, “Why risk such a response [referring to
errors] by a child if we have evidence-based instructional procedures (CTD, PTD, most
to least prompting) that can and do reduce error during the early stages of learning?” (p.
237). Others have argued against the use of error correction procedures with claims that
allowing the learner to err leads to more errors (Mueller et al., 2007), makes teaching
sessions aversive for learners (LaVigna & Donnellan, 1986), or produces negative
emotional responding (LaVigna & Donnellan, 1986). Given the common use of DTT
within language instruction for individuals diagnosed with ASD and the disagreement
about allowing errors, empirical investigations have included comparisons of errorless
learning and error correction procedures.
Leaf et al. (2010) compared simultaneous prompting to no-no-prompt procedures to
teach three children diagnosed with ASD a variety of skills (e.g., math skills, listener
behavior, responses to wh– questions). Simultaneous prompting consisted of delivering an
instruction and a controlling prompt (i.e., a prompt that resulted in the learner responding
with 100% accuracy in a prior assessment) on each trial. No-no-prompt procedures
consisted of allowing two consecutive incorrect responses to occur before a prompted
trial occurred. After the first incorrect response, the researcher stated “no” and re-presented
the trial. If the second presentation of the trial also resulted in an incorrect response, the
researcher said “no” and repeated the same trial, but the instruction was accompanied by a
controlling prompt. The relative effectiveness of the two approaches was evaluated using a
parallel treatment design. Participants reached the mastery criterion for all of the skills
taught with no-no-prompt procedures but for only 10% of the skills taught using simul-
taneous prompting. Participants also maintained more of the skills taught with the no-no-
prompt procedure compared to the skills taught with simultaneous prompting.
Fentress and Lerman (2012) compared most-to-least prompting to no-no-prompt
procedures to teach a variety of skills (e.g., listener behavior, imitation, matching) to
4 The Analysis of Verbal Behavior (2020) 36:1–20

four children diagnosed with autism. Most-to-least prompting involved a full physical
prompt (hand-over-hand prompt), followed by a partial physical prompt, and then a less
intrusive prompt (Fentress & Lerman, 2012). No-no-prompt procedures occurred
similarly to Leaf et al. (2010), with the inclusion of a prompt-fading system. That is,
when prompts were provided on the third trial, those prompts were faded similarly to
how prompts were faded in the most-to-least condition. The relative effectiveness of
both procedures was evaluated using an alternating-treatments design. Results of the
study indicated that both procedures were effective, but the most-to-least prompting
procedure reduced the frequency of learner errors and resulted in better maintenance,
whereas the no-no-prompt procedure was found to be more efficient (i.e., participants
reached the mastery criterion faster with the no-no-prompt procedure).
Leaf et al. (2014a) compared an error correction procedure to most-to-least
prompting to teach listener behavior (referred to as receptive labels in the study) to
two children diagnosed with ASD. The error correction procedure consisted of the
teacher presenting an instruction and providing the participant with an opportunity to
respond independently. Correct responses resulted in praise and brief access to a toy.
Incorrect responses resulted in the therapist saying, “No, that’s not it,” and then
gesturing to the correct stimulus while saying, “This is the [correct response].” The
most-to-least prompting procedure involved a three-step prompting hierarchy
consisting of a point prompt, a positional prompt, and the instruction alone. The
researchers used an adapted alternating-treatments design embedded in a multiple-
baseline design to evaluate the effectiveness of the prompting procedures. The results
indicated that both procedures were effective (i.e., both participants reached the mastery
criterion for all targets), similar in terms of efficiency with respect to the number of
sessions to mastery, and similar with respect to the number of incorrect responses
during teaching trials.
Leaf et al. (2014b) compared error correction to most-to-least prompting to teach
two children diagnosed with ASD tact relations. The error correction procedure
consisted of the researcher presenting an instruction (e.g., “What is it?”) and allowing
the learner the opportunity to respond independently. Correct responses resulted in
praise, and incorrect responses resulted in corrective feedback and informative feed-
back (e.g., “No, it’s Kermit.”). The most-to-least prompting hierarchy consisted of (a)
the controlling prompt (i.e., stating the correct response), (b) a multiple-choice question
(e.g., “Is it Fozzie, Animal, or Beaker?”), and (c) the instruction alone (i.e., no prompt).
Correct responses resulted in praise, and incorrect responses resulted in no feedback
and movement to a more assistive prompt on the next trial. The results of an
alternating-treatments design embedded in a multiple-baseline design demonstrated
that the error correction procedure was slightly more effective (i.e., one set did not
reach the mastery criterion in the most-to-least condition) and more efficient (i.e., the
mastery criterion was reached in fewer sessions in the error correction condition) and
resulted in a higher percentage of correct teaching trials. Mixed results were observed
across both approaches with respect to maintenance of the skills taught.
Although there have been several studies documenting the effectiveness of error
correction as the sole intervention (see Cariveau et al., 2018, for a review), evaluating
procedural variations of error correction (e.g., Carroll et al., 2018), and comparing error
correction procedures to errorless learning procedures (e.g., Leaf et al., 2014b), the
debate continues as to the conditions under which each procedure should be selected or
The Analysis of Verbal Behavior (2020) 36:1–20 5

avoided. One potential reason for the continued debate could be the use of single-
subject experimental designs. Although these are a hallmark of behavior-analytic
research, single-subject experimental designs have been criticized for lacking external
validity and failing to demonstrate causal relationships, and some do not consider them
true experimental designs (Kazdin, 2011). Nonetheless, single-subject experimental
designs do offer important benefits (e.g., repeated observations across time, continuous
assessment, participants serving as their own baseline) and will continue to be neces-
sary and useful. For example, single-subject experimental designs conducted thus far
help identify the conditions under which errorless learning procedures or error correc-
tion procedures are more or less effective at the individual level. However, group
designs could help extend the previous research and improve our understanding of
errorless learning and error correction procedures for individuals diagnosed with ASD
in a variety of ways.
First, Smith et al. (2007), when describing the design of research studies for
individuals diagnosed with ASD, suggested that group studies, such as randomized
clinical trials, are the authoritative test of the efficacy of an intervention. If this is the
case, a group study evaluating errorless learning and error correction procedures could
provide an answer to the debate over which is most effective or efficient. Second, Smith
et al. noted that group designs can “enable investigators to analyze variables such as
subject characteristics that may be associated with favorable or unfavorable responses
to intervention” (p. 361). This may be especially desirable given the common associ-
ation between punishment-based procedures, such as error correction, and undesirable
behavior (e.g., aggression; Ferster & DeMeyer, 1962; Gast, 2011a; Lerman &
Vorndran, 2002; Sulzer-Azaroff & Mayer, 1991). Finally, group designs are commonly
the standard in other related fields (Smith, 2012; Smith et al., 2007). It may be the case
that conducting a group design evaluating errorless learning and error correction
procedures would help applied behavior analysis, as it relates to the treatment of
ASD, disseminate research related to the behavioral science of language to a larger
audience. Therefore, the purpose of this study was to compare the relative effectiveness
of an error correction procedure to an errorless learning procedure (i.e., most-to-least
prompting) using a randomized clinical trial (RCT) to teach tact relations to 28 children
diagnosed with ASD.

Method

Participants

Participants were 28 children diagnosed with ASD. To be included in this study,

children had to meet the following criteria: (a) the child had an independent diagnosis
of ASD; (b) the child had an Expressive One-Word Picture Vocabulary Test, Fourth
Edition (EOWPVT; Martin & Brownell, 2010), standard score above 75; (c) the child
had a Peabody Picture Vocabulary Test, Fourth Edition (PPVT; Dunn & Dunn, 2007),
standard score above 75; (d) the child had a previous history with DTT; (e) caregivers
consented for their child to participate in the study; (f) the child would benefit from
instructional strategies for tact relations, or tact relations were a current goal; and (g)
praise had been identified to function as a reinforcer. The inclusion criteria were
6 The Analysis of Verbal Behavior (2020) 36:1–20

selected in an effort to recruit a large number of participants for whom the interventions
employed and targets selected would likely be appropriate. Participants were recruited
from the agency in which the study was conducted and through a mass e-mail service.
Following obtaining caregivers’ consent, the first 26 participants were randomly
assigned (via random.org) to one of two groups: (a) error correction or (b) errorless
learning (i.e., most-to-least prompting). All children had a history of receiving inter-
vention in applied behavior analysis that most likely included the use of prompting or
error correction. However, the researchers found that it was not possible to accurately
and completely identify each participant’s complete history with each of the procedures
evaluated within this study. The scores of the first 26 children yielded statistically
significant differences on the EOWPVT; therefore, the final two participants were
assigned into each of the conditions so there would be no differences across the two
groups.
Table 1 provides individual participant characteristics and group means. A paired t
test was conducted to evaluate preintervention demographic variables across the two
conditions. The results indicated no significant difference in the age of participants
between the error correction condition (M = 5.071; SD = 1.859) and the errorless
learning condition (M = 5.786; SD = 1.847), t(26) = −1.020, p = .317. There was no
significant difference in the Vineland Adaptive Behavior Scales, Second Edition
(VABS; Sparrow, Cicchetti, & Balla, 2005), scores of participants between the error
correction condition (M = 87.417; SD = 7.786) and the errorless learning condition (M
= 83.231; SD = 13.361), t(23) = .354; p = .354. There was no significant difference in
the PPVT standard score of participants between the error correction condition (M =
107; SD = 21.071) and the errorless learning condition (M = 101.357; SD = 16.213),
t(26) = .794; p = .434. There was no significant difference in the EOWPVT standard
score of participants between the error correction condition (M = 114.929; SD =
20.764) and the errorless learning condition (M = 104.071; SD = 13.826), t(26) =
1.628; p = .115.

Setting

Sessions were conducted in rooms in a private agency that provides behavioral

intervention to individuals diagnosed with ASD. The rooms consisted of a table
designed for a child, two child-sized chairs, educational materials (e.g., books, toys,
whiteboard), and adult furniture (e.g., desks, couch, desk chairs).

Stimuli

The same stimuli were used for all participants and conditions to best evaluate any
differences between the error correction and errorless learning procedures used. That is,
if different stimuli were used for each participant or each procedure, it is possible that
the stimuli used could be partly, or wholly, responsible for any differences. Therefore,
the same target stimuli were used in both conditions and consisted of six pictures of
unknown comic book characters. The stimuli were selected based on common themes
within popular media. Given the frequent release of movies and TV shows related to
comic books, it was determined that comic book characters may have some immediate
utility and would be maintained by the respective verbal communities of all
The Analysis of Verbal Behavior (2020) 36:1–20 7

Table 1 Participant Demographics Across the Two Groups

Participant Group Age VABSa PPVT EOWPVT

EC1 EC 9 yr 8 mo — 98 122
EC2 EC 4 yr 7 mo 80 132 145
EC3 EC 5 yr 6 mo 84 94 114
EC4 EC 5 yr 0 mo 103 120 145
EC5 EC 5 yr 0 mo 91 120 117
EC6 EC 8 yr 7 mo 87 117 110
EC7 EC 7 yr 9 mo 95 135 132
EC8 EC 8 yr 0 mo — 143 145
EC9 EC 8 yr 0 mo 96 92 101
EC10 EC 4 yr 2 mo 80 82 89
EC11 EC 4 yr 1 mo 79 107 100
EC12 EC 3 yr 8 mo 91 84 111
EC13 EC 4 yr 6 mo 80 97 95
EC14 EC 6 yr 8 mo 83 77 83
Average 6 yr 1 mo 87 107 114
EL1 EL 8 yr 7 mo 74 109 97
EL2 EL 7 yr 4 mo — 106 118
EL3 EL 6 yr 11 mo 91 130 121
EL4 EL 5 yr 1 mo 94 78 83
EL5 EL 6 yr 5 mo 69 77 94
EL6 EL 5 yr 2 mo 79 90 106
EL7 EL 5 yr 2 mo 71 81 84
EL8 EL 6 yr 2 mo 93 118 119
EL9 EL 7 yr 6 mo 116 100 107
EL10 EL 8 yr 4 mo 70 92 104
EL11 EL 9 yr 8 mo 80 99 89
EL12 EL 9 yr 9 mo 91 111 101
EL13 EL 3 yr 10 mo 80 109 107
EL14 EL 5 yr 11 mo 74 119 127
Average 6 yr 10 mo 83 101 104

Note. VABS = Vineland Adaptive Behavior Scales, Second Edition; PPVT = Peabody Picture Vocabulary
Test, Fourth Edition; EOWPVT = Expressive One-Word Picture Vocabulary Test, Fourth Edition.
a Does not include scores from three participants (i.e., EC1, EC8, and EL2); however, this measure did not

determine inclusion in the study.

participants. Although this could create a possibility of exposure outside of experimen-

tal sessions, we felt the social validity of the stimuli used outweighed this possible risk.
The six comic book characters were randomly divided into three sets of two. The two
targets in the first set were Iron Fist and Daredevil. The two targets in the second set
8 The Analysis of Verbal Behavior (2020) 36:1–20

were Black Bolt and Hawkfire. The two targets in the third set were Yellowjacket and
Carnage.

Dependent Measures

The main dependent measure was participant responding during full probes, which
occurred prior to and following intervention. Participant responding was categorized as
independent correct, incorrect, or no response. During pre- and postprobes, participant
responding was categorized as a correct, incorrect, or no response. An independent
correct response was defined as engaging in the vocal response that corresponded to
the presented stimulus within 5 s of its presentation (e.g., saying “Iron Fist” in the
presence of the picture of Iron Fist after the interventionist said, “Who is this?”). An
independent incorrect response was defined as engaging in a vocal response that did
not correspond to the presented stimulus within 5 s of its presentation. No response was
defined as not engaging in any vocal response within 5 s of the stimulus presentation.
The second dependent measure was the number of stimulus sets for which each
participant reached the mastery criterion as assessed through daily probes. The mastery
criterion was defined as the learner engaging in 100% independent correct responses on
all trials across three consecutive daily probes. If this mastery criterion was not met
following 20 intervention sessions, the next set was introduced. If any participant did
not reach the mastery criterion for a set, they were taught following the completion of
the study. During daily probes, response categories and definitions were the same as in
full probes (i.e., independent correct, incorrect, and no response).
The third dependent measure assessed participant responding during teaching ses-
sions across the two conditions. Again, we used independent correct and incorrect
responses as previously described during the error correction and errorless learning
conditions. Additionally, during the errorless learning condition, participant responding
was also categorized as prompted correct or incorrect. A prompted correct response
was defined as engaging in the vocal response that corresponded to the presented
stimulus within 5 s of its presentation following a prompt. The percentage of prompted
correct responding was measured by dividing the number of trials with prompted
correct responses by the total number of trials per session and multiplying by 100. A
prompted incorrect response was defined as engaging in a vocal response that did not
correspond to the presented stimulus within 5 s of its presentation following a prompt
(e.g., saying “Joe” in the presence of the picture of Iron Fist after the interventionist
vocally stated, “Iron”). The percentage of prompted incorrect responding was measured
by dividing the number of trials with prompted incorrect responses by the total number
of trials per session and multiplying by 100. All responses during teaching were scored
using a datasheet that outlined the trial order. We also measured the occurrence of
aberrant behavior during teaching trials. Aberrant behavior was defined as the partic-
ipant engaging in any aggression (e.g., hitting or attempting to hit the researcher), self-
injurious behavior (e.g., hitting oneself), stereotypic behavior (e.g., hand flapping),
crying, yelling, or other (i.e., an aberrant behavior not predefined). If aberrant behavior
occurred, the researcher marked the trial during which aberrant behavior occurred, as
well as the topography of the aberrant behavior.
The final dependent measure was an evaluation of efficiency, assessed via daily
probes. To determine the efficiency of each condition, we evaluated the total number of
The Analysis of Verbal Behavior (2020) 36:1–20 9

sessions until a participant reached the mastery criterion for each set or until teaching
had concluded for all three sets (i.e., if the mastery criterion had not been reached
within 20 sessions). It should be noted that this number was capped at 20. Specifically,
if the participant had not reached the mastery criterion for a set within 20 sessions, the
next set was introduced.

Probes

Full probes Full probe sessions were used to make pre- and postcomparisons of the
effectiveness of each of the conditions. Two full probe sessions for target stimuli were
conducted immediately prior to and following intervention. Full probe trials began with
the interventionist holding up a stimulus and providing an instruction for the participant
to respond (e.g., “Who is it?”). The participant was given 5 s to respond. Neutral
feedback (e.g., “Thanks.”) was provided regardless of accuracy, and no prompts were
provided on any trial. Each stimulus was presented three times for a total of 18 trials.
The order of the presentation of the stimuli was randomized prior to each full probe.
Interventionists conducting full probes prior to teaching were kept blind to which
condition the participant was assigned; however, they were not blind to which condi-
tion the participant was assigned during full probes following intervention.

Daily probes With the exception of the first intervention session, daily probes were
conducted prior to each intervention session across both conditions. Daily probes were
similar to full probes, but only the two stimuli currently being targeted during inter-
vention sessions were included (e.g., Black Bolt and Hawkfire). Each stimulus was
presented in a random order three times for a total of six trials during daily probes.
Neutral feedback (e.g., “Thank you,” “OK”) was provided regardless of participant
responses across all trials. The mastery criterion for each stimulus set was a participant
responding correctly on 100% of trials during a daily probe for three consecutive daily
probe sessions. Once a participant reached the mastery criterion for one set, the next set
was introduced. If the mastery criterion was not met after 20 sessions, intervention for
that set was discontinued and the next set was introduced.

General Procedure

Sessions occurred once a day for 2 to 5 days a week across both conditions depending
on participants’ schedules (there were no noticeable differences between the two groups
with respect to scheduling). Sessions lasted between 5 and 10 min. A session consisted
of either a full probe, to assess responding prior to or after intervention, or a daily probe
followed by intervention based on the condition to which the participant was assigned.
Across both intervention conditions, each stimulus within the targeted set was present-
ed 10 times, in a random order, for a total of 20 trials in each intervention session.

Intervention

Error correction The error correction procedure in this study was identical to the
procedure used by Leaf et al. (2014b), which combined components of less intrusive
10 The Analysis of Verbal Behavior (2020) 36:1–20

error correction procedures (i.e., differential reinforcement for correct responses and
error statements; Cariveau et al., 2018) with verbally stating the correct response. This
type of error correction was selected based on its low level of intrusiveness, documented
effectiveness, and common use within the agency in which this study took place. Each
trial began with the researcher presenting the stimulus (i.e., picture of the comic book
character) and providing an instruction (e.g., “What is his/her name?”). If the participant
engaged in an independent correct response within 5 s, the researcher provided praise
(e.g., “Nailed it!”). Praise was selected because conditioning praise as a reinforcer was
part of all of the participants’ programming at some point, and, as a result, it had been
reported by the participants’ case managers to be an effective reinforcer during the
participants’ sessions. If the participant engaged in an incorrect response or no response,
the researcher provided corrective and informative feedback (e.g., “No, it is Daredev-
il.”). Any aberrant behavior was ignored if it was nondangerous (e.g., crying, yelling)
and redirected if it was dangerous (e.g., aggression, self-injurious behavior). However,
at no point in time did any dangerous topographies of aberrant behavior occur.

Errorless learning Most-to-least prompting was used to minimize errors within the
errorless learning condition. The researchers selected most-to-least prompting based
on the research indicating that it is often more effective and efficient than other errorless
prompting systems (Day, 1987; McDonnell & Ferguson, 1989; Miller & Test, 1989).
The most-to-least prompting procedure used within this study consisted of a three-step
prompt hierarchy: (a) a full echoic prompt (i.e., stating the full comic book character’s
name; e.g., “Carnage”), (b) a partial echoic prompt (i.e., stating part of the comic book
character’s name; e.g., “Car”), and (c) no prompt. The criterion to move to a less
assistive prompt was a correct response on two consecutive teaching trials. The
criterion to move to a more assistive prompt was an incorrect response on any teaching
trial. Each stimulus within a set was on its own prompt hierarchy, and movement on the
prompting hierarchy for each target was independent of responding on the other target
within the set. On subsequent sessions, the researcher began with the prompt level from
the previous session. If the participant engaged in an independent or prompted correct
response within 5 s, the researcher provided praise (e.g., “You got it!”). If the partic-
ipant engaged in an independent or prompted incorrect response or no response, the
researcher provided corrective feedback (e.g., “No.”). Any aberrant behavior was
ignored if it was nondangerous (e.g., crying, yelling) and redirected if it was dangerous
(e.g., aggression, self-injurious behavior). However, at no point in time did any
dangerous topographies of aberrant behavior occur.

Interobserver Reliability

A second researcher independently (in vivo or via videotapes) recorded participant

responding during 45.8% of pre- and postprobes, 43.9% of daily probes, 38.6% of error
correction teaching sessions, and 47.5% of errorless learning conditions. Interobserver
agreement (IOA) was calculated by totaling the number of agreements divided by the
number of agreements plus disagreements and multiplying by 100. Agreements consisted
of trials in which both observers scored the same participant response (i.e., correct,
incorrect, prompted correct, prompted incorrect, no response, and aberrant behavior;
The Analysis of Verbal Behavior (2020) 36:1–20 11

e.g., if both observers scored that aggression occurred on the same trial). Disagreements
consisted of trials in which both observers scored different participant responses (i.e.,
correct, incorrect, prompted correct, prompted incorrect, no response, and aberrant
behavior; e.g., if one observer scored a correct response and the other observer scored
a prompted correct response). IOA across all sessions and participants during pre- and
postprobes was 100%, during daily probes IOA was 99.8% (range 67%–100% across
sessions), during error correction IOA was 99.4% (range 85%–100% across sessions),
and during errorless learning IOA was 99.9% (range 95%–100% across sessions).

Treatment Fidelity

Treatment fidelity was taken during 45.2% of pre- and postprobes, 44.6% of daily
probes, 33% of error correction sessions, and 47.1% of errorless learning sessions.
Correct researcher behaviors on each probe trial (i.e., pre-, post-, and daily probes)
consisted of (a) presenting the stimulus (i.e., picture of the comic book character), (b)
providing an instruction (e.g., “What is his/her name?”), (c) providing up to 5 s for the
participant to respond, and (d) providing neutral feedback regardless of accuracy.
Correct researcher behavior during each error correction teaching trial consisted of (a)
presenting the stimulus (i.e., picture of the comic book character), (b) providing an
instruction (e.g., “What is his/her name?”), (c) providing up to 5 s for the participant to
respond, (d) providing praise (e.g., “You got it!”) following a correct response, and (e)
providing corrective and informative feedback (e.g., “No, it is Daredevil.”) following an
incorrect response or no response. Correct researcher behavior during each errorless
learning teaching trial consisted of (a) presenting the stimulus (i.e., picture of the comic
book character), (b) providing an instruction (e.g., “What is his/her name?”), (c)
providing the correct level of prompt, (d) providing up to 5 s for the participant to
respond, (e) providing praise (e.g., “You got it!”) following correct and prompted correct
responses, and (f) providing corrective feedback (e.g., “Nope.”) following an incorrect
response or no response. Treatment fidelity was calculated by dividing the number of
trials during which the researcher engaged in all of the correct steps by the total number
of trials and multiplying by 100. Treatment fidelity was 99.9% (range 94.4%–100%
across sessions) during pre- and postprobes, 99.9% (range 83.3%–100% across ses-
sions) during daily probes, and 100% during error correction teaching trials and errorless
learning trials.

Design

The two intervention approaches were assessed using a two-group RCT. An RCT differs
from a randomized control trial in that each group receives treatment and comparisons are
made on the relative effectiveness of each treatment (Abbott, 2014). This design could be
compared to an alternating-treatments design (Barlow & Hayes, 1979) in that both are
designed to compare the effects of two or more interventions, procedures, or approaches.

Statistical Analyses

Initial analyses of differences between conditions were conducted via an independent-

samples t test to determine the potential for statistical differences across participants in
12 The Analysis of Verbal Behavior (2020) 36:1–20

the two groups in terms of independent correct responding, independent incorrect

responding, prompted correct responding, prompted incorrect responding, total correct
responding (i.e., independent and prompted), total incorrect responding (i.e., indepen-
dent and prompted), and aberrant behavior. An independent-samples t test is designed
to compare means between two groups where there are different participants in each
group. If, however, the Shapiro–Wilk test of normality indicated the data were not
amenable to an independent-samples t test, a Mann–Whitney rank sum test was used
instead. This is a common approach if the variances are not equal, which requires a
variance-stabilizing transformation or a modification of the t test. The Mann–Whitney
rank sum test is a nonparametric test to compare outcomes between two independent
groups. All statistical analyses were conducted using SigmaPlot, Version 13.0 (Systat
Software, San Jose, CA).

Results

Full Probes

Figure 1 displays the results for the main dependent measure, the mean percentage of
correct responding for all participants during full probes. Prior to intervention, the mean
percentage of correct responding was 0% (as planned, to ensure no prior learning
history) across the groups for the targeted stimuli with no difference between groups.
Therefore, it was only necessary to evaluate postperformance data. Following inter-
vention, the mean percentage of correct responding was 85.7% and 77.1% for partic-
ipants in the error correction and errorless learning conditions, respectively. When
comparing postperformance on full probes between the two conditions, a Mann–
Whitney U test indicated there was a significant difference, U = 53, p = .039, in correct
responding for participants in the error correction condition when compared to partic-
ipants in the errorless learning condition.

Number of Stimulus Sets With Which Participants Reached the Mastery Criterion

A total of 42 sets was introduced across the participants in each condition (i.e., three sets
per participant with 14 participants in each condition). All but one participant (i.e., EC12,
who reached the mastery criterion on two of the three sets) reached the mastery criterion in
the error correction condition for all three sets, for a total of 97.6% of sets mastered (i.e., 41
out of 42 sets). Similarly, all but one participant (i.e., EL6, who reached the mastery
criterion for one of the three sets) reached the mastery criterion in the errorless teaching
condition for all three sets, for a total of 95.2% of sets mastered (i.e., 40 out of 42 sets).

Responding During Teaching

Table 2 provides a summary of responding during intervention sessions across the two
conditions. The mean percentage of independent correct responding was 90.8% and
75.3% across the error correction and errorless learning conditions, respectively. The
results of an independent-samples t test indicated there was a significant difference in
independent correct responding between the error correction condition (M = 90.8; SD =
The Analysis of Verbal Behavior (2020) 36:1–20 13

100 Pre
Post

Percentage Correct 80

60

40

20

*
1

9
10

11

13

ER 4
E
12
EC

EC

EC

EC

EC

EC

EC

EC

EC

A EC1

G
EC

EC

EC

A
EC

V
100

80
Percentage Correct

60

40

20

0
*
1

9
10

11

12

13

ER 4
E
EL

EL

EL

EL

EL

EL

EL

EL
6*

A EL1

G
EL

EL

EL

EL

A
EL

Fig. 1. The mean percentage of independent correct responding for each participant and condition during pre-
and postprobes. The number of asterisks indicates the number of sets in which the participant did not reach the
mastery criterion. EC = error correction; EL = errorless learning

5.575) and the errorless learning condition (M = 75.3; SD = 11.786), t(26) = 4.456; p <
.001. The mean percentage of independent incorrect responding was 9.2% and 2.9%
across the error correction and errorless learning conditions, respectively. A Mann–
Whitney U test indicated there was a significant difference, U = 19, p < .001, in
independent incorrect responding for participants in the error correction condition when
compared to participants in the errorless learning condition.
The mean percentage of prompted correct and prompted incorrect responding in the
errorless learning condition was 19.3% and 2.6%, respectively. Data on prompted
responses in the error correction condition are not reported, given they could not occur
in that condition based on the procedures within that condition. The mean percentage of
overall correct responding (i.e., independent and prompted) was 90.8% and 94.6%
across the error correction and errorless learning conditions, respectively. A Mann–
Whitney U test indicated there was a significant difference, U = 52, p = .036, in overall
correct responding for participants in the errorless learning condition when compared to
participants in the error correction condition. This is not surprising, however, given that
it was not possible for prompted correct responses to occur in the error correction
condition. The mean percentage of overall incorrect responding was 9.2% and 5.5%
across the error correction and errorless learning conditions, respectively. A Mann–
14 The Analysis of Verbal Behavior (2020) 36:1–20

Table 2 The Percentage of Responding During Teaching Conditions Across the Two Conditions

Condition Independent Independent Prompted Prompted Total Total Aberrant

Correct Incorrect Correct Incorrect Correct Incorrect Behavior

Error 90.8% 9.2% 90.8% 9.2% 0.07%

correction
Errorless 75.3% 2.9% 19.3% 2.6% 94.6% 5.5% .55%
learning

Whitney U test indicated there was a significant difference, U = 52.5, p = .039, in

overall incorrect responding for participants in the error correction condition when
compared to participants in the errorless learning condition.
The mean percentage of aberrant behavior was 0.07% and 0.55% across the error
correction and errorless learning conditions, respectively. A Mann–Whitney U test
indicated there was no significant difference, U = 77.5, p = .169, in aberrant behavior
when comparing participants in the error correction condition to participants in the
errorless learning condition.

Sessions to Mastery

Figure 2 displays the number of sessions to mastery for each set of stimuli across all
participants in both conditions’ groups, as well as the average number of sessions to
mastery for each set of stimuli within each condition. The top panel displays the results
for participants in the error correction condition, and the bottom panel displays the
results for participants in the errorless learning condition. Asterisks next to a partici-
pant’s number indicate the participant did not reach mastery criterion on one or more
sets. The total number of sessions, then, represents the number of sessions to reach the
mastery criterion or the termination criterion (i.e., 20 sessions), whichever came first. A
total of 73, 68, and 89 sessions were required across the 14 participants in the error
correction condition for Sets 1, 2, and 3, respectively. A total of 79, 84, and 72 sessions
were required across the 14 participants in the errorless learning condition for Sets 1, 2,
and 3, respectively. Overall, participants required an average of 16 sessions to reach the
mastery criterion across all three stimulus sets in both conditions. A Mann–Whitney U
test indicated there was no significant difference, U = 87.5, p = .643, in the average
number of sessions required to reach the mastery criterion when comparing participants
in the error correction condition to participants in the errorless learning condition.

Discussion

The purpose of this study was to compare the relative effectiveness of an error
correction procedure to an errorless learning procedure (i.e., most-to-least prompting)
using an RCT to teach tact relations to 28 children diagnosed with ASD. In general, the
results of this study are consistent with previous research on error correction and
errorless learning procedures. Both procedures were effective with respect to stimulus
sets mastered and resulted in significant changes from pre- to postprobes. Second, both
The Analysis of Verbal Behavior (2020) 36:1–20 15

20
SET 1
SET 2
SET 3
15
Number of Sessions

10

*
1

9
10

11

13

ER 4
E
12
EC

EC

EC

EC

EC

EC

EC

EC

EC

A EC1

G
EC

EC

EC

A
EC

V
20
SET 1
SET 2
SET 3
15
Number of Sessions

10

0
*
1

9
10

11

12

13

ER 4
E
EL

EL

EL

EL

EL

EL

EL

EL
6*

A EL1

G
EL

EL

EL

EL

A
EL

Fig. 2. The total number of sessions to reach the mastery criterion for each participant for each set and the
average per set across conditions. The number of asterisks indicates the number of sets in which the participant
did not reach the mastery criterion

procedures were efficient with respect to the number of sessions required to reach the
mastery criterion, with no significant differences between the two conditions. Third,
participants in both conditions engaged in minimal rates of aberrant behavior, with no
significant differences between the two conditions.
However, there were differences worth mentioning. First, the results indicated that
performance on postprobes for participants assigned to the error correction condition
was significantly better when compared to the errorless learning condition. Second, the
results indicated that independent correct responding during teaching for participants in
the error correction condition was significantly higher than in the errorless learning
condition. These results are consistent with previous research comparing error correc-
tion procedures to errorless learning procedures. Researchers have compared error
correction procedures to errorless learning procedures to teach a variety of behaviors
to individuals diagnosed with ASD and have found error correction procedures to be
16 The Analysis of Verbal Behavior (2020) 36:1–20

either more effective, more efficient, or both more effective and efficient than errorless
learning procedures (e.g., Fentress & Lerman, 2012; Leaf et al., 2010).
This study contributes to the existing literature on error correction and errorless
learning procedures through the use of a group design. To our knowledge, this is the
first study to evaluate error correction and errorless learning procedures using an RCT.
Previous studies have evaluated these procedures in isolation or through comparison
using single-subject experimental designs, which remains an important research en-
deavor. This study adds to the natural progression of research described by Smith
(2012) and Smith et al. (2007), in which group research is informed through previous
studies a using single-subject methodology. As such, the use of an RCT can help
disseminate research related to the behavioral science of language to a larger audience
and contribute to the external validity of behavior-analytic procedures for teaching tact
relations to individuals diagnosed with ASD (Campbell & Stanley, 1963; Kazdin,
2011).
The findings of this study, as well as other studies, have implications for clinical
practice with respect to ASD intervention. First, when using similar instructional
preparations (e.g., introducing targets in sets of two), the use of the error correction
procedure evaluated in this study may lead to acquisition with more independent
correct responses. Second, concerns about errors during the acquisition of tact relations
being detrimental to the individual or to the learning process appear unsupported based
on these results. Although it appears that the unacceptability of errors is accepted
dogma, the results of the present study indicated that error correction procedures can
result in more independent correct responding than errorless learning procedures can.
Third, it appears that error correction, as a punishment procedure, can be used without
evoking undesired aberrant behavior commonly associated with punishment-based
procedures. In this study, the occurrence of aberrant behavior (e.g., aggression, crying,
self-injury) was comparable and extremely rare across both conditions. Moreover, the
rare instances of aberrant behavior involved topographies that would be considered
nondangerous (e.g., yelling). Therefore, the use of error correction procedures should
not be avoided simply because punishment-based procedures are commonly associated
with undesired side effects, as that is inconsistent with the present study, as well as with
others (e.g., Leaf et al., 2019; Lovaas & Simmons, 1969; Matson & Taras, 1989).
Despite the contributions of this study to the literature base and its clinical implications,
there are several limitations that could be addressed in future evaluations. First, this study
included participants with sophisticated vocal-verbal repertoires. It remains unclear wheth-
er similar results would be obtained with participants who have more limited vocal-verbal
repertoires or more cognitive impairments (e.g., IQ < 70). Future RCTs evaluating error
correction and errorless learning procedures should include individuals differing from
those included in this study with respect to scores on standardized assessments, cognitive
functioning (e.g., IQ), and the presence of challenging behavior (e.g., aggression).
Second, this study evaluated an error correction procedure and an errorless learning
procedure to teach only one skill (i.e., tacting, sometimes referred to as expressive
labeling). Therefore, it remains unknown if the results would be replicated when
teaching other skills (e.g., listener behavior, sometimes referred to as receptive label-
ing). Despite this limitation, previous studies have compared error correction proce-
dures to errorless learning procedures to teach other skills (e.g., Leaf et al., 2010);
however, these studies used single-subject experimental designs. As such, a potentially
The Analysis of Verbal Behavior (2020) 36:1–20 17

fruitful area for future research would be using an RCT to compare error correction
procedures to errorless learning procedures to teach other skills.
Third, this study only evaluated one method of errorless learning (i.e., most-to-least
prompting) and one method of error correction. As previously noted, there are several
errorless learning techniques, including stimulus fading, stimulus shaping, response
prevention, delayed prompting, superimposition with stimulus fading, and superimposi-
tion with stimulus shaping. Similarly, there are several error correction procedures not
evaluated here, such as demonstration, remove and re-present, and multiple response
repetition (Cariveau et al., 2018). There are also other prompting systems that use
techniques other than the one evaluated within this study, such as CTD, PTD, simulta-
neous prompting, stimulus shaping, and stimulus fading. Therefore, future researchers
should continue to evaluate the relative effectiveness of other error correction and errorless
learning procedures using group designs.
Fourth, it is likely that within clinical settings, error correction and errorless learning
procedures are not used in isolation from one another but are, rather, combined in one
teaching package (e.g., flexible prompt fading; Soluaga, Leaf, Taubman, McEachin, &
Leaf, 2008). Given that some have discussed that “near-errorless” learning is a more
descriptive term for errorless learning (e.g., Fillingham, Hodgson, Sage, & Ralph, 2003),
clinicians still need to respond in some way to errors. For instance, error correction could
be used in combination with a sufficient frequency of proactive prompting to guard against
the occurrence of excessive errors. Future researchers should evaluate the relative effec-
tiveness of various combinations of errorless learning and error correction procedures.
Finally, the methods employed in this study do not permit an evaluation of the possible
mechanisms responsible for the outcomes. Participants in the error correction condition
could have engaged in more independent correct responses as a result of corrective
feedback. It is likely that this corrective feedback functioned as a conditioned punisher.
This could have increased the likelihood of correct responses to avoid the corrective
feedback. In a sense, the participants engaged in signaled avoidance following contacting
the contingencies surrounding incorrect responses. Alternatively, no presumed condi-
tioned punishers were in effect contingent on incorrect responses during the errorless
learning condition. Both conditions resulted in the same outcome for correct responses.
Therefore, perhaps the combination of reinforcing and punishing contingencies created
the ideal context for more independent correct responses. Future studies utilizing single-
subject designs will likely be necessary to fully evaluate the mechanisms responsible for
the effectiveness of the two procedures evaluated within this study.
Despite these limitations, this study provides one of the few group designs evaluating
behavioral approaches to language interventions. Given the importance of group designs
outside of the field of behavior analysis (Kazdin, 2011; Smith, 2012; Smith et al., 2007),
behavior-analytic researchers should continue to provide research using designs that speak
to a larger audience (Smith, 2012; Smith et al., 2007). We hope that this study functions as
a motivating operation for other behavior-analytic researchers to design and conduct group
research designs evaluating other procedures based on the scientific foundation of behav-
ior analysis. The results of such studies will have many important implications for
clinicians and researchers and improve the quality of services available to those in need.

Author Note Justin B. Leaf, Autism Partnership Foundation, Seal Beach, California, and
Endicott College Institute for Behavioral Studies; Joseph H. Cihon, Autism Partnership
18 The Analysis of Verbal Behavior (2020) 36:1–20

Foundation, Seal Beach, California, and Endicott College Institute for Behavioral Studies;
Julia L. Ferguson, Autism Partnership Foundation, Seal Beach, California; Christine M.
Milne, Autism Partnership Foundation, Seal Beach, California, and Endicott College
Institute for Behavioral Studies; Ronald Leaf, Autism Partnership Foundation, Seal Beach,
California; John McEachin, Autism Partnership Foundation, Seal Beach, California.

Funding No funding was received for this study.

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 research committee and/or national research committee with the
1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent Informed consent was obtained from the parents of all individual participants included
in the study.

References

Abbott, J. H. (2014). The distinction between randomized clinical trials (RCTs) and preliminary feasibility and
pilot studies: What they are and are not. Journal of Orthopaedic & Sports Physical Therapy, 44, 555–558.
https://doi.org/10.2519/jospt.2014.0110
Barbera, M. L., & Rasmussen, T. (2007). The verbal behavior approach: How to teach children with autism
and related disorders. London, UK: Kingsley.
Barlow, D. H., & Hayes, S. C. (1979). Alternating treatments design: One strategy for comparing the effects of
two treatments in a single subject. Journal of Applied Behavior Analysis, 12, 199–210. https://doi.
org/10.1901/jaba.1979.12-199
Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Boston,
MA: Houghton Mifflin Company.
Cariveau, T., La Cruz Montilla, A., Gonzalez, E., & Ball, S. (2018). A review of error correction procedures
during instruction for children with developmental disabilities. Journal of Applied Behavior Analysis, 3,
217. https://doi.org/10.1002/jaba.524
Carroll, R. A., Joachim, B. T., St. Peter, C. C., & Robinson, N. (2015). A comparison of error-correction
procedures on skill acquisition during discrete-trial instruction. Journal of Applied Behavior Analysis, 48,
257–273. https://doi.org/10.1002/jaba.205
Carroll, R. A., Owsiany, J., & Cheatham, J. M. (2018). Using an abbreviated assessment to identify effective
error-correction procedures for individual learners during discrete-trial instruction. Journal of Applied
Behavior Analysis, 51, 482–501. https://doi.org/10.1002/jaba.460
Day, H. M. (1987). Comparison of two prompting procedures to facilitate skill acquisition among severely
mentally retarded adolescents. American Journal of Mental Deficiency, 91, 366–372.
Demchak, M. (1990). Response prompting and fading methods: A review. American Journal of Mental
Retardation, 94, 603–615.
Dunn, L. M., & Dunn, D. M. (2007). Peabody picture vocabulary test (4th ed.; PPVT-IV). San Antonio, TX:
Pearson. https://doi.org/10.1037/t15144-000
Fentress, G. M., & Lerman, D. C. (2012). A comparison of two prompting procedures for teaching basic skills
to children with autism. Research in Autism Spectrum Disorders, 6, 1083–1090 https://doi.org/10.1016/j.
rasd.2012.02.006.
Ferster, C. B., & DeMeyer, M. K. (1962). A method for the experimental analysis of behavior of autistic
children. The American Journal of Orthopsychiatry, 32, 89–98. https://doi.org/10.1111/j.1939-0025.1962.
tb00267.x
The Analysis of Verbal Behavior (2020) 36:1–20 19

Fillingham, J. K., Hodgson, C., Sage, K., & Ralph, L. (2003). The application of errorless learning to aphasic
disorders: A review of theory and practice. Neuropsychological Rehabilitation, 13, 337–363. https://doi.
org/10.1080/09602010343000020
Gast, D. L. (2011a). An experimental approach for selecting a response-prompting strategy for children with
developmental disabilities. Evidence-Based Communication Assessment and Intervention, 5, 149–238.
https://doi.org/10.1080/17489539.2011.637358
Gast, D. L. (2011b). A rejoinder to Leaf: What constitutes efficient, applied, and trial and error. Evidence-
Based Communication Assessment and Intervention, 5, 234–238. https://doi.org/10.1080
/17489539.2012.689601
Green, G. (1996). Early behavioral intervention for autism: What does research tell us? In C. Maurice, G.
Green, & S. C. Luce (Eds.), Behavioral intervention for young children with autism: A manual for parents
and professionals (pp. 29–44). Austin, TX: Pro-Ed.
Green, G. (2001). Behavior analytic instruction for learners with autism advances in stimulus control
technology. Focus on Autism and Other Developmental Disabilities, 16, 72–85. https://doi.org/10.1177
/108835760101600203
Greer, R. D., & Ross, D. E. (2008). Verbal behavior analysis: Inducing and expanding complex communi-
cation in children with severe language delays. Boston, MA: Allyn & Bacon.
Grow, L. L., Kodak, T., & Clements, A. (2017). An evaluation of instructive feedback to teach play behavior
to a child with autism spectrum disorder. Behavior Analysis in Practice, 10, 313–317. https://doi.
org/10.1007/s40617-016-0153-9
Kayser, J. E., Billingsley, F. F., & Neel, R. S. (1986). A comparison of in-context and traditional instructional
approaches: Total task, single trial versus backward chaining, multiple trials. Journal of the Association
for Persons With Severe Handicaps, 11, 28–38. https://doi.org/10.1177/154079698601100104
Kazdin, A. E. (2011). Single-case research designs: Methods for clinical and applied settings (2nd ed.).
Oxford, UK: Oxford University Press.
Kodak, T., Fuchtman, R., & Paden, A. (2012). A comparison of intraverbal training for children with autism.
Journal of Applied Behavior Analysis, 45, 155–160. https://doi.org/10.1901/jaba.2012.45-155
LaVigna, G. W., & Donnellan, A. M. (1986). Alternatives to punishment: Solving behavior problems with
non-aversive strategies. New York, NY: Irvington.
Leaf, J. B., Alcalay, A., Leaf, J. A., Tsuji, K., Kassardjian, A., Dale, S., et al. (2014a). Comparison of most-to-
least to error correction for teaching receptive labeling for two children diagnosed with autism. Journal of
Research in Special Education Needs, 16, 217–225. https://doi.org/10.1111/1471-3802.12067
Leaf, J. B., Leaf, J. A., Alcalay, A., Dale, S., Kassardjian, A., Tsuji, K., et al. (2014b). Comparison of most-to-
least to error correction to teach tacting to two children diagnosed with autism. Evidence-Based
Communication Assessment and Intervention, 7, 124–133. https://doi.org/10.1080
/17489539.2014.884988
Leaf, J. B., Sheldon, J. B., & Sherman, J. A. (2010). Comparison of simultaneous prompting and no-no
prompting in two-choice discrimination learning with children with autism. Journal of Applied Behavior
Analysis, 43, 215–228. https://doi.org/10.1901/jaba.2010.43-215
Leaf, J. B., Townley-Cochran, D., Cihon, J. H., Mitchell, E., Leaf, R., Taubman, M., & McEachin, J. (2019).
A descriptive analysis of the use of punishment-based techniques with children diagnosed with autism
spectrum disorder. Education and Training in Autism and Developmental Disorders. Advance online
publication. https://eric.ed.gov/?id=EJ1217376
Ledford, J. R., Gast, D. L., Luscre, D., & Ayres, K. M. (2008). Observational and incidental learning by
children with autism during small group instruction. Journal of Autism and Developmental Disorders, 38,
86–103. https://doi.org/10.1007/s10803-007-0363-7.
Lerman, D. C., & Vorndran, C. M. (2002). On the status of knowledge for using punishment: Implications for
treating behavior disorders. Journal of Applied Behavior Analysis, 35, 431–464. https://doi.org/10.1901
/jaba.2002.35-431.
Lovaas, O. I. (1987). Behavioral treatment and normal educational and intellectual functioning in young
autistic children. Journal of Consulting and Clinical Psychology, 55, 3–9. https://doi.org/10.1037/0022-
006X.55.1.3.
Lovaas, O. I. (2003). Teaching individuals with developmental delays: Basic intervention techniques. Austin,
TX: Pro-Ed.
Lovaas, O. I., & Simmons, J. Q. (1969). Manipulation of self-destruction in three retarded children. Journal of
Applied Behavior Analysis, 2, 143–157. https://doi.org/10.1901/jaba.1969.2-143.
Majdalany, L., Wilder, D. A., Smeltz, L., & Lipschultz, J. (2016). The effect of brief delays to reinforcement
on the acquisition of tacts in children with autism. Journal of Applied Behavior Analysis, 49, 411–415.
https://doi.org/10.1002/jaba.282.
20 The Analysis of Verbal Behavior (2020) 36:1–20

Martin, N. A., & Brownell, R. (2010). Expressive one-word picture vocabulary test (4th ed.; EOWPVT-4).
Novato, CA: Academic Therapy Publications.
Matson, J. L., & Taras, M. E. (1989). A 20 year review of punishment and alternative methods to treat problem
behaviors in developmentally delayed persons. Research in Developmental Disabilities, 10, 85–104.
https://doi.org/10.1016/0891-4222(89)90031-0.
McDonnell, J., & Ferguson, B. (1989). A comparison of time delay and decreasing prompt hierarchy strategies
in teaching banking skills to students with moderate handicaps. Journal of Applied Behavior Analysis, 22,
85–91. https://doi.org/10.1901/jaba.1989.22-85.
McGhan, A. C., & Lerman, D. C. (2013). An assessment of error-correction procedures for learners with
autism. Journal of Applied Behavior Analysis, 46, 626–639. https://doi.org/10.1002/jaba.65.
Miller, U. C., & Test, D. W. (1989). A comparison of constant time delay and most-to-least prompting in
teaching laundry skills to students with moderate retardation. Education and Training in Mental
Retardation, 24, 363–370.
Moore, R., & Goldiamond, I. (1964). Errorless establishment of visual discrimination using fading procedures.
Journal of the Experimental Analysis of Behavior, 7, 269–272. https://doi.org/10.1901/jeab.1964.7-269.
Mueller, M., Palkovic, C. M., & Maynard, C. S. (2007). Errorless learning: Review and practical application
for teaching children with pervasive developmental disorders. Psychology in the Schools, 44, 691–700.
https://doi.org/10.1002/pits.20258.
Petursdottir, A. I., & Carr, J. E. (2011). A review of recommendations for sequencing receptive and expressive
language instruction. Journal of Applied Behavior Analysis, 44, 859–876. https://doi.org/10.1901
/jaba.2011.44-859.
Plaisance, L., Lerman, D. C., Laudont, C., & Wu, W. L. (2016). Inserting mastered targets during error
correction when teaching skills to children with autism. Journal of Applied Behavior Analysis, 49, 251–
264. https://doi.org/10.1002/jaba.292.
Schilmoeller, G. L., Schilmoeller, K. J., Etzel, B. C., & LeBlanc, J. M. (1979). Conditional discrimination after
errorless and trial-and-error training. Journal of the Experimental Analysis of Behavior, 31, 405–420.
https://doi.org/10.1901/jeab.1979.31-405.
Skinner, B. F. (1957). Verbal behavior. Englewood Cliffs, NJ: Prentice Hall.
Smith, T. (1999). Outcome of early intervention for children with autism. Clinical Psychology: Science and
Practice, 6, 33–49.
Smith, T. (2001). Discrete trial training in the treatment of autism. Focus on Autism and Other Developmental
Disabilities, 16, 86–92. https://doi.org/10.1177/108835760101600204.
Smith, T. (2012). Evolution of research on interventions for individuals with autism spectrum disorder:
Implications for behavior analysts. The Behavior Analyst Today, 35, 101–113. https://doi.org/10.1007
/BF03392269.
Smith, T., Scahill, L., Dawson, G., Guthrie, D., Lord, C., Odom, S., et al. (2007). Designing research studies
on psychosocial interventions in autism. Journal of Autism and Developmental Disorders, 37, 354–366.
https://doi.org/10.1007/s10803-006-0173-3.
Soluaga, D., Leaf, J. B., Taubman, M., McEachin, J., & Leaf, R. (2008). A comparison of flexible prompt
fading and constant time delay for five children with autism. Research in Autism Spectrum Disorders, 2,
753–765. https://doi.org/10.1016/j.rasd.2008.03.005.
Sparrow, S. S., Cicchetti, D. V., & Balla, D. A. (2005). Vineland adaptive behavior scales (2nd ed.; Vineland-
II). Circle Pines, MN: American Guidance Service. https://doi.org/10.1037/t15164-000.
Sulzer-Azaroff, B., & Mayer, R. G. (1991). Behavior analysis for lasting change. Fort Worth, TX: Harcourt
Brace College Publishers.
Sundberg, M. L., & Partington, J. W. (1998). Teaching language to children with autism or other develop-
mental disabilities. Pleasant Hill, CA: Behavior Analysts.
Townley-Cochran, D., Leaf, J. B., Leaf, R., Taubman, M., & McEachin, J. (2017). Comparing error correction
procedures for children diagnosed with autism. Education and Training in Autism and Developmental
Disabilities, 52, 91–101.
Walker, G. (2008). Constant and progressive time delay procedures for teaching children with autism: A
literature review. Journal of Autism and Developmental Disorders, 38, 261–275. https://doi.org/10.1007
/s10803-007-0390-4.
Worsdell, A. S., Iwata, B. A., Dozier, C. L., Johnson, A. D., Neidert, P. L., & Thomason, J. L. (2005). Analysis
of response repetition as an error-correction strategy during sight-word reading. Journal of Applied
Behavior Analysis, 38, 511–527. https://doi.org/10.1901/jaba.2005.115-04.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.