You are on page 1of 19

Behavioral Interventions

Behav. Intervent. 25: 109–127 (2010)


Published online 25 February 2010 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/bin.299

RESPONSE REPETITION AS AN ERROR-CORRECTION


PROCEDURE FOR SIGHT-WORD READING:
A REPLICATION AND EXTENSION

Kendra L. Marvin, John T. Rapp*, Michelle T. Stenske, Nairim R. Rojas,


Greg J. Swanson and Sara M. Bartlett
St. Cloud State University, St. Cloud, MN, USA

We evaluated the effects of response repetition (RR) as an error-correction procedure for increasing
sight-word reading for four individuals using a multiple baseline design. The results for each participant
showed that correct responding increased following the introduction of the RR procedure. In addition,
responding for three of four participants showed evidence of stimulus generalization, response
generalization, or both. Likewise, we found that comparable behavior changes were produced when
a participant’s paraprofessional implemented RR in a typical classroom. These findings are briefly
discussed in terms of procedures that promote behavior acquisition and generalization via negative
reinforcement. Copyright # 2010 John Wiley & Sons, Ltd.

In recent years, the need to improve reading skills of young students has become a
primary area of focus for our nation’s public education system (e.g., National Center
for Education Statistics, 2002; National Reading Panel, 2000). Although a number of
recent behavior-analytic studies have evaluated the effects of combinations of
antecedent- and consequent-based procedures on reading performance in typically
functioning children (e.g., Daly, Bonfiglio, Mattson, Persampieri, & Foreman-Yates,
2005; Eckert, Ardoin, Daly, & Martens, 2002), very few of these studies have
evaluated the specific role of error-correction procedures for increasing correct
responses. Typically, error-correction procedures are implemented in conjunction
with contingent praise (or other potentially reinforcing events) for correct responding.
Magee and Ellis (2006) evaluated college students’ performance on a matching
task when praise was provided for correct responses and error correction (i.e., a trainer
demonstrated the correct response) was provided for (a) no responses (i.e., the
absence of an active response within 10 s of instruction), (b) incorrect responses, (c)
both (a) and (b), and (d) neither (a) nor (b). In general, Magee and Ellis found that

*Correspondence to: John T. Rapp, St. Cloud State University, Education Building, A261, 720 Fourth Avenue
South, St. Cloud, MN 56301-4498, USA. E-mail: jtrapp@stcloudstate.edu

Copyright # 2010 John Wiley & Sons, Ltd.


110 K. L. Marvin et al.

participants acquired the correct responses with fewer training trials when error
correction was implemented for no responses, incorrect responses, or both; however,
scores on retention probes, which were conducted 1 week after achieving the mastery
criterion, were higher for individuals with whom error correction was not
implemented for incorrect responses. In a follow-up experiment, Magee and Ellis
found that nine of ten participants achieved the mastery criterion in fewer trials when
they were permitted to mand for error correction than when error correction was not
provided for incorrect responses. Unlike the previous experiment, however, the
authors found that performance on the retention probes was high and comparable for
responses that were acquired with or without error correction. These findings suggest
that error-correction procedures facilitate more rapid acquisition of new responses.
Nevertheless, the generality of the findings from the Magee and Ellis study is
somewhat limited because the training task required each participant to select one
correct stimulus from an array of ten stimuli. Therefore, acquiring one correct
response increased the probability of acquiring another correct response by reducing
the number of unknown available alternatives. In addition, participants were only
required to point at the correct response. Thus, it is not clear if the same results would
be produced for training tasks involving an expressive response within an unrestricted
number of alternative operants (e.g., providing the correct verbal response when
presented a sight-word).
Worsdell, Iwata, Dozier, Johnson, Neidert, and Thomason (2005) speculated that
error-correction procedures may increase correct responding by, in part, providing
additional opportunities for the correct response to occur in the presence of a
respective antecedent. To evaluate this possibility, Worsdell et al. conducted a series
of studies to evaluate the effects of an error-correction procedure of studies involving
response repetition (RR) for incorrect responses during a sight-word reading task for
several adults diagnosed with mental retardation and other developmental disabilities.
The results of the first experiment showed that participants acquired more sight-
words when incorrect responses were followed by multiple-response (MR) repetition
than when incorrect responses were followed by single-response (SR) repetition.
This finding was inconsistent with results from a study by Ferkis, Belifiore, and
Skinner (1997) wherein both SR and MR methods produced comparable increases in
sight-word reading for three participants. The results of Worsdell et al.’s second
experiment showed that participants acquired more sight-words when a continuous
schedule of MR repetition was implemented as a consequence for incorrect responses
as opposed to an intermittent schedule of MR repetition. In the final experiment of
this investigation, the authors found that error-correction procedures with MR
repetition using either a relevant word (i.e., repeating the correct word) or an
irrelevant word (i.e., repeating a word that was not related to the correct response)
increased the participants’ correct sight-word reading; however, some participants

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 111

showed better acquisition when relevant words were used in the MR repetition
procedure. This finding suggests that response acquisition produced with the MR
procedure is likely attributable to the learner’s avoidance of the stimulation imposed
by error correction, rather than additional opportunities for the learner to practice
the correct response. Based on the results of these experiments, Worsdell et al.
suggest that error-correction procedures should (a) be delivered on a continuous basis
and (b) involve at least five repetitions of the relevant word (i.e., the word for which
the error occurred).
Although the literature suggests that error-correction procedures involving RR
produce relatively rapid acquisition of sight-word reading, it is not clear whether such
procedures promote generalized increases in sight-word reading. Stimulus
generalization is typically defined as the spread of the effect of reinforcement from
a discriminative stimulus to other stimuli, whereas response generalization is
typically defined as the spread of the effect of reinforcement to other behavior for
which reinforcement has not been provided (e.g., Catania, 2007; Martin & Pear,
2007). As it pertains to acquisition of academic skills such as reading, interventions
that promote both stimulus generalization and response generalization are highly
desirable. For example, following intervention, an individual who was taught to read
sight-words from flashcards should also be able to read those same words within the
context of a sentence that is written on a sheet of paper or on the page of a book.
Likewise, an individual who learns to read the sight-word ‘what’ may also be able to
read the sight-word ‘when’. To our knowledge, no study has evaluated the extent to
which error correction with RR procedures, alone, promotes generalized changes in
individuals’ reading repertoires.
Although the results of the Worsdell et al. (2005) investigation suggest that
acquisition of correct responding was attributable to negative reinforcement in the
form of avoiding the error-correction procedure, the results are potentially limited for
at least two reasons. First, Worsdell et al. evaluated the effects of the MR repetition
procedures using a multiple baseline across participants design wherein the
dependent measure was the cumulative number of words mastered by the respective
participant. Although this experimental design was appropriate for determining
whether SR or MR repetition generated faster acquisition of sight-words, it was
limited insofar as it did not evaluate the extent to which the acquisition of a given
word or set of words produced increases in untargeted words (i.e., response
generalization) or correct responses to targeted words in untrained contexts (i.e.,
stimulus generalization). Second, Worsdell et al. did not evaluate the extent to which
the error-correction procedures could be successfully implemented by direct-care
staff or school personnel (e.g., paraprofessionals). Even if a procedure is empirically
shown to facilitate generalized behavior change, such change will not occur unless
change agents implement the procedure correctly. Thus, researchers should evaluate

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
112 K. L. Marvin et al.

the extent to which classroom trainers can implement RR procedures with


comparable effectiveness (Ferkis et al., 1997).
The purpose of this study was to replicate and extend the results of the Worsdell
et al. (2005) study by evaluating the effects of the MR repetition procedure for error-
correction using a concurrent multiple baseline across sets (of sight-word cards)
design with multiple participants. In this way, we evaluated the extent to which
responses that were acquired using MR procedures generalized to untrained settings,
untrained responses, or both.

METHODS

Participants, Setting, and Materials


Four children participated in this study. Grant was a 9-year-old male who was
diagnosed with Asperger’s disorder. Tony was a 7-year-old male who was diagnosed
with Attention Deficit Hyperactivity Disorder and was taking Adderall, Prozac, and
Clonidine. Leah was an 11-year-old female who was an English-as-a-second-
language student; she did not have a formal diagnosis. Dan was a 12-year-old male
who was diagnosed with moderate mental retardation. Each participant was selected
based on teacher or parent referral for conduct problems in the classroom, a need for
additional assistance in the area of reading, or both. Dan received educational services
in a special education classroom. Grant and Tony attended a general education
classroom and received one-to-one support from a paraprofessional throughout the
school day. Leah attended a general education classroom and did not receive
additional educational support.
Teaching sessions were conducted by graduate and undergraduate students in a
designated training room in the respective participant’s school or home (Dan only).
The primary observer conducted the session while seated across a table or desk from
the participant and a secondary observer sat within 1 m of the participant. Some of the
training sessions were video-taped to allow a secondary observer to score the session
at a later time. Each session consisted of two to four sets of sight-words. Sessions were
conducted once or twice per day, 2–4 days/week with each participant. Session length
varied from 15 to 20 min across participants.

Response Measurement and Reliability


The percent of correct responses was recorded during both the sight-word and
reading (generalization) assessments. The sight-words were obtained from a list
outlining the 100 sight-words that children should know upon the completion of first

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 113

grade (Instant Fry Word List). The words were arbitrarily divided into three or four
sets of 10 (Tony and Dan) to 20 (Leah and Grant) cards; the cards were presented in
those sets for the entire assessment. A sight-word that appeared in one set did not
appear in any other set (i.e., there were no duplicate cards). Sight-words were
individually printed in approximately 36-point font size on 400  600 white index cards.
Observers recorded data on correct and incorrect responses in vivo or via tape
recordings using a paper-and-recording method on a trial-by-trial basis. A correct
response was defined as orally reading the word provided on the respective index card
within 5 s of presentation. An incorrect response was defined as failing to provide the
correct oral response within 5 s of presentation (this included nonresponding). The
number of correctly read words was divided by the total number of words in each set
then multiplied by 100% to arrive at the percent of correct responses for each set in
every session. A second observer collected data during at least 33% of the total
sessions for each participant. Interobserver agreement (IOA) scores were calculated
on a trial-by-trial basis by comparing data (i.e., a score of correct or incorrect for a
given trial) collected by the primary observer to data collected by the secondary
observer. For each session, the number of agreements was divided by the total number
of trials and multiplied by 100%. The scores for each session were totaled and then
divided by the number of sessions for which IOA scores were collected. The mean
IOA scores were 95% or higher for each participant.

Procedures
Prior to each session, participants were brought to a designated training room. An
experimenter instructed the participant to sit at a desk or table and then she or he
began the session. All instances of problem behavior were ignored. After each
session, the experimenter escorted that participant back to his or her classroom or
ongoing activity (Dan only).

Experimental Design
The effects of RR as an error-correction procedure on correct reading of sight-
words were evaluated using in a multiple baseline across sets of sight-words design
for Grant, Tony, and Dan and the combination of a multiple baseline across sets and
reversal design for Leah. Within this design, stable responding was defined as 15%
or less variability across at least three consecutive data points without an increasing
or decreasing trend. Thus, the decision to implement the RR intervention for a given
set was based on the three or four most recent data points with that set. A given set
of cards was considered mastered when the participant achieved 80% correct
responding or higher for two of three consecutive sessions. We used this liberal

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
114 K. L. Marvin et al.

definition based on the assumption that additional training procedures might be


necessary to increase correct responding to 100% for some participants. Baseline
conditions were conducted until stable responding emerged in one or more sets and
then the intervention was implemented with one of those sets. Once responding with
the set in the intervention phase was mastered or stable, the intervention was
subsequently implemented for another stable set (it was not necessary for a previous
set to be mastered before intervention was provided on a subsequent set). With the
exception of Dan, this process continued for each participant until (a) all the sets
were in the intervention phase or (b) a substantial increase in the correct reading of
words in untargeted sets (in the baseline phase) was noted (i.e., direct intervention
was not required).

Procedures
Prior to exposing Leah to the conditions described below, we implemented
prompting and reinforcement procedures, which were based in part on studies by
Cuvo and Klatt (1992) and Coleman-Martin and Heller (2004) for teaching correct
sight-word reading and spelling, respectively, to increase her correct sight-word
reading. During the initial baseline condition, the trainer provided verbal praise for
correct answers and said ‘okay’ following incorrect answers. The first treatment phase
involved a zero-second delay phoneme prompt where a trainer presented a card to
Leah while simultaneously saying the word. After this prompt, alone, did not increase
the number of correct answers, another condition was implemented where a zero-
second delay phoneme prompt was provided and verbal praise was delivered for
correct imitation of the modeled (prompted) response. Subsequently, another
condition was implemented wherein the trainer provided a zero-second delay full-
word prompt and verbal praise for correct responding. In the next phase, the trainer
provided the full-word prompt for each card after a 3 s constant-time delay and
delivered praise for correct responding. After the procedures described above
produced only marginal increases in Leah’s correct, independent responding, we
exposed her to the RR procedure described below. After the RR procedure was shown
to increase correct reading of sight-words with Leah, the effects of the procedure were
evaluated with subsequent participants who were referred to our project.

Baseline (BL)
The procedures differed slightly across participants during this phase. For each
participant, an experimenter held up a flashcard and asked ‘what is this?’ Brief verbal
praise was provided if the correct answer was given orally (reading) within 5 s of the
initial presentation. If the participant self-corrected an incorrect response within 5 s of

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 115

the initial presentation, the word was counted as correct (this rarely occurred). If the
participant responded incorrectly or did not respond within 5 s of the card
presentation, the experimenter said ‘okay’. After both correct and incorrect responses,
the trainer immediately presented the next flashcard. After each session, the
experimenter shuffled the flashcards so that the words would not be presented in the
same order during the next session.
Because each participant (with the exception of Dan) typically engaged in problem
behavior when provided reading tasks in class, additional procedures were
implemented to decrease the probability that these individuals would verbally or
physically refuse to attend training sessions. For Leah, Tony, and Dan, sessions were
initially alternated between reading sight-words and having free access to preferred
items. Once a history of four or five sessions was established in this manner, sessions
with training and free access to toys were conducted at a 3:1 ratio. Because Grant had
a history of engaging in escape-maintained aggression and property destruction (as
indicated by the results of a brief functional analysis), additional procedures involving
noncontingent reinforcement (NCR) were implemented to increase his motivation to
participate in the training sessions. Specifically, tokens were delivered to Grant on a
response-independent basis throughout each session. That is, tokens were provided to
Grant after completing each set of sight-words, regardless of the number of his correct
and incorrect responses. At the end of each training session, Grant was permitted to
exchange each token for approximately 5 min of toy play.

Response Repetition (RR)


During this condition, the trainer continued to provide brief verbal praise for
correct responses; however, contingent on an incorrect response the trainer required
the participant to repeat the correct response five times (Worsdell et al., 2005).
Specifically, the experimenter prompted the participant to ‘say, ‘‘correct word’’’. This
process was then repeated four additional times, spaced 2–3 s apart, until the
participant stated the correct word five times while looking at the corresponding card.
Participants were typically compliant with this process, but occasionally attempted to
repeat the word too quickly or without looking at the card. In both scenarios, the
participant was redirected to follow the trainer’s instructions. As in the baseline
condition, each set of cards was shuffled after each session and the same procedures
were used to maintain the participants’ compliance with attending sessions.
After Grant mastered the initial five sets of sight-words, his paraprofessional
conducted training sessions with three additional sets of words in Grant’s classroom.
Prior to conducting training in the classroom, the paraprofessional observed several of
the initial sessions, which were conducted by the experimenter, and intermittently
collected data Grant’s performance. After achieving 100% reliability with the

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
116 K. L. Marvin et al.

experimenter for three consecutive sessions, the paraprofessional began conducting


sessions with Grant using the additional word sets. The paraprofessional also served
as the primary observer for the sessions for which she was the trainer (i.e., she
collected data as she conducted the trials) and an experimenter served as the
secondary observer for approximately 70% of the sessions. The experimenter
provided brief verbal praise to the paraprofessional for accurately recording data. The
paraprofessional erroneously implemented the RR intervention for set 7 while Grant’s
correct responding was on an increasing trend.

Generalization Assessment
The extent to which increases in sight-word reading that occurred during training
sessions generalized to other reading contexts was evaluated for Leah and Tony.
Leah’s classroom teacher conducted three standard reading assessments during the
course of her participation in the training. For each assessment, Leah was presented a
list of 100 words that she was required to read aloud to her teacher who subsequently
reported Leah’s performance to an experimenter. The word list contained high content
overlap (HCO) with the words that were used in the training sets (see Daly et al.,
2005). Specifically, HCO passages or lists contain a high percentage of words from
instructional phases. For Tony, an experimenter generated various reading passages
that each contained approximately 120 words from either trained (with RR) or
untrained (baseline) sets. Unlike Leah, there were several baseline sets for which
Tony was not exposed to the RR procedure during the course of the study; words from
those set comprised the untrained passages. By contrast, words from sets that were
trained with RR were included in the RR passages. For both types of passages, access
to a preferred item and praise or a break and praise were provided on a fixed ratio 30
schedule for reading words aloud (correctly or incorrectly) to an experimenter. A
secondary observer collected data for Tony’s reading in the same manner as described
above, however, IOA measures were not collected with Leah because the assessment
was conducted by her classroom teacher. The teacher was not informed about Leah’s
performance during training sessions until the end of the study. Due to their limited
availability, generalization assessments were not conducted with the other
participants.

RESULTS

Figure 1 shows the percentage of correct words read by Leah across set 4 (first
panel), set 3 (second panel), set 2 (third panel), set 1 (fourth panel), set 5 (fifth panel),
and the classroom generalization assessment (sixth panel). As previously mentioned,

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 117

Figure 1. Percent of sight-words read correctly by Leah across baseline and response repetition phases
for set 4 (first panel), set 3 (second panel), set 2 (third panel), set 1 (fourth panel), and set 5 (fifth panel).
Omitted data points in the baseline phase for set 4 denote the absence of the opportunity for independent
responding. Percent of sight-words read correctly by Leah during three classroom assessments that were
conduct by her teacher (sixth panel).

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
118 K. L. Marvin et al.

initial interventions involving prompts and contingent praise (sessions 4–56) yielded
a slight increase in Leah’s correct responding in set 4 (missing data points in the
baseline phase involved simultaneous prompts for which independent responses
could not be emitted). After RR was implemented (beginning in session 58) her
correct responding increased to 70% in set 4, but was relatively unchanged in the
other sets. When RR was introduced for set 3, correct responding steadily increased in
set 3 and increased further in set 4, but remained relatively unchanged in the other
sets. After Leah’s correct responding stabilized at approximately 50% with Set 3, RR
was introduced in Set 2. As Leah’s correct responding increased in Set 2, her correct
responses decreased slightly in Sets 3 and 4 for a five sessions, but her correct
responding ultimately increased in all three sets. When RR was implemented for
Set 1, Leah’s correct responding increased to 90–100% in sets 4, 3, 2, and 1. We used
set 5 to further evaluate the effects of RR on Leah’s correct responding with sight-
words by providing RR and praise in sessions 174–179, praise only (baseline) in
sessions 180–182, and RR and praise again in sessions 183–192. The results for set
5 indicate that Leah’s correct responding was at or below 20% during baseline,
steadily increased to 50% during RR, remained at 50% for three consecutive sessions
when only praise was provided for correct responses, and slowly increased to 90%
during the subsequent RR phase. Results for the classroom generalization assessment
(sixth panel) show that Leah’s correct responses increased from 50% in the first
assessment to over 90% in the third assessment.
Figure 2 depicts the percent of correctly read sight-words for Grant when training
sessions were conducted by an experimenter across set 3 (first panel), set 4 (second
panel), set 2 (third panel), and set 5 (fourth panel). Following the introduction of RR
in set 3, Grant’s correct responding was initially variable, but it gradually increased
across sessions. Grant’s correct responding in set 4 and set 2 also increased and
stabilized following the introduction of RR for incorrect responses in those sets.
For set 5, Grant’s correct responding was stable at 15% for the first three sessions.
After RR was implemented for set 3, his correct responding increased slightly during
set 5. After RR was implemented for set 4, Grant’s correct responding gradually
increased again; however, correct responding did not increase in set 5 when RR
was implemented for set 2. Subsequently, RR was implemented for set 5 and
Grant’s correct responding increased to approximately 95% for each set by the end
of training.
Figure 3 shows the percent of correct responses by Grant across set 7 (upper panel),
set 9 (center panel), and set 10 (lower panel) when training sessions were conducted
by Grant’s paraprofessional in the classroom. The results show that Grant’s correct
responding was at or below 40% during baseline for each of the three sets and,
subsequently, correct responding increased to 90% for each set when the
paraprofessional implemented RR. Regardless of who served as the trainer, Grant’s

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 119

Figure 2. Percent of sight-words read correctly by Grant across baseline and response repetition phases
for set 3 (first panel), set 4 (second panel), set 2 (third panel), and set 5 (fourth panel) when an
experimenter served as the trainer.

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
120 K. L. Marvin et al.

Figure 3. Percent of sight-words read correctly by Grant across baseline and response repetition phases
for set 7 (upper panel), set 9 (center panel), and set 10 (lower panel) when a paraprofessional served as
the trainer.

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 121

correct responses increased to 90–100% for each of the seven sets in Figures 2 and 3
after RR was implemented.
Figure 4 shows that the percent of correctly read sight-words by Tony across set 4
(first panel), set 6 (second panel), set 5 (third panel), set 7 (fourth panel), and the
generalization assessment (fifth panel). Baseline levels of correct responding for each
set were at 40% or below and, subsequently, correct responding increased to 80% or
higher in each set after RR was implemented. Although RR was not implemented in
set 7, the percentage of correctly read sight-words also increased above 90% in the
final session. Specifically, the largest increase in Tony’s correct responding in set 7
occurred after he was exposed to RR for set 4. The results of generalization
assessment showed that the percentage of correctly read words by Tony was higher for
the six HCO passages containing sight-words that acquired using RR (M ¼ 83.3%)
than for the six HCO passages containing sight-words from baseline sets
(M ¼ 68.3%).
Figure 5 shows the percent of correctly read sight-words by Dan across set 3 (first
panel), set 2 (second panel), set 1 (third panel), and set 4 (fourth panel). Consistent
with the other participants, the results show that following stable responding in sets
3, 2, and 1, the systematic implementation of RR for incorrect responding
increased the percent of correctly read sight-words by Dan. Unlike the other
participants, however, the data for Dan’s responding do not provide evidence of
generalization of correct sight-word reading across sets. Specifically, responding in
set 4 remained relatively stable and unchanged when RR was implemented for each
of the other sets.

DISCUSSION

These results show that error-correction with RR increased sight-word reading


performance in all four participants. In addition, the results suggest that correct
responding generalized to untrained words for Tony and Grant and untrained contexts
for Leah and Tony. Furthermore, the results for Grant indicate that school support staff
could correctly implement RR for incorrect responses with minimal training.
As suggested by Worsdell et al. (2005), the behavior changes that were produced
with RR may be attributable to positive punishment for incorrect responses. That is,
by emitting the correct response in the presence of a written sight-word, participants
avoided the additional stimulation imposed by the RR procedure for an incorrect
response. Alternatively, it is possible that repeated practice of the correct response
increased correct responding; however, the results for Leah suggest that praise,
feedback, and additional opportunities to practice yielded only modest increases in

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
122 K. L. Marvin et al.

Figure 4. Percent of sight-words read correctly by Tony across baseline and response repetition (RR)
phases for set 4 (first panel), set 6 (second panel), set 5 (third panel), and set 7 (fourth panel). Percent of
sight-words read correctly by Tony during passages containing words from either baseline or RR sets
(fifth panel).

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 123

Figure 5. Percentage of sight-words read correctly by Dan across baseline and response repetition
phases for set 3 (first panel), set 2 (second panel), set 1 (third panel), and set 4 (fourth panel).

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
124 K. L. Marvin et al.

her correct responding. By contrast, her correct responses rapidly increased following
the implementation of RR.
The results of this study replicate and extend previous studies by Barbetta, Heward,
and Bradley (1993), Ferkis et al. (1997), and Worsdell et al. (2005) by showing that
correct responding produced via negative reinforcement generalized in the form of
increased correct responding to (a) untrained words that were presented in the training
setting and (b) trained words that were presented in a novel settings. Specifically,
stimulus generalization was evident in the form of increased correct responding on
reading passages for Tony and word lists for Leah, both of which involved words in
smaller font size and in the context of other words (as opposed to individual
presentation of a word on a flash card). Additionally, response generalization was
evident in the form of increased correct responding for untrained words in baseline
sets for Leah, Tony, and Grant.
Anecdotally, each participant (with the exception of Dan) emitted behaviors that
were indicative of increased motivation to provide correct responses. For example,
Leah and Grant often asked the experimenter to use the RR procedures during sets
that were still the in the baseline phase and, in some sessions, verbally objected when
the procedure was not used. Magee and Ellis (2006) found that participants acquired
correct responses with fewer training trials in a condition wherein the learner was
permitted to mand for a prompt (i.e., an error correction procedure) than in a condition
wherein only praise was provided for correct responses (error correction was not
provided for incorrect responses). The finding that participants learned faster with the
error-correction (prompt) procedure suggests that receiving error-correction for
incorrect responses may be a reinforcing event for learners. However, it is not clear
whether individuals need a history of not receiving error correction for incorrect
responses before manding for error correction functions as a reinforcing event.
Some other findings from this study warrant discussion. First, there was some
indirect evidence that Leah’s acquisition of words in one set disrupted stimulus
control (i.e., produced interference) for previously acquired words in other sets.
For example, data from individual trials indicated that Leah could no longer read
the sight-word ‘who’ from set 3 after learning to read the sight-word ‘what’ from set
2. It is possible that the more proximal reinforcement history for making a specific
blend in the presence of the visual stimulus ‘wh’ overshadowed the more distal
reinforcement history for making a different blend in the presence of that same
stimulus. Leah ultimately learned to discriminate between the two words after
additional exposure to the RR procedure for the ‘who’ card.
Second, the results for set 5 (Figure 1, fifth panel) provide further support that
praise alone was not sufficient for increasing Leah’s correct responding. Specifically,
the results for set 5 show that correct responding was low and stable during the initial
baseline phase, moderate and increasing during the first RR phase, moderate and

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 125

stable (i.e., unchanged) during the second baseline, and high and increasing during the
second RR phase. Thus, Leah’s correct responding only increased during RR phases.
Although additional manipulations would have been necessary to determine the
relative contribution of praise in the maintenance of Leah’s correct responding, it is
possible that her correct responding would have maintained in the second baseline
phase without contingent praise. Regardless of the effects of praise for maintaining
correct responses, the results for set 5 with Leah suggest that correct responses were
acquired only when RR was implemented for incorrect responses.
In conjunction with the results from previous studies (e.g., Barbetta et al., 1993;
Ferkis et al., 1997; Worsdell et al., 2005), the results of this study suggest that RR may
be a useful training procedure in school settings for at least two reasons. First, as
indicated by the results of this study, the procedure may promote stimulus
generalization, response generalization, or both. Second, the procedure is consistent
with technology that is often implemented in typical classroom settings wherein the
learner is required to practice or engage in a compensatory response contingent on
incorrect responses. Although the results of this study suggest that contingent praise
did not directly contribute to the acquisition of sight-wording, it may be appropriate
for trainers to provide students with both noncontingent and contingent praise for
participating in the training sessions.
Several limitations to the current study should also be discussed. First, providing
individuals with preferred items noncontingently or contingently for participating in
the training sessions (on an intermittent schedule) may have increased motivation to
respond correctly, decreased the aversiveness of the trainer’s demands, or both.
Nevertheless, the additional procedures were implemented in both baseline and RR
phases and correct responding did not increase until RR was introduced. Moreover,
similar procedures involving NCR have been used in other studies that evaluated
acquisition of discrete responses (e.g., Berens & Hayes, 2007). Thus, it seems
unlikely that the additional procedures directly contributed to the participants’
increased performance.
A second potential limitation is that we used a stability criterion of three
consecutive sessions with less than 15% variability to guide our decisions regarding
phase changes. The results suggest that this criterion, which was adopted to minimize
the overall amount of time each participant was out of class for training sessions, may
have been too liberal to thoroughly evaluate the effects of RR. For example, for Leah
we implemented the RR intervention for set 3 (see Figure 1; second panel) when
responding was stable (based on our criteria), but below mastery, in set 4 (first panel).
As a result, it is not clear if the subsequent increases in set 4 were a function of
additional exposure to RR for that specific set, the implementation of RR for a
different set (i.e., generalization), or a combination of both. In short, our stability
criterion may not have accounted for lengthy transition states that were generated by

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
126 K. L. Marvin et al.

the RR intervention. Likewise, Leah’s prior exposure to prompting and reinforcement


procedures for sight-words in set 4 may have promoted subsequent acquisition with
RR. Nevertheless, functional control of RR on correct sight-word reading was clearly
demonstrated with each participant’s behavior.
A third potential limitation is that we did not evaluate the differential effects of SR
error-correction on the participants’ sight-word reading. Given the results of the
Ferkis et al. (1997) study, it is possible that SR error-correction would have produced
comparable outcomes for each participant with less training time. Nevertheless, a
recent study by Ardoin, Williams, Klubnik, and McCall (2009) found that error
correction involving the rereading of passages (those containing incorrectly read
words) six times produced greater generalization on HCO passages during 1 week
maintenance assessments than having the participants reread the passages only three
times. Thus, it is possible our use of the MR error-correction procedure promoted
greater generalization of correct reading for each participant; however, the present
study did not specifically address this question.
The results of this study also provide some directions for future research. First,
given that participants in this study often asked the experimenters to implement the
RR procedures, future research should evaluate participants’ preference for RR versus
baseline conditions or RR versus interventions that are based on positive
reinforcement (see Heal & Hanley, 2007). Second, researchers should also evaluate
the effects of RR as an error-correction procedure for other academic skills such as
basic math computation. Finally, future research should separate the effects of the
trainer modeling of correct responses from the learner’s practice of engaging in the
correct response during RR by having the trainer repeat the correct response without
imitation by the learner. This manipulation will help researchers determine if practice
is a necessary component for acquiring correct responding.

REFERENCES

Ardroin, S. P., Williams, J. C., Klubnik, C., & McCall, M. (2009). Three versus six rereading of practice
passages. Journal of Applied Behavior Analysis, 42, 375–380.
Barbetta, P. M., Heward, W. E., & Bradley, D. M. (1993). Relative effects of whole-word and phonetic-
prompt error correction on the acquisition and maintenance of sight-words by students with
developmental disabilities. Journal of Applied Behavior Analysis, 26, 99–110.
Berens, N. M., & Hayes, S. C. (2007). Arbitrarily applicable comparative relations: Experimental
evidence for a relational operant. Journal of Applied Behavior Analysis, 40, 45–71.
Catania, A. C. (2007). Learning (4th ed.) New York: Sloan Publishing.
Coleman-Martin, M., & Heller, K. W. (2004). Using a modified constant prompt-delay procedure
to teach spelling to students with physical disabilities. Journal of Applied Behavior Analysis, 37,
469–480.

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin
Response repetition as an error-correction 127

Cuvo, A. J., & Klatt, K. P. (1992). Effects of community-based, videotape, and flash-card instruction of
community-referenced sight-words on students with mental retardation. Journal of Applied Behavior
Analysis, 25, 499–512.
Daly, E. J. III, Bonfiglio, C. M., Mattson, T., Persampieri, T. M., & Foreman-Yates, K. (2005). Refining
the experimental analysis of academic skills deficits. Part I: An investigation of variables that affect
generalized oral reading performance. Journal of Applied Behavior Analysis, 38, 485–497.
Eckert, T. L., Ardoin, S. P., Daly, E. J. III, & Martens, B. K. (2002). Improving oral reading fluency: A
brief experimental analysis of combining an antecedent intervention with consequences. Journal of
Applied Behavior Analysis, 35, 271–281.
Ferkis, M. A., Belifiore, P. J., & Skinner, C. H. (1997). The effects of response repetition on sight word
acquisition for students with mild disabilities. Journal of Behavioral Education, 7, 307–324.
Heal, N. A., & Hanley, G. P. (2007). Evaluating preschool children’s preferences for motivational
systems during instruction. Journal of Applied Behavior Analysis, 40, 249–261.
Magee, S. K., & Ellis, J. (2006). The role of error-correction procedures in the reinforcement of errors.
Behavioral Interventions, 21, 205–226.
Martin, G., & Pear, J. (2007). Behavior modification: What it is and how to do it (8th ed.) New Jersey:
Pearson Prentice Hall.
National Center for Education Statistics. (2002). National assessment of educational progress [online].
Available at http://nces.ed.gov/nationsreportcard/.
National Reading Panel. (2000). Teaching children to read: An evidence-based assessment of the
scientific research literature on reading and its implications for reading instruction. Retreived
February 15, 2009, from http://www.nichd.nih.gov/publications/nrp/smallbook.htm.
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.

Copyright # 2010 John Wiley & Sons, Ltd. Behav. Intervent. 25: 109–127 (2010)
DOI: 10.1002/bin

You might also like