Behavioral Interventions

Behav. Intervent. 15: 1±20 (2000)

SOCIAL INTERACTION SKILLS FOR CHILDREN WITH

AUTISM: A SCRIPT-FADING PROCEDURE FOR
NONREADERS

Cynthia L. Stevenson, Patricia J. Krantz and Lynn E. McClannahan*

Princeton Child Development Institute, Princeton, NJ, USA

Although children with autism often learn to answer questions and make requests, many do not
initiate or pursue conversation with others. In this study, audiotaped scripts were introduced and
then systematically faded to teach four boys with autism to converse with a target adult. A
multiple-probe design across participants was used to assess the number of scripted and un-
scripted interactions during Baseline I, Baseline II, Teaching, and Maintenance phases. The
intervention procedures increased unscripted interaction and the e€ects were maintained for 10±
92 sessions. Previous research has documented the favorable e€ects of fading written scripts for
children with reading skills. The current investigation demonstrates the e€ectiveness of
audiotaped scripts and script fading for children with autism who are nonreaders. Copyright
# 2000 John Wiley & Sons, Ltd.

Impaired social interaction skills are characteristic of children with autism.

Spontaneous speech±speech that is controlled by nonverbal cuesÐis typically
de®cient or absent (Charlop, Schreibman, & Thibodeau, 1985). Often, language
training begins with discrete-trial instruction that features verbal prompts, but
reliance on verbal cues from others may produce a very limited set of stimuli to
which youngsters can respond (Halle, Baer, & Spradlin, 1981; McClannahan &
Krantz, 1997). Sundberg and Partington (1999) observed that the tight stimulus
control established by discrete-trial drills may favor rote responding and inhibit
spontaneous verbal behavior. Nevertheless, discrete-trial training is important,
especially early in intervention, in order to establish attending skills and diminish
responses incompatible with learning (Etzel & LeBlanc, 1979).
This dilemma is not easily resolved. Although incidental teaching and other
naturalistic language-training procedures may promote generalized and spon-
taneous language use (McGee, Krantz, Mason, & McClannahan, 1983; McGee,
Krantz, & McClannahan, 1985), the frequency of a child's initiations may limit

* Correspondence to: Lynn E. McClannahan, Princeton Child Development Institute, 300 Cold Soil Road,
Princeton, NJ 08540, USA. E-mail address: njpcdi@earthlink.net

Copyright # 2000 John Wiley & Sons, Ltd.

2 C. L. Stevenson et al.

the number of learning opportunities, and initiations may be absent when

prepotent reinforcers are not visible or immediately available. Further, because
incidental-teaching episodes evolve into discrete-trial training if the teacher
requests too many elaborations, it may be dicult to achieve the give-and-take of
ordinary conversation.
Thus, we are left with the question of how to help children with autism initiate
and engage in communicative, reciprocal, turn-taking responses that continue
for several turns, and that pertain to a shared topic (Chadsey-Rusch, 1991).
Scripted conversations and script-fading procedures represent one e€ort to
address this problem.
In an early use of scripts, Krantz, Zalenski, Hall, Fenske, and McClannahan
(1981) taught two ®ve-year-old children with autism to answer questions about
temporally remote events by rehearsing scripted responses. Answers to questions
later posed at home were rehearsed at school, and answers to questions posed at
school were rehearsed at home. A multiple baseline design across language tasks
demonstrated that the children learned to provide as many as ®ve unprompted
answers to questions about events that happened at an earlier time, using
sentences and paragraphs, but their descriptions of past events occurred only in
response to parents' and teachers' questions.
In 1989, Charlop and Milstein used video modeling to present scripted
conversations to three boys with autism, ages 6±7, who attended an after-school
program. After training, the children's conversational skills generalized to
di€erent toys, unfamiliar persons, and di€erent settings; further, unmodeled
responses increased after video modeling was implemented. It may be noted,
however, that conversations were not initiated by the boys, but by therapists,
unfamiliar adults, or siblings who asked them questions.
Goldstein and Cisar (1992) used script training to teach three preschoolers
with characteristics of autism and six children without disabilities to engage in
appropriate social interaction. Triads (one child with a disability and two
typically developing children) received training on one sociodramatic script at a
time. The scripted nonverbal and verbal behaviors, taught with a most-to-least
hierarchy of verbal prompts, were relevant to a carnival setting, a magic show,
and a pet shop. Teachers, who did not observe script training, were asked to
maintain stable rates of prompts and praise statements during free-play sessions
before and after script training. The participants displayed increases in social
and communicative interaction in each of the three activities only subsequent to
training on each of the scripts. The investigators reported that the ``use of the
multiple baseline design demonstrated clearly that teacher prompting by itself
did not result in improved social interaction'' ( p. 278). Teachers' promptsÐ
``reminding children what behavior had just occurred, whose turn was next, what

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 3

the theme was, and what their role was'' ( p. 269)Ðwere present throughout the
study, but teachers did not model scripted responses.
In 1993, Krantz and McClannahan used written scripts that were system-
atically faded from end to beginning to promote peer initiations. Four children
with autism, ages 9±12, were taught to read and then say ten di€erent scripts
about recently completed, current, and future activities. Scripts were faded in ®ve
phases, from end to beginning, by deleting words. As scripts were faded, both
unscripted initiations and responses to others' initiations increased. Although
there were only ten scripts, the children typically made many more than ten
initiations per session and recombined elements of the scripts to produce new,
unprompted and unscripted statements and questions. However, this use of
scripts was limited to youngsters with reading skills.
In a more recent investigation, Krantz and McClannahan (1998) used script-
fading procedures to increase the unscripted social interaction of three boys with
autism (ages 4, 4, and 5 years) who had minimal reading skills. Prior to the study,
the children were taught to read the words ``Look'' and ``Watch me'', and during
the teaching condition these textual cues were embedded in their photographic
activity schedules. After they learned to use the scripts, the boys' unscripted
interactions and conversational elaborations increased, and these gains were
maintained when the original recipient of interaction was replaced by a new
recipient. The textual cues ``Look'' and ``Watch me'' were then faded from end to
beginning by cutting away portions of the cards on which they appeared; after
the ®nal fading step, the children's unscripted interactions maintained and then
generalized to new activities that had never been the topic of teaching. This study
and its predecessor (Krantz & McClannahan, 1993) documented promising
procedures that enabled children with autism to initiate and participate in social
exchanges without verbal prompts from adults.
Our previous investigations assessed the e€ects of written scripts. The present
study was designed to examine the use of audiotaped scripts and script fading to
teach conversational skills to young people with autism who were nonreaders.
Interaction was never initiated by the conversational partner, and verbal
prompts, instructions, and questions were not used.

METHOD

Participants

The participants were four boys who attended the Princeton Child
Development Institute's day school and intervention program for 5.5 h per

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
4 C. L. Stevenson et al.

day, 5 days per week. All met the criteria for autism noted in the Diagnostic and
Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Associ-
ation, 1994) and autism was diagnosed by one or more outside agencies before
they were enrolled. Informed consent was obtained from each participant's
parent(s) prior to the study.
Rick, Mike, Brett, and John, ages 12, 15, 13, and 10 respectively, had been
enrolled in the program for 7±12 years, and required continuous supervision and
ongoing intervention. Age-equivalent scores on the Peabody Picture Vocabulary
Test were 6±9 and 5±0 for Rick and Brett (Form L), and 6±2 and 2±7 for Mike
and John (Form M). Age-equivalent scores on the Vineland Adaptive Behavior
Scales were 4±8, 5±0, 4±0 and 3±2 for Rick, Mike, Brett, and John respectively.
The youths had acquired limited expressive language repertoires and had
learned to mand (e.g., ``I want drink'', ``May I go to the bathroom?''); to greet
others (e.g., ``Hi, Mom''); and to use polite words and phrases such as ``please''
and ``thank you''. When verbally prompted, they responded to instructions or
questions with words, phrases, or simple sentences (e.g., ``Apple'', ``eat apple'',
or ``The boy is eating'', but they rarely engaged in spontaneous conversation,
except to make requests.
Prior to the study, the boys had learned to follow photographic activity
schedules (Krantz, MacDu€, & McClannahan, 1993; MacDu€, Krantz, &
McClannahan, 1993), and one youth had learned to sequence his activities by
choosing photographs from a display board and mounting them in his schedule.
The participants used activity schedules throughout the school day and followed
their schedules relatively independently, even when activities and materials were
resequenced (McClannahan & Krantz, 1999).

Setting and Materials

The study was conducted in a 3  4 m classroom. A desk and chair were

located in the center of the room; another desk was located below a display
board and next to a bookcase. The recipient of interaction (a familiar teacher) sat
in a corner of the room; a 2.5 cm tape line on the ¯oor marked a distance of 1 m
from her. Observers, separated by a moveable partition, were located in a corner
adjacent to the recipient.
The materials used in this study were 25 nonsocial activities (e.g., handwriting
worksheets, puzzles, and play materials such as Boggle Junior) and ®ve social
activities (Language Master cards). When the cards were placed in a slot in the
Language Master (Bell & Howell, No. 1732B), previously recorded scripts were
played. Social activities were represented by ®ve photographs of Language

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 5

Master cards; because these photographs were identical, they did not suggest any
particular content of conversation, but merely cued participants to approach the
recipient of interaction and initiate interaction. Nonsocial activities were
represented by photographs of speci®c academic and leisure materials, such as
a picture of a vehicle puzzle or a picture of a tracing task. Photographs were
encased in plastic baseball-card holders and attached with Velcro to a yellow
foam display board. The 25 materials for nonsocial activities were randomly
assigned to ®ve groups and these groups were systematically rotated across
sessions. Also, photographs depicting the ®ve Language Master cards (the social
activities) and ®ve academic or leisure activities (the nonsocial activities) were
systematically rotated across ten positions on the display board. In all conditions
except Baseline I, an activity schedule book (a 15  18 cm binder containing ten
yellow pages covered with plastic page protectors and with small Velcro circles in
the center of each page) was placed on the desk below the display board.
Materials for nonsocial activities were located on a bookcase and the position
of materials was systematically rotated across sessions. During Teaching, the
Language Master and Language Master cards were placed on a legal-size
clipboard on the recipient's lap. The order of presentation of Language Master
cards was systematically rotated across sessions, thus changing the sequence of
audiotaped scripts.

Pre-investigation Teaching and Assessment of Imitation Skills

Prior to the study, each youth received training on imitating four- and ®ve-
word audiotaped Language Master scripts (e.g., ``I like your sweater'', ``Bob is a
teacher''), none of which was used during the experiment. Training sessions
occurred once per day and were discontinued prior to baseline. Each youth
achieved at least 80% accuracy on imitating four- and ®ve-word scripts before
participating in the study. Rick met criterion ®rst, and John was last to achieve
criterion.

Dependent Variables
Interaction
Scripted 1, scripted 2 and unscripted interactions were de®ned as under-
standable statements or questions (i.e., verbal productions that included a noun
or pronoun and a verb) that were not prompted by another person asking a
question (e.g., ``Where do you go to school?'') or giving an instruction (e.g., ``Say
Ð Ð'' or ``Tell me about Ð Ð''). Statements or questions were scored as

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
6 C. L. Stevenson et al.

interactions if they were directed to the recipient by using her name or by

orienting the face toward her while standing within 1 m of her; and if they were
separated from the participant's previous vocalizations by a change in topic (i.e.,
a change in subject or object). Interactions included statements or questions such
as ``I like candy'', ``Do you have a dog?'' and ``Pete plays Uno''.
Scripted interaction 1 was de®ned as verbal productions that matched the most
recently played Language Master script, with the exception that conjunctions,
articles, prepositions, and pronouns could be altered or deleted (e.g., substituting
``and'' for ``or'', or dropping ``the''); verbal tense could be changed; singular or
plural endings could be altered; and the recipient's name could be added. Thus,
``Do you have the pet?'' was scored as scripted 1, although the script was ``Do
you have a pet?''; ``You like to eat pizza'' was scored as scripted 1 although the
script was ``I like to eat pizza''; and ``Pete like to play Nintendo'' was scored
although the script was ``Pete likes to play Nintendo''. These rules were adopted
because, prior to the study, it was noted that all four participants made frequent
errors relevant to the correct use of conjunctions, articles, prepositions,
pronouns, verb tense, and plurals.
The ®ve scripts used in the study included the three examples noted above, as
well as ``What is your favorite food?'' and ``My school is PCDI''. Scripts were
selected because they identi®ed topics of conversation and could serve as social
initiations, because the participants could pronounce (or approximate pronun-
ciation) of the target words, and because the boys' instructors reported that they
comprehended the words included in the scripts.
Scripted interaction 2 was de®ned as statements or questions that matched a
Language Master script (with the exceptions noted above), but not the most
recently played Language Master card. Playing one Language Master script and
then saying a script played earlier in the session or in a prior session indicated
that a boy could now use a script that was not the most recently played to open
or continue a conversation.
Unscripted interaction referred to verbal productions that di€ered from the
scripts by more than conjunctions, articles, prepositions, pronouns, singular or
plural endings, changes in verb tense, or the addition of the recipient's name. The
statement ``I like to eat spaghetti'' was scored as unscripted because the script
was ``I like to eat pizza''; likewise, the question ``Do you have a cat?'' was
unscripted because the script was ``Do you have a pet?''
Non-interaction included a repetition (i.e., a participant repeated his question
or statement verbatim or altered it only by adding the recipient's name); echolalic
responses to the observers' audiotape or other ambient noise; and single words
(e.g., ``Candy''). Other verbal productions not scored as interaction were
greetings and ``good-byes'' (e.g., ``Hi, Ann'' and ``Bye, Ann''); polite statements

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 7

such as ``Thank you'', ``Excuse me'', and ``Please, Ann''; responses to questions
or instructions (e.g., ``What?'', ``Okay'', ``No'', and ``Yes I do''); and requests
such as ``I want bathroom'' and ``May I have a tickle?'' These exclusions were
adopted because the participants had already acquired skills in greeting people,
using polite phrases, and manding. Although many such non-interactions
appeared in conversation during the study (and were not discouraged), they were
not scored as unscripted.
In addition, if a participant used a statement that included `understood'
nouns, pronouns, or verbs (e.g., ``Nice talking to you'') and then immediately
repeated the statement using the understood component (e.g., ``It's been nice
talking to you''), the repetition was scored as non-interaction. Finally, verbal
productions prompted by the teacher or the recipient of interaction were scored
as non-interaction. These exclusions were adopted in order to achieve a
conservative measure of unscripted interaction.

Measurement Procedures and Experimental Design

Number of interactions was measured using a continuous event-recording
system (in 1-min intervals), during which observers recorded verbatim all verbal
productions that occurred when a participant's face was oriented toward the
recipient and he was standing within the taped area. Numbered scripts were
listed at the top of the data sheet, and when a Language Master card was played,
observers recorded the number of that script. An audiotape signaled each
successive 1-min interval; at the signal, observers drew lines on their data sheets
and began recording in the next interval. Observers also recorded whether social
and nonsocial activities were prompted by the teacher (by engaging in verbal
behavior, or using graduated guidance, spatial fading, or increased proximity) or
by the recipient (by asking questions, giving directions, or using graduated
guidance or spatial fading).
Transcribed responses were later scored as scripted 1, scripted 2, unscripted, or
non-interaction. A multiple-probe design across four participants was used to
evaluate levels of interaction during Baseline I, Baseline II, Teaching, and
Maintenance.

Experimental Conditions
Baseline I
The recipient of interaction (a familiar teacher) and observers were stationed
in adjacent corners of the room; they occupied these locations throughout the

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
8 C. L. Stevenson et al.

study. During Baseline I, the instructor remained in a corner opposite the

observers. Materials for ®ve nonsocial tasks were present (thus, there was
something for the youths to do and, potentially, to talk about), but activity
schedule book, Language Master, and Language Master cards were absent.
In Baseline I and all subsequent conditions, the recipient observed a boy's
activities, smiled when he oriented towards her, and remained continuously
available for conversation. She was asked not to initiate interaction, give
instructions, or ask questions, but to respond to a boy's verbal productions with
comments that might be of interest to him, and then to wait expectantly. For
example, if a boy said ``My school is PCDI'', the recipient might respond with a
statement relevant to that topic, such as ``Dad brought you to school today'' or
``You ride your bike at PCDI'' or ``Larry is your teacher''.
Baseline sessions began when a participant entered the room and ended after
10 consecutive minutes. At the end of each session in this and all other con-
ditions, the participant went to his usual classroom and engaged in a preferred
activity. No programmed rewards were delivered during any session throughout
the study.

Baseline II

Baseline II occurred under the same conditions as Baseline I, except that an

activity schedule book with ten pages and a foam display board with photo-
graphs depicting ®ve nonsocial activities were provided; as noted earlier, photo-
graphs and corresponding materials were rotated across sessions. Because the
boys had previously learned to follow photographic activity schedules, this
condition assessed whether a familiar structure of activities would promote
social interaction. Baseline II was identical to the Maintenance phase.

Teaching

When Teaching began, all materials were present. The activity schedule book
was on a desk, and the display board above the desk included ®ve photographs
of academic and leisure (nonsocial) activities; materials relevant to completing
these activities were arranged on the book shelves. The display board also
included ®ve photographs of Language Master cards, and Language Master and
cards were placed on a clipboard on the recipient's lap. This location was selected
in order to promote approaching the recipient.
The instructor stood behind the participant and used graduated guidance to
assist him in (a) opening the schedule book, (b) selecting a photograph from the

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 9

display board, (c) mounting the selected photograph on the page to which the
schedule was open, (d) obtaining the depicted materials, (e) completing the
activity (i.e., approaching the recipient, running a card through the Language
Master, and saying the statement or question most recently played on the
Language Master, or emitting the responses required to complete a nonsocial
activity such as a worksheet or puzzle), and ( f) returning materials to their
original locations, returning to the schedule book, and turning the page. The
instructor made every e€ort to use only the amount of guidance necessary to
achieve correct responses. For example, if a boy did not reach toward the display
board, his hand was guided toward the board, but not toward any particular
stimulus; only if he did not then make a selection was he guided to select a
photograph.
During subsequent sessions, graduated guidance was replaced by spatial
fading; that is, the teacher gradually changed the location of manual prompts
from hands to forearms to elbow to shoulder (Cooper, 1987); this prompt-fading
procedure quickly enabled the boys to select their own sequences of activities.
Next, spatial fading was replaced by shadowing, and then proximity to the
participant was decreased. This most-to-least prompts hierarchyÐgraduated
guidance, spatial fading, shadowing, and decreased proximityÐachieved a
relatively errorless teaching procedure.
In order to prevent errors, the instructor reinstated the previous level of
prompting at which the participant was successful. For example, if the instructor
was shadowing a boy and observed that he was about to make an incorrect
response, such as beginning to reach toward materials that did not correspond to
a photograph, she returned to spatial fading. If she was using spatial fading and
noted that a participant who was about to turn a page of the schedule book was
holding two pages rather than one, she returned to graduated guidance.
Similarly, the instructor returned to the prior level of prompting in order to
correct errors. If she was shadowing and a boy made an incorrect response (e.g.,
returned materials to the bookshelf but put them on top of other materials,
rather than in the empty space that they previously occupied), she returned to
spatial fading; and if that was not sucient to correct the error, she moved back
to graduated guidance. If the instructor was using spatial fading and a
participant made an error (e.g., poked holes in a worksheet rather than writing
on it), she returned to a graduated guidance. Whether preventing or correcting
errors, as soon as a youth's performance stabilized, prompts were again faded in
the previously noted sequence.
The instructor was completely silent throughout Teaching, never using verbal
prompts. If a boy ran a card through the Language Master and did not repeat the
audiotaped script, he was manually guided to run the card through the Language

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
10 C. L. Stevenson et al.

Master again (and again, if necessary). All four participants said target scripts
after running cards through the Language Master one to three times (but
typically one time), and verbal prompts were never required.
Throughout all conditions, the recipient of interaction responded to
participants with elaborations of their statements or questions (e.g., if the
participant said ``Do you like pizza?'' the recipient made comments such as
``Yes, I like cheese pizza'' or ``Yes, I go to Pizza Hut''), but never gave
instructions or asked questions. After recipient and participant had completed
approximately four exchanges, the recipient modeled a closing statement such as
``It's been nice talking to you'', ``Thanks for telling me that'', or ``Talk to you
later''. However, early in Teaching, all four boys began to use these statements,
with the result that they terminated conversation soon after it began. Thus, in
approximately two exchanges per session, the recipient used a continuation
statement such as, ``I'd like to tell you something else'' or ``I thought of another
thing to tell you''. Teaching sessions ended when a participant completed the ®ve
social and ®ve nonsocial activities that constituted the activity schedule. Early in
Teaching, when the boys were learning to follow the activity schedule and say the
scripts, some sessions lasted as long as 15 minutes; by the end of this condition,
sessions lasted 10 minutes or less.
Script fading began after at least one session in which no prompts were
delivered during nonsocial activities, and after three consecutive sessions in
which a participant said the Language Master scripts without prompts. Scripts
were faded from end to beginning by deleting words (see Table 1), and, as they
were faded, pictorial stimuli representing social activities were also faded by
cutting away portions of the photographs of Language Master cards. Only one
fading step was introduced in any session. In Step 7, when neither Language
Master nor Language Master cards were present, the remaining portions of the
pictorial stimuli were removed, leaving baseball-card holders with plain yellow
backgrounds. Finally, in Step 8, baseball-card holders and half of the display
board were removed; the remaining board showed only ®ve nonsocial activities
(exactly as in Baseline II).

Maintenance

Maintenance began after fading was completed and after three consecutive
sessions in the Teaching condition in which no prompts were delivered. No
prompts were delivered during Maintenance, and the instructor remained in a
corner of the classroom. Sessions ended after a participant completed his activity
schedule (10 minutes or less).

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 11

Table 1. Fading procedure

Fading step Recorded script Pictorial stimuli on display board

(e.g., ``What is your favorite food?'') (Language Master card)

Step 1 Last word of scripts was deleted (e.g.,

Five-sixths of depicted Language
``What is your favorite'') Master cards was visible
Step 2 Last two words of scripts were deleted
Two-thirds of depicted Language
(e.g., ``What is your'') Master cards was visible
Step 3 All but ®rst two words of scripts were
One-half of depicted Language Master
deleted (e.g., ``What is'') cards was visible
Step 4 All but ®rst word of scripts was deleted
One-half of depicted Language Master
(e.g., ``What'') cards was visible
Step 5 Language Master cards were blank One-third of depicted Language Master
cards was visible
Step 6 Language Master and clipboard were One-sixth of depicted Language Master
removed cards was visible
Step 7 Language Master cards were removed Pictures of Language Master cards were
removed
Step 8 Scripts were completely faded Empty baseball-card holders and one-
half of the display board were removed.
(The remaining half of the display board
was the same as that used in Baseline II.)

Interobserver Agreement

Interobserver agreement was obtained for (a) the presence or absence of

prompts, (b) observers' verbatim recording of participants' verbal productions,
and (c) the number of scripted 1, scripted 2, unscripted, and non-interactions.

Prompted Responses

In-classroom observers recorded whether social and nonsocial activities were

prompted or unprompted by the instructor. Interobserver agreement on the
presence or absence of prompts was 100% throughout the study.

Verbatim recording of verbal productions

Independent observers in the classroom recorded a boy's verbal productions

verbatim within 1-min intervals. Agreement on content of interaction was scored
if both in-classroom observers sequentially recorded responses with matching
nouns or pronouns (with the exception that singular or plural endings could
di€er) and with matching verbs (with the exception that verb tense could di€er)
in the same interval. A ``slippage rule'' pertained to any entry recorded as the

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
12 C. L. Stevenson et al.

Table 2. Percentage interobserver agreement on verbatim recording of scripted interaction 1,

scripted interaction 2, unscripted interaction, and non-interaction

Scripted 1 Scripted 2 Unscripted Non-interaction

Participant Mean/Range Mean/Range Mean/Range Mean/Range

Rick 100 99/50±100 87/70±100 65/0±100

Mike 100/80±100 100 98/70±100 90/50±100
Brett 100 96/67±100 93/70±100 80/0±100
John 100 100 86/67±100 85/33±100

®rst interaction in the next interval. The slippage rule meant that agreement was
scored for interactions that were not recorded in the same interval if matching
responses were the last entry in an interval for one observer and the ®rst entry in
the next interval for the other observer. Agreement on non-interaction was
scored if both observers recorded a response that did not include a noun or
pronoun and a verb. Percentage interobserver agreement was calculated by
dividing the number of agreements by the number of agreements plus dis-
agreements, and multiplying by 100. Interobserver agreement on verbatim
recording of interactions was obtained for 75%, 66%, 51%, and 47% (mean
60%) of sessions during Baseline I, Baseline II, Teaching, and Maintenance
respectively; these data are shown in Table 2.
Mean interobserver agreement on the three primary variablesÐscripted 1,
scripted 2, and unscripted interactionÐwas at acceptable levels. Lower means
and ranges for non-interaction matched the experimenters' informal observation
that these verbal productions were often low volume and poorly articulated.

Number of Scripted 1, Scripted 2, Unscripted and Non-interactions

Independent observers coded completed data sheets to indicate whether

transcribed responses were scripted 1, scripted 2, unscripted, or non-interactions.
Data sheets contained a numbered list of all scripts used during the sessions and
a list of all nonsocial activities present. In-classroom observers identi®ed
activities and scripts by writing the name of the activity or the number of the
script in the left column of the data sheet; verbatim recording was done in the
right column; and the number of each interval was written in the center column.
Agreement on number of interactions required that independent scorers
sequentially code the same type of interaction in the same interval. If two
verbal productions were recorded on the same line, they were counted as two
interactions if each contained a noun or pronoun and a verb (e.g., ``I drink Coke.
I like Coke.''). In the case of two non-interactions written on the same line, they

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 13

Table 3. Percentage interobserver agreement on coding of scripted interaction 1, scripted

interaction 2, unscripted interaction, and non-interaction

Scripted 1 Scripted 2 Unscripted Non-interaction

Participant Mean/Range Mean/Range Mean/Range Mean/Range

Rick 100 88/0±100 97/80±100 89/85±100

Mike 97/75±100 100 98/75±100 95/90±100
Brett 98/80±100 95/0±100 96/75±100 97/93±100
John 95/75±100 99/75±100 96/0±100 88/80±100

were scored as separate productions only if they ended with periods, question
marks, or exclamation points and if the ®rst word began with a capital letter
(e.g., ``Coke. 'Bye''). Interobserver agreement on coding of interactions was
obtained for 100%, 70%, 82% and 88% (mean 85%) of sessions during Baseline
I, Baseline II, Teaching, and Maintenance respectively (see Table 3).

RESULTS

Figure 1 shows the number of scripted 1, scripted 2, and unscripted

interactions scored for each participant in each condition, with the exception
that scripted 1 is not shown after scripts were faded, because Language Master
and cards were no longer present. During Baselines I and II, Rick, Mike, and
Brett did not interact with the recipient. Three unscripted interactions (``Call
people'', ``I'm done'', and ``Sit down with a chair'') were scored for John during
the ®rst session in Baseline I but no other interactions were scored for him in
Baselines I and II.
In ®ve to nine sessions, the youths' performances met criterion for saying all
®ve scripted statements or questions without prompts, after which fading began.
Fading steps 1±8 were completed in eight sessions for Rick and Brett, and
11 sessions for Mike and John. None of the boys exhibited a scripted 2 inter-
action before the ®fth fading step was introduced, and these remained
comparatively low throughout the study. The mean number of unscripted
interactions during teaching was 15 (range ˆ 0±23) for Rick, 17 (range ˆ 9±23)
for Mike, 16 (range ˆ 0±23) for Brett, and 13 (range ˆ 0±27) for John.
During Maintenance, when no prompts were delivered, unscripted interaction
remained high and stable; Rick made 15±35 (mean ˆ 21), Mike made 16±26
(mean ˆ 23), Brett made 20±31 (mean ˆ 27), and John made 21±25
(mean ˆ 23) unscripted statements or questions.

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
14 C. L. Stevenson et al.

Figure 1. Number of scripted 1, scripted 2, and unscripted interactions by each participant during
Baseline I, Baseline II, Teaching, and Maintenance conditions. Arrows 1±8 indicate fading steps.

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 15

No non-interactions were scored for Rick, Mike, or Brett during Baselines I

and II. John's non-interactions ranged from 9 to 25 (mean ˆ 12) during Baseline
I, and from zero to 12 (mean ˆ 3) in Baseline II. His non-interactions included
verbal productions such as ``Bus'', ``Cheese sticks'', and ``May I have a
tickle?'' The mean number of non-interactions recorded during Teaching was
two for Rick (range ˆ 0±15), four for Mike (range ˆ 1±14), nine for Brett
(range ˆ 0±36), and 17 for John (range ˆ 0±42). During Maintenance, the
mean number of non-interactions was three, four, two and four for Rick, Mike,
Brett, and John respectively.

DISCUSSION

Baseline data documented the participants' de®cits in social interaction skills.

During Baseline I, three of the four participants engaged solely in nonsocial
activities. Rick selected one to three di€erent nonsocial activities per session, and
completed them at the desk. When a paper-and-pencil task was available, he
consistently selected it and made marks on the paper as if coloring. During every
session in Baseline I and II, Mike stood next to the instructor but said nothing.
Brett selected a nonsocial activity and completed it (typically incorrectly) from a
squatting position on the ¯oor. When he ®nished the activity, he returned
materials to the shelf, selected the same activity, and repeated this sequence for
the duration of each baseline session. Although the recipient of interaction was
present and attentive throughout baseline sessions, these participants never
talked to her.
John interacted three times during Baseline I, session one, and also said several
non-interactions. During six of seven sessions in Baseline I, and 11 of 17 sessions
in Baseline II, he engaged in verbal behavior that was excluded from the
de®nition of interaction, and most of these verbal productions were mands. For
example, he said, ``May I have tickle?'' ``(I want) Noogie'', and ``May I have
squeeze?'' When not making requests, he walked in circles around the room.
Aside from the ®rst baseline session for John, none of the participants inter-
acted with the recipient during Baseline I, when nonsocial activities were present,
or during Baseline II when nonsocial activities and an activity schedule were
present. Only after teaching began did the participants begin to use unscripted
statements and questions.
In teaching, all participants mastered the ®ve scripts in nine sessions or fewer.
Rick and Brett completed the fading steps in eight consecutive sessions, and
Mike and John had no more than one session between steps; thus, the teaching
procedure was relatively time ecient. After the scripts were faded, Rick, Mike,

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
16 C. L. Stevenson et al.

and Brett continued to utter a few scripted 2 responses. John ceased using the
scripts after the ®fth fading step. All participants said 15 or more unscripted
statements or questions per session after script fading was completed. In add-
ition, all four boys emitted 15 or more unscripted responses per session in
Maintenance, when no prompts were delivered.
During teaching, the participants engaged in conversation on a variety of
topics. Content analysis revealed that Rick and Brett had recurrent topics of
conversation (e.g., going to McDonalds or watching movies), most of which were
modeled by the recipient early in this condition; they altered scripted responses
to engage in these unscripted interactions. For example, Rick changed the script
``What is your favorite food?'' to ``What is your favorite movie?'' and Brett
changed the script ``I like to eat pizza'' to ``I like to roller-skate''.
As noted earlier, the recipient modeled closing statements after approximately
four verbal exchanges. Thus, all four boys quickly learned to end conversations
by saying ``Bye'' or making other types of closing statements, such as ``Talk to
you later'' or ``Nice talking to you''. By the ®rst session in teaching for Mike and
John, and the second and third session for Rick and Brett respectively, the
participants began to end conversations on their own initiative, usually before
four exchanges had occurred. When this happened, the recipient sometimes used
a continuation statement such as ``I thought of something else''. The pattern of
ending interactions after only one or two exchanges may have been a result of the
recipient's models, may have served as escape responses that terminated social
interactions that the participants found dicult, or may simply underline the
boys' severe conversational skill de®cits. In retrospect, it may be important to
provide more experience with conversation before modeling closing statements.
The boys' choices of social and nonsocial activities were measured after the
third teaching session, and after each participant completed all fading steps
(after sessions 41, 59, 83, and 92). Choice was de®ned as the number of social and
nonsocial tasks completed when a participant was presented with ten social and
ten nonsocial activities and a schedule that allowed him to select ten of 20
available activities. Observers recorded the number of social and nonsocial tasks
completed, using a continuous event recording system in 1-min intervals.
Results varied by participant. Rick chose only nonsocial activities. Mike
alternately chose nonsocial and social activities. In both assessments for Brett,
he ®rst selected ®ve social activities and then ®ve nonsocial tasks, an
arrangement that he also consistently followed in teaching sessions. Finally,
John picked eight nonsocial and two social tasks during the ®rst probe and three
nonsocial and seven social tasks in the second assessment. Interobserver agree-
ment on number and order of social and nonsocial tasks completed was 100%
for all participants in all sessions. The data on activity choice were idiosyncratic

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 17

across participants, and only Rick's choices implied a preference for nonsocial
activities, the pattern that one might predict based on the de®nition of autism.
The measurement procedure was sensitive to participants' repetitions of their
own statements and questions, but did not assess whether the recipient's
responses were echoed. In order to examine echolalic responses, three Teaching
and Maintenance sessions (sessions 61, 75, and 91) were audiotaped and scored
by independent observers. One hundred percent agreement was obtained for
transcribing and scoring audiotaped responses for all participants in each taped
session. The analysis showed that Rick said 21, 20, and 28 non-echolalic state-
ments or questions during the three taped sessions. Mike said ®ve non-echolalic
responses in each session, Brett exhibited eight, 11, and nine non-echolalic
statements per session, and John produced ®ve non-echolalic responses in each
audiotaped session. Although three participants often echoed the recipient, they
were, nevertheless, responding to her, practising turn taking, and practising
statements that were sometimes used in later conversation.
Echoic responses may be functional in promoting the conversation skills of
young people with autism, just as they have been shown to contribute to
the acquisition, generalization, and maintenance of receptive naming skills
(Charlop, 1983; Leung & Wu, 1997). Echoics frequently occur in infant
vocalizations and in the speech of persons learning a second language (Pierce &
Epling, 1995). It is not surprising that they are often observed in the speech of
children with autism whose verbal repertoires consist primarily of mands, and
who are making ®rst attempts to engage in conversation about shared topics.
Echoing another person's statement or question reinstates a stimulus (Bijou,
1993), perhaps facilitating the child's next response. Both anecdotal observation
and transcriptions of audiotaped sessions indicated that participants not only
repeated the recipent's statements, but also used them appropriately at other
times, and in later conversation.
As the study progressed, the boys looked more competent as interaction
partners; pauses before they initiated conversation and before they responded to
the recipient's statements appeared to be of shorter duration. Like the parents of
typical young children described by Hart and Risley (1999), the recipient ``made
e€ective the children's utterances that made sense, tolerated the mistakes of a
novice, (and) modeled and demonstrated the skills to be learned . . .'' ( p. 186).
Thus, the procedures used in the investigation set occasions for the participants
to contact and practice many di€erent models of appropriate, contextual
language, and to practice conversational turn taking that was not interrupted by
verbal prompts.
At the conclusion of the study, a new set of scripts was introduced in the boys'
classroom. Social activities were embedded in their school activity schedules

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
18 C. L. Stevenson et al.

(McClannahan & Krantz, 1999), and their regular teachers were trained to use
the intervention procedures included in the study. For example, teachers who
were recipients were trained not to ask questions or give directions, but to
provide contextual language models and to fade scripts at relevant times.
Classroom conversations were audiotaped and later transcribed and scored, and
the de®nition of unscripted interaction was revised to exclude echoic responses;
nevertheless, clinical data showed substantial increases in unscripted responses.
In addition, four di€erent recipients were often available, and it was noted that
the youths quickly learned to seek out a di€erent recipient if a particular teacher
was busy, suggesting that their interaction skills generalized across persons.
Although the study did not formally measure generalization of social
interaction skills across persons or settings, the assessment of unscripted inter-
action provided a di€erent measure of generalizationÐthe occurrence of target
responses in training and nontraining conditions (Stokes & Baer, 1977). In
Baselines I and II, Teaching, and Maintenance, interaction was assessed before,
during, and after scripts were introduced and faded, and the data clearly show
that after scripts were faded and prompts were absent, interaction continued at
levels well above baseline.
It is dicult, in a journal article, to convey how the participants sounded when
they conversed with the recipient. Did they sometimes sound stilted and robotic?
The answer is unquestionably yes (and so do novice psychology lecturers,
travellers to Paris who have just completed a beginning course in French, and
many others who are learning new skills that require additional practice). If,
instead of saying a script, a participant changed only a noun (e.g., substituted
``Do you have a cat?'' for ``Do you have a pet?'') was that credited as unscripted
interaction? Yes, it was. These participants were essentially silent during
baseline; before the study, they did not initiate unless verbally prompted; and
their age-equivalent scores on a measure of receptive language ranged from 2±7
to 6±9.
What did they sound like when they conversed with the recipient? As in
previous studies (Krantz & McClannahan, 1993, 1998), we noted that responses
that were initially imitative often reappeared in spontaneous, generative speech.
For example, after saying the script, ``Pete likes to play Nintendo'', Brett
frequently imitated the recipient, who modeled statements such as ``You like to
go camping with Dad'', ``You sleep in a tent'', or ``You go to the museum''.
After scripts were faded, Brett made unscripted contextual statements such as ``I
go camping with a tent'', and ``I like to go to the museum with Dad''.
Building conversational language is a key issue in autism intervention.
Although many young people learn to mand and to verbally respond to others'
instructions or questions, their expressive language repertoires often appear to be

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
Script-fading procedure 19

under the control of very speci®c stimuliÐverbal prompts from parents or

teachers. The use of scripts and script-fading procedures may be helpful in
achieving broader control by stimuli such as the availability of a conversation
partner and the partner's comments. The use of audiotaped scripts makes these
procedures available to young people who are nonreaders. Systematic assessment
of the generalization of interaction skills to new settings, recipients, and scripts,
and investigation of the role of echoics in the development of conversational
skills, await further research.

ACKNOWLEDGEMENTS

We wish to thank Margaret A. MacDu€ for her competent adherence to the

role of recipient of interaction, and Victoria L. Ford for her expert help with data
collection.

REFERENCES

American Psychiatric Association. 1994. Diagnostic and Statistical Manual of Mental Disorders
(4th edn). Washington, DC.
Bijou SW. 1993. Behavior Analysis of Child Development (2nd rev. edn) Context: Reno, NV.
Chadsey-Rusch J. 1991. Communication training. In Handbook of Mental Retardation (2nd edn),
JL Matson, JA Mulick (eds). Pergamon: New York; 424±435.
Charlop MH. 1983. The e€ects of echolalia on acquisition and generalization of receptive labeling
in autistic children. Journal of Applied Behavior Analysis 16: 111±126.
Charlop MH, Milstein JP. 1989. Teaching autistic children conversational speech using video
modeling. Journal of Applied Behavior Analysis 22: 275±285.
Charlop MH, Schreibman L, Thibodeau MG. 1985. Increasing spontaneous verbal responding
in autistic children using a time delay procedure. Journal of Applied Behavior Analysis 18:
155±166.
Cooper JO. 1987. Stimulus control. In Applied Behavior Analysis, JO Cooper, TE Heron, WL
Heward (eds). Merrill: Colombus, OH; 315.
Etzel BC, LeBlanc, JM. 1979. The simplest treatment alternative: the law of parsimony applied to
choosing appropriate instructional control and errorless-learning procedures for the dicult-
to-teach child. Journal of Autism and Developmental Disorders 9: 361±382.
Goldstein H, Cisar CL. 1992. Promoting interaction during sociodramatic play: teaching scripts
to typical preschoolers and classmates with disabilities. Journal of Applied Behavior Analysis 25:
265±280.
Halle JW, Baer DM, Spradlin JE. 1981. Teachers' generalized use of delay as a stimulus control
procedure to increase language use in handicapped children. Journal of Applied Behavior
Analysis 14: 389±409.

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)
20 C. L. Stevenson et al.

Hart B, Risley TR. 1999. The Social World of Children Learning to Talk. Brookes: Baltimore,
MD.
Krantz PJ, MacDu€ MT, McClannahan LE. 1993. Programming participation in family activities
for children with autism: parents' use of photographic activity schedules. Journal of Applied
Behavior Analysis 26: 137±139.
Krantz PJ, McClannahan LE. 1993. Teaching children with autism to initiate to peers: e€ects of a
script-fading procedure. Journal of Applied Behavior Analysis 26: 121±132.
Krantz PJ, McClannahan LE. 1998. Social interaction skills for children with autism: a script-
fading procedure for beginning readers. Journal of Applied Behavior Analysis 31: 191±202.
Krantz PJ, Zalenski S, Hall LJ, Fenske EC, McClannahan LE. 1981. Teaching complex language
to autistic children. Analysis and Intervention in Developmental Disabilities 1: 259±297.
Leung JP, Wu KI. 1997. Teaching receptive naming of Chinese characters to children with autism
by incorporating echolalia. Journal of Applied Behavior Analysis 30: 59±68.
MacDu€ GS, Krantz PJ, McClannahan LE. 1993. Teaching children with autism to use
photographic activity schedules: maintenance and generalization of complex response chains.
Journal of Applied Behavior Analysis 26: 89±95.
McClannahan LE, Krantz PJ. 1997. In search of solutions to prompt dependence: teaching
children with autism to use photographic activity schedules. In Environment and Behavior, DM
Baer, EM Pinkston (eds). Westview: Boulder, CO; 271±278.
McClannahan LE, Krantz PJ. 1999. Activity Schedules for Children with Autism: Teaching
Independent Behavior. Woodbine: Bethesda.
McGee GG, Krantz PJ, Mason D, McClannahan LE. 1983. A modi®ed incidental teaching
procedure for autistic youth: acquisition and generalization of receptive object labels. Journal of
Applied Behavior Analysis 16: 329±338.
McGee GG, Krantz PJ, McClannahan LE. 1985. The facilitative e€ects of incidental teaching on
preposition use by autistic children. Journal of Applied Behavior Analysis 18: 17±31.
Pierce WD, Epling WF. 1995. Behavior Analysis and Learning. Prentice-Hall: Englewood Cli€s,
NJ.
Stokes TF, Baer DM. 1977. An implicit technology of generalization. Journal of Applied Behavior
Analysis 10: 349±367.
Sundberg ML, Partington JW. 1999. The need for both discrete trial and natural environment
language training for children with autism. In Autism: Behavior-Analytic Perspectives, PM
Ghezzi, WL Williams, JE Carr (eds). Context: Reno, NV; 139±156.

Copyright # 2000 John Wiley & Sons, Ltd. Behav. Intervent. 15: 1±20 (2000)