Journal of Autism and Developmental Disorders, Vol. 28, No.

3, 1998

Increasing Speech Intelligibility in Children with Autism

Robert L. Koegel,1 Stephen Camarata,2 Lynn Kern Koegel,1 Ayala Ben-Tall,1 and
Annette E. Smith1

Accumulating studies are documenting specific motivational variables that, when combined
into a naturalistic teaching paradigm, reliably influence the effectiveness of language teaching
interactions for children with autism. However, the effectiveness of this approach has not
yet been assessed with respect to improving speech intelligibility. The purpose of this study
was to systematically compare two intervention conditions, a Naturalistic approach (which
incorporated motivational variables) vs. an Analog (more traditional, structured) approach,
with developmentally similar speech sounds equated within and across conditions for each
child. Data indicate that although both methods effectively increased correct production of
the target sounds under some conditions, functional use of the target sounds in conversation
occurred only when the naturalistic procedures were used during intervention. Results are
discussed in terms of pivotal variables that may produce improvements in speech sounds
during conversational speech.
KEY WORDS: Speech intelligibility; naturalistic approach; analog approach; autism; pivotal response.

INTRODUCTION ample, incorporating the child's choice, lead, and in-

terest in the selection of target stimuli during lan-
Communication difficulties are common in chil- guage intervention have been shown to increase
dren with autism, and intervention attempts have learning in children with autism and other types of
been difficult. Early approaches to intervention for disabilities with expressive language delays (Cama-
teaching speech often required massive numbers of rata, Nelson, & Camarata, 1994; Hart & Risley, 1992;
trials presented repeatedly in an analog teaching L. K. Koegel & Koegel, 1995; Warren & Kaiser,
paradigm. Three major difficulties encountered were 1986; Yoder, 1987). Task variation and variation of
that: (a) gains were extremely slow (often requiring the stimulus materials (Dunlap, 1984; Dunlap &
many thousands of trials to teach a single word); (b) Koegel, 1980), and using natural reinforcers (R. L.
when speech gains occurred they were often minimal; Koegel & Williams, 1980; Saunders & Sailor, 1979;
and (c) the children typically were unmotivated to Skinner, 1979) also seem to be important variables
be involved in the teaching sessions, frequently ex- that positively influence the effectiveness of interven-
hibiting severe disruptive behaviors (L. K. Koegel & tions. Additionally, reinforcing a child's communica-
Koegel, 1995; Lovaas, 1977). As a result, more recent tive attempts, rather than only reinforcing correct
research has focused on variables that increase the motor speech responses in a narrowly defined shap-
child's responsivity to the communication task. ing procedure, has been shown to enhance learning
Accumulating studies are documenting specific (R. L. Koegel, O'Dell, & Dunlap, 1988; O'Dell, Dun-
lap, & Koegel, 1983; O'Dell & Koegel, 1981). A
motivational variables that reliably influence the ef-
number of researchers have reported an especially
fectiveness of language teaching interactions. For ex-
effective intervention outcome in terms of efficiency
'Univeisity of California, Santa Barbara, California. of language learning across communicative environ-
Vanderbilt University, Nashville, Tennessee. ments, when the above procedures are combined

241
0162-3257/98/06004241S15.IXM) © 1998 Plenum Publishing Corporation
242 Koegel, Camarata, Koegel, Ben-Tall, and Smith

(Camarata, 1996; Camarata & Nelson, 1992; Cama- range; (b) documented evidence of a minimum of a
rata etal, 1994; Halle, Marshall, Spradlin, 1979; Kai- 12-month expressive speech and language delay, as
ser, 1993; R. L. Koegel, O'Deil, & Koegel, 1987; measured by a battery of standardized language tests
McGee, Krantz, & McClannahan, 1985; Warren & and by language samples; and (c) speech intelligibil-
Gazdag, 1990). These procedures closely parallel the ity often unintelligible to conversational partners and
learning parameters available in the ambient natu- functioning below expected levels for same age peers,
ralistic linguistic environment, and therefore, many as measured by both the Arizona Articulation Profi-
have referred to the general approach as "naturalis- ciency Scale (AAPS; Fudala & Reynolds, 1986) and
tic." by language samples collected across a number of
Another potentially important application of settings and contexts.
this naturalistic intervention methodology relates to Child 1 was 5 years 6 months old at the start of
children whose speech is largely unintelligible the study and was enrolled in a regular education
(Camarata, 1993; R. L. Koegel et at, 1988; Moerk, classroom at a private school with special education
1992). For example, R. L. Koegel et al. (1988) sug- support services. On the battery of standardized
gested that reinforcing verbal attempts improved tests, including the Peabody Picture Vocabulary
speech sound production in children with autism. Test—Revised (PPVT-R; Dunn & Dunn, 1981), the
Further, the children who participated in that study Expressive One Word Picture Vocabulary Test—Re-
were rated as happier, more enthusiastic, more in- vised (EOWPVT-R; Gardner, 1990), and the Clinical
terested, and better behaved during sessions when Evaluation of Language Fundamentals—Revised
their verbal attempts to speak were reinforced. This (CELF; Semel, Wiig, & Secord, 1980), his language
literature suggests that embedding motivational vari- was estimated to be at the 4 year 3 month level,
ables such as those described within the naturalistic slightly over a year below his chronological age level.
interventions may also be effective for improving in- His AAPS total score was 89.5 (18th percentile), with
telligibility of children with numerous speech sound 13 phonological errors, beginning at the 3-year-old
production errors. Therefore, the purpose of this level. Additionally, he produced numerous speech
study was to systematically compare two intervention production errors when speaking in multiword utter-
conditions, a naturalistic versus an analog teaching ances, and was frequently unintelligible during con-
paradigm, with speech sounds equated within and versation.
across conditions for each child. Child 2 was 6 years 0 months old at the start of
the study, and attended a regular education class-
room with special education support services in a
METHOD public school. His parents requested additional serv-
ices due to his delay in language skills and lack of
Participants intelligibility in speech production. On a battery of
standardized tests (PPVT-R, EOWPVT-R, Assess-
Five children, four boys and one girl, partici- ment of Children's Language Comprehension
pated in this study. Four of the children were of (ACLC; Foster, Giddan, & Stark, 1983), and the
European American descent, one was of Asian Communication Domain of the Vineland Adaptive
American descent, and all were monolingual English Behavior Scales (VABS; Sparrow, Balla, & Cicchetti,
speakers. All five of the children were diagnosed with 1984); he functioned from l'/2 to 3 years below his
autism in accordance with the diagnostic criteria for chronological age level. On the AAPS his total score
autism as defined in the DSM-IV (American Psychi- was 82.5 (below the 3rd percentile), with 20
atric Association [APA], 1994) by outside agencies phonological errors, beginning at the 3-year-old level.
and referred to our clinic for treatment due to poor Child 3 was 7 years 6 months old at the start of
speech intelligibility. All children were enrolled in the study, and was referred to our clinic due to be-
speech/language services through the public schools, havior and communication concerns. He attended
however, services were coordinated so that the chil- both a regular education class with special education
dren did not receive any speech sound intervention support services and a special day class for children
during the course of this study. In addition, the chil- with severe disabilities. A functional analysis indi-
dren met the following criteria for inclusion in this cated that his poor articulation skills and the inability
study: (a) estimated hearing acuity within the normal of others to understand his speech attempts led to
 Increasing Speech Intelligibility in Autism 243

many inappropriate behaviors. On a battery of stand- developing sounds (Le., beyond 48 months). While
ardized tests, including the PPVT-R, the EOWPVT- some authors place /t/ at a lower developmental
R, the Test of Early Language Development (TELD; level, others include it with later developing sounds
Hresko, Reid, & Hammill, 1981), and the VABS, he (i.e., Sander, 1972). Therefore, in the present study,
typically functioned 2 3/4 years below age level. On /t/ was assigned to both treatments to ensure coun-
the AAPS his total score was 72 (below the 1st per- terbalancing. Target sounds were then arbitrarily as-
centile), with 25 phonological errors, beginning at the signed to one of the two experimental treatment
2 year 6 month age level. conditions. This resulted in the target sounds being
Child 4 was 3 years 8 months old at the start of approximately equal and unsystematically assigned to
the study, and was having difficulty communicating conditions, and some identical sounds being assigned
with others in his regular education preschool class- to both treatment conditions within and across chil-
room due to poor speech intelligibility. On a battery dren. The specific sounds assigned to each child were
of standardized tests, including the PPVT-R, the as follow. For Child 1 sounds were "v" (as in van)
EOWPVT-R, and the VABS, he functioned from 7 in initial position in words; "th" (as in throw) in in-
months to 1 year below age level. His total AAPS itial position in words; "th" (as in bath) in final po-
score was 78 (6th percentile), with 25 phonological sition in words. For Child 2 sounds were "1" (as in
errors, beginning at the 2-year-old age level. like) in initial word position; "r" (as in rock) in initial
Child 5 was 5 years 0 months old at the start of word position, and "th" (as in thumb) in initial word
the study, and was referred to our clinic by her par- position. For Child 3 sounds were "s" (as in sand)
ents, who reported an inability to understand their in initial word position; "ch" (as in chew) in initial
daughter's speech. She attended both a regular edu- word position; and "1" (as in like) in initial word po-
cation class with special education support services sition. For Child 4 sounds were "t" (as in bat in final
and a special day class for children with communi- word position, "g" (as in bag) in final word position,
cative disabilities. On a battery of standardized tests, and "th" (as in bath) in finafword position. For Child
including the PPVT-R, the EOWPVT-R, the TELD, 5, sounds were "j'Vdy (as in juice) in initial word po-
sition, "t" as in bat) in final word position, and "th"
the ACLC, and the VABS, she functioned at about
(as in that) in initial word position.
the 2 year 6 month age level, approximately 21/2 years
below her chronological age level. Her AAPS total
score was 73.5 (4th percentile), with 23 phonological Procedure
errors, beginning at the I'/i-year-old level. Addition-
ally, when she attempted to speak in multiword ut- Within each condition in the ABA design, a
terances, almost all of her speech was unintelligible. baseline phase preceded the assigned experimental
(Naturalistic vs. Analog) treatment condition. Chil-
dren typically participated in each intervention con-
Design
dition (Analog or Naturalistic) for a minimum of 20
sessions. However, conditions were considered com-
An ABA design (Barlow & Hersen, 1984) was
plete if the child was responding in a stable manner
employed with order of conditions counterbalanced
at above 80% correct productions of the target sound
to control for order effects. Target sounds were during conversation for approximately four consecu-
equated across conditions for each child by selecting tive sessions. In addition, consistent with the Human
a pool of sounds for treatment that were produced Subjects Protocol, conditions were discontinued fol-
incorrectly and that were developmentally similar. lowing four consecutive sessions of repeated verbal
Developmentally similar was defined as appearing and nonverbal expressions of discomfort with the
within a similar chronological age within the devel- task during a session.
opmental literature (i.e., Goldman & Fristoe, 1986;
Prather, Hedrick, & Kern, 1975; Sander, 1972; Tem-
plin, 1957). For example, /b/, /d/, /n/, /m/, /p/, A/, and Baseline
/h/ would be considered as developmentally similar
because all are mastered (90% correct criterion) by Prior to the implementation of each of the ex-
36 months of age. Similarly, /r/, /!/, M, Is/, M, IQI, perimental treatment conditions, baseline measures
Aty, /V, and voiced "th" would be considered later were obtained for each of the targeted sounds during
244 Koegel, Camarata, Koegel, Ben-Tall, and Smith

language samples (as described below in the Depend- licensed speech pathologists, with both observers in-
ent Measures section). Throughout the experiment, dependently agreeing that the steps and intervention
measures were obtained in three settings: (a) a sepa- procedures in Table I were implemented as described
rate room in the clinic building with a person other in all instances.
than the clinician; (b) at home with family members; Analog Condition. Intervention was imple-
and (c) at school with peers. These sessions were mented as follows. The clinician first worked on pro-
videotaped for later analysis so as not to interfere duction of the target sound in isolation until an 80%
with the natural conversational flow. Conversational correct production criterion on 20 trials was
partners in the clinic and the home settings were in- achieved. To do this, the clinician modeled the target
structed to simply engage in con- versation with the sound and asked the child to repeat it. If the child
child while playing or participating in an activity to- responded correctly, the clinician gave the child so-
gether. Language samples in the school setting were cial praise and a desired object or edible reward. If
obtained while the children were interacting with the child's response was incorrect, the clinician pro-
peers, such as free time, recess, and lunch time. No vided the child with visual and motor placement cues
guidelines or instructions were provided. (e.g., "Put your tongue between your teeth like this."
[to make the "th" sound]). In this step and all sub-
sequent steps the child was rewarded as described
Treatment above on both prompted and other correct responses
using a shaping paradigm. That is, each sound pro-
Following each baseline assessment, one of the duction had to be as good or better than the previous
experimental treatment conditions (Analog vs. Natu- sound production in order to be reinforced. After the
ralistic) was implemented in accordance with its as- child was able to imitate the sound correctly at an
signed (counterbalanced) order within the ABA 80% level, the clinician then asked the child to pro-
design. In both conditions, treatment sessions were duce the target sound spontaneously without a model
conducted twice a week for 45 min per session by a ("Say it by yourself."). When the child was able to
clinician with graduate level training in each of the produce the target sound spontaneously at the 80%
treatment procedures employed in the study. Treat- criterion level, production of the target sound was
ment sessions were observed for fidelity of imple- taught in a word. To do this, one pool of stimulus
mentation for each clinician in each condition by two items consisting of 20 pictures and photographs of

Table I. Differences Between the Analog and the Naturalistic Conditions

Analog Naturalistic

Stimulus items Chosen by clinician Chosen by child

Pictures/photographs Objects and toys containing target sounds
containing target sounds Selection of items of high interest to child
Steps Begin with sounds in isolation; drill Begin with word production of item
until mastered
Sounds presented sequentially in words, Productions are immediately imbedded in words,
phrases, and sentences; drill until mastered phases, and sentences in naturalistic conversational
at each level play interactions until mastered
Interaction Clinician models sound production Clinician models word following child attempts
Direct feedback (e.g., motor placement Clinician and child naturally (play) interact with
cues) stimulus items
Response-reinforcer Child verbally reinforced for production A broad shaping contingency was employed so
contingency of correct responses in a shaping that both correct responses and verbal attempts
paradigm, with successive approximations were reinforced, and included a correct model of
reinforced, and feedback provided for the target sound
motor placement
Consequences Social reinforcers, desired objects, and Social reinforcers, and natural reinforcers (e.g.,
edible reinforcers opportunity to play with stimulus item)
 Increasing Speech Intelligibility in Autism 245

objects containing the target sound was chosen by the word, and waited for the child to make a correct re-
clinician. The clinician held up the picture and mod- sponse or attempt before giving the child the rein-
eled the sound production and gave prompts as de- forcing item.
scribed above (if the child responded incorrectly).
Again, the child received social reinforcers, token re-
inforcers, and edible reinforcers as described above. Dependent Measures
When the child reached an 80% criterion, the child
was asked to produce the sound in a word sponta- Data were recorded on the children's correct
neously, using the pictures as stimuli to evoke the production of the target sounds under two language
word. This same procedure was then repeated with sampling conditions: (a) Prior to each treatment ses-
the target sound contained in phrases, a sentence, sion, language samples were recorded for a minimum
and then in gradually increasing numbers of sen- of 10 min, with the requirement that there be at least
tences. Table I itemizes the steps and procedures em- 6 productions of the target sound in each sample.
ployed in the Analog condition, and compares them During each sample, the child interacted with a fa-
directly with the Naturalistic treatment condition de- miliar person other than the clinician. Conversational
scribed below. partners were instructed to simply engage in conver-
Naturalistic Condition. Intervention was imple- sation with the child while playing or participating in
mented as follows. The clinician began working on an activity together, (b) In addition, periodic lan-
production of the target sound in words occurring guage samples were obtained within the child's home
during natural interactions, without working on the while the child interacted with family member(s) and
sound in isolation. To do this, of a pool of 20 stimulus within the child's school while the child interacted
items was selected for each sound. In contrast to the with classmates, with the requirement that there be
analog condition where stimulus items were selected at least 50 utterances, and at least 3 productions of
only with respect to their containing the targeted the target sound in each sample, (c) Finally, to assess
sound, regardless of the child's interest in the item, whether there were improvements in subjective rat-
three criteria were used for selection of stimulus ings of the children's overall intelligibility during un-
items in the naturalistic condition: (a) the label of structured conversational interactions, as a function
the stimulus item contained the target sound (e.g., of improvements in speech production of the target
"v" in the word van) or a verb directly related to the sounds in the language samples, pre and post ratings
use of the stimulus item contained the target sound were made by listeners who were unfamiliar with the
(e.g., "r" for roll the ball); (b) the stimulus items children and naive to the experimental conditions
were selected for use during each session only if they and hypotheses. For each child, a 6-point Likert scale
had high interest value for each particular child; and was utilized to rate four 5-min segments selected
(c) objects and toys were selected that would provide from the pre and post language samples while the
naturally reinforcing consequences when the item children were engaging in conversational interac-
was accessed by the child. Further, to increase mo- tions. To eliminate order effects and experimenter
tivation, opportunities for child choice of activity with drift, the observers viewed the tapes in a random or-
the stimulus items were made available within the der. Scores could range from 0 (not intelligible, very
session (catch the football, throw the football, etc.) difficult to understand) to 5 (very intelligible, easy to
using the pool of high interest toys and objects. Dur- understand). The Appendix presents the full rating
ing each session, the clinician modeled the target scale.
sound in words, phrases, and sentences as they natu- Language samples were videotaped, using a
rally came up in the play interactions. A broad shap- small wireless microphone, in order to prevent data
ing contingency was employed, so that following the recording from interfering with the pace of the con-
child's correct productions of the target sound or at- versation. Words containing the target sound were
tempts to produce the target sound in words, phrases transcribed by individuals who were familiar with
or sentences, the child was naturally reinforced with each child. Correct and incorrect production of the
the opportunity to play with the stimulus items, target sound was then recorded from the transcripts.
paired with social reinforcers. When the child did not Correct productions were defined as target sounds
attempt to produce the target sound, the clinician produced without substitution or distortion. Incorrect
modeled the correct production of the sound in the speech productions were defined as the child omit-
246 Koegel, Camarata, Koegel, Ben-Tall, and Smith

ting or distorting the target phoneme, or substituting different room in the clinic building, the child's
another phoneme in its place. Percentage correct home, and the child's school setting.
production of the target sound was then computed Within the context of the ABA design, the treat-
for each session. ment condition for a new target sound was then
changed from the naturalistic condition to the analog
condition for Child 1. Correct production of the tar-
Reliability get sound remained at or near zero in the language
samples throughout the analog treatment condition.
Reliability measures were obtained by two inde- The results indicate little or no correct conversa-
pendent data recorders (who were graduate or un- tional use of the target sound occurred in any of the
dergraduate students with previous training in the three language sample settings. Consistent with the
recording of speech production) for 68 (i.e., 20%) of ABA design, a target sound was then treated within
the language samples, with reliability measures occur- the naturalistic treatment condition. Similar to the
ring in all conditions for all children. To control for first naturalistic phase, the correct production of the
experimenter bias, at least one of the observers was sound within the language samples increased rapidly
naive to the experimental condition in over 50% of and reached an initially high level of over 80% cor-
the sessions. Reliability was calculated on a trial by rect production within four sessions.
trial basis according to the formula, agreements times These results are replicated with Children 2, 3,
100 divided by agreements plus disagreements. Using 4, and 5. That is, for all sounds, high levels of correct
one transcript as the master and one as the reliability conversational use of the target sounds only occurred
measure, agreements were defined as the two ob- when treatment was conducted within the naturalistic
servers recording identical speech production sounds condition. Treatment for Children 1 and 2 began with
on a given trial. Disagreements were defined as one the naturalistic condition followed by the analog con-
observer recording the speech production as correct, dition with a return to the naturalistic condition. For
and the other observer recording it as incorrect. In- Children 3, 4, and 5, treatment began with the analog
terrater reliability scores averaged 87%, with a range condition followed by the naturalistic condition with
of 81-89% across children (kappa was .686, corre- a return to the analog condition. Again, the pattern
sponding to a "very good" level of agreement for was consistent for all of the children. The language
categorical data; Fleiss, 1981; Landes & Koch, 1977). sample results indicate that these children exhibited
On the speech intelligibility rating scales, reliability low levels or no correct conversational use of the tar-
measures were obtained for 10 (i.e., 50%) of the ses- get sound during the language samples when treat-
sions. Both observers recorded exactly the same score ment was conducted using the analog procedures. In
for 6 of the 10 ratings, a 1-point difference for 3 ses- contrast, all of the children showed high levels of cor-
sions, and a 2-point difference for 1 session. rect conversational use of the targeted sounds in the
language samples when treatment was conducted us-
ing the naturalistic procedures.
It is also noteworthy that Child 5 exhibited large
RESULTS increases in disruptive and avoidance behavior during
the analog conditions (refusing to enter the room or
The results from the language samples, showing sit in the chair, verbally expressing "No cards," etc.)
functional use of the targeted speech sounds, are so that those analog sessions were terminated to be
shown in Fig. 1. Baseline measures indicate that the in compliance with the Human Subjects Protocol (see
target sounds were produced incorrectly prior to Participants section). Such avoidance responding and
treatment in each condition for each target sound. expressions of discomfort by the children never oc-
Within the ABA design, treatment for Child 1 began curred during the naturalistic sessions.
with the naturalistic treatment condition for his first In summary, regardless of the order of condi-
target sound. As indicated in the figure, his produc- tions, the data indicate low levels or no correct con-
tion of the target sound improved throughout this versational use of the targeted speech sounds during
phase, reaching 100% correct production by the end the language samples when treatment was conducted
of the condition in all three settings in which lan- within the analog condition. However, large gains,
guage samples were taken. These settings included a usually to near 100% correct conversational use of
 Increasing Speech Intelligibility in Autism 247

Fig. 1. Percentage correct production of target sounds within language samples measured during each treatment phase. Circles indicate
language samples in a different room in the clinic building with a person other than the clinician while the squares and diamonds represent
samples at home with family members and at school with peers, respectively. Symbols with hatch marks reflect that insufficient opportunity
to use words containing the target sound occurred during the language sample.
248 Koegel, Camarata, Koegel, Ben-Tall, and Smith

the target sound, during the language samples oc- showing a 1- to 2-point improvement in speech intel-
curred when treatment was conducted in the natu- ligibility. Note that before the intervention was imple-
ralistic condition. These results were consistent for mented, naive observers reported that the children
all five children, regardless of the target sounds or were only "sometimes intelligible" or "mostly not in-
of the order of experimental conditions. Thus, the telligible," and that it typically required "some strain"
naturalistic procedures repeatedly resulted in greater or "much strain" to understand their speech. In con-
treatment gains in comparison to the analog proce- trast, following the intervention, 4 of the 5 children
dures. were reported to be in the "sometimes" to "mostly"
intelligible range, requiring "minimal strain" or "usu-
ally does not require strain" to understand their
Treatment Gains in the Analog Condition
speech. Child 5, who was the one exception, was noted
by observers to speak primarily in one-word utterances
Because the analog condition required working
during the pre measures, and mostly in multiword ut-
on the target sound in systematically increasing dif-
terances during the post measures. These improve-
ficulty of steps (sound in isolation, sound in words,
ments are substantial, considering the numerous
etc.), it was necessary to keep data on the children's
phonological errors each child demonstrated prior to
progress on these steps. The outcomes of these data
implementation of the intervention.
are interesting given the above results showing that
within the analog condition the children only re-
sponded correctly at low levels during the language DISCUSSION
samples plotted in Fig. 1. Table II shows the criterion
level reached with the clinician present in the treat- The children who participated in this study
ment sessions for each sound within the analog con- clearly demonstrated greater gains in functional
dition. The results of the target sound acquisition speech use when the naturalistic intervention proce-
showed that in almost every instance the children dures were employed as compared to the analog in-
learned to produce speech sounds at some level dur- tervention procedures. This finding may be
ing the analog condition within the clinical treatment interpreted in a number of ways. First, previous re-
sessions. For example, Children 1 and 3 reached cri- search in the area of language intervention has dem-
terion at the sentence and multiple sentence levels, onstrated that learning is greater and disruptive
and Children 4 and 5 reached criterion for correct behavior occurs at lower levels when naturalistic pro-
production at the word and/or phrase levels. Thus, 4 cedures focusing on motivation are incorporated (L.
of the 5 children made substantial gains within the K. Koegel & Koegel, 1995; L. K. Koegel, Koegel, &
analog clinical intervention sessions. However, refer- Surratt, 1992; R. L. Koegel et al., 1987; Lovaas,
ring back to Fig. 1, in no case did this progress result 1977). The results of this study are consistent with
in substantial functional use of the targeted sounds those findings, showing that the naturalistic proce-
in their conversational interactions during the lan- dures were also more effective in establishing accu-
guage sample sessions. Thus, both treatment condi- rate sound production. Thus, the motivational
tions typically were effective in producing acquisition components of the naturalistic teaching paradigm
of the target sound at some level. However, as Fig. may have resulted in the children being more actively
1 shows, only the naturalistic condition resulted in
functional use of the targeted sound during conver-
sation. Table II. Criterion Level Reached During Treatment Sessions
for Each Sound Within the Analog Condition

Overall Intelligibility of the Children's Speech Analog condition

Child First Second
Results for the children's overall speech intelligi- 1 Sentence Not in exp. design
bility for the 2 pre and 2 postvideotaped language 2 No criterion level reached Not in exp. design
sample segments for each child are shown in Table III. 3 Multiple sentence Multiple sentence
Following the improvements in their speech sounds 4 Phrase Words
during conversational speech in the naturalistic inter- 5 Words No criterion level
reached
vention conditions, the children typically were rated as
 Increasing Speech Intelligibility in Autism 249

Table III. Overall Intelligibility Ratings for Pre and Post common stimuli, using sufficient exemplars, and us-
Conversational Language Sample Segments for Each Childa
ing commonly occurring stimuli as clinically impor-
Overall intelligibility rating tant variables for producing widespread treatment
Child Preintervention Postintervention gains. These themes were enhanced during the natu-
ralistic procedures employed in this study.
1 2(1) 4(4)
2 3
Finally, reinforcing the children's attempts to
2 2(1) 4(3) produce the speech sounds in the target words during
2 3 the naturalistic condition may also have influenced
3 3(3) 4(4) responding. Hovell, Schumaker, and Sherman (1978)
3 3 have pointed out that parents of children without dis-
4 1(1) 4(4)
1 4
abilities frequently reinforce their children's attempts
5 2(0)* 2(2) c to imitate parent models of speech, and that the re-
26 y sulting high rate of imitative behavior may facilitate
"Parentheses contain the reliability observer's ratings.
speech acquisition (also see Moerk, 1972). It is in-
^Observers noted that conversation consisted of many one word teresting to note that these improvements were ob-
utterances. served without any direct motor training on targets
cObservers noted that conversation consisted of mostlyly multiword (as in Camarata, 1993). This finding calls into ques-
utterances. tion the long-standing assumption that speech inter-
vention uniformly requires analog application of
motor training to be successful (Bernthal & Bankson,
involved in the intervention and less likely to exhibit 1993; Camarata, 1995; Klein, 1996; Swift, 1918, Van
disruptive avoidance behaviors, increasing the effec- Riper, 1938). At least for the children with autism
tiveness of teaching. who participated in this study, superior performance
Second, it may be important to note that when was evidenced during the naturalistic condition,
the naturalistic procedures were implemented, the which did not incorporate motor speech sound drills.
children with autism participating in the present study Given that the children demonstrated learning
demonstrated improved speech in a variety of settings, within naturalistic teaching procedures, one may
including a separate room in the clinic, in their homes, wonder why such learning does not occur in the am-
and in their schools. It is possible that using the tar- bient environment. It may be necessary to present
geted speech sound within the context of natural in- the teaching stimuli in a salient, focused manner that
teractions produced more widespread exposure to a occurs relatively rarely in the ambient learning envi-
natural reinforcer, rather than an arbitrary reinforcer ronment (Camarata, 1995, 1996; Nelson, 1989). Be-
that had no direct relationship to the target sound. cause the naturalistic intervention presented in this
This direct response-reinforcer relationship may have study included multiple models of correct speech,
been emphasized by providing the child with access to within the context of simple language structures,
a desired item that contained the targeted speech there was a high degree of salience for the speech
sound contingent upon a correct response or attempt targets. This increased salience, within meaningful
at the speech sound. Thus, repeated exposure to a fa- motivational contexts may have been sufficient to
vorable response-reinforcer contingency would have generate improved speech in the absence of motor
been created, a condition which has been shown to training. In summary, the present study suggests that
improve motivation and most types of learning (L. K. a naturalistic intervention, implemented systemati-
Koegel & Koegel, 1995; Seligman, 1972). cally, is an effective method for improving speech in-
Similarly, the use of stimulus items such as pre- telligibility in children with autism.
ferred toys rather than artificial stimuli such as pic-
ture cards used in the analog condition, and the
presentation of the teaching within the context of APPENDIX
natural (play) interactions during the naturalistic ses-
sions, may have been variables that increased the Overall Speech Intelligibility Rating Scale Admin-
likelihood of functional use of the speech sounds out- istered During the Children's Conversations
side of the teaching setting. Also, Stokes and Baer
(1977) have discussed variables such as programming 0 = Not intelligible, very difficult to understand
250 Koegel, Catnarata, Koegel, Ben-Tall, and Smith

1 = Mostly not intelligible, requires much strain to Dunlap, G. (1984). The influence of task variation and mainte-
nance tasks on the learning and affect of autistic children.
understand Journal of Experimental Child Psychologist, 37, 41-64.
2 = Sometimes intelligible, requires some strain to Dunlap, G., & Koegel, R. L (1980). Motivating autistic children
understand through stimulus variation. Journal of Applied Behavior Analy-
sis 13, 619-627.
3 = Sometimes intelligible, requires minimal strain Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary
to understand Test—Revised. Circle Pines, MN: American Guidance Service.
4 = Mostly intelligible, and usually does not require Foster, R., Giddan, J. J., & Stark, J. (1983). Assessment of chil-
dren's language comprehension. Palo Alto, CA: Consulting Psy-
strain to understand chologists Press.
5 = Very intelligible, easy to understand Fleiss, J. (1981). Statistical methods for rates and proportions. New
York: Wiley.
FudaJa, J. B., & Reynolds, W M. (1986). Arizona Articulation Pro-
Please rate each 5-minute segment according to the ficiency Scale (2nd ed.). Los Angeles: Western Psychological
described scale, and record in the appropriate place Services.
on the form. Gardner, M. F. (1990). Expressive One-Word Picture Vocabulary
Test—Revised. Novato, CA: Academic Therapy Publications.
Goldman, R., & Fristoe, M. (1986). Goldman-Fristoe test of ar-
ticulation competence. Circle Pines, MN: American Guidance
ACKNOWLEDGMENTS Service.
Halle, J. W, Marehall, A. M., & Spradlin, J. E. (1979). Time delay:
A technique to increase language use and facilitate generali-
This project was supported in part by Grant zation in retarded children. Journal of Applied Behavior Analy-
sis, 12, 431-439.
MH-28210 from the National Institute of Mental Hart, B., & Risley, T. R. (1992). American parenting of language-
Health awarded to R. L. Koegel, Grant G0087C0234 learning children: Persisting difference in family-child inter-
awarded to R. L. Koegel and L. K. Koegel, Grant actions observed in natural home environments.
Developmental Psychology, 28, 1096-1105.
H023C30070 awarded to S. Camarata, L. K. Koegel, Hovell, M. F., Schumaker, J. B., & Sherman, J. A. (1978). A com-
and R. L. Koegel from the U. S. Department of Edu- parison of parents' models and expansions in promoting chil-
cation, and by an endowment to S. Camarata from dren's acquisition of adjectives. Journal of Experimental Child
the Scottish Rite Foundation of Nashville. Psychology, 25, 41-57.
Hresko, W R, Reid, D. K., & Hammill, D. D. (1981). The Test
of Early Language Development, Austin, TX: PRO-ED.
Kaiser, A. R (1993). Understanding human behavior: Problems of
science and practice: Response. Journal of the Association for
REFERENCES Persons with Severe Handicaps, 18, 24-242.
Klein, E. S. (1996). Phonological/traditional approaches to articu-
American Psychiatric Association. (1994). Diagnostic and statistical lation therapy: A retrospective group comparison. Language,
manual of mental disorders, (4th ed.). Washington, DC: Speech & Hearing Services in the Schools, 27, 314-329.
Author. Koegel, L. K., & Koegel, R. L. (1995). Motivating communication
Barlow, D. A., & Hersen, M. (1984). Single case experimental de- in children with autism. In E. Schopler & G. B. Mesibov
signs. New York: Pergamon. (Eds.), Learning and cognition in autism: Current issues in
Bernthal, J., & Bankson, N. (1993). Articulation disorders (3rd ed.). autism. New York, NY: Plenum Press.
Englewood Cliffs, NJ: Prentice-Hall. Koegel, L. K., Koegel, R. L., & Surratt, A. (1992). Language in-
Camarata, S. M. (1993). The application of naturalistic conversa- tervention and disruptive behavior in preschool children with
tion training to speech production in children with speech dis- autism. Journal of Autism and Developmental Disorders, 22,
abilities. Journal of Allied Behavior Analysis, 26, 173-182. 141-153.
Camarata, S. M. (1995). A rationale for naturalistic speech intel- Koegel, R. L, O'Dell, M. C, & Dunlap, G. (1988). Producing
ligibility intervention. In M. E. Fey, J. Windsor, & S. F. War- speech use in nonverbal autistic children by reinforcing at-
ren (Eds.), Language intervention: Preschool through tempts. Journal of Autism and Developmental Disorders, 18,
elementary years (Communication and language intervention 525-538.
series. Vol. 5). Baltimore, MD: Paul H. Brookes. Koegel, R. L., O'Dell, M. C, & Koegel, L. K. (1987). A natural
Camarata, S. M. (1996). On the importance of integrating natu- language teaching paradigm for nonverbal autistic children.
ralistic language, social intervention, and speech-intelligibility Journal of Autism and Developmental Disorders, 17, 187-200.
training. In L. K. Koegel, R. L. Koegel, & G. Dunlap (Eds.), Koegel, R. L, & Williams, J. A. (1980). Direct versus indirect re-
Positive behavioral support: Including people with difficult be- sponse-reinforcer relationships in teaching autistic children.
havior in the community. Baltimore, MD: Paul H. Brookes. Journal of Abnormal Child Psychology, 8, 537-547.
Camarata, S. M., & Nelson, K. E. (1992). Treatment efficiency as Landes, J., & Koch, G. (1977). The measurement of observation
a function of target selection in the remediation of child lan- agreement for categorical data. Biometrics, 33, 159-174.
guage disorders. Clinical Linguistics and Phonetics, 6,167-178. Lovaas, O. I. (1977). The autistic child: Language development
Camarata, S. M., Nelson, K. E., & Camarata, S. M. (1994). Com- through behavior modification. New York, NY: Irvington.
parison of conversational-recasting and imitative procedures McGee, G. G., Krantz, E J., & McClannahan, L. E. (1985). The
for training grammatical structures in children with specific facilitative effects of incidental teaching on preposition use
language impairment. Journal of Speech and Hearing Research, by autistic children. Journal of Applied Behavior Analysis, 18,
37, 1414-1423. 17-31.
Increasing Speech Intelligibility in Autism 251

Moerk, E. (1972). Principles of interaction in language learning. Semel, E., Wiig, E. H., & Secord, W. (1980). Clinical evaluation
Merrill-Palmer Quarterly, IS, 229-257. of language fundamentals—Revised. New York: Psychological
Moerk, E. (1992). A first language taught and learned. Baltimore, Corp., Harcourt, Brace, Jovanovich.
MD: Paul H. Brookes. Sparrow, S. S., Balla, D. A., & Cicchetti, D. V (1984). Vineland
Nelson, K. E. (1989). Strategies for first language teaching. In M. Adaptive Behavior Scales. Circle Pines, MN: American Guid-
L Rice & R. L. Schiefelbusch (Eds.), The teachability of Ian- ance Service.
guage. Baltimore, MD: Paul H. Brookes. Stokes, T E, & Baer, D. M. (1977). An implicit technology of
O'Dell, M. C, Dunlap, G., & Koegel, R. L. (1983). The importance generalization. Journal of Applied Behavior Analysis, 10, 349-
of reinforcing verbal attempts during speech training with non- 367.
verbal children. Paper presented at the annual meeting of the Swift, W. (1918). Speech defects in school children. Cambridge, MA:
American Psychological Association, Anaheim, CA. Riverside.
O'Dell, M. C, & Koegel, R. L. (1981). The differential effects of Templin, M. (1957). Certain language skills in children. (Institute
two methods of promoting speech in non-verbal autistic children. on Child Welfare Monograph Series 26). Minneapolis: Uni-
Paper presented at the 1981 annual meeting of the American versity of Minnesota.
Speech-Language Hearing Association, Los Angeles. Van Riper, C. (1938). Speech correction. Englewood Cliffs, NJ:
Prather, E., Hedrick, D., & Kern, C (1975). Articulation devel- Prentice-Hall.
Warren, S. F, & Gazdag, G. (1990). Facilitating early language
opment in children aged two to four years. Journal of Speech development with milieu intervention procedures. Journal of
and Hearing Disorders, 40, 179-191. Earfy Intervention, 14, 62-86.
Sander, E. (1972). When are speech sounds learned? Journal of Warren, S. E, & Kaiser, A. P. (1986). Incidental language teaching:
Speech and Hearing Disorders, 37, 55-63. A critical review. Journal of Speech and Hearing Disorders, 51,
Saunders, R., & Sailor, W (1979). A comparison of three strategies 291-299.
of reinforcement on two-choice language problems with se- Yoder, P J. (1987). Relationship between degree of infant handi-
verely retarded children. AAESPH Review, 4, 323-333. cap and clarity of infant cues. American Journal of Mental De-
Seligman, M. E. (1972). Learned helplessness. Annual Review of ficiency, 92, 639-641.
Medicine, 23, 207-412. Yoder, P. J. (1990). The theoretical and empirical basis of early
Skinner, N. F. (1979). Learned helplessness: Performance as a amelioration of developmental disabilities: Implications for
function of task significance. Journal of Psychology, 102, 77-82. future research. Journal of Early Intervention, 14, 27-42.