Randomized Comparison of

Augmented and Nonaugmented

Language Interventions for Toddlers
With Developmental Delays and
Their Parents

MaryAnn Romski
Rose A. Sevcik
Purpose: This study compared the language performance of young children with
Lauren B. Adamson
developmental delays who were randomly assigned to 1 of 3 parent-coached
Melissa Cheslock
language interventions. Differences in performance on augmented and spoken
Ashlyn Smith word size and use, vocabulary size, and communication interaction skills were
R. Michael Barker examined.
Roger Bakeman Method: Sixty-eight toddlers with fewer than 10 spoken words were randomly
Georgia State University, Atlanta assigned to augmented communication input (AC-I), augmented communication
output (AC-O), or spoken communication (SC) interventions; 62 children completed
the intervention. This trial assessed the children’s symbolic language performance
using communication measures from the language transcripts of the 18th and
24th intervention sessions and coding of target vocabulary use.
Results: All children in the AC-O and AC-I intervention groups used augmented and
spoken words for the target vocabulary items, whereas children in the SC intervention
produced a very small number of spoken words. Vocabulary size was substantially
larger for AC-O and AC-I than for SC groups.
Conclusions: This study found that augmented language interventions that include
parent coaching have a positive communication effect on young children with
developmental delays who begin with fewer than 10 spoken words. Clinical
implications suggest that augmented communication does not hinder, and actually
aids, speech production abilities in young children with developmental delays.
KEY WORDS: augmentative/alternative communication, developmental delay,
parents, intervention, language disorders

S
evere spoken language impairments, early in life, have far-reaching
consequences for a child’s long-term development. When a young
child with developmental delay is unable to speak, his or her ability
to express wants and needs, interact socially, gain information about the
complex world in which we live, and benefit from educational contexts is
significantly compromised (Romski, Sevcik, Hyatt, & Cheslock, 2002).
Early language interventions that remediate communication difficulties
are critical to successful developmental outcomes for the child and may
enhance a family’s ability to support their child’s communication devel-
opment (e.g., American Speech-Language-Hearing Association [ASHA],
2008; Guralnick, 2001).

350 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010 • D American Speech-Language-Hearing Association
Augmentative and Alternative Wilcox, 2008; Cress & Marvin, 2003). With the aid of an
augmented language system, the hypothesis is that
Communication Interventions functional symbolic communication skills may emerge
One intervention approach that has been employed earlier than they would through other therapeutic ap-
with children with severe spoken language impairments proaches that focus on prelinguistic skills or speech
and developmental delays is augmentative and alterna- alone. There are only a few research reports in the lit-
tive communication (AAC). AAC systems include manual erature that address the use of AAC in early language
signs and gestures as well as speech-generating devices intervention. Using their research findings with school-age
(SGDs) that produce digitized and/or synthesized speech children (Romski & Sevcik, 1996) as a foundation, Sevcik,
(ASHA, 2002). AAC can play multiple roles in speech and Romski, and Adamson (2004), for example, reported that
language development, including providing an alter- a 4-year-old child with severe developmental disabilities
native output mode and stimulating the development and fewer than 10 spoken words was able to use the
of speech (Hustad & Shapley, 2003; Romski & Sevcik, System for Augmenting Language (SAL) that included
2005). Research has shown that AAC interventions sup- an SGD at home with his parents and sibling as well as
port communication development for school-age chil- during speech-language therapy. The SAL intervention
dren with developmental disabilities (e.g., Beukelman & combined the use of augmented input and output. The
Mirenda, 2005; Nigam, Schlosser, & Lloyd, 2006; Reichle, intervention resulted in increases in the child’s commu-
Light, & Beukelman, 2002; Romski & Sevcik, 1996; nication attempts in both settings and facilitated the
Sigafoos, Drasgow, & Schlosser, 2003). It is important development of a few spoken words.
to note that research has shown that AAC interventions For young children to develop functional communi-
have not hindered speech development for these chil- cation skills, they must be able to comprehend and pro-
dren (e.g., Millar, Light, & Schlosser, 2006). duce language so that they can take on the roles of both
listener and speaker. When AAC interventions are be-
Early Language Interventions gun early in life, the relationship between comprehen-
sion and production is unknown. A number of reviews
The early intervention literature reports that chil- (Blockberger & Sutton, 2003; Reichle et al., 1998; Sevcik
dren are likely to make the greatest gains when inter- & Romski, 2002) have discussed the importance of this
vention is begun during the early stages of development relationship but suggest that we still know very little
(e.g., ASHA, 2008; Guralnick, 2001). For very young chil- about how the two interface to facilitate language de-
dren with developmental delays and concomitant spo- velopment through augmented means.
ken communication impairments, developing functional
When an SGD is available, the focus of early aug-
communication skills is an immediate goal. One ap-
mented language intervention is often on physical access
proach to facilitating language development is to focus
to the device and production of spoken communication
on teaching prelinguistic skills (e.g., Fey et al., 2006;
messages using the SGD (Beukelman & Mirenda, 2005).
Warren et al., 2008; Yoder & Warren, 2002) in order to es-
These interventions have typically focused on having the
tablish a strong intentional communicative foundation,
child produce communications using the SGD with a
including vocalizations and gestures on which to build
range of prompts (Romski et al., 2002). Although SGD
spoken language skills. Another approach is to empha-
production permits beginning language learners to use
size the acquisition of spoken words (Wilcox & Shannon,
language to influence their world (Bates, 1993), it does
1998). Unfortunately, even when these intentional com-
not incorporate the language input a child would receive
munication skills are in place or when a few words are
during development when a caregiver provides spoken
acquired, many children do not smoothly make the tran-
language models. Brady (2000) reported two case stud-
sition from intentional communication to expressive
ies demonstrating that an intervention focused on teach-
language skills (Wilcox & Shannon, 1998). Without a
ing the use of an SGD in joint activity routines to request
symbolic means to communicate while the child is slowly
items also improved comprehension skills for children
(and sometimes unsuccessfully) learning to speak, the
less than 6 years of age with severe intellectual dis-
child, as well as his or her family, becomes frustrated,
abilities. Both participants learned to request six dif-
and behavioral difficulties may emerge (Reichle, Halle,
ferent objects and evidenced increased comprehension
& Drasgow, 1998; Reichle et al., 2002).
skills for the names of the objects requested.
Another approach used to provide experience with
Early Augmented Language Interventions symbols that does not require production is augmented
Early language intervention that includes AAC ap- input (Romski & Sevcik, 2003; Sevcik, Romski, Watkins,
proaches has only recently begun to be considered (e.g., & Deffebach, 1995). It is characterized as the incoming
Beukelman & Mirenda, 2005; Campbell, Milbourne, & communication/language from a child’s communicative

Romski et al.: Early Language Interventions 351

partner that includes speech and is supplemented by children. These studies also highlight the role of the
AAC symbols, the speech output produced by an SGD partner in communication interactions as they provide
when the symbol is activated, and the environmental language models and input to the child. There have been
context (Romski & Sevcik, 2003). Young children learn to no investigations to date, however, that compare the
comprehend language before they begin to produce lan- effects of providing augmented input versus augmented
guage, though production emerges very soon thereafter output on symbolic language learning with very young
(Reichle et al., 1998; Romski & Sevcik, 2005). Romski children with developmental delays.
and Sevcik have argued that like very young typical
children, children receiving augmented language inter- Family Participation in Early
vention should have the opportunity to be exposed to a
mode for language production before they are asked to Language Intervention
use it. Using augmented input may place less pressure Another aspect of intervention for very young chil-
on the dyad and at the same time mirror the typical dren is family participation (Guralnick, 2001; Sevcik &
developmental experience. This point becomes particu- Romski, 2007). Because the majority of the young child’s
larly salient when the targeted intervention participants daily activities are with the parents, an intervention
are very young children. that provides ways for parents to enhance their commu-
Variations of this approach include aided language nication interactions with the child is essential during
stimulation, a broader term somewhat analogous to this developmental period (Kaiser & Hancock, 2003;
total communication and simultaneous communication Kent-Walsh & McNaughton, 2005). There are a number
(Goossens’, 1989), aided language modeling (Drager et al., of approaches that have been employed to teach partners
2006), and aided AAC modeling (Binger & Light, 2007). to utilize communication strategies. Binger, Kent-Walsh,
Aided language modeling provides AAC symbol models Berens, DelCampo, and Rivera (2008), for example, used
on a language board (without speech output) during in- an eight-step cognitive strategy approach to successfully
teractive naturalistic play activities. Similarly, aided teach three Latino parents to use an AAC modeling
AAC modeling provides models of symbols and/or sym- approach to support their children’s multisymbol mes-
bol combinations with the expressed goal of the child sage productions. Another approach that has been im-
producing them. plemented in early intervention is parent coaching (Kaiser
A few recent studies have examined the impact of & Hancock, 2003; McCollum & Yates, 1994). A parent
these input interventions on preschool children’s symbol coaching strategy teaches parents how to interact and
comprehension and production. Harris and Reichle use new tools with their children. There is some research
(2004) examined the use of aided language stimulation that suggests this strategy can work in early language
for 3 preschool children with moderate intellectual dis- intervention (e.g., Hemmeter & Kaiser, 1994; Kaiser
abilities who were functionally nonspeaking with fewer et al., 1996), but there has been minimal use with AAC
than 30 words by using a multiple-probe design across interventions.
symbol sets/activities. All 3 children displayed increased
symbol comprehension and production. Drager et al. (2006)
examined the effectiveness of aided language modeling
Study Questions and Hypotheses
on symbol comprehension and production in two 4-year- In the present study, we compared the symbolic lan-
old preschool children with autism who used fewer guage development of children who were randomly as-
than 30 functional words. Although both children sub- signed to one of three early parent-coached language
stantially increased symbol comprehension and elicited interventions (one spoken communication [SC] and aug-
symbol production, the rate of symbol comprehension mented communication input [AC-I] and output [AC-O]).
development was faster than the rate of symbol pro- Specifically, we asked what differences there were in per-
duction development, indicating that aided language formance on the target vocabulary size and use (in spoken
modeling may be a viable intervention for teaching com- and augmented words) and on communication interac-
prehension and, to a lesser extent, symbol production for tion skills across the three language interventions. The
preschoolers with autism. Using a multiple-probe de- primary hypothesis was that all children would benefit
sign, Binger and Light (2007) examined how aided AAC from the three interventions but that the two interven-
modeling affected the production of symbol combinations tion groups who received augmented language compo-
in 5 preschool children who had an expressive vocabulary nents would have better language outcomes than the
of at least 25 symbols/words. Of the 5 children, 4 con- spoken language intervention group. Of the two aug-
sistently learned to produce symbol combinations when mented language interventions, we expected that given
provided with models. These studies suggest that pro- the presenting age and language skills of the children,
viding varying forms of augmented models increases the AC-I intervention would have stronger effects than the
symbol comprehension and /or production for preschool AC-O intervention.

352 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
 percent of the 62 children scored below the 1st percentile
Method on the MSEL.
Participant Recruitment and Identification Each child and his or her parent were randomly as-
Participants were recruited for the study from ap- signed to one of the three language interventions. Ran-
proximately 45 sources (e.g., early intervention services, dom assignment was stratified for etiology and MSEL
private speech-language pathologists [SLPs], clinical composite score. Table 1 provides a description of the
psychologists, developmental pediatricians, and pediat- participants by group assignment at the beginning of the
ric neurologists) in the metropolitan Atlanta area who interventions on chronological age; the MSEL standard
regularly see children who could meet the participant score on visual reception, fine motor skills, and receptive
selection criteria. Participant selection criteria included and expressive language; the MSEL composite t score;
an age range of between 24 and 36 months; a significant the Sequenced Inventory of Communication Develop-
risk for speech and language impairment, which was ment (SICD; Hendrick, Prather, & Tobin, 1984) receptive
operationally defined as not having begun to talk (as in- and expressive language age in months; the MacArthur
dicated by a vocabulary of at most 10 intelligible spoken Communicative Development Inventories (Fenson et al.,
words and a score of less than 12 months on the Ex- 1993) receptive and expressive vocabulary size; and the
pressive Language Scale of the Mullen Scales of Early Clinical Assessment of Language Comprehension (Paul
Learning [MSEL; Mullen, 1995]); at least primitive in- & Miller, 1995) raw score out of 9 on the Emerging Lan-
tentional communication abilities; upper extremity gross guage Scale. These intervention experiences were sup-
motor skills that permitted the child to touch the symbols plemental to other clinical services the children received.
on the SGD; a primary etiology other than delayed speech At the onset of the study, 74% of the children were en-
and language impairment, deafness/hearing impairment, rolled in state-funded early intervention services, and
or autism; and English as the primary language spoken 87% of the children received individual speech-therapy
at home. Parents received information about the study services. None of the children participated in a parent-
from a recruitment source and, if interested in partici- coached language intervention.
pating in the project, individually contacted project re- Six children (9%; 2 AC-I, 2 AC-O, and 2 SC) with-
searchers to discuss the child’s profile and schedule an drew from the intervention (reasons included time demand,
initial appointment with the principal investigator and
the project’s coordinating SLP. Table 1. Variable means at baseline by intervention group.

Intervention group

Participants Variable AC-I AC-O SC h2G p

Across the 5-year period of the study, 93 parents Chronological age (months) 29 30 29 .01 .73
contacted project researchers about participating in the MSEL visual reception (SS) 29 32 32 .008 .78
study. Of these 93 families, 13 children did not meet the MSEL fine motor (SS) 26 24 28 .03 .36
participant selection criteria (usually the child had more MSEL receptive (SS) 28 29 27 .005 .87
than a 10 spoken word vocabulary), and an additional MSEL expressive (SS) 22 22 21 .02 .57
12 children who met the criteria did not complete the MSEL composite 60 59 59 .002 .94
SICD receptive (months) 18 20 19 .01 .67
preintervention assessment. Sixty-eight children were
SICD expressive (months) 12 13 13 .01 .70
randomly assigned to an intervention by a co-investigator
CDI receptive (vocabulary size) 114 144 155 .03 .43
who did not participate in the day-to-day interaction CDI expressive (vocabulary size)a 11 14 8 .03 .44
with the children. CALC (score out of 9) 8.0 8.6 7.7 .01 .72
Sixty-two children (91% of those children who began
the intervention; 43 boys, 19 girls; mean chronological Note. N = 62: 21 in AC-I, 20 in AC-O, and 21 in SC. The effect
size statistic (h2G) is the proportion of error variance accounted for by
age = 29.60 months, range from 21 to 40 months) com-
group per a one-way analysis of variance (ANOVA). AC-I = augmented
pleted the intervention. All but one of these children
communication input; AC-O = augmented communication output;
completed all 24 intervention sessions. One child in SC SC = spoken communication; MSEL = Mullen Scales of Early Learning;
completed 23 sessions; we used the child’s 23rd session SS = standard score; SICD = Sequenced Inventory of Communication
as the final session in our data analyses. The sample in- Development; CDI = MacArthur Communicative Development Inventories;
cluded 18, 7, and 37 children from African American, CALC = Clinical Assessment of Language Comprehension.
Asian, and Caucasian backgrounds, respectively. Medical a
CDI expressive vocabulary included an animal sounds category.
etiology included genetic syndromes (e.g., Down syn- Words in this category were not included in the calculation of fewer than
drome), seizure disorders, cerebral palsy, or unknown 10 words. Clinical observation and the MSEL expressive score were
conditions. All children had hearing and vision within utilized to ensure that the children were producing fewer than 10 words.
normal limits as described in intake reports. Eighty-eight

Romski et al.: Early Language Interventions 353

travel distance, and perceived lack of benefit from the randomly assigned to deliver all three interventions so
intervention to which they were assigned). They ended that each interventionist delivered on average 10 inter-
their participation on average in the sixth session of the ventions across the three intervention protocols. There
intervention (range = second to ninth session). They were usually two interventionists on staff at any one
were evenly distributed across the three interventions time during the study. They could not be masked to the
and did not differ on their preintervention MSEL scores intervention assignment but were masked to the re-
from those who remained. search questions to be evaluated. The project ASHA-
Sixty-two parents (58 mothers, 4 fathers) with a certified SLP with 13 years of early intervention clinical
mean chronological age of 37.33 years (SD = 4.73; range experience supervised a minimum of 30% of the inter-
from 31 to 45 years of age) participated in this study. The vention sessions as per guidelines for supervision of SLP
parent who participated was the one who could commit assistants (ASHA, 2004).
to participating in all of the assessment and intervention Each intervention had four components (target vo-
sessions with the child at the laboratory and at home. cabulary, parent coaching, mode, and strategies). Table 2
The sample included 18, 5, and 39 parents from African compares the three interventions across these compo-
American, Asian, and Caucasian backgrounds, respec- nents and includes examples for each.
tively. Six of the parents had completed at least 12th Target vocabulary. To equate target vocabulary use
grade or had obtained a general equivalency diploma, across interventions, a set of vocabulary items was in-
8 had some college, 25 graduated from college, 21 had at dividually chosen for each child. With guidance from the
least some postgraduate education, and 2 did not project’s SLP, each parent suggested vocabulary that
complete the question. Thirty-one of the parents stayed (a) the child did not comprehend or produce in speech or
at home with their children and did not work (50%; 9 AC-I, sign, ( b) would be motivating for the child during each of
13 AC-O, and 9 SC). the three 10-min routines, and (c) was appropriate for
use at home during similar routines. At the onset of the
Research Design intervention, each child then had a set of target vocab-
ulary items appropriate to play (e.g., blocks, play, and
Each child and his or her parent participated in pre- car), book reading (e.g., book, open, and truck), and snack
intervention assessments and were then randomly as- routines (e.g., spoon, yogurt, cookie, and juice), and a few
signed to an intervention. During the preintervention vocabulary items that were appropriate across routines
phase of the study, participants completed an assessment (e.g., my turn, more, all done, and what’s this?). Target
battery (including descriptive, cognitive, and language/ words were represented as spoken words or augmented
communication measures) that was conducted by the words depending on the intervention condition. For the
project’s SLP, who was masked to the child’s group as- AC-I and AC-O interventions, target vocabulary aug-
signment until after the assessments were complete. mented words were represented on the SGD using Pic-
Parents received a written report of the child’s perfor- ture Communication Symbols (Mayer-Johnson, 1981).
mance after the assessment. To ensure that children in all interventions were given
experience with each of the targeted vocabulary items,
interventionists and parents were encouraged to use all
Interventions target vocabulary in every session. They had an index
Parent–child pairs were randomly assigned to one of card with all the child’s targeted vocabulary available
three parent-implemented communication interventions— during each session so that they could ensure that all of
SC, AC-I, or AC-O—and participated in 24 intervention the vocabulary items were used during a session. Over
sessions. Eighteen sessions (usually 9 weeks) took place the course of the intervention, additions to the target
in the laboratory setting, and 6 sessions (usually 3 weeks) vocabulary were made when a child began to use a ma-
took place in the child’s home so that the parent could jority of his or her already assigned target vocabulary
generalize the strategies to routines at home. Sessions as determined by the project’s SLP, interventionist, and
were 30 min in length and consisted of three 10-min parent.
blocks of play, book reading, and snack in that specific Parent coaching. Parents received an intervention
order. Each 10-min block was monitored with a timer to protocol manual with weekly materials that included
ensure equal time per routine across interventions. goals for the parent, the interventionist, and the child
Six female interventionists (mean age = 25.6 years), (Romski & Cheslock, 2000). For the first 8 sessions, the
with at least a bachelor’s degree in psychology or com- parent and the project’s SLP observed, through a one-
munication, were taught to deliver all three interven- way mirror, as the interventionist implemented the three
tions across the study. The six interventionists worked routines with the child. During these observations, the
for the project an average of 21.8 months and were project’s SLP highlighted general and specific strategies

354 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
Table 2. Components of the three interventions.

Component AC-I AC-O SC

Target vocabulary Individualized target vocabulary of Individualized target vocabulary of Individualized target vocabulary of
visual-graphic symbols + spoken visual-graphic symbols + spoken spoken words with use of ALL
words with use of ALL target words with use of ALL target target vocabulary during each
vocabulary during each session. vocabulary during each session. session. I/P had a card with all
I/P had a card with all target I/P had a card with all target target vocabulary listed.
vocabulary listed. vocabulary listed.
Mode I/P uses SGD to provide Child uses SGD to communicate. I/P and child use speech to
communication input to child. communicate.
Strategies I/P provides vocabulary models to I/P provides verbal and/or I/P provides verbal prompts so
child using the device; symbols are hand-over-hand prompts that the child produces spoken
positioned in the environment to so that the child produces words.
mark referents; I/P reinforces the communication using the SGD.
child’s productive communications.
Parent coaching I provides coaching for P. I provides coaching for P. I provides coaching for P.

Sample interaction Adult (A) and child (C: Emily) Adult (A) and child (C: Johnny) are Adult (A) and child (C: Lem)
are having snack. playing with blocks. are playing.
A: Mmm. A: Look Johnny. A: Let’s play with the truck.
A: Now what do you want? A: Here are the blocks. A: Look (A points to mouth).
A: COOKIE or CRACKER. A: Tell mama build. A: Look.
C vocalizes unintelligible and holds C: PLAY. A: /t/ /t/
out hand. A : Yep, we’re playin’. C: XX (vocalizes unintelligibly).
A: Cookie or cracker? A: Tell mama build (A taps on SGD). A: Truck.
C: CRACKER. A: Tell me build. C: XX.
A: Good. You want a cracker. C: BUILD (A provides A: Right?
A: OK. (A gives the cracker to Emily.) hand-over-hand assistance.) A: Look at my face.
A: That tastes good. A: Alright.

Note. Words in caps indicate speech-generating device (SGD) use. I = interventionist; P = parent; XX = unintelligible vocalization. The interventionist
treatment implementation manual is available from the authors upon request (Romski & Cheslock, 2000).

that the interventionist implemented with the child and lightweight, modestly priced SGDs that employed
answered the parent’s questions about the interactions. direct selection to access the device, had a digitized
Beginning in the 9th session, the parent joined the ses- voice, and were adequate for the needed target vocab-
sion for the last 10 min (snack), implemented the interven- ulary size represented via visual graphic symbols.
tion routines, and received coaching by the interventionist. The SGD inventory included CheapTalk, Communi-
General coaching included, for example, instructing par- cation Builder, GoTalk, TechSpeak, and TechTalk. If a
ents to shorten their utterance length, to give the child a child gestured or signed during the intervention, it
choice between two toys or books or snacks, and to pause was accepted as part of the child’s general communi-
and slow the interaction down so that the child had a cation, but there were no signed target vocabulary
chance to communicate. Specific coaching focused on items.
how to implement each intervention (e.g., model SGD Intervention strategies. All three interventions were
use in AC-I, use hand-over-hand prompts in AC-O, focus presented in a naturalistic fashion individualized for
on producing speech sounds in SC). Beginning in the each child rather than a specific number of trials or
16th session, the parent led the remaining sessions, in- opportunities. They shared some general intervention
cluding the last 6 sessions at home. The interventionist strategies, such as age-appropriate naturalistic rou-
continued to provide coaching as needed. Intervention tines, environmental arrangement, provision of choices,
sessions were videotaped through the one-way mirror or and pausing to “encourage” communication. It is the
from a discrete location at home. mode in which the strategies were implemented and
Mode. For mode, SC intervention used speech in- whether a response was required that differentiated the
put and output. The AC-I and AC-O interventions used interventions.

Romski et al.: Early Language Interventions 355

 SC served as a contrast condition to AC-I and AC-O. Table 3. Means (standard deviations) for proportion of intervention
In SC, the interventionist and the parent visually and features appropriately implemented across interventions and
verbally prompted the child to produce spoken words for session type.
an individualized target set of vocabulary. For example,
the interventionist would prompt the child to “tell me Session type
using your mouth” while pointing to her or the child’s Interventionist Parent Parent
mouth. Sound stimulation cues (e.g., “say /m/ more”) Intervention only supported led Overall
were provided as necessary.
SC 0.93 (0.06) 0.94 (0.04) 0.92 (0.07) 0.93 (0.06)
In AC-I, the interventionist and the parent modeled
AC-I 0.95 (0.06) 0.95 (0.07) 0.90 (0.10) 0.93 (0.08)
augmented and spoken word use of the individualized AC-O 0.93 (0.07) 0.93 (0.06) 0.92 (0.08) 0.93 (0.07)
target vocabulary via an SGD, and symbols also were Overall 0.94 (0.06) 0.94 (0.06) 0.91 (0.08)
positioned in the environment to mark referents. No
direct demand was made for the child to communicate. If
the child used an augmented or spoken word, s/he was
groups were .80, .83, and .92 for AC-I, AC-O, and SC,
provided with natural reinforcers and the activity con-
respectively. Kappas for the agreement between the two
tinued. If the child did not attempt communication at
ratings across sessions within an intervention were .89,
that time, the activity continued without encouraging
.74, and .82 for early (1–8), middle (9–16), and late
the child to produce an augmented word.
(17–24) sessions, respectively. The overall kappa was
In AC-O, the interventionist and the parent used an .83. Fleiss (1981) characterizes kappas of .60 to .75 as
SGD and individualized target vocabulary represented good and over .75 as excellent.
via symbols and spoken words to have the child produce
augmented and spoken words. The interventionist and
the parent visually, verbally, and physically prompted Transcript Creation
the child to produce augmented words for an individu-
Language transcripts, using the Systematic Analy-
alized target set of vocabulary. For example, the inter-
sis of Language Transcripts (SALT; Miller & Chapman,
ventionist would prompt the child to “tell me what you
1985), were created to characterize parent–child com-
want using the device” while generally pointing in the
munications during 30-min videotaped observations of
direction of the SGD. If the child did not activate the
the parent and child at three points in time: preinter-
symbol on the device, hand-over-hand prompting was
vention and during the 18th session (in the laboratory)
used so that the child produced the symbol.
and the 24th session (at home). At preintervention, the
parent and child interacted for 30 min in 10-min blocks
Reliability of Intervention Implementation of play, book reading, and snack. No target vocabulary
items were available during this interaction; thus, target
To ensure that the three interventions were imple-
vocabulary use was zero at the preintervention session.
mented reliably across interventionists and time, the
At Sessions 18 and 24, the parent and child interacted
Treatment Implementation Rating Scale (TIRS; Romski,
during the 30-min intervention session that included
Sevcik, Adamson, Cheslock, & Smith, 2007) was devel-
three 10-min blocks of play, book reading, and snack.
oped. The TIRS assessed how accurately each interven-
tion was implemented across participants. It included Nine transcribers, masked to the questions of in-
13 yes/no items that encompassed common components terest, were trained using a transcription and coding
across the three interventions, components specific to manual (Romski & Cheslock, 2004). Transcribers were
each intervention, and a scoring key specific to each in- not informed of the intervention group to which the child
tervention condition. Seven independent observers, was assigned; however, there was no way to mask the
masked to the questions of interest, were taught to re- presence of an SGD. Before transcribing and coding
liably utilize the TIRS to code the intervention sessions. began on the actual samples, transcribers learned to use
A total of 310 sessions (21% of each child’s sessions, dis- SALT on a set of specific videotapes until they demon-
tributed across the 24 sessions) were viewed and rated in strated at least 90% agreement using SALT conventions
a random order by one of seven independent observers. with a preestablished standard.
The features of the interventions were implemented SALT defines an utterance “as a sentence or phrase,
appropriately across session type (interventionist only, reflecting a single thought. The primary goal is to doc-
parent-supported, and parent-led) and intervention (SC, ument ‘thought completion’.” Determining what con-
AC-I, and AC-O; see Table 3). Of these video-recorded stitutes an utterance is simple when a turn consists of
sessions, 20% (62) were randomly selected, viewed, and a single sentence. As a general rule, turns consisting
rated independently by a second trained rater. Kappas for of several sentences may be broken down into single
the agreement between the two ratings across intervention sentences, each transcribed as a separate utterance.

356 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
Falling/rising intonation and the presence of a pause agreement in the number of target words identified
helped determine the completion of a thought. To in- across the two versions of the transcripts, and the num-
clude SGD symbol use as utterances, our transcription ber of target words identified correlated very highly
manual added, “If a person takes a communicative turn between each version of the transcript (r = .99, p < .001).
that consists only of a nonverbal communicative act, Kappas for the agreement between the two coders were
this act should be recorded as an utterance” (Romski & .98, .97, and 1.00 for AC-I, AC-O, and SC, respectively.
Cheslock, 2004). Utterances were transcribed as unintel- The overall weighted kappa was .97 and characterized
ligible (XX is the SALT convention) when the transcriber as excellent by Fleiss (1981).
was unable to understand the speaker’s vocalization after The SALT program was used to automatically cal-
listening to the utterance three times. An utterance with culate 6 child and 3 adult communication interaction mea-
any unintelligible parts was considered unintelligible. sures from the reliable transcripts. The child measures
To ensure that the transcripts were reliable, three included the number of augmented and/or spoken words
independent reviews were conducted. A second trained employed during the session, the mean length of utterance
transcriber independently reviewed every transcript in morphemes ( MLUm; calculated by dividing the total
while watching the videotaped interaction and made number of morphemes the child or parent used in a
corrections and changes as appropriate according to transcript by the total number of speaker utterances),
SALT conventions and the transcription manual. A third type/token ratio (calculated by dividing the total number
trained transcriber then independently error checked the of different words spoken by the child [i.e., types] by the
augmented and spoken word codes assigned to the target total number of words spoken by the child [i.e., tokens]),
words in each transcript to ensure that the appropriate total turns (SALT defined a speaker turn as one or more
codes were assigned to the appropriate target words as consecutive utterances by a single speaker; these were
defined by the transcription manual. Finally, a fourth summed to determine total turns), mean length of turn
trained transcriber independently reviewed each tran- in utterances (calculated by dividing the number of ut-
script while watching the videotaped interaction and made terances used by the speaker by the number of turns
appropriate changes to ensure maximum accuracy. taken by the speaker), and proportion of intelligible ut-
terances (calculated by dividing the number of speaker
Child and Parent Language intelligible utterances by the total number of speaker
utterances). The parent measures calculated were MLUm,
and Communication Measures mean length of turn in utterances, and total turns.
As the SALT transcripts were created, the tran-
scriber also coded each transcript for target augmented
and spoken word use by the child and the parent or in-
terventionist. The transcriber was given a list of the Results
target vocabulary for the specific child. Target augment- Preintervention Group Comparisons
ed word use was defined as touching a visual-graphic
symbol representing a target vocabulary item on an SGD Comparisons of the three groups were conducted to
that produced a digitized equivalent of the English word. assess possible group differences on a set of variables
Target spoken word use was defined as the production of (see Table 1). One-way analyses of variance (ANOVAs)
a combination of sounds that was consistently used by revealed weak and insignificant differences on all the
the child and interpreted by the listener as a target variables. This finding indicates that the three groups
word. Only spontaneous (nonimitative or nonprompted) were well matched in terms of the variables that were
augmented and spoken word use was included in the measured before intervention.
analyses. In the infrequent event that a child used two
different symbols or a symbol and a spoken word, sepa-
rate codes (one for each target vocabulary item) were Length of Intervention Implementation
employed. A code was placed at the utterance level to The median number of actual weeks required for a
denote a combination of symbols or symbols and words, parent–child pair to complete the intervention was 16,
but this code is not analyzed in this study. 16, and 15 for AC-I, AC-O, and SC, respectively (ranges
To determine reliability for the coded augmented and were 10–23, 12–23, and 10–28, respectively, except for
spoken target vocabulary words, 20% of the 124 tran- one AC-I outlier at 51 weeks). Delays were due primarily
scripts (n = 24) were randomly selected, balanced across to child illness, family vacation periods, or family sched-
intervention and Sessions 18 and 24, to compare the uling conflicts. There were no significant correlations
number of available target words identified by the between the length of intervention in weeks and the chil-
independent transcribers in the third and final version dren’s target vocabulary, defined as the combined use of
of the transcripts. The transcribers demonstrated 86% augmented and spoken words for a target vocabulary item,

Romski et al.: Early Language Interventions 357

at the 18th or 24th session (r = .012 and .0, respectively; expected, the AC-O and SC groups did not differ at Ses-
p = .90 and .76). sion 18, with means for spoken input for AC-I, AC-O, and
SC of 185, 303, and 293, respectively. At Session 24, the
means for AC-I, AC-O, and SC spoken input were 211,
Available Target Vocabulary Items 328, and 256, respectively. All three groups were signifi-
At the onset of the study, the mean number of target cantly different, with AC-I receiving the least spoken in-
vocabulary items was 15 for all three groups. At Sessions put and AC-O receiving the most spoken input.
18 and 24, the mean target vocabulary size increased
to 19 and 22 or 19 and 23, for the children in the AC-I Augmented Word Use by AC-I
and AC-O interventions, respectively. The mean target
vocabulary for children in the SC group increased to 16 and AC-O Children
for both Sessions 18 and 24. Variability in target vo- The mean number of augmented words used and their
cabulary size was much greater for AC-I and AC-O chil- percentage of the target vocabulary are given in Table 4.
dren at Sessions 18 and 24 (SD = 10.9–14.5) than for AC-I All children in the AC-I and AC-O intervention groups
and AC-O children at baseline and SC children at all used at least one augmented word for the target vocab-
sessions (SD = 2.0–2.7) due to 2 AC-I and 4 AC-O chil- ulary items at Sessions 18 and 24 except for 1 AC-I child
dren with extreme values (e.g., at 24 sessions, the max- who did not use any augmented words at the 24th ses-
imum was 52, 57, and 21 available target vocabulary sion. A Session × Group ANOVA showed that augmented
items for AC-I, AC-O, and SC children, respectively). word use was higher for AC-O than for AC-I children but
The target vocabulary input was calculated by tal- did not change significantly from Session 18 in the lab-
lying the number of target vocabulary augmented and oratory to Session 24 at home (h2S < .01, p > .5, for
spoken words used by the adult from the transcripts number and percentage). This finding suggests that both
separately at Sessions 18 and 24. A Group × Session × augmented intervention groups maintained their aug-
Mode (augmented, spoken) ANOVA showed no signifi- mented word use when the intervention moved from the
cant main effects for group or session, F(1, 59) < 1, p > .50; laboratory to the home.
and F(2, 59) = 2.56, p = .09, respectively. There also was
a significant main effect for mode, F(1, 59) = 277.35, Spoken Word Use by AC-I, AC-O,
p < .001, h2M = .83; and significant two-way interac-
tions between group and session, F(2, 59) = 4.14, p = .02,
and SC Children
h2G × S = .12; and group and mode, F(2, 59) = 40.12, Many children did not use any spoken words at Ses-
p < .001, h2G × M = .58. These effects were qualified by a sions 18 or 24, especially in the SC intervention (see
significant three-way Group × Session × Mode interac- Table 5). The percentage of children using spoken words
tion, F(2, 59) = 5.11, p < .01, h2G × S × M = .15. A Tukey ad exceeded 50% only for the AC-O intervention. Four AC-I
hoc analysis showed that, as expected, children in the children, 8 AC-O children, and 1 SC child used spoken
AC-I group received significantly more augmented input words during both Sessions 18 and 24. Because few
than did children in the AC-O group. The means for children used spoken words at both sessions, group dif-
augmented input were 148 and 1 and 142 and 2 for the ferences were tested separately for each session; as
AC-I and AC-O interventions at Sessions 18 and 24, shown in Table 5, none were significant, although at both
respectively. Ideally, AC-O input should have been zero; sessions the mean number of spoken words was highest
the very small means are input produced in error. An for the AC-O and lowest for the SC intervention. Another
additional Tukey ad hoc analysis revealed that at Ses- analysis considered only those 4 AC-I and 8 AC-O chil-
sion 18, the AC-I group received significantly less spoken dren who used spoken words at both sessions; their
input than did children in the AC-O and SC groups. As Session 18 and Session 24 means were 2.75 and 5.50 for

Table 4. Augmented word use by children in the AC-I and AC-O interventions.

Session 18 Session 24

Variable AC-I AC-O AC-I AC-O h2G p

No. of augmented words used 9.3 (6.4) 14.4 (9.5) 9.4 (7.0) 14.9 (10.0) .11 .04
% of target augmented words 50 (23) 68 (19) 50 (25) 66 (26) .14 .02

Note. Scores are means (with standard deviations in parentheses); n = 21 and 20 for AC-I and AC-O groups, respectively;
h2G is the AC-I versus AC-O effect size per a Group × Session ANOVA. Scores at the start of intervention were zero.

358 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
 Table 5. Spoken word use by children in the AC-I, AC-O, and SC interventions.

Session 18 Session 24

Variable AC-I AC-O SC h2G AC-I AC-O SC h2G

Children using spoken words 10 12 5 — 6 12 7 —

No. of different spoken words used 2.3 (1.8) 3.3 (4.9) 1.4 (0.5) .05 4.2 (4.7) 4.8 (6.7) 2.3 (2.2) .04
% target spoken words 14 (13) 13 (10) 9 (4) .03 21 (23) 15 (13) 15 (14) .03

Note. No. of different spoken words used and % target spoken words are means for children who used at least one word (with
standard deviations in parentheses); n = 10, 12, and 5 for Session 18, and 6, 12, and 7 for Session 24 for the AC-I, AC-O, and SC
groups, respectively. h2Gs are group effect sizes per one-way ANOVAs for Sessions 18 and 24 separately, p > .50 for all.

the AC-I group and 4.50 and 6.50 for the AC-O group, Child and Parent
respectively. Per repeated measures ANOVAs for each
group, effect sizes were moderate but, due to the small n,
Communication Interaction
statistically insignificant, ph2 = .20 and .17, p = .46 and Child and parent MLUm, mean length of turn ( ML
.27, for the AC-I and AC-O groups, respectively. turn), and number of total turns and child type/token
The odds that children in the AC-I and AC-O inter- ratio, utterance intelligibility, and utterance rate were
ventions would use spoken words at Session 18 was 2.9 assessed for the three intervention groups. Total child
and 4.8 greater than the odds that children in the SC and parent turns and child utterance rate intercorre-
intervention would, although only the latter was statis- lated at .95 or better for all sessions; nonetheless, even
tically significant (95% confidence intervals [CIs] = 0.78– though these three variables can be viewed as essen-
10.9 and 1.25–18.4, p = .11 and .02, respectively). tially the same variable, means for all three are given
Corresponding odds ratios for children in the AC-I or here for descriptive purposes. Additionally, these three
AC-O interventions at Session 24 were 0.8 and 3.0 variables all correlated at .50 or better with parent ML
(95% CIs = 0.21–3.0 and 0.84–10.7, p = .74 and .09), turn. Intervention Group × Session ANOVAs with ses-
respectively. sion as a repeated measure ( baseline, Session 18, Ses-
sion 24) showed significant session effects for all of these
variables except parent MLUm (see Table 6). Tukey ad hoc
Vocabulary Size analyses revealed that Sessions 18 and 24 differed from
Because augmented words were the functional baseline ( p < .05) for the remaining eight variables.
equivalents of spoken words for the children in the AC-I With the exception of child type/token ratio and child ut-
and AC-O interventions, we tallied the combined use terance intelligibility (discussed in the next paragraph),
of augmented and spoken words for a target vocabulary
item to gain a common measure of “vocabulary size” across
Figure 1. Vocabulary size means for augmented communication
all three interventions. For children in the AC-I and AC-O
input (AC-I), augmented communication output (AC-O), and spoken
interventions, a child was given credit for only one vo-
communication (SC) intervention groups at Sessions 18 and 24. Black
cabulary item if a vocabulary item was represented by indicates that only an augmented word was used, white that only a
an augmented and a spoken word. Considering both aug- spoken word was used, and gray that both an augmented and
mented and spoken words increased the scores shown spoken word were used.
in Table 5 by about one item or a couple of percentage
points and showed the same pattern of AC-I versus AC-O
difference but little difference due to session. An ANOVA
that included scores for all children (including zeros)
indicated group effects for both Sessions 18 and 24,
F(2, 59) = 25.6 and 20.5, h2 = .47 and .41, respectively;
p < .001 for both. A Tukey ad hoc test revealed that all
three groups differed significantly on vocabulary size at
Session 18; the mean for SC was significantly lower than
both AC-I and AC-O, and AC-I was also significantly
lower than AC-O. Another Tukey ad hoc test revealed
that the SC group mean was significantly lower than
both the AC-O and AC-I group means at Session 24 (see
Figure 1).

Romski et al.: Early Language Interventions 359

 Table 6. Child and parent communication measures across sessions.

Session

Variable Baseline 18 24 ph2S p

Child MLUm 0.91 (0.17) 0.98 (0.08) 0.99 (0.09) .15 <.001
Child ML turn 1.10 (0.09) 1.19 (0.16) 1.18 (0.18) .14 <.001
Child type/token ratio .07 (.07) .12 (.08) .14 (.08) .26 <.001
Child utterance intelligibility .09 (.14) .35 (.28) .41 (.29) .59 <.001
Child utterance rate 3.40 (2.48) 5.57 (3.06) 5.22 (3.22) .26 <.001
Child total turns 91 (65) 134 (76) 131 (78) .26 <.001
Parent MLUm 3.37 (0.62) 3.42 (0.57) 3.46 (0.56) .02 .38
Parent ML turn 12.05 (13.35) 6.82 (5.41) 7.05 (6.67) .17 <.001
Parent total turns 91 (60) 128 (67) 133 (71) .27 <.001

Note. Scores are means (with standard deviations in parentheses); n = 62 for each session. ph2S is the session effect size from
a Group × Session ANOVA. MLUm = mean length of utterance in morphemes; ML turn = mean length of turn.

group effects for the other variables were weak and not on the communication of very young children with de-
statistically significant (h2G ranged from .02 to .06, velopmental delays and fewer than 10 spoken words at
p > .10), and except for child type/token ratio and the onset of intervention. Using a random assignment to
utterance intelligibility, no Group × Session interac- intervention group, we found that all children in both
tions were significant. AC-O (the production intervention) and AC-I (the com-
Child type/token ratio and child utterance intelligi- prehension intervention) used augmented words for
bility correlations were .25, .67, and .61 at baseline, Ses- communication, and more children in both augmented
sion 18, and Session 24, respectively. The group effect was interventions produced target spoken words than did
significant for both (h2G = .30 and .49; p < .001 for both), and the children in the spoken language intervention. Con-
as just noted, the session and group main effects were trary to our initial hypothesis, children who were ran-
qualified with a Group × Session interaction (h2G × S = domly assigned to AC-O gained more than did children
.087 and .31, p = .04 and < .001, for child type/token ratio assigned to AC-I. Nevertheless, it is noteworthy that
and utterance intelligibility, respectively), although this in both augmented interventions, children made signifi-
was marginal for type/token ratio. For type/token ratio, cant gains in target augmented word use after just 18 ses-
baseline differed from Sessions 18 and 24 for all groups, sions of parent-coached intervention and maintained their
but the difference was considerably greater for the AC-I target augmented word use at 24 sessions after the inter-
and AC-O groups than for the SC group—thus the signif- vention was transferred to the children’s homes. These
icant Group × Session interaction. Specifically, baseline, results lend support to previous findings that parents
Session 18, and Session 24 type/token ratio means were are able to implement augmented language interventions
.09, .16, and .16 for the AC-I intervention and .07, .14, (e.g., Binger et al., 2008; Romski et al., 2007). Parent coach-
and .17 for the AC-O intervention but .05, .06, and .08 for ing is also a viable approach by which to integrate parents
the SC intervention. Finally, for utterance intelligibility, into an augmented intervention.
group means averaged over sessions differed significantly With respect to the development of target spoken
from each other (.31, .44, and .10 for AC-I, AC-O, and SC, word vocabulary, the effects were modest for children in
respectively). Moreover, baseline differed from Session 18 all three interventions. A small number of children pro-
only for AC-I and AC-O, whereas Session 24 differed for duced a small percentage of their target word vocabulary
all groups—hence the significant Group × Session inter- in spoken words. The children who received AC-O, how-
action. Specifically, baseline, Session 18, and Session 24 ever, were more likely to produce spoken words at 18 ses-
utterance intelligibility means were .07, .42, and .48 for sions and maintain them at 24 sessions than children in
the AC-I intervention and .15, .53, and .63 for the AC-O either the AC-I or SC interventions. This result supports
intervention but .06, .10, and .13 for the SC intervention. the extant research with older children with developmen-
tal delays (Millar et al., 2006) and the clinical literature
showing that AAC intervention does not hinder the
Discussion development of speech for young children (Cress & Marvin,
2003). It is also consistent with the results of Yoder and
The results of this study suggest that parent-coached Stone’s (2006) randomized comparison of Responsive
augmented language interventions have a positive effect Education and Prelinguistic Milieu Teaching ( RPMT;

360 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
Yoder & Warren, 2002) and the Picture Exchange Com- outcome when either augmented input or output is the
munication System ( PECS; Bondy & Frost, 1994) with intervention focus. Like the process observed in typi-
young children with autism. They reported that PECS, cal children developing first words, an integrated use
which employed symbols but no SGD, had a more rapid of comprehension and production may permit children
effect on spoken word production than did RPMT. Sim- with developmental delays who are not speaking to ob-
ilar to our findings, however, the increases in spoken serve and to actively engage in the communicative pro-
word production were also very modest. In fact, these cess while also taking on the role of speaker. Again, like
findings suggest that AAC intervention actually facil- very young typical children, children with developmen-
itates speech better than spoken language alone, albeit tal delays who use limited to no speech should have the
very slowly. It is not known how the development of speech opportunity to be exposed to a mode for language pro-
would continue if the duration of the three interventions duction as they are also beginning to use it. Given typical
were increased. development and our own research findings about achieve-
When augmented and spoken target words were ment patterns with school-age children (Romski & Sevcik,
combined to create vocabularies, it became even more 1996), it may be important to alter the focus of augmented
evident that the children who received AC-O and AC-I language interventions with very young children to com-
interventions had acquired more expressive language bine input and output with mode. As Reichle et al. (1998)
than the children who received the contrast SC inter- and Romski and Sevcik (1996) have speculated, children
vention. This result highlights the important role AAC may be using spoken input to attach to symbol output.
interventions can play at a very early age for children Parents of the children in the spoken and augmented
who are having difficulty with speech and language de- output interventions provided more spoken input to
velopment. Target vocabulary size also increased for the their children than did parents in augmented input inter-
children in the AC-O and AC-I interventions because vention. The children in the augmented output interven-
they were learning an expressive vocabulary (via an- tion, however, were able to take on the role of speaker,
other mode) and their target vocabulary needed to be which may then essentially change the dynamic between
expanded to meet their developing communication the parent and the child.
skills.
Although on average the intervention took a month
longer to complete than we originally estimated, this ex-
Limitations of This Study
tended time frame did not (positively or negatively) There are a number of study limitations that must
affect the children’s performance. It is striking that in a be considered. First, in this study, we employed a con-
relatively short period of time the children in the AC-I trast group design with random assignment to group
and AC-O interventions made the communication where we compared the two augmented interventions
progress they did. Both augmented language interven- with a standard spoken language intervention. We did
tions provided a way for the children to communicate not employ a nontreated control group, so we cannot rule
with their parents via augmented and spoken words out the role of maturation in these findings. We are
after only 18 sessions of intervention and to maintain currently conducting a study that compares AC-O with a
them at home at 24 sessions. Having a way to com- hybrid intervention ACIO (a combination of AC-O and
municate increased child utterance intelligibility and AC-I) and uses a wait-list control design to determine the
type/token ratio and facilitated parent–child interac- effects of time on communication outcomes. On a related
tion. Although parent MLUm remained stable across the point, 6 children we originally intended to treat withdrew
intervention, turns changed. Parents in all three inter- from the study before outcome measures were collected on
ventions took shorter but more turns. them, and they were not included in the analyses. These
Children who were given augmented input in the children were equally distributed across the three inter-
form of augmented and spoken words did develop aug- ventions, and their preintervention profiles were com-
mented and spoken word skills, but children who were parable to the children who completed the study. It is
asked to produce augmented words spoke more. This com- thus unlikely that they would have altered the results
parison suggests that the relationship between compre- substantially.
hension and production of augmented and spoken words is Second, this sample was a self-selected group of
more complex than we initially hypothesized. Blockberger parents and children from the broad area of metropol-
and Sutton (2003) discussed the importance of this rela- itan Atlanta. Although parents who chose to participate
tionship but suggested that very little is known about in this study were from diverse backgrounds, they were
how the two interface to facilitate language development generally highly educated and on average in their 30s.
through augmented means. Our present work begins to They were able and willing to commit approximately
suggest the complexity of the relationship in terms of 2 hr of time every week for at least 12 weeks to participate

Romski et al.: Early Language Interventions 361

in the interventions. We may not be able to generalize Clinical Implications
these findings to a sample of less educated and/or much
younger parents who could not make a commitment to Clinically, these findings support employing an aug-
this amount of intervention time. Kaiser and Hancock mented language approach that includes parent coach-
(2003) have argued that parent-coached interventions ing as part of early intervention before 3 years of age.
must be initiated when parents are interested in par- Not only do interventions like AC-O and AC-I provide a
ticipating, consider being involved a high priority, have way for a child to almost immediately communicate,
sufficient time and energy to participate, are willing to they do not hinder the development of spoken words. Un-
make a relatively long-term commitment, and are sup- like some other AAC approaches, using an SGD with
ported by their families. It may be that the parents in our very young children also can provide a way for a child to
study met these criteria and that this is one of the rea- be independently intelligible to his or her communica-
sons they were successful. Variations in this type of in- tion partners while he or she is developing spoken words.
tervention protocol that may support a broader range of
parents are important to consider. In addition, varia- Conclusions
tions in the intervention protocol also may be needed to
generalize this type of intervention to preschool class- In conclusion, young children with significant de-
rooms and specific groups of children (e.g., children at velopmental delays learned to communicate better with
risk for autism spectrum disorder; Romski, Sevcik, Smith, their parents after each of the three parent-implemen-
et al., 2009). ted interventions than they did at the beginning of the
intervention. The children in the AC-O and AC-I inter-
Third, all three interventions used a naturalistic ventions developed a word vocabulary even though their
format and not a trial-based format. While the natural- spoken word skills were modest at best. Children in the
istic format permitted the interaction to flow and the SC intervention learned to use about the same percent-
adult to follow the child’s lead, it did not provide a age of spoken words as the children in the AC-I and AC-O
specified number of trials or opportunities for the child interventions, but they did not have another conventional
to communicate during the three 10-min routines. Al- way to communicate available to them. The parent-
though the available data do not rule out that the aug- coached augmented language interventions did not hinder
mented interventions provided more opportunities to spoken language development and provided a way for
communicate than the spoken intervention, the spoken the child to develop communication skills in the absence
and augmented input the children received was fairly of substantial gains in spoken words.
comparable as evidenced by the lack of a significant
group main effect for target vocabulary input.
Acknowledgments
Future Research Directions The research was funded by National Institutes of Health
Grant DC-03799. Special thanks go to all the families who
Our findings about language and communication participated in this study. Earlier versions of this article were
are promising and strongly suggest that additional re- presented at the Gatlinburg Conference on Research and
search is needed to examine the many factors that may Theory in MRDD, San Diego, CA, March 2006, and the biennial
have influenced these outcomes. Currently, we are ex- meeting of the International Society for Augmentative and
amining changes in shared engagement in communi- Alternative Communication in Dusseldorf, Germany, August
cation development for this same sample of children 2006. We thank Tanya Kobek, Rebekah Walker, Ramona
Blackman Jones, Sara Dowless, Mia Ligon, and Laura Carey
(Adamson, Romski, Bakeman, & Sevcik, 2009). Because
for their assistance in implementing the protocols, creating
this intervention was parent coached, we also are ex-
the transcripts, and collecting the data.
amining the effects of the interventions on parent stress
(Smith, Romski, Sevcik, Adamson, & Bakeman, 2009)
as well as parent perception of child communication de- References
velopment (Romski, Sevcik, Adamson, et al., 2009). We
have followed these children into preschool (Romski & Adamson, L. B., Romski, M. A., Bakeman, R. B., & Sevcik,
Sevcik, 2008) and more recently elementary school and R. A. (2009). The effects of augmented language intervention
on the development of joint engagement. Manuscript sub-
will be able to illustrate growth trajectories of their lan- mitted for publication.
guage development. In addition, it may be important to
American Speech-Language-Hearing Association.
examine whether a longer intervention protocol would (2002). Augmentative and alternative communication:
result in stronger spoken word effects. Finally, the rela- Knowledge and skills for service delivery. The ASHA Leader,
tionship between mode and comprehension and production 7(Suppl. 22), 97–106.
is an important one to pursue (Barker, Sevcik, & Romski, American Speech-Language-Hearing Association.
2007). (2004). Guidelines for the training, use, and supervision of

362 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
 speech-language pathology assistants [Guidelines]. Avail- Fleiss, J. L. (1981). Statistical methods for rates and propor-
able from www.asha.org/policy. tions. New York: Wiley.
American Speech-Language-Hearing Association. (2008). Goossens’, C. (1989). Aided communication intervention
Roles and responsibilities of speech-language pathologists before assessment: A case study of a child with cerebral palsy.
in early intervention: Technical report [Technical report]. Augmentative and Alternative Communication, 5, 14–26.
Available from www.asha.org/policy. Guralnick, M. (2001). Early childhood inclusion: Focus on
Barker, R. M., Sevcik, R. A., & Romski, M. A. (2007, change. Baltimore: Brookes.
November). AAC language intervention and receptive lan- Harris, M., & Reichle, J. (2004). The impact of aided language
guage ability. Poster session presented at the Annual stimulation on symbol comprehension and production in
Convention of the American Speech-Language-Hearing children with moderate cognitive disabilities. American
Association, Boston. Journal of Speech-Language Pathology, 13, 155–167.
Bates, E. (1993). Comprehension and production in early Hemmeter, M. L., & Kaiser, A. P. (1994). Enhanced milieu
language development: Comments on Savage-Rumbaugh teaching: Effects of parent-implemented language inter-
et al. Monographs of the Society for Research in Child vention. Journal of Early Intervention, 18, 269–289.
Development, 58(3–4, Serial No. 233), 222–242.
Hendrick, D., Prather, E., & Tobin, A. (1984). Sequenced
Beukelman, D., & Mirenda, P. (2005). Augmentative and Inventory of Communication Development. Los Angeles:
alternative communication: Management of severe commu- Western Psychological Services.
nication impairments (3rd ed.). Baltimore: Brookes.
Hustad, K., & Shapley, K. (2003). AAC and natural speech
Binger, C., Kent-Walsh, J., Berens, J., DelCampo, S., & in individuals with developmental disabilities. In J. Light,
Rivera, D. (2008). Teaching Latino parents to support the D. Beukelman, & J. Reichle ( Eds.), Communicative
multi-symbol message productions of their children who competence for individuals who use AAC ( pp. 41–62).
require AAC. Augmentative and Alternative Communica- Baltimore: Brookes.
tion, 24, 323–338.
Kaiser, A., & Hancock, T. (2003). Teaching parents new
Binger, C., & Light, J. (2007). The effect of aided AAC skills to support their young children’s development. Infants
modeling on the expression of multi-symbol messages by and Young Children, 16, 9–21.
preschoolers who use AAC. Augmentative and Alternative
Communication, 23, 30–43. Kaiser, A. P., Hemmeter, M. L., Ostrosky, M., Fischer,
R., Yoder, P., & Keefer, M. (1996). The effects of teaching
Blockberger, S., & Sutton, A. (2003). Toward linguistic com- parents to use responsive interaction strategies. Topics in
petence: Language experiences and knowledge of children Early Childhood Special Education, 16, 375–406.
with extremely limited speech. In J. Light, D. Beukelman, &
J. Reichle ( Eds.), Communicative competence for individuals Kent-Walsh, J., & McNaughton, D. (2005). Communication
who use AAC ( pp. 63–106). Baltimore: Brookes. partner instruction in AAC: Present practices and future
directions. Augmentative and Alternative Communication,
Bondy, A. S., & Frost, L. A. (1994). PECS: The Picture Ex- 21, 195–204.
change Communication System training manual. Cherry
Hill, NJ: Pyramid Educational Consultants. Mayer-Johnson. (1981). Picture Communication Symbols.
Stillwater, MN: Author.
Brady, N. (2000). Improved comprehension of object names
following voice output communication aid use: Two case McCollum, J., & Yates, T. (1994). Dyad as focus, triad as
studies. Augmentative and Alternative Communication, 16, means: A family-centered approach to supporting parent–
197–204. child interactions. Infants and Young Children, 6, 54–63.

Campbell, P., Milbourne, S., & Wilcox, J. (2008). Adapta- Millar, D., Light, J., & Schlosser, R. (2006). The impact of
tions interventions to promote participation in natural augmentative and alternative communication intervention
settings. Infants and Young Children, 21, 94–106. on the speech production of individuals with developmental
disabilities: A research review. Journal of Speech, Lan-
Cress, C., & Marvin, C. (2003). Common questions about guage, and Hearing Research, 49, 248–264.
AAC services in early intervention. Augmentative and
Alternative Communication, 19, 254–272. Miller, J., & Chapman, R. (1985). Systematic Analysis of
Language Transcripts [Computer software]. Madison:
Drager, K. D. R., Postal, V. J., Carrolus, L., Catellano, M., University of Wisconsin.
Gagliano, C., & Glynn, J. (2006). The effect of aided lan-
guage modeling on symbol comprehension and production Mullen, E. (1995). Mullen Scales of Early Learning. Circle
in 2 preschool children with autism. American Journal of Pines, MN: AGS.
Speech-Language Pathology, 15, 112–125. Nigam, R., Schlosser, R., & Lloyd, L. (2006). Concomitant
Fenson, L., Dale, P., Reznick, S., Thal, D., Bates, E., use of the matrix strategy and the mand-model procedure in
Hartung, J., et al. (1993). MacArthur Communicative De- teaching graphic symbol combinations. Augmentative and
velopment Inventories: User’s guide and technical manual. Alternative Communication, 22, 160–177.
San Diego, CA: Singular. Paul, R., & Miller, J. (1995). The Clinical Assessment of
Fey, M., Warren, S., Brady, N., Finestack, L., Bredin-Oja, Language Comprehension. Baltimore: Brookes.
S., Fairchild, M., et al. (2006). Early effects of responsivity Reichle, J., Halle, J., & Drasgow, E. (1998). Implementing
education/prelinguistic milieu teaching for children with augmentative communication systems. In A. Wetherby,
developmental delays and their parents. Journal of Speech, S. Warren, & J. Reichle ( Eds.), Transitions in prelinguistic
Language, and Hearing Research, 49, 526–547. communication ( pp. 417–436). Baltimore: Brookes.

Romski et al.: Early Language Interventions 363

Reichle, J., Light, J., & Beukelman, D. (2002). Exemplary Sevcik, R. A., & Romski, M. A. (2007). Children, families,
practices for beginning communicators: Implications for clinicians, and AAC. Perspectives on Augmentative and
AAC. Baltimore: Brookes. Alternative Communication, 5, 7–9.
Romski, M. A., & Cheslock, M. (2000). Parent-coached Sevcik, R. A., Romski, M. A., & Adamson, L. B. (2004).
language intervention protocol manual. Unpublished Augmentative communication and preschool children: Case
manuscript, Georgia State University, Atlanta. example and research directions. Disability and Rehabili-
Romski, M. A., & Cheslock, M. (2004). Parent-coached tation, 26, 1323–1329.
language intervention transcription and coding manual. Sevcik, R. A., Romski, M. A., Watkins, R., & Deffebach, K.
Unpublished manuscript, Georgia State University, (1995). Adult partner-augmented communication input to
Atlanta. youth with mental retardation using the system for augment-
Romski, M. A., & Sevcik, R. A. (1996). Breaking the speech ing language (SAL). Journal of Speech and Hearing Research,
barrier: Language development through augmented means. 38, 902–912.
Baltimore: Brookes. Sigafoos, J., Drasgow, E., & Schlosser, R. (2003). Strate-
Romski, M. A., & Sevcik, R. A. (2003). Augmented input: gies for beginning communicators. In R. Schlosser ( Ed.), The
Enhancing communication development. In J. Light, efficacy of augmentative and alternative communication:
D. Beukelman, & J. Reichle ( Eds.), Communicative compe- Toward evidence-based practice ( pp. 323–346). New York:
tence for individuals who use AAC ( pp. 147–162). Baltimore: Academic Press.
Brookes. Smith, A., Romski, M. A., Sevcik, R. A., Adamson, L. B., &
Romski, M. A., & Sevcik, R. A. (2005). Early intervention Bakeman, R. (2009). Examining parental stress and paren-
and augmentative communication: Myths and realities. tal perceptions of children’s communication development
Infants and Young Children, 18, 174–185. following language intervention. Manuscript in preparation.

Romski, M. A., & Sevcik, R. A. (2008, August). Communi- Warren, S., Fey, M., Finestack, L., Brady, N., Bredin-Oja,
cation growth patterns of young children with developmental S., & Fleming, K. (2008). A randomized trial of longitudinal
disabilities who received early augmented language inter- effects of low-intensity responsivity education/prelinguistic
ventions. Paper presented at the biennial meeting of the milieu teaching. Journal of Speech, Language, and Hearing
International Society for Augmentative and Alternative Research, 51, 451–470.
Communication, Montreal, Quebec, Canada. Wilcox, J., & Shannon, M. (1998). Facilitating the transi-
Romski, M. A., Sevcik, R. A., Adamson, L. B., Cheslock, tion from prelinguistic to linguistic communication. In
M., & Smith, A. (2007). Parents can implement AAC A. Wetherby, S. Warren, & J. Reichle ( Eds.), Transitions
interventions: Ratings of treatment implementation across in prelinguistic communication ( pp. 385–416). Baltimore:
early language interventions. Early Childhood Services, 1, Brookes.
100–115. Yoder, P., & Stone, W. (2006). A randomized comparison of
Romski, M. A., Sevcik, R. A., Adamson, L. B., Cheslock, the effect of two prelinguistic communication interventions
M., Smith, A., & Bakeman, R. (2009). Measuring parent on the acquisition of spoken communication in preschoolers
perception of early communication development and inter- with ASD. Journal of Speech, Language, and Hearing
vention. Manuscript submitted for publication. Research, 49, 698–711.

Romski, M. A., Sevcik, R. A., Smith, A., Barker, R. M., Yoder, P., & Warren, S. (2002). Effects of prelinguistic milieu
Folan, S., & Barton-Hulsey, A. (2009). The system for teaching and parent responsivity education on dyads involv-
augmenting language: Implications for children with autism ing children with intellectual disabilities. Journal of Speech,
spectrum disorder. In P. Mirenda, T. Iacono, & J. Light Language, and Hearing Research, 45, 1158–1174.
( Eds.), AAC and autism ( pp. 219–245). Baltimore: Brookes.
Romski, M. A., Sevcik, R. A., Hyatt, A. M., & Cheslock, M. Received August 10, 2008
(2002). A continuum of AAC language intervention strate-
Accepted June 1, 2009
gies for beginning communicators. In J. Reichle, D. R.
Beukelman, & J. C. Light ( Eds.), Exemplary practices for DOI: 10.1044/1092-4388(2009/08-0156)
beginning communicators: Implications for AAC ( pp. 1–23). Contact author: MaryAnn Romski, Department of
Baltimore: Brookes. Communication, Georgia State University, Atlanta,
Sevcik, R. A., & Romski, M. A. (2002). The role of language GA 30302-4000. E-mail: mromski@gsu.edu.
comprehension in establishing early augmented conversa-
tions. In J. Reichle, D. Beukelman, & J. Light ( Eds.),
Implementing an augmentative communication system:
Exemplary strategies for beginning communicators
( pp. 453–474). Baltimore: Brookes.

364 Journal of Speech, Language, and Hearing Research • Vol. 53 • 350–364 • April 2010
 Randomized Comparison of Augmented and Nonaugmented Language
Interventions for Toddlers With Developmental Delays and Their Parents

MaryAnn Romski, Rose A. Sevcik, Lauren B. Adamson, Melissa Cheslock, Ashlyn

Smith, R. Michael Barker, and Roger Bakeman
J Speech Lang Hear Res 2010;53;350-364
DOI: 10.1044/1092-4388(2009/08-0156)

The references for this article include 9 HighWire-hosted articles which you can
access for free at: http://jslhr.asha.org/cgi/content/full/53/2/350#BIBL

This article has been cited by 2 HighWire-hosted article(s) which you can access for
free at:
http://jslhr.asha.org/cgi/content/full/53/2/350#otherarticles

This information is current as of April 13, 2011

This article, along with updated information and services, is

located on the World Wide Web at:
http://jslhr.asha.org/cgi/content/full/53/2/350