Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Research

Effectiveness of the Picture Exchange


Communication System (PECS) on
Communication and Speech for Children With
Autism Spectrum Disorders: A Meta-Analysis

Michelle Flippin
Stephanie Reszka
Linda R. Watson
University of North Carolina at Chapel Hill

Purpose: The Picture Exchange Communication intervention for facilitating communication in


System (PECS) is a popular communication- children with ASD ages 1–11 years. Small to
training program for young children with autism moderate gains in communication were demon-
spectrum disorders (ASD). This meta-analysis strated following training. Gains in speech were
reviews the current empirical evidence for PECS small to negative.
in affecting communication and speech out- Conclusions: This meta-analysis synthesizes
comes for children with ASD. gains in communication and relative lack of gains
Method: A systematic review of the literature on made in speech across the PECS literature for
PECS written between 1994 and June 2009 was children with ASD. Concerns about maintenance
conducted. Quality of scientific rigor was assessed and generalization are identified. Emerging
and used as an inclusion criterion in computation of evidence of potential preintervention child char-
effect sizes. Effect sizes were aggregated sepa- acteristics is discussed. Phase IV was identified
rately for single-subject and group studies for as a possibly influential program characteristic
communication and speech outcomes. for speech outcomes.
Results: Eight single-subject experiments
(18 participants) and 3 group studies (95 PECS
participants, 65 in other intervention/control) were Key Words: autism, Picture Exchange
included. Results indicated that PECS is a prom- Communication System, communication
ising but not yet established evidence-based intervention, speech

D
elayed development of language is an early and per- in treatment outcomes is likely to be a function of individual,
sistent marker of autism (Lord, Risi, & Pickles, 2004). pretreatment characteristics (Lord et al., 2005). To date, there
Although estimates vary, approximately 25% of chil- are no clear guidelines for clinicians in determining which
dren with autism do not develop functional speech (Volkmar, communication strategy is likely to be effective for an indi-
Lord, Bailey, Schultz, & Klin, 2004). Thus, the need for early vidual child with ASD. The purpose of this meta-analysis is
and effective communication training for children with autism to review the current empirical evidence for one communi-
is clear. However, given the myriad communication-training cation strategy, the Picture Exchange Communication System
strategies available, it is difficult for clinicians to choose the (PECS). PECS is a manualized program for teaching children
most effective intervention for an individual child with autism to use an exchange-based communication system that has
spectrum disorders (ASD). Variability in language outcomes been a common intervention choice for nonverbal children
for children with ASD makes it difficult to predict which with ASD in clinical and school settings. In this meta-analysis,
children will go on to develop speech and which children will the quality of the current research evidence for the PECS
require augmentative and alternative communication. In ad- method is assessed. Additionally, pretreatment characteris-
dition, differential response to intervention for individual tics of children who responded to PECS training are identified
children with ASD makes it difficult for clinicians to identify across studies to generate a profile of a young child with
the best method of communication training for any one partic- ASD for whom PECS may be an appropriate and effective
ular child. Given the heterogeneous nature of ASD, variability communication intervention, and program characteristics
178 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010 • A American Speech-Language-Hearing Association
that may support the development of speech in children TABLE 1. Six phases of Picture Exchange Communication
through the PECS program are discussed. System (PECS) training.

Phase Description
Brief Overview of PECS
I Training begins on a single picture of highly desired item.
PECS is a behaviorally based pictorial communication Student picks up picture of desired item and releases into
system designed for children with social-communicative def- communicative partner’s hand in exchange for desired
icits. Using PECS, expressive communication skills are tar- item. Communicative partner gives the item to the child
while naming the item (e.g., “car ”). Two adults (i.e., the
geted through the training of requests and, later, comments. communicative partner and a physical prompter) are used
As Bondy and Frost (1994, p. 3) describe, “Children using during this phase.
PECS are taught to approach and give a picture of a desired II A communication book is introduced, and increased distance
item to a communicative partner in exchange for that item. By is placed between the child and communicative partner.
Child is required to get picture from her communication
doing so, the child initiates a communicative act for a con- book and travel to communicative partner to request item.
crete item within a social context.” It is important to note that To increase spontaneity and persistence, placement of
the term PECS does not refer generally to all exchange-based picture symbol is varied in the book. Also, generalization
pictorial communication interventions (i.e., exchanging a across a variety of trainers, contexts, and reinforcers is
photograph or line drawing for a corresponding real item); introduced at this phase.
III Child is required to discriminate between two picture
rather, PECS is a specific, manualized intervention protocol. symbols (highly desired vs. nondesired item to gradually
In the PECS program, a child’s expressive communication multiple desired items). Correspondence checks are
abilities are shaped via the use of reinforcement, delay, and done to ensure that child is truly requesting preferred
generalization across trainers and settings. PECS training item.
consists of six phases (see Table 1). In Phase I: The Physical IV Child uses a sentence starter (“I want”) to make a request by
building and exchanging a 2-picture-sequence sentence
Exchange, two trainers physically prompt the child to exchange strip with “I want” symbol plus picture symbol for preferred
a single picture for a preferred item, without distractor pic- item. Communicative partner provides verbal model “I
tures. In Phase II: Expanding Spontaneity, a communication wantI” and pauses before labeling the requested item
book is introduced, and increased distance is placed between and handing sentence strip and requested item back to
child. Communicative partner differentially reinforces
the child and communicative partner. The child is required any vocal attempt.
to get a picture symbol from his or her communication book V Communicative partner introduces the verbal prompt “What
and travel to the communicative partner to request an item. do you want?” Over time, a delay is inserted between
Placement of picture symbols is varied in the book, and gen- the verbal prompt and an additional gesture prompt
eralization is targeted across a variety of trainers and contexts. toward the “I want” picture symbol. The child begins
answering the question before her communicative
In Phase III: Picture Discrimination, the child discriminates partner uses the gestural prompt.
between two picture symbols (first between a highly desired VI Comments are trained as child exchanges sentence strips
and a nondesired item and then between two desired items). In to respond to partner’s communicative questions (i.e.,
Phase IV: Sentence Structure, the child makes a request by “What do you see?” “What do you want?” “What do
you have?”).
building and exchanging a two-picture-sequence sentence
strip with an “I want” symbol plus the picture symbol for the Note. From PECS Training Manual (Frost & Bondy, 2002). Adapted
preferred item. In Phase IV, after the child has requested by with permission.
giving the sentence strip, the communication partner provides
the verbal model “I wantI” and uses a time delay before
labeling the requested item and handing the sentence strip and request preferred items immediately upon starting the first
requested item back to the child. In Phase V: Responding to phase of PECS. Also, because PECS is a behavioral approach
“What Do You Want?” the communicative partner introduces that shapes communication via the use of reinforcers, it is
the verbal prompt “What do you want?” As Phase V interven- motivating for children. Finally, there is little initial training
tion continues, a time delay is inserted between the verbal required for trainers to implement PECS (i.e., usually a 2-day
prompt and an additional gestural prompt toward the “I want” workshop).
picture symbol. Eventually, the child begins answering the Limitations of PECS include the restricted range of com-
question before his or her communicative partner uses the municative functions targeted in the approach. The communi-
gestural prompt. Finally, in Phase VI: Responsive and Spon- cative repertoire trained in PECS consists primarily of requests
taneous Commenting, comments are trained via the exchange (Phases I–V). A response form that the PECS developers
of a sentence strip in response to the communicative partner’s label “comments” is trained in the final phase (Phase VI).
questions (i.e., “What do you see?” and “What do you have?” However, comments as defined in the program are not true,
contrasted with the request cue “What do you want?”). self-initiated comments. Rather, the comments are responses
Ease of implementation for both children and interven- to adult prompts, learned via shaping and rewards. PECS
tionists is one of the features that helped PECS to become a does not include steps in the program to specifically guide the
widely popular social-communication-training system for chil- child to initiate with picture communication symbols to share
dren with ASD. For example, children do not need to master interests with others (i.e., the type of communicative acts
prerequisite skills (e.g., eye contact, gestures, and verbal imi- generally considered as “commenting”). In addition, other
tation) prior to beginning PECS training. Rather, children with communicative functions that develop early among typically
relatively limited skills begin exchanging picture symbols to developing children, such as protests and refusals, are not

Flippin et al.: Effectiveness of PECS 179


directly targeted in PECS. For example, although PECS trains syntheses of the evidence for PECS, they do not provide as-
discrimination between preferred and nonpreferred items, the sessments of the quality of the studies included, or quantita-
standard PECS protocol does not provide a built-in method for tive analyses of results.
directly indicating that an item is not wanted. Finally, as PECS is In contrast, Schlosser and Wendt (2008) employed both
an aided communication technique, preparation of picture quality assessment and quantitative analyses in their system-
symbols can be labor intensive for clinicians, and the need to atic review of the effects of augmentative and alternative
carry a communication book with limited vocabulary choices communication intervention on communication and speech
can be restrictive for children using the PECS approach. production for children with autism. Although the independent
In the 14 years since its inception, PECS has been widely variables examined in that review were not limited to PECS,
adopted in special education settings for children with ASD. articles that met criteria for review included seven studies (six
Descriptive reports have suggested that the PECS method single-subject experiments and one group design) in which
leads to increased functional communication in a relatively PECS was the independent variable. The authors proposed that
short time period and, sometimes, speech development in PECS was highly effective for teaching requesting to children
children with ASD. For example, in a retrospective report on with autism. In contrast, the authors reported mixed results
longitudinal data for children with autism who were 5 years of for speech outcomes, with two studies yielding percentage of
age and younger, Bondy and Frost (1994) documented that nonoverlapping data (PND) considered highly effective for
more than half of the 66 participating children were able to promoting speech, one study ineffective, and one study mixed
use speech without pictures to communicate within a year of (highly effective for one participant and ineffective for another);
starting PECS. Similar gains in the use of PECS symbols to however, effect sizes (ESs) were not aggregated. Further-
communicate and in the development of spontaneous speech more, since the publication of that review, several additional
have been documented across other case studies and quasi- group and single-subject experiments have been published
experimental reports (e.g., Magiati & Howlin, 2003; Schwartz, that add to the evidence base for the PECS approach. Most
Garfinkle, & Bauer, 1998; Webb, 2000). Further, these in- recently, Preston and Carter (2009), in their examination of
creases in communication and speech were documented to the efficacy of PECS in 10 single-subject and three group de-
occur over a relatively short time period, within 6 to 14 months signs, reached similar conclusions to Schlosser and Wendt
of starting PECS (Schwartz et al., 1998; Webb, 2000). Al- (2008) that PECS is effective in providing a means of commu-
though these reports offer anecdotal support for the effec- nication for individuals with little or no speech but that the
tiveness of PECS, the quality of evidence is limited due data are not as clear for speech outcomes following PECS
to lack of experimental control. Empirical studies on the training. Assessment of quality and calculation of ESs were
acquisition and impact of PECS strengthen the quality of completed for individual studies in that review; however, par-
evidence related to the effectiveness of PECS and have ticipant criteria were not limited to children with a diagnosis
added to the evidence base for this approach. of autism, and independent variables were not limited to
standard PECS protocol. Rather, the review included studies
Previous Reviews of PECS with participants ranging in age from early childhood to adult-
hood, studies of participants with comorbid diagnoses (e.g.,
Several recent reviews have employed varying scopes and autism and attention deficit/hyperactivity disorder [ADHD],
methodologies to examine the efficacy of PECS for commu- blindness, and deafness), and studies that involved modifi-
nication and/or speech outcomes for children and adults with cations of the PECS approach. Thus, despite the contributions
various disabilities. For instance, Lancioni and colleagues of these systematic reviews of literature, to date no meta-
(2007) reviewed the literature on the effectiveness of pictorial analysis has examined the effectiveness of the standard PECS
exchange communication interventions (including PECS) protocol on communication and speech outcome variables
and voice output communication aids in promoting requesting for children with ASD.
among children with developmental disabilities. In that review,
the authors concluded that 170 out of 173 children who par-
ticipated in studies of PECS or other pictorial exchange com-
munication interventions showed improvement in their ability Research Questions and Method
to make requests. The independent variables in that review, The evidence base for PECS includes both single-subject
however, were not limited to PECS, and participant inclusion designs and group research. To thoroughly evaluate the evi-
criteria were not specific to autism. dence for this intervention, both types of studies must be
In a more narrowly focused synthesis of 13 studies of PECS, considered in the meta-analysis. Calculation and aggregation
Tien (2008) reached similar conclusions that PECS appears to of ESs for meta-analysis is well established for group designs
be an effective intervention for improving functional commu- and is becoming increasingly accepted practice in single-
nication skills for participants with ASD. In a broad review of subject experiments (Schlosser & Sigafoos, 2008). Although
all of the empirical articles published on the PECS approach, there is not a standard metric, the calculation of ESs in single-
Sulzer-Azaroff, Hoffman, Horton, Bondy, and Frost (2009) subject designs allows for the comparison of treatment out-
synthesized data from 34 peer-reviewed studies to conclude comes within and between individuals as well as comparison
that PECS is an effective means of training functional com- of the relative strength of treatments (Beeson & Robey, 2006).
munication for individuals with impaired or no speech. That In the autism literature, specifically, meta-analysis of single-
review was not limited to specific participant ages, diagnoses, subject studies has contributed to the evidence base for several
or outcome variables. While these reviews provide important popular interventions, including social skills training (Bellini,

180 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


Peters, Benner, & Hopf, 2007; Schneider, Goldstein, & 2. What are the effects of PECS on speech outcomes for
Parker, 2008), video modeling (Bellini & Akullian, 2007), children with ASD?
and self-management (Suk-Hyang, Simpson, & Shogren,
2007). Although there is not a way to aggregate ESs across
single-subject and group designs, including single-subject Search Strategy
research in a meta-analysis allows for a more comprehensive
Studies were identified for inclusion in this meta-analysis
understanding of the effect of an intervention (Kavale, Mathur,
through a three-step process. First, searches were conducted
Forness, Quinn, & Rutherford, 2000). In fact, several studies
to identify conceptual and empirical articles related to PECS
have employed meta-analytic techniques for both single-
published between 1994 and June 2009 that included at least
subject and group designs to determine the effectiveness of
one child participant (18 years or younger) with ASD, in
interventions for children with externalizing behaviors and
the following databases: PsycInfo, ERIC, Academic Search
disruptive behavior disorders (Maughan, Christiansen, Jenson,
Premier, and Cumulative Index to Nursing and Allied Health
Olympia, & Clark, 2005), ADHD (Fabiano et al., 2009), and
Literature (CINAHL). Table 2 describes the search terms and
autism (Schlosser & Wendt, 2008). Given the volume of
yields. Search terms used were Picture Exchange Communi-
single-subject publications, the current meta-analysis includes
both single-subject and group studies of the PECS approach. cation System, autism, PECS, and Bondy and Frost. Reverse
searches were then conducted to identify articles that cited the
This meta-analysis provides an evaluation of the quality of
seminal article on PECS (Bondy & Frost, 1994). Second, re-
empirical evidence, using appropriate quality indicators for
single-subject or group designs. In addition, ESs are aggre- sults were cross-referenced with articles listed on the research
page of the official PECS Web site (www.pecs.com/research.
gated within each type of design to examine the effectiveness
htm). Finally, reference lists were examined from three rele-
of PECS for both speech and communicative outcomes spe-
vant literature reviews (Preston & Carter, 2009; Schlosser &
cifically for young children with ASD. This study addressed
Wendt, 2008; Sulzer-Azaroff et al., 2009) to identify additional
two research questions:
studies that were not captured through previous searches. The
1. What are the effects of PECS on communication outcomes systematic search returned 54 articles. Of the 54 articles iden-
for children with ASD? tified, only empirical studies (single-subject experiments or

TABLE 2. Search strategies across general-purpose databases.

Database Search strategy Yield (references) Revised search

PsycInfo (DE “Picture Exchange Communication System”) 48


(DE “Picture Exchange Communication System” AND DE “autism”) 45 added AND “autism”
(DE “Picture Exchange Communication System” OR DE “PECS”) 154 added OR “PECS”
(DE “Picture Exchange Communication System” OR DE “PECS” 51 added AND “autism”
AND DE “autism”)
(DE “PECS”) 150
(DE “PECS” AND DE “autism”) 45 added AND “autism”
(Reverse search for articles citing Bondy & Frost, 1994) 98
Academic Search (DE “Picture Exchange Communication System”) 36
Premier (DE “Picture Exchange Communication System” AND DE “autism”) 24
(DE “Picture Exchange Communication System” OR DE “PECS”) 3,295 added OR “PECS”
(DE “Picture Exchange Communication System” OR DE “PECS” 71 added AND “autism”
AND DE “autism”)
(DE “PECS”) 3,141 added AND “autism”
(DE “PECS” AND DE “autism”) 28 added AND “autism”
(Reverse search for articles citing Bondy & Frost, 1994) 32
ERIC (DE “Picture Exchange Communication System”) 33
(DE “Picture Exchange Communication System” AND DE “autism”) 24 added AND “autism”
(DE “Picture Exchange Communication System” OR DE “PECS”) 82 added OR “PECS”
(DE “Picture Exchange Communication System” OR DE “PECS” 81 added AND “autism”
AND DE “autism”)
(DE “PECS”) 76
(DE “PECS” AND DE “autism”) 23 added AND “autism”
CINAHL (DE “Picture Exchange Communication System”) 52
(DE “Picture Exchange Communication System” AND DE “autism”) 48 added AND “autism”
(DE “Picture Exchange Communication System” OR DE “PECS”) 192 added OR “PECS”
(DE “Picture Exchange Communication System” OR DE “PECS” 56 added AND “autism”
AND DE “autism”)
(DE “PECS”) 188
(DE “PECS” AND DE “autism”) 49 added AND “autism”
(Reverse search for articles citing Bondy & Frost, 1994) 13

Note. DE = descriptor; ERIC= Education Resources Information Center; CINAHL= Cumulative Index to Nursing and Allied Health Literature.

Flippin et al.: Effectiveness of PECS 181


group design) published in English in a peer-reviewed journal were extracted if reported; if not, reported statistics from which
were considered for inclusion in this synthesis. The remaining ESs could be calculated were extracted. For single-subject
30 studies were reviewed for inclusion criteria by at least two studies, data points on communication and speech outcome
of the authors. Interrater reliability for inclusion of studies variables were extracted from plotted graphs to calculate ESs.
was 93.3%. In the rare instances of disagreement, the decision to In the study by Angermeier and colleagues (2008), in which
include or exclude an article was reached by consensus. Studies baseline data were reported separately for each of the
included in this meta-analysis met the following criteria: trained PECS phases, ESs calculated for communicative
Participants. This meta-analysis included children younger outcomes were averaged across the trained phases of PECS
than 18 with a diagnosis of autism, ASD, or pervasive devel- for each participant. For the study by Charlop-Christy,
opmental disorder-not otherwise specified (PDD-NOS). Carpenter, Le, LeBlanc, and Kellet (2002), which employed
Five studies on the use of PECS with adults (Chambers & a multiple-baseline-across-settings design, ES was averaged
Rehfeldt, 2003; Rehfeldt & Root, 2005; Rosales & Rehfeldt, across the two contexts (academic and play sessions) in
2007; Stoner et al., 2006; Ziomek & Rehfeldt, 2008) were which speech outcomes were measured. Finally, to satisfy
not included. In addition, five studies of participants diag- the independence assumption of meta-analysis, for four
nosed with neurological or developmental disorders other studies in which multiple outcome variables were reported
than autism (Bock, Stoner, Beck, Hanley, & Prochnow, 2005; for speech (Charlop-Christy et al., 2002; Ganz, Simpson, &
Kern, Gallagher, Staosta, Hickman, & George, 2006) or with Corbin-Newsome, 2008; Yoder & Stone, 2006a, 2006b)
both autism and a comorbid physical or neurological diag- and communication (Howlin, Gordon, Pasco, Wade, &
nosis such as ADHD (Adkins & Axelrod, 2002), blindness Charman, 2007), variables were averaged so that only one
(Lund & Troha, 2008), and deafness (Malandraki & Okalidou, ES per construct (i.e., speech or communication) was included
2007) were not included. for each study in calculation of the composite ES (Hedges
PECS implementation. The independent variable in this & Olkin, 1985). Included outcome variables and summaries
meta-analysis was one or more of the six manualized PECS of the data for group and single-subject designs included are
phases. Five studies with independent variables that were presented in Tables 3–6.
components or modifications of the manualized PECS meth-
odology (Cummings & Williams, 2000; Dooley, Wilczenski,
& Torem, 2001; Ganz, Cook, Corbin-Newsome, Bourgeois, Quality of Evidence
& Flores, 2005; Ganz, Sigafoos, Simpson, & Cook, 2008; The quality of evidence for the PECS approach was evalu-
Yokoyama, Naoi, & Yamamoto, 2006) were not included. ated via the use of rating scales adapted from guidelines by
For the alternating treatment single-subject study conducted Horner et al. (2005) and Wolf (1978) for single-subject de-
by Angermeier and colleagues (Angermeier, Schlosser, Luiselli, signs (see Appendix A) and by Gersten et al. (2005) for group
Harrington, & Carter, 2008), which compared the effec- designs (see Appendix B). Quality indicator items were as-
tiveness of PECS in two conditions (PECS with Picture Com- signed a rating of adequate, partially adequate, or inadequate/
munication Symbols and PECS using Blissymbols), only data unknown and a score of 2, 1, or 0 points, respectively. Each
from PECS trials using Picture Communication Symbols article was independently coded for quality assessment by
(as recommended in the PECS manual) were included. For the first two authors. Interrater reliability was calculated via
the study by Tincani, Crozier, and Alazetta (2006), data were point-to-point agreement on the 11 included articles and
included for Study I, which used the standard PECS protocol. ranged from 84.2% to 100%. Composite scores were tallied
Outcome variables. The dependent variables for the meta- as a percentage of points assigned out of total possible points
analysis were (a) communicative behaviors (e.g., requests for each article. Scores at 90% and above were considered
and initiations) and (b) speech or vocalization. Two studies to be strong quality of evidence. Scores ranging from 75% to
with outcomes that were nonlanguage behaviors—that is, 90% were considered to be adequate quality of evidence, and
aggressive behavior (Frea, Arnold, & Vittimberga, 2001) and scores of 74% and below were considered to be inadequate
problem solving (Marckel, Neef, & Ferreri, 2006)—were not quality of evidence.
included in this meta-analysis. Finally, one study of PECS Overall, the quality of evidence for PECS was considered
in which symbols were modified with braille for use with a to be adequate (see Tables 3–6). Quality assessment ratings
visually impaired therapist was excluded because the graphed of group studies ranged from strong to inadequate levels of
data of communication outcomes in that study were provided evidence (one study of strong quality and two of inadequate
for the therapist and not the child participants (Charlop, quality) for the effectiveness of PECS on communicative and
Malmberg, & Berquist, 2008). speech abilities of children with autism. For group designs,
Articles selected for review. A total of 11 studies (eight the quality of evidence for PECS would be strengthened by
single-subject experiments and three group designs) met improved measurement and reporting of treatment fidelity.
criteria for inclusion. For the purpose of this meta-analysis, For example, in two of the three group studies in this meta-
findings from articles that analyzed data from the same par- analysis, treatment fidelity was not reported. Without measures
ticipants (i.e., Carr & Felce, 2007a, 2007b; Yoder & Stone, of treatment fidelity, it is difficult to know whether PECS was
2006a, 2006b) were collapsed into a single study. delivered in the appropriate manner and dosage; thus, it is
Data extraction and decisions. Data extracted from pub- difficult to demonstrate a functional relationship between the
lished studies of PECS included participant diagnoses and intervention and the communication and speech outcomes.
ages, experimental design, outcome variables, phases of Quality assessment ratings of the single-subject experi-
PECS trained, and reported results. For group studies, ESs ments ranged from strong to adequate (three studies of strong

182 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


TABLE 3. Summary of communication outcomes and quality ratings for single-subject designs.

Ages PECS Treatment Quality rating


Reference ( years;months) Design phases fidelity (Y/N) Dependent variables Results reported (% scores) PND (%) ES G/M

Angermeier et al. 9;0, 7;0, 6;0, 8;0 Alternating treatments I–III Y % independent PECS Gains in PECS exchanges S (90) 93.7 1.04 M
(2008) and multiple baselines exchanges for all 4 participants
across participants
Beck et al. (2008) NR Alternating treatments I–III Y % accuracy PECS Gains in PECS exchanges A (78) — — G
(3 preschoolers) exchanges for all 3 participants
with autism
Ganz & Simpson 5;8, 7;2, 3;9 Changing criterion I–IV N % independent PECS Gains in PECS exchanges A (83) — — NR
(2004) exchanges for all 3 participants
Ganz, Simpson, & 4;5, 3;1, 4;1 Multiple baselines across I–IV N % independent PECS Mastery for 2/3 participants A (82) 64.4 0.71 NR
Corbin-Newsome participants exchanges in independent PECS
(2008) exchanges
Kravitz et al. (2002) 6;0 Multiple baselines across I–III N Spontaneous language Increases in spontaneous A (76) 64.8 0.16 NR
settings (home, (icons and language (icons and
centers, journal) verbalizations) some verbal)
Tincani (2004) 5;10, 6;8 Alternating treatments I–III Y % independent mands Gains in mands for both S (93) 100 0.30 G
participants; greater
gains with signs for
1 participant
Tincani et al. 10;2, 11;9 Multiple baselines across I–IV Y % independent mands Gains in mands for both S (93) 98.4 0.34 G
(2006; Study I) participants participants

Note. PND = percentage of nonoverlapping data points; ES = effect size; G/M = generalization or maintenance measures included; S = strong quality of evidence; NR = not reported;
A = adequate quality of evidence. Dash indicates incalculable due to absence of baseline phase.

TABLE 4. Summary of communication outcomes and quality ratings for group designs.

Ages PECS Treatment Dependent Quality rating


Reference Participants ( years;months) Design phases fidelity (Y/N) variables Results reported (% scores) ES G/M

Carr & Felce 24 PECS, 3;0–7;0 Group I–III N Initiations Initiations increased for children in PECS I (50) 0.95 NR
(2007a, 2007b) 17 control (nonrandom; group but not for control
nonmatched)
Howlin et al. (2007) 26 immediate, 3;9–8;9 RCT NR N Initiations; Significant increase in frequency of I (45) 0.65 M
28 delayed, 5;2–9;5 PECS usage initiations and PECS usage; gains not
29 control 5;1–10;0 maintained on follow-up
Yoder & Stone 17 PECS, 3;0–7;0 RCT I–VI Y Requesting No main effects for PECS on requesting; S (100) — —
(2006a, 2006b) 19 RPMT PECS more effective than RPMT on
requesting for children with low JA

Note. I = inadequate quality of evidence; RCT = randomized controlled trial; RPMT = Responsive Education and Prelinguistic Milieu Teaching; JA = joint attention. Dash indicates incalculable

Flippin et al.: Effectiveness of PECS


from data presented.

183
TABLE 5. Summary of speech outcomes and quality ratings for single-subject designs.

Ages PECS Treatment Quality rating


Reference (years;months) Design phases fidelity (Y/N) Dependent variables Results reported (% scores) PND (%) ES G/M

Beck et al. (2008) NR Alternating treatments I–III Y Total utterances; % intelligible Gains in speech outcomes A (78) — — NR
(3 preschoolers) utterances; % spontaneous for 1/3 participants with
intelligible utterances; autism
different words
Charlop-Christy 3;8, 5;9, 12;0 Multiple baselines I–III N Spontaneous speech; MLU Gains in spontaneous A (83) 59.4 –0.05 M
et al. (2002) across participants; speech, MLU for all
multiple baselines 3 participants
across settings
(home, centers,
journal)
Ganz & Simpson 5;8, 7;2, 3;9 Changing criterion I–IV N Average spoken words Gains in average spoken A (82) — — NR
(2004) per trial words for all
3 participants
Ganz, Simpson, & 4;5, 3;1, 4;1 Multiple baselines I–IV N Word approximations and No gains in word A (80) 6.5 0.01 NR
Corbin-Newsome across participants intelligible words approximations or
(2008) intelligible words
Tincani (2004) 5;10, 6;8 Alternating treatments I–III Y Word vocalizations Gains in words for both S (93) 100 0.26 NR
participants, although

184 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


greater gains with
sign for 1 participant
Tincani et al. 10;2 Multiple baselines I–IV Y Vocal approximations Gains in vocal S (93) 12 –1.45 NR
(2006; Study I) across participants approximations for
1 participant during
Phase IV

Note. Dash indicates incalculable due to absence of baseline phase. MLU = mean length of utterance.
TABLE 6. Summary of speech outcomes and quality ratings for group designs.

Ages PECS Treatment Dependent Quality rating


Reference Participants (years;months) Design phases fidelity (Y/N) variables Results reported (% scores) ES G/M

Howlin et al. (2007) 26 immediate, 3;9–8;9 RCT NR N Frequency No significant change in I (45) 0.05 M
28 delayed, 5;2–9;5 of speech frequency of words
29 control 5;1–10;0
Yoder & Stone 17 PECS, 3;0–7;0 RCT I–VI Y Nonimitative Overall, PECS more S (100) 0.57 M
(2006a, 2006b) 19 RPMT words; effective than RPMT
nonimitative in facilitating words
spoken acts and spoken acts;
PECS especially
effective for children
with low JA and high
object exploration;
advantages for PECS
group at Time 1 not
maintained at Time 2

Flippin et al.: Effectiveness of PECS


185
quality and five of adequate quality). Quality of the single- for three studies that used alternating treatment (Angermeier
subject experiments would be strengthened by inclusion of et al., 2008; Beck, Stoner, Bock, & Parton, 2008) or changing
measures of maintenance, generalization, and social validity. criterion designs (Ganz & Simpson, 2004) without a base-
Follow-up measures to assess maintenance and generaliza- line phase. Interpretation of PND scores as outlined by Scruggs
tion were reported in only 50% of the single-subject studies, and colleagues (Scruggs, Mastropieri, Cook, & Escobar, 1986)
and social validity was measured in only two of the eight is as follows: <50% = unreliable treatment; 50%–70% =
single-subject designs. Failure to measure maintenance, gen- questionable effectiveness; 70%–90% = fairly effective;
eralization, and social validity in the PECS single-subject >90% = highly effective.
literature is consistent with other single-subject investigations After individual ESs were calculated, composite ESs for
in the autism literature. For example, Odom and colleagues the speech and communication outcome variables were com-
(2003) reported that only 35% of single-subject intervention puted independently for single-subject and group designs.
experiments with individuals with ASD measured mainte- For single-subject multiple-baseline designs across partici-
nance and only 41% reported generalization measures. Know- pants, ESs are reported as the mean ES for all participants. As
ing whether speech and communication gains made during the study quality is a focus of this meta-analysis, a threshold of
PECS program are maintained after stopping PECS interven- at least adequate quality was set for inclusion of a study in
tion is important for clinicians in deciding on a communication- the computation of aggregate ESs (Pettiti, 2000). All eight of
training approach for children with autism. the single-subject experiments met criteria of at least adequate
quality. For group designs, however, only one of the three
group experiments met quality criteria. Thus, for the group
Analysis Procedures studies, individual ESs are described here, and aggregate ESs
For group designs, ESs for posttest group differences were were not calculated.
reported as given or calculated from published data. For one
study that reported results using the nonparametric Mann–
Whitney U statistic (Carr & Felce, 2007a, 2007b), an equiv- Synthesized Results
alent of Cohen’s d was calculated as outlined by Rosnow Findings were synthesized to evaluate the effectiveness of
and Rosenthal (2008). The dequivalent metric represents the the PECS approach on communication and speech outcomes
“value of d analogous to the result for an exactly normally for children with autism. Results are reported separately for
distributed outcome with N/2 units in each group” (Rosnow single-subject studies and group designs, as ESs from single-
& Rosenthal, 2008, p. 382). For another study (Howlin et al., subject experiments tend to be inflated when compared with
2007), Cohen’s d was calculated from the published odds those of group designs (Swanson & Sachse-Lee, 2000), and
ratios as outlined by Chinn (2000). Standard interpretation there is no method to combine ESs from different experimental
of Cohen’s d (Cohen, 1988) was used to consider the strength designs (Kavale et al., 2000).
of evidence for group design studies (i.e., <0.2 = trivial;
0.2–0.5 = small; 0.5–0.8 = moderate; >0.8 = strong).
For single-subject experiments, two methods of calculating Communication Outcomes
ES were used, as there is not currently an established metric Communication outcomes examined in the PECS litera-
for ES in single-subject experiments (Beretvas & Chung, 2008). ture include the frequency of PECS exchanges, requests
In the first analysis, data from published graphs were entered (Angermeier et al., 2008; Beck et al., 2008; Ganz & Simpson,
and analyzed using the software program ITSACORR (Crosbie, 2004; Ganz, Simpson, & Corbin-Newsome, 2008; Howlin
1993). ITSACORR conducts an analysis of the single-subject et al., 2007; Kravits, Kamps, Kemmerer, & Potucek, 2002;
data by fitting least-squares trend lines to baseline and treat- Tincani, 2004; Tincani et al., 2006), and initiations (Carr &
ment phases while controlling for autocorrelation and short Felce, 2007a, 2007b; Howlin et al., 2007). Increases in com-
data sets (Maughan et al., 2005). ITSACORR was chosen munication outcomes following PECS training were demon-
over other methods of ES calculation (e.g., standardized dif- strated across both single-subject and group studies. Overall,
ference approach) because of the validity of this approach for the single-subject experiments, the mean weighted ES
for graphs with zero slope in the baseline phase (Huitema, was 0.51. The confidence interval did not include zero, 95% CI
2004) and undefined standard deviations (i.e., less than two [0.04, 0.67]; therefore, the assumption can be made that
data points in baseline or all baseline data points equal to zero; PECS treatment had a significant impact on communication
Beeson & Robey, 2006; Kratochwill, Stoiber, & Gutkin, outcomes. The mean PND of 84.3%, 95% CI [68.4%, 100.1%],
2000). ITSACORR provides an F statistic as well as tests of indicates that PECS is a “fairly effective” treatment for im-
change in intercept and change in slope. The y-intercept proving communication for young children with autism. How-
t statistic given by ITSACORR and the number of data points ever, evidence of the maintenance and generalization of
per phase were used to calculate ESs as Glass’s delta, as communication gains made through PECS intervention is
outlined by Maughan et al. (2005). limited in the single-subject literature. In fact, maintenance
In addition to the ITSACORR method of data analysis for of communication gains was measured in only one of the
single-subject designs, the studies were also analyzed using seven single-subject studies that examined communication
the PND method, in which the highest data point in baseline outcomes. In that study, Angermeier and colleagues (2008)
phase was identified and the percentage of data points during reported that all three participants maintained mastery level
intervention exceeding this level was calculated (Scruggs, of communicative exchanges at the Phase II level of PECS
Mastropieri, & Castro, 1987). PND could not be calculated immediately after completing Phase II training. Furthermore,

186 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


generalization of communication gains achieved through et al., 2002; Howlin et al., 2007) as well as more specific
PECS was measured in only three single-subject studies. measures of the frequency of nonimitative spoken acts and
Beck et al. (2008) measured generalization across settings words (Yoder & Stone, 2006a, 2006b), word vocalizations
for two of the four participants in their study comparing the (Tincani, 2004), word approximations (Ganz, Simpson, &
effectiveness of PECS and VOCA. Probes conducted in the Corbin-Newsome, 2008), vocal approximations (Tincani
children’s homes demonstrated limited generalization of et al., 2006), and average number of words spoken per trial
communication gains made through PECS 1 week after (Ganz & Simpson, 2004). In the five single-subject experi-
school-based training. One participant used PECS to request ments examining speech, outcomes varied widely across
only once during the sessions, while the other participant the 12 participating children (ES ranged from –1.45 to 1.37).
demonstrated only three requests using PECS (vs. 17 re- It is interesting to note that several single-subject studies
quests when using a voice output device). reported overall gains in speech based on visual inspection
In contrast, Tincani (2004) measured generalization across and interpretation of the data. However, as a group, the mean
trainers in an alternating treatment design. In that study, PECS weighted ES for speech outcome variables across single-
was more effective than sign language in promoting com- subject studies was negligible (0.17), and the CI included
munication for one child; this child demonstrated picture zero, 95% CI [–0.01, 0.36]. The mean PND was 44.7%, [1.8%,
exchanges at levels similar to those achieved in training (93% 87.6%], indicating questionable effectiveness of the PECS
on average) during generalization sessions conducted with approach for increasing speech for young children with autism
a new communication partner. Similar results were reported across the single-subject literature.
by Tincani and colleagues (2006), in that both participants Furthermore, maintenance of speech gains was measured
produced levels of independent mands during four generaliza- in only one of the five single-subject studies examining speech
tion probes with a new communication partner (i.e., the child’s outcomes. Charlop-Christy and colleagues (2002) reported
teacher) that were similar to those achieved during Phase IV 1-year follow-up data for one participant who demonstrated
training. All generalization probes were higher than levels of maintenance of speech gains made through PECS at levels
communication achieved during baseline phase. similar to those achieved during PECS training.
For the group designs, the mean weighted ES could not Finally, generalization of speech gains made through PECS
be aggregated, as ES was not calculable for communication was also measured in only one of the five single-subject studies
outcomes for the PECS intervention in one of the three group examining speech outcomes. Tincani and colleagues (2006)
studies (Yoder & Stone, 2006a, 2006b) examining commu- reported that for one participant who made gains in speech
nication outcomes, and the two group studies reporting cal- through PECS, four generalization probes implemented by the
culable ES data did not meet quality inclusion criteria. With child’s teacher demonstrated levels of vocalization similar to
regard to maintenance of communicative gains, concerns were those achieved in training. However, only two of the four
raised by one group study. Howlin and colleagues (2007) generalization probes were above the child’s baseline levels
reported that the increases in rate of initiations and PECS us- of vocalization prior to starting PECS training.
age that made up immediate treatment effects were not main- Results of speech outcomes following PECS training were
tained over time, and upon follow-up, ESs were negligible, also variable across the group experiments. In a follow-up
0.04, 95% CI [–0.61, 0.1], and 0.24, 95% CI [–0.42, 0.92], analysis of their study, Carr and Felce (2007b) noted that five
respectively. Maintenance of communication gains for the of 17 children in PECS group showed increases in speech;
PECS intervention group was not reported in the study by however, ESs for speech outcomes in this study were not
Carr and Felce (2007a, 2007b) and not calculable from pre- calculable from the published data. Children in the Howlin
sented data for Yoder and Stone (2006a, 2006b). In addition, et al. (2007) study achieved a negligible ES for speech (0.05).
generalization of communicative gains was not reported for In contrast, moderate effects for speech outcome variables
any of the three group studies of PECS. Thus, overall, small to (i.e., frequency of nonimitative spoken acts, ES = 0.63, and
moderate effects were demonstrated for the short-term efficacy number of different nonimitative words, ES = 0.50) were
of PECS in improving communicative outcomes for young reported by Yoder and Stone (2006a, 2006b) at posttest in
children with autism across the single-subject experiments their randomized controlled trial comparing PECS with
and group studies of PECS. Evidence for maintenance and Responsive Education and Prelinguistic Milieu Teaching
generalization of communicative gains achieved through (RPMT). In that study, PECS was found overall to be the
PECS training, however, is limited and mixed. more effective treatment for increasing number of non-
imitative spoken communication acts, t(34) = 2.30, p = .03,
and the number of different nonimitative words used, t(34) =
Speech Outcomes 2.10, p = .04. However, posttest advantages in speech real-
Although the developers of the program describe PECS as ized by the PECS group were not maintained at follow-up
an augmentative communication system and are careful not 6 months after intervention. In fact, ES for all speech out-
to claim that PECS improves speech for children with ASD, come measures at 6 months after intervention was negligible
speech is a frequently measured outcome variable in the (ES = 0). Thus, the variable effects in speech outcomes follow-
PECS literature. Clinicians and parents alike desire informa- ing PECS illustrate that the program is not equally effective
tion regarding the effectiveness of PECS in influencing speech in improving speech outcomes for all children with autism.
outcomes for young children with autism. Speech outcomes Variability in response to PECS may be due to preinterven-
were measured broadly across the PECS studies and included tion characteristics of individual children. Available evidence
the general measures of spontaneous speech (Charlop-Christy related to this possibility is examined in the following section.

Flippin et al.: Effectiveness of PECS 187


Child Characteristics: Moderating Variables Object Exploration
and Preintervention Characteristics A second potential preintervention variable that may mod-
Given the heterogeneous nature of the population of chil- erate the effects of PECS on speech outcomes is object ex-
dren with ASD, it is important for both clinical intervention ploration. Differential response to treatment based on a child’s
and research to determine which characteristics make indi- initial levels of object exploration was documented by Yoder
vidual children more likely to benefit from PECS, as well as and Stone (2006a, 2006b). In their study, Yoder and Stone defined
to identify which components of the PECS program are most object exploration as the number of different toys a child
effective. Therefore, in addition to determining the overall touched during a play assessment. For children with initially
effectiveness of the PECS communication approach, further higher levels of object exploration (+1 SD above the group
analysis was conducted across group studies and single-subject mean), PECS was shown to be a more effective treatment than
experiments to identify modifying variables and preinter- RPMT in increasing the frequency of nonimitative speech at
vention characteristics of children who responded to PECS posttest. Initial levels of object exploration were also shown to
training. For group studies, examination of variables that be important in maintaining gains in speech made through PECS.
moderate the effectiveness of PECS provides a strong statis- In fact, only children with high object exploration maintained
tical tool which allows us to identify particular child char- speech gains at the 6-month follow-up, relative to the alternate
acteristics that have a measurable and direct influence on treatment group. Thus, object exploration may be a variable
speech and communication outcomes. For single-subject that moderates speech outcomes achieved by children with
experiments, examination of preintervention participant char- autism through the use of PECS, but this finding has not yet
acteristics allows us to describe which children gained the been replicated across studies or by independent researchers.
most from PECS, at the level of individual participant, al-
though a direct relationship between preintervention charac- Imitation
teristics and speech and language outcomes is not identifiable.
Three child characteristics emerged as potential pretreatment Imitation is a third potential preintervention characteristic
identifiers of a child with autism who is likely to respond to that may influence both speech and communication outcomes
PECS intervention: (a) limited joint attention, (b) relatively for children participating in PECS intervention. Motor imitation
stronger object exploration, and (c) limited motor imitation. was examined as a preintervention characteristic that influenced
the acquisition and effectiveness of PECS in a single-subject
study comparing the effectiveness of sign language and PECS
Joint Attention on the training of requests. In that study, Tincani (2004) ex-
Joint attention refers to three-point, systematic sharing amined the pretreatment ability of two participants to imitate
of attention between (a) an individual, (b) an object or event, 27 hand, arm, and finger movements, similar to those required
and (c) another individual (Tomasello, 1995). Evidence from to communicate via sign language. As was predicted by the
one group design study suggests that PECS may be a more author, sign language was more effective than PECS in facil-
effective communication intervention than RPMT for chil- itating independent requests for a participant with relatively
dren with initially low rates of joint attention. In Yoder and higher motor imitative ability. In contrast, PECS was the more
Stone’s study (2006a, 2006b), exploratory analysis of pre- effective treatment for improving communication for a partic-
treatment variables revealed that initial rates of joint atten- ipant with relatively lower motor imitative ability. With re-
tion were a strong predictor of whether a child made greater gard to speech outcomes, however, sign language was the
communication (i.e., requesting) gains from either PECS or more effective intervention for both participants, regardless
RPMT. Initiating joint attention was operationally defined of initial levels of motor imitation.
as an “intentional communication act about an object that Verbal imitation has also emerged as a potentially influ-
attempts to get the adults to comment, laugh, smile, show at- ential preintervention child characteristic. Charlop-Christy
tention, or give a label” (Yoder & Stone, 2006b, p. 430). For and colleagues (2002) described the three participants in their
children who began an intervention with initially lower fre- study, who each made gains in both communication and
quencies of joint attention (i.e., one instance or less of initiat- speech through PECS, as having some ability to verbally imi-
ing joint attention across two assessment procedures), PECS tate before beginning PECS training. The authors postulate
was the more successful approach in targeting communica- that participants’ baseline level of verbal imitation may have
tion outcomes. Joint attention also emerged as a potentially accounted for the ability to produce spontaneous speech dur-
influential preintervention characteristic in a single-subject ing and after training. Assessment measures of preinterven-
experiment by Charlop-Christy and colleagues (2002), which tion verbal imitation were not described, however. It is important
documented that the largest overall gains in both speech to emphasize that results from these two single-subject ex-
and communication were made by the youngest participant, periments may not generalize to the overall population of
who began the study with no spontaneous speech, and rela- children with ASD. Nevertheless, motor and verbal imitation
tively weaker social communicative behaviors (i.e., joint at- may be a variable of interest in future studies of the effec-
tention, eye contact, and toy play). Thus, joint attention may tiveness of PECS on communication and speech outcomes
be one child characteristic to be explored as a moderating for children with autism. In fact, given that early imitation
variable of main effects in group designs and as a preinterven- skills have been shown to be a strong predictor of later speech
tion characteristic to describe for participants in single-subject development in children with autism (Charman et al., 2003;
experiments. Toth, Munson, Meltzoff, & Dawson, 2006), it is interesting

188 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


that imitation has not been more frequently examined as a models, the use of time delay, or some combination of strat-
pretreatment characteristic across the autism communication egies that may contribute to an increase in speech production
intervention literature. during PECS Phase IV. Alternatively, it may be that those
Thus, the current evidence base for the effects of modera- children who reach the criterion of 80% mastery in Phases I–III
tors and preintervention characteristics is limited, but the are the same children who are likely to go on to develop
measurement and analysis of preintervention variables in future speech. There is some support in the PECS literature for the
studies would be invaluable in informing clinical decisions impact of time delay on increased speech production (Charlop-
about the selection of effective communication interventions Christy et al., 2002; Tincani, 2004; Tincani et al., 2006). For
for individual children with ASD. For future group designs, example, after the introduction of a 4-s delay of reinforcement,
examination of the potential moderating impact of preinter- one child in Tincani’s study (2004) increased percentage of
vention variables such as joint attention, object exploration, word vocalizations from a low of 30.1% during PECS Phase III
and motor/verbal imitation on the effects of PECS on commu- training to an average of 90%. Tincani and colleagues (2006)
nicative and speech outcomes is warranted. In future single- further documented the effectiveness of using a 3–5-s delay
subject experiments, although statistical examination of in an alternating treatment design, with one participant who
moderators is not possible given the constraints of the re- demonstrated greater use of speech during those phases that
search design, reporting preintervention assessment results employed a delay versus baseline phases. Thus, there is some
for these and other participant characteristics would allow for limited evidence that the verbal modeling, time delay, or some
systematic replication of studies and provide pilot data for combination introduced during Phase IV of PECS may play
larger group studies to help identify the profile of a child with a role in the acquisition of spontaneous speech for those chil-
ASD who is likely to make communication and/or speech dren who go on to develop speech through the PECS program,
gains through PECS. but few studies have addressed these questions, limiting the
evidence available. Further research to determine whether in-
Program Characteristics: Phase IV Intervention cluding verbal models and time delay strategies earlier in the
PECS program results in an earlier speech gains would have
and Speech Outcomes important clinical implications.
Another interesting and potentially clinically useful find-
ing of this analysis is the impact of training PECS Phase IV
on spontaneous speech for those children who went on to Summary, Limitations, and Future Directions
develop speech through PECS. In their seminal report, Bondy This meta-analysis synthesized findings from both group
and Frost (1994) documented that although speech develop- and single-subject experiments to evaluate the quality of the
ment was not a specific aim of their approach, vocalizations research evidence and the effectiveness of the PECS approach
generally developed in the later phases of PECS training on speech and communicative abilities for children with au-
(IV to VI). Phase IV of PECS is unique in that it introduces tism. With one group design of strong quality and seven single-
both verbal modeling and time delay strategies. In Phase IV, subject experiments of at least adequate quality documenting
the trainer models the phrase “I wantI” and briefly pauses gains in communication following PECS training, the body
before handing the requested object to the child. Evidence of evidence for the PECS approach demonstrates that PECS is
from several single-subject studies lends empirical support a promising, although not yet established, evidence-based
to Bondy and Frost’s (1994) initial reports of increased speech practice for promoting communication in children with au-
during later PECS phases (Charlop-Christy et al., 2002; Ganz tism. In contrast, the evidence for gains in speech achieved
& Simpson, 2004; Tincani, 2004). through PECS is not as strong. Fewer studies have examined
Because single-subject experiments document the partic- speech outcomes (one group study of strong quality and four
ipants’ changes in speech across each phase of PECS, it was single-subject studies of at least adequate quality), and ESs
possible to compare the magnitude of gains in speech between have varied widely across studies. Given the nine controlled,
Phases III and IV for those single-subject experiments that empirical studies of at least adequate quality that were iden-
reported speech outcomes through Phase IVof PECS (Ganz & tified in this meta-analysis, PECS is relatively well-researched
Simpson, 2004; Ganz, Simpson, & Corbin-Newsome, 2008; as compared with other popular communication interven-
Tincani, 2004). This analysis resulted in a composite ES for tions for children with autism. Considering that there are only
speech outcomes across Phase IVof the single-subject studies nine such studies, however, our results on the effectiveness of
of 0.73. This lends support to evidence that for children who the PECS approach should be viewed with caution.
go on to develop speech, Phase IV is important. In contrast, Overall, the data do provide evidence that PECS has small
for those same children, communication gains between to moderate effects in improving communication skills for
Phases III and IV were found to be negative (ES = –0.37). children with autism. In addition, PECS may improve speech
Large ESs for speech development and negative effects for ability for some children with autism relative to speech prior
communication outcomes between Phases III and IV suggest to intervention or gains made as a function of other interven-
that as these children become increasingly more verbal, speech tion approaches, but further research is needed related to speech
begins to replace the picture-based exchanges of PECS. development as an outcome of PECS intervention. The find-
Thus, it serves to highlight that Phase IV may be particularly ing of a negligible aggregated ES for speech outcomes across
important for development of speech. the single-subject experiments and small to negative ESs for
However, questions remain as to whether speech gains group studies is an important outcome of this meta-analysis
realized in Phase IV were due to the introduction of verbal and demonstrates the value of using synthesized results to

Flippin et al.: Effectiveness of PECS 189


uncover trends in the composite data that are not readily appar- verbal modeling and time delay components earlier in the
ent by visual inspection and narrative review. These results program than Phase IV would result in earlier gains in sponta-
also provide information on some preintervention child char- neous speech or whether it is necessary to master Phases I–III
acteristics and program characteristics that may be influenc- as a foundation to developing speech in those children who
ing speech and communicative outcomes for children with do go on to develop speech.
autism undergoing PECS training, but studies addressing In summary, the overall effectiveness of the PECS approach
these questions are currently few in number. for communication outcomes with children with ASD is prom-
Despite the overall adequate quality of evidence for the ising although not yet established. Evidence for the effec-
effectiveness of the PECS approach and the potentially clin- tiveness of the approach on speech outcomes is not as strong.
ically useful identification of child and intervention charac- PECS may be more beneficial for promoting speech in chil-
teristics for examining which children are most likely to have dren with a specific developmental profile (i.e., low joint
favorable speech and/or communication outcomes through attention, low motor imitation, and high object exploration)
PECS, several concerns and questions remain. One impor- than some alternative approaches, at least as a short-term
tant concern is the lack of measurement and reporting of effect. Further empirical research is now needed to directly
treatment fidelity for two of the three group studies of PECS examine the generalizablity of speech and communication
and four of the eight single-subject experiments. Without gains following PECS and to examine and identify child
clear documentation of how PECS was implemented, includ- characteristics that may account for the variance in outcomes.
ing descriptions of personnel training and measurement of Based on this meta-analysis, we recommend that researchers
adherence to the manualized PECS protocol, the functional include measures of treatment fidelity, maintenance, and gen-
relationship between the intervention and communication eralization in future studies of PECS. We also support the
and speech outcomes at both posttest and during generaliza- need for additional studies examining the effectiveness of
tion is unclear. In addition, there is insufficient evidence to PECS on speech and communication outcomes for children
determine whether PECS results in generalization and long- with ASD as compared to other interventions. In addition, the
term maintenance of communication and speech gains for evidence base for PECS will be enhanced if researchers ex-
children with ASD. Two group studies raised concerns re- amine the potential influence of moderators and preinter-
garding the maintenance of communication and speech gains vention characteristics described in this synthesis (i.e., joint
made through PECS, and only a few single-subject studies attention, object exploration, and motor and verbal imitation)
reported maintenance or generalization measures. Clinically, and other relevant pretreatment variables. Implementing these
this is a concern, as it is important for interventionists to know recommendations will strengthen the evidence base of this
whether gains made in PECS can be maintained without con- approach and aid clinicians in making informed treatment de-
tinued support, across varying contexts. Furthermore, given cisions for individual children with ASD.
that there is only one published group design comparing
PECS with a specific alternative treatment, it is still unknown Acknowledgment
whether the PECS approach results in better communication
and speech outcomes than other available autism interven- The authors thank Necdet Karasu for his advice on calculating
effect sizes in single-subject research.
tions. Thus, several concerns and questions remain about the
overall effectiveness of PECS for communication and speech
outcomes for children with ASD. References
Questions also remain regarding the influence of modera- References marked with an asterisk indicate studies included in
tors on the effectiveness of PECS and the identification of the meta-analysis.
pretreatment characteristics of children who are more likely to Adkins, T., & Axelrod, S. (2002). Topography versus selection-
respond to PECS. Given the small number of studies providing based responding: Comparison of mand acquisition in each
the relevant pretreatment data on participants, the evidence modality. The Behavior Analyst Today, 2, 259–266.
is inconclusive regarding which of the currently identified *Angermeier, K., Schlosser, R. W., Luiselli, J. K., Harrington, C.,
& Carter, B. (2008). Effects of iconicity on requesting with
factors (i.e., joint attention, motor and verbal imitation, and the Picture Exchange Communication System in children with
object exploration) contribute the most to the relative effec- autism spectrum disorder. Research in Autism Spectrum Dis-
tiveness of PECS on speech and communication outcomes. orders, 2, 430–446.
In addition, the three preintervention characteristics described *Beck, A. R., Stoner, J. B., Bock, S. J., & Parton, T. (2008).
in this synthesis likely do not exhaust the possible initial charac- Comparison of PECS and the use of VOCA: A replication. Edu-
teristics of children with ASD influencing response to PECS. cation and Training in Developmental Disabilities, 43, 198–216.
Finally, Phase IV of PECS and two strategies introduced Beeson, P. L. M., & Robey, R. R. (2006). Evaluating single-
in that phase (i.e., verbal modeling and time delay) were subject treatment research: Lessons learned from the aphasia
discussed as possible influences on speech acquisition for literature. Neuropsychology Review, 16, 161–169.
those children who went on to develop speech through PECS Bellini, S., & Akullian, J. (2007). A meta-analysis of video mod-
eling and video self-modeling interventions for children and
training. It is still unclear, however, as to whether the doc- adolescents with autism spectrum disorders. Exceptional Chil-
umented increases in spontaneous speech in Phase IV are dren, 73, 264–287.
related to the introduction of verbal modeling, time delay, or Bellini, S., Peters, J. K., Benner, L., & Hopf, A. (2007). A meta-
some combination, or whether those children who are able to analysis of school-based social skills interventions for children
achieve success in Phases I–III are simply more likely to go with autism spectrum disorders. Remedial & Special Education,
on to develop speech. Also unclear is whether introducing 28, 153–162.

190 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


Beretvas, S. N., & Chung, H. (2008). A review of meta-analyses Ganz, J. B., Sigafoos, J., Simpson, R. L., & Cook, K. E. (2008).
of single-subject experimental designs: Methodological issues Generalization of a pictorial alternative communication system
and practice. Evidence-Based Communication Assessment and across instructors and distance. AAC: Augmentative & Alter-
Intervention, 2, 129–141. native Communication, 24, 89–99.
Bock, S. J., Stoner, J. B., Beck, A. R., Hanley, L., & Prochnow, J. *Ganz, J., & Simpson, R. (2004). Effects on communicative re-
(2005). Increasing functional communication in non-speaking questing and speech development of the picture exchange com-
preschool children: Comparison of PECS and VOCA. Educa- munication system in children with characteristics of autism.
tion and Training in Developmental Disabilities, 40, 264–278. Journal of Autism and Developmental Disorders, 34, 395–409.
Bondy, A. S., & Frost, L. A. (1994). The Picture Exchange Com- *Ganz, J., Simpson, R. L., & Corbin-Newsome, J. (2008). The
munication System. Focus on Autistic Behavior, 9, 1–19. impact of the Picture Exchange Communication System on
*Carr, D., & Felce, J. (2007a). Brief report: Increase in production requesting and speech development in preschoolers with autism
of spoken words in some children with autism after PECS train- spectrum disorders and similar characteristics. Research in
ing to Phase III. Journal of Autism and Developmental Dis- Autism Spectrum Disorders, 2, 157–169.
orders, 37, 780–787. Gersten, R., Fuchs, L. S., Compton, D., Coyne, M., Greenwood,
*Carr, D., & Felce, J. (2007b). The effects of PECS teaching to C., & Innocenti, M. S. (2005). Quality indicators for group
Phase III on the communicative interactions between children with experimental and quasi-experimental research in special educa-
autism and their teachers. Journal of Autism and Developmental tion. Exceptional Children, 71, 149–164.
Disorders, 37, 724–737. Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-
Chambers, M., & Rehfeldt, R. (2003). Assessing the acquisition analysis. Orlando, FL: Academic Press.
and generalization of two mand forms with adults with severe Horner, R. H., Carr, E. G., Halle, J., McGee, G., Odom, S., &
developmental disabilities. Research in Developmental Dis- Wolery, M. (2005). The use of single-subject research to iden-
abilities, 24, 265–280. tify evidence-based practice in special education. Exceptional
Charlop, M. H., Malmberg, D. B., & Berquist, K. L. (2008). An Children, 71, 165–179.
application of the Picture Exchange Communication System *Howlin, P., Gordon, R. K., Pasco, G., Wade, A., & Charman, T.
(PECS) with children with autism and a visually impaired therapist. (2007). The effectiveness of Picture Exchange Communication
Journal of Developmental and Physical Disabilities, 20, 509–525. System (PECS) training for teachers of children with autism:
*Charlop-Christy, M. H., Carpenter, M., Le, L., LeBlanc, L. A., A pragmatic, group randomised controlled trial. Journal of
& Kellet, K. (2002). Using the Picture Exchange Communication Child Psychology and Psychiatry, 48, 473–481.
System (PECS) with children with autism: Assessment of PECS Huitema, B. E. (2004). Analysis of interrupted time-series ex-
acquisition, speech, social-communicative behavior, and problem periments using ITSE: A critique. Understanding Statistics, 3,
behavior. Journal of Applied Behavior Analysis, 35, 213–231. 27–46.
Charman, T., Baron-Cohen, S., Swettenham, J., Baird, G., Kavale, K. A., Mathur, S. R., Forness, S. R., Quinn, M. M., &
Drew, A., & Cox, A. (2003). Predicting language outcome in Rutherford, R. B. (2000). Right reason in the integration of
infants with autism and pervasive developmental disorders. The group and single subject research in behavioral disorders.
International Journal of Language and Communication Dis- Behavioral Disorders, 25, 142–157.
orders, 38, 265–285. Kern, L., Gallagher, P., Staosta, K., Hickman, W., & George,
Chinn, S. (2000). A simple method for converting an odds ratio to M. (2006). Longitudinal outcomes of functional behavioral
effect size for use in meta-analysis. Statistics in Medicine, 19, assessment-based intervention. Journal of Positive Behavioral
3127–3131. Interventions, 8, 67–78.
Cohen, J. (1988). Statistical power analysis for the behavioral Kratochwill, T. R., Stoiber, K. C., & Gutkin, T. B. (2000). Em-
sciences (2nd ed.). Hillsdale, NJ: Erlbaum. pirically supported interventions in school psychology: The
Crosbie, J. (1993). Interrupted time-series analysis with brief role of negative results in outcome research. Psychology in the
single-subject data. Journal of Consulting & Clinical Psychol- Schools, 37, 399–413.
ogy, 61, 966–974. *Kravits, T. R., Kamps, D. M., Kemmerer, K., & Potucek, J.
Cummings, A. R., & Williams, W. (2000). Visual identity matching (2002). Brief report: Increasing communication skills for an
and vocal imitation training with children with autism: A surprising elementary-aged student with autism using the Picture Exchange
finding. Journal on Developmental Disabilities, 7, 109–122. Communication System. Journal of Autism and Developmental
Dooley, P., Wilczenski, F. L., & Torem, C. (2001). Using an ac- Disorders, 32, 225–230.
tivity schedule to smooth school transitions. Journal of Positive Lancioni, G. E., O’Reilly, M. F., Cuvo, A. J., Singh, N. N.,
Behavior Interventions, 3, 57–61. Sigafoos, J., & Didden, R. (2007). PECS and VOCAs to enable
Fabiano, G. A., Pelham, W. E., Coles, E. K., Gnagy, E. M., students with developmental disabilities to make requests: An
Chronis-Tuscano, A., & O’Connor, B. C. (2009). A meta- overview of the literature. Research in Developmental Disabili-
analysis of behavioral treatments for attention-deficit / ties, 28, 468–488.
hyperactivity disorder. Clinical Psychology Review, 29, 129–140. Lord, C., Risi, S., & Pickles, A. (2004). Trajectory of language
Frea, W., Arnold, C., & Vittimberga, G. (2001). A demonstration of development in autistic spectrum disorders. In M. L. Rice &
the effects of augmentative communication on the extreme aggres- S. F. Warren (Eds.), Developmental language disorders: From
sive behavior of a child with autism within an integrated preschool phenotypes to etiologies (pp. 1–38). Mahwah, NJ: Erlbaum.
setting. Journal of Positive Behavior Intervention, 3, 194–198. Lord, C., Wagner, A., Rogers, S., Szatmari, P., Aman, M.,
Frost, L., & Bondy, A. (2002). PECS training manual. Newark, Charman, T., . . . Yoder, P. (2005). Challenges in evaluating
DE: Pyramid Educational Consultants. psychosocial interventions for autistic spectrum disorders.
Ganz, J. B., Cook, K. E., Corbin-Newsome, J., Bourgeois, B., & Journal of Autism and Developmental Disorders, 35, 695–708.
Flores, M. (2005). Variations on the use of a pictorial alternative Lund, S. K., & Troha, J. M. (2008). Teaching young people who
communication system with a child with autism and develop- are blind and have autism to make requests using a variation
mental delays. Teaching Exceptional Children Plus, 1(6). on the Picture Exchange Communication System with tactile
Retrieved from http://escholarship.bc.edu /education /tecplus/ symbols: A preliminary investigation. Journal of Autism and
vol1/iss6/3. Developmental Disorders, 38, 719–730.

Flippin et al.: Effectiveness of PECS 191


Magiati, I., & Howlin, P. (2003). A pilot evaluation study of the Suk-Hyang, L., Simpson, R. L., & Shogren, K. A. (2007). Effects
Picture Exchange Communication System (PECS) for children and implications of self-management for students with autism:
with autistic spectrum disorders. Autism: The International A meta-analysis. Focus on Autism and Other Developmental
Journal of Research & Practice, 7, 297–320. Disabilities, 22, 2–13.
Malandraki, G. A., & Okalidou, A. (2007). The application of Sulzer-Azaroff, B., Hoffman, A., Horton, C., Bondy, A., &
PECS in a deaf child with autism: A case study. Focus on Autism Frost, L. (2009). The Picture Exchange System (PECS): What
and Other Developmental Disabilities, 22, 23–32. do the data say? Focus on Autism and Other Developmental
Marckel, J. M., Neef, N. A., & Ferreri, S. J. (2006). A preliminary Disabilities, 24, 89–103.
analysis of teaching improvisation with the Picture Exchange Swanson, H. L., & Sachse-Lee, C. (2000). A meta-analysis of
Communication System to children with autism. Journal of single-subject-design intervention research for students with
Applied Behavior Analysis, 39, 109–115. LD. Journal of Learning Disabilities, 33, 114–136.
Maughan, D. R., Christiansen, E., Jenson, W. R., Olympia, D., Tien, K. C. (2008). Effectiveness of the Picture Exchange Commu-
& Clark, E. (2005). Behavioral parent training as a treatment nication System as a functional communication intervention for
for externalizing behaviors and disruptive behavior disorders: individuals with autism spectrum disorders: A practice-based
A meta-analysis. School Psychology Review, 34, 267–286. research synthesis. Education and Training in Developmental
Odom, S. L., Brown, W. H., Frey, T., Karasu, N., Smith-Canter, Disabilities, 43, 67–76.
L. L., & Strain, P. S. (2003). Evidence-based practices for young *Tincani, M. (2004). Comparing the Picture Exchange Commu-
children with autism: Contributions for single-subject design re- nication System and sign language for children with autism.
search. Focus on Autism and Other Developmental Disabilities, Focus on Autism and Other Developmental Disabilities, 19,
18, 166–175. 152–163.
*Tincani, M., Crozier, S., & Alazetta, L. (2006). The Picture Ex-
Pettiti, D. B. (2000). Meta-analysis, decision analysis, and cost-
change Communication System: Effects on manding and speech
effectiveness analysis: Methods for quantitative synthesis in
development for school-aged children with autism. Education
medicine. New York, NY: Oxford University Press.
and Training in Developmental Disabilities, 41, 177–184.
Preston, D., & Carter, M. (2009). A review of the efficacy of the Tomasello, M. (1995). Joint attention as social cognition. In
Picture Exchange Communication System intervention. Journal C. Moore & P. J. Dunham (Eds.), Joint attention and its origins
of Autism and Developmental Disorders, 39, 1471–1486. and role in development (pp. 103–130). Hillsdale, NJ: Erlbaum.
Rehfeldt, R., & Root, S. (2005). Establishing derived requesting Toth, K., Munson, J., Meltzoff, A. N., & Dawson, G. (2006).
skills in adults with severe developmental disabilities. Journal of Early predictors of communication development in young chil-
Applied Behavior Analysis, 38, 101–105. dren with autism spectrum disorder: Joint attention, imitation,
Rosales, R., & Rehfeldt, R. A. (2007). Contriving transitive condi- and toy play. Journal of Autism and Developmental Disorders,
tioned establishing operations to establish derived manding skills 36, 993–1005.
in adults with severe developmental disabilities. Journal of Volkmar, F. R., Lord, C., Bailey, A., Schultz, R. T., & Klin, A.
Applied Behavior Analysis, 38, 101–105. (2004). Autism and pervasive developmental disorders.
Rosnow, R., & Rosenthal, R. (2008). Assessing the effect size of Journal of Child Psychology and Psychiatry, 45, 135–170.
outcome research. In A. M. Nezu & C. M. Nezu (Eds.), Evidence- Webb, T. (2000). The talking goes on—the Picture Exchange
based outcome research: A practical guide to conducting ran- Communication System. Special Children, 130, 30–34.
domized controlled trials for psychosocial interventions Wolf, M. (1978). Social validity: The case for subjective measure-
(pp. 379–401). New York, NY: Oxford University Press. ment or how applied behavior analysis is finding its heart.
Schlosser, R. W., & Sigafoos, J. (2008). Meta-analysis of single- Journal of Applied Behavioral Analysis, 11, 203–214.
subject experimental designs: Why now? Evidence-Based Com- *Yoder, P., & Stone, W. L. (2006a). A randomized comparison of the
munication Assessment & Intervention, 2, 117–119. effect of two prelinguistic communication interventions on the
Schlosser, R. W., & Wendt, O. (2008). Effects of augmentative acquisition of spoken communication in preschoolers, with ASD.
and alternative communication intervention on speech produc- Journal of Speech, Language, and Hearing Research, 49, 698–711.
tion in children with autism: A systematic review. American *Yoder, P., & Stone, W. L. (2006b). Randomized comparison
Journal of Speech-Language Pathology, 17, 212–230. of two communication interventions for preschoolers with autism
Schneider, N., Goldstein, H., & Parker, R. (2008). Social skills spectrum disorders. Journal of Consulting and Clinical Psy-
interventions for children with autism: A meta-analytic ap- chology, 74, 426–435.
plication of percentage of all non-overlapping data (PAND). Yokoyama, K., Naoi, N., & Yamamoto, J. (2006). Teaching verbal
Evidence-Based Communication Assessment and Intervention, behavior using the Picture Exchange Communication System
2, 152–162. (PECS) with children with autism spectrum disorder. Japanese
*Schwartz, I. S., Garfinkle, A. N., & Bauer, J. (1998). The Pic- Journal of Special Education, 43, 485–503.
ture Exchange Communication System: Communicative out- Ziomek, M. M., & Rehfeldt, R. A. (2008). Investigating the ac-
comes for young children with disabilities. Topics in Early quisition, generalization, and emergence of untrained verbal
Childhood Special Education, 18, 144–159. operants for mands acquired using the Picture Exchange Com-
Scruggs, T. E., Mastropieri, M. A., & Castro, G. (1987). The munication System in adults with severe developmental dis-
quantitative synthesis of single subject research: Methodology abilities. Analysis of Verbal Behavior, 24, 15–30.
and validation. Remedial & Special Education, 8, 24–33.
Scruggs, T. E., Mastropieri, M. A., Cook, S. B., & Escobar, C. Received March 19, 2009
(1986). Early intervention for children with conduct disorders: Revision received June 30, 2009
A quantitative synthesis of single-subject research. Behavioral Accepted January 25, 2010
Disorders, 11, 260–271. DOI: 10.1044/1058-0360(2010/09-0022)
Stoner, J. B., Beck, A. R., Bock, S. J., Hickey, K., Kosuwan, K.,
& Thompson, J. R. (2006). The effectiveness of the Picture Contact author: Michelle Flippin, UNC—Chapel Hill Division
Exchange Communication System with nonspeaking adults. of Speech & Hearing Sciences, Bondurant Hall, CB 7190,
Remedial & Special Education, 27, 154–165. Chapel Hill, NC 27599. E-mail: mflippin@unc.edu.

192 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


Appendix A (p. 1 of 2)
Quality Indicators for Single-Subject Research

Flippin et al.: Effectiveness of PECS 193


Appendix A (p. 2 of 2)
Quality Indicators for Single-Subject Research

194 American Journal of Speech-Language Pathology • Vol. 19 • 178–195 • May 2010


Appendix B
Quality Indicators for Group Research

Flippin et al.: Effectiveness of PECS 195


Copyright of American Journal of Speech-Language Pathology is the property of American Speech-Language-
Hearing Association and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

You might also like