Content uploaded by Cheryl Gabig
Author content
All content in this area was uploaded by Cheryl Gabig
Content may be subject to copyright.
LSHSS
Verbal Working Memory and Story
Retelling in School-Age
Children With Autism
Cheryl Smith Gabig
Lehman College/City University of New York, Bronx
Autism is a complex neurodevelopmental disorder that
is characterized by deficits in language, communi-
cation, and social interaction, as well as a restricted
range of interests (American Psychiatric Association [APA], 1994).
A core diagnostic feature of autism is a lack of spoken language
or a significant delay in the development of language and commu-
nication, frequently accompanied by universal speech and language
characteristics that are unique to a diagnosis of autism. For ex-
ample, early research on language and communication deficits
in children with autism documented a wide range of characteristic
features of speech and language that are specific to autism, includ-
ing the presence of echolalia; atypical prosody of speech; pronoun
reversals; stereotyped, ritualistic language; restricted word mean-
ings; and idiosyncratic words or neologisms (Tager-Flusberg, 1996).
However, recent research on language and communication in chil-
dren with autism has moved beyond documenting the universal
characteristics of language and communication behaviors specific to
a diagnosis of autism and has focused more on characterizing the
range and variability of the language profiles and communication
ability seen in children with autism (Kjelgaard & Tager-Flusberg,
2001; Tager-Flusberg, 2003). There is a high degree of variabil-
ity in the clinical manifestation of the disorder, especially within
the communication and language domains. Some children with
autism never develop functional speech or language and remain
nonverbal; others use speech with well-developed syntactic ability.
In the latter case, the communication and language disorder may be
more subtle or confined to the use of conversational language in
social interactions (Kjelgaard & Tager-Flusberg, 2001; Tager-
Flusberg, Paul, & Lord, 2005).
In an attempt to better understand the characteristics and range
of communication difficulties in children with autism, some in-
vestigators have theorized that variations and deficits in verbal
working memory and complex information processing ability may
be causal explanations for the clinical expression of the disorder
(Gathercole & Alloway, 2006; Minshew & Goldstein, 1998, 2001;
Russell, Jarrold, & Henry, 1996; Williams, Goldstein, & Minshew,
2006). The notion that deficits in working memory capacity con-
tribute to individual differences in language skill acquisition has
been noted by researchers who were investigating working memory
function in children with language impairments but without autism
ABSTRACT: Purpose: This study examined verbal working memory
and language ability in 15 school-age children with autism using
3 verbal working memory tasks and 1 story recall task.
Method: Three measures of verbal working memory—nonword
repetition, memory for digits span, and sentence imitation—were
given to children with autism and age-matched controls. Verbal
working memory measures were chosen to reflect increasing levels
of cognitive–linguistic complexity. Story retelling was measured
using The Renfrew Bus Story (J. Cowley & C. Glasgow, 1994) and
was scored for the percentage of propositions recalled and the
average utterance length.
Results: A profile of verbal working memory deficits was seen
in children with autism, with poorer performance on more
complex verbal memory tasks. Performance on the 3 verbal
memory tasks was independent of articulation ability. For the group
with autism, receptive vocabulary was associated with sentence
imitation and story recall but not with nonword repetition or digits
span. Sentence imitation was related to story recall, but the
relationship disappeared when the effect of vocabulary was
removed.
Conclusions: Vocabulary and language processing demands affect
the performance of children with autism on tasks of verbal memory
and story retelling. Results are viewed within a connectionist
framework of verbal working memory.
KEY WORDS: verbal working memory, story retelling, autism
LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008 *American Speech-Language-Hearing Association
0161-1461/08/3904-0498
498
(Ellis Weismer, Evans, & Hesketh, 1999; Montgomery, 1995,
2000). A defining feature of verbal working memory is its capacity
to temporarily store and manipulate varying levels of linguistic
material. Working memory differs from short-term memory in that
it involves both the storage and processing of information; short-
term memory is the temporary storage of material in specific do-
mains. Working memory is influenced by a limited capacity reserve
that is linked theoretically to attention (Cowan et al., 2005) or to the
central executive—a regulatory control mechanism that supports
working memory (Baddeley, 1986, 2003). Failure to meet the pro-
cessing demands of complex language activities in verbal working
memory will result in deficits in learning, both academic and
linguistic (Gathercole & Alloway, 2006). Thus, deficits in the ver-
bal working memory of children with autism would be associated
with deficits in language learning and competence. This report
examines the role of verbal working memory function in children
with autism and the relationship between verbal working memory
and complex language processing.
Speech Articulation and Language Characteristics
of Children With Autism
Communication disorders are a core feature of autism. Recent
research has emerged that characterizes the nature and range of
speech and language ability in children with autism. There is sig-
nificant heterogeneity in the receptive and expressive language
skills of children with autism, but subgroups of children are being
identified based on their performance on standardized measures
of speech, language, and cognitive ability.
In an extensive examination of speech and language functioning
in a group of 89 verbal children between 4 and 14 years of age with
autism, Kjelgaard and Tager-Flusberg (2001) measured a broad
range of language ability using standardized measures including
the Peabody Picture Vocabulary Test—III (PPVT–III; Dunn &
Dunn, 1997), Expressive Vocabulary Test (EVT; Williams, 1997),
Goldman Fristoe Test of Articulation (GFTA; Goldman & Fristoe,
1986), Clinical Evaluation of Language F undamentals —III
(CELF–III; Semel, Wiig, & Secord, 1995), and CELF—Preschool
(Wiig, Secord, & Semel, 1992). In addition, the researchers included
a measure of phonological working memory using the Nonsense
Word Repetition subtest (NWRT) from the NEPSY (Korkman,
Kirk, & Kemp, 1998), a neuropsychological assessment battery.
Of the original 89 children included in the study, 44 (49%) were
able to complete the core subtests for age on the CELF, and this
subset of children was divided into three subgroups: normal lan-
guage learners with an overall language quotient in the normal
range (SS ≥85), borderline language learners (SS = 70–84), and
impaired language learners (SS < 70). The group identified as
normal language learners scored within the average range on all
speech and language areas measured, including receptive vocabu-
lary (PPVT ), expressive vocabulary (EVT), speech articulation
(GFTA), and higher order semantic and morphosyntactic abilities
(CELF). In addition, the normal language group demonstrated
average ability to repeat novel phonological sequences (NWRT).
The group identified as borderline in language learning had def-
icits in all aspects of language including receptive and expres-
sive vocabulary and higher order semantic and morphosyntactic
abilities, as well as below average performance in phonological
working memory. However, speech articulation ability was
within the average range of performance for age for the borderline
language group. Children with the most language impairment, al-
though demonstrating spontaneous language ability, showed global
language deficits across all measures of language with the exception
of speech articulation, which was in the average range of perfor-
mance for chronological age. Significant deficits in phonological
working memory were also identified in the language impaired
group. Kjelgaard and Tager-Flusberg noted three trends in the pat-
tern of language functioning in the borderline and impaired sub-
groups of children, including higher scores on the two vocabulary
measures (PPVT, EVT) than on the higher order semantic and
morphosyntactic abilities measure (CELF), normal performance in
speech articulation (GFTA), and below average performance on
phonological working memory (NWRT). The profiles of the two
language impaired subgroups of children with autism suggested that
lexical knowledge is generally better developed than higher order
morphosyntactic and semantic knowledge, and that basic speech
articulation ability is developmentally appropriate, at least at the
single-word level.
A recent study by Shriberg et al. (2001) provided a different
view of phonological ability in individuals with autism and an-
other pervasive developmental disorder, Asperger syndrome (AS).
Shriberg et al. investigated the speech and prosody characteristics
of adolescents and young adults who had been diagnosed with
either high-functioning autism (HFA) or AS. Each group revealed
a percentage of correct consonant production between 79.9% and
100%, although one third of the HFA and AS participants had
clinically significant distortion errors on sibilant and liquid con-
sonants and rhotic vowels. These individuals did not demonstrate
other types of speech articulation errors such as omission or sub-
stitution of phonemes. However, errors in prosody and phrasing in
their utterances were present, which interfered with the clarity of the
message being conveyed.
Regardless of the range and variability of language skills and
phonological working memory in children with autism, deficits in
the pragmatic use of language are a defining feature of autism
(Kjelgaard & Tager-Flusberg, 2001; Tager-Flusberg et al., 2005).
All children with autism demonstrate impairments in pragmatic
aspects of language, including restricted range of functional speech
acts (Stone & Caro-Martinez, 1990), decreased conversational
ability, and impaired narrative discourse (Losh & Capps, 2003;
Norbury & Bishop, 2003; Tager-Flusberg, 1995). Deficits in the
pragmatic function of language are so pervasive in the clinical
population of autism that these deficits distinguish autism from
other developmental language delays (Rice, Warren, & Betz, 2005;
Tager-Flusberg, 2003).
Research on the nature of language impairment in children with
autism reveals that, when language is impaired, a hierarchy of
language deficit is seen. This hierarchy is, in order of least to most
impaired, single-word speech articulation (average performance)
(Kjelgaard & Tager-Flusberg, 2001); single-word lexical knowl-
edge (minimally-moderately impaired) (Kjelgaard & Tager-Flusberg,
2001; Tager-Flusberg, 2003); and higher order morphosyntactic
skill, narrative discourse, and pragmatic competence (most im-
paired) (Rice et al., 2005; Tager-Flusberg, 1999). Speech articula-
tion is essentially spared (Kjelgaard & Tager-Flusberg, 2001;
Tager-Flusberg, 2003), with the exception of prosody, phrasing,
and consonant distortions on later acquired phonemes (Shriberg
et al., 2001). The question is: What cognitive mechanisms may
account for the range and patterns of language impairment that
are seen in children with autism? Because verbal working memory
Smith Gabig: Working Memory and Story Retelling 499
is considered to be an essential cognitive function in the processing
of complex language (Gathercole & Alloway, 2006; Just & Carpenter,
1992; Minshew & Goldstein, 1998, 2001; Williams et al., 2006),
it is logical to assume that variations in verbal working memory
would be associated with patterns of language functioning seen in
children with language impairment and autism.
Verbal Working Memory in Children With Autism
The majority of research on verbal working memory in children
with autism has focused on documenting the children’s perfor-
mance on varying tasks of verbal working memory relative to either
age- or cognitive-age-matched control groups. Children with autism
demonstrate poor performance across a variety of tasks that are
used to measure verbal working memory. For example, recent
research by Kjelgaard and Tager-Flusberg (2001) reported that
some children who had been diagnosed with both autism and
normal nonverbal intelligence showed poor performance on a task
of nonword repetition, which is a task of phonological working
memory (Gathercole, Willis, Baddeley, & Emslie, 1994). Other
studies have shown that children with autism were able to recall
word lists as well as typically developing (TD) children ( Russell
et al., 1996) and as well as children with learning impairment
(Minshew & Goldstein, 1998) but had more difficulty recalling
semantically organized word strings or syntactic sequences ( Tager-
Flusberg, 1991). Beyond the recall of words and phrases, work-
ing memory deficits are seen during the recall of more complex
verbal information such as sentences and stories (Tager-Flusberg,
1995; Williams et al., 2006).
There is some evidence that the complexity of the to-be-
remembered material influences the outcome during recall for chil-
dren with autism (Fein et al., 1996; Williams et al., 2006). Cognitive
theorists define complexity as the number of elements that are
contained in the stimulus materials as well as the number of cogni-
tive processes that are involved in the task (Minshew & Goldstein,
1998). When presented with tasks of digit span, sentence repetition,
or story recall, preschool children with autism were found to have the
least difficulty with memory for digit span, more with sentences,
and the most with stories (Fein et al., 1996). Similar findings were
reported in a recent study of memory profiles of older HFA children
between 8 and 16 years of age. Using the Wide Range Assessment
of Memory and Learning (WRAML; Sheslow & Adams, 1990),
researchers found no significant difference in performance on a task
of number/letter recall between an older HFA group and a cognitive/
age-matched control group, although significant between-groups
differences were found for the recall of sentences and stories (Williams
et al., 2006). This research on verbal memory task complexity
and performance outcomes for children with autism has focused on
either preschool children who were younger than 5;0 (years;months)
or children age 8;0 and older. Less is known about the verbal
working memory and language processing abilities of school-age
children between 5;0 and 7;11 with autism who have achieved
some functional competence in verbal expression.
It seems intuitive that individuals would find number/letter
sequences easier to recall than sentences and stories because these
sequences do not contain higher level linguistic information coded
within syntactic organizational framing or within story schema.
Both sentence and story memory require the activation of concepts
and schemas outside the realm of immediate short-term memory
in order to process and recall the material. The contribution of
long-term memory representations and the influence of the central
executive ( Baddeley, 1986) during the performance of complex
language tasks such as sentence repetition and story recall has
been noted by many investigators (Gathercole & Baddeley, 1993;
Minshew & Goldstein, 1998; Tager-Flusberg, 1995; Williams,
Goldstein, Carpenter, & Minshew, 2005; Williams et al., 2006).
Theoretical Models of Verbal Working Memory
There are several competing theories of verbal working memory
that may explain differences in performance across memory tasks
and language ability for children with autism. The capacity theory
of working memory (Just & Carpenter, 1992) views working
memory as a temporary storage space or processing arena that is
able to interact with stored declarative and procedural knowledge
in long-term memory for the purpose of language comprehension.
Under this theory, a limited set of processing resources is avail-
able in working memory for the demands of the verbal task,
resulting in decreased comprehension and poor task performance.
Verbal working memory tasks, although speech based, present
different levels of language complexity and require different levels
of cognitive manipulation, placing competing demands on work-
ing memory and its capacity reserve. Within this framework of
working memory, the demands of the information processing task
and the efficiency of the temporary workspace can exceed the
capacity limitations in working memory, leading to a loss of infor-
mation. In this view, children with autism would demonstrate
working memory deficits for more complex verbal material because
of ineffective processing and reduced functional capacity reserve.
An alternative hypothesis to capacity theory that may explain
poor performance on verbal working memory tasks by children
with autism is the possibility that difficulty in the passive temporary
storage of linguistic information itself may contribute to poor ver-
bal working memory performance. The phonological working
memory theory of Baddeley (1986) provides a conceptual model
for this hypothesis with the addition of a phonological loop. Baddeley
conceptualized a model of verbal working memory containing a
central executive control component and a phonological loop. The
central executive was conceptualized as a limited capacity work-
space that is responsible for attention and regulatory control during
human information processing while the phonological loop allows
for the temporary storage of phonological representations. Baddeley
distinguished between two components of the loop: one for the
temporary storage of phonological representations and the other for
the subvocalization and rehearsal of material during verbal memory
tasks. In this view, the phonological loop functions to support the
brief temporary storage of verbal information while allowing a
subvocalization control process to refresh verbal material in the
loop, a source of re-input. Limitations in verbal working memory
performance by children with autism may be due to an inability
either to instantiate adequate phonological representation of the
stimulus material or to apply subvocal rehearsal strategies to aide
in the maintenance of the phonological representation.
Finally, a connectionist theory of verbal working memory
and language processing (MacDonald & Christiansen, 2002) may
explain the challenges faced by children with autism. The influence
of previous experience during performance of working memory
tasks is a central feature of the theory of connectionism, and this
approach to language processing and working memory diverges
significantly from each of the previous accounts of the structure
500 LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008
and function of verbal working memory. In connectionist theory,
language processing is the product of verbal input passing though
a biological network, a multilayer system of nodes or stored in-
formational knowledge clusters. Language comprehension is the
result of the passing of activation through the multilayer network
as connections are made between and among nodes. In the con-
nectionist account, language knowledge is not separated from the
locus of processing; it is not relegated to a separate storage arena
for accessing structure, function, and meaning. Likewise, work-
ing memory is not a separate construct of human information
processing or an independent computational storage space, but the
network itself. In this view, knowledge and experience play a
significant role in that network efficiency and outcome is enhanced
by experiencing similar input. Under this theory, verbal working
memory deficits for children with autism would be influenced by
the complexity of language input, the level of complexity or sophis-
tication of the network itself, and the interaction of the component
properties of the network. Table 1 summarizes the key features
of each of these theoretical accounts of verbal working memory
relative to the structure, function, and role of experience.
Purpose of the Study
The primary purpose of this study was to measure the verbal
working memory ability of verbal school-age children with autism
using standardized memory tasks chosen to reflect increasingly
more complex levels of language and processing demands—
nonword repetition, memory for digits span (forward), and sentence
imitation—and to compare findings to an age-matched TD con-
trol group. A secondary purpose was to examine the relationship
between performance on verbal working memory tasks and lan-
guage processing ability by children with autism using a language
task that places significant cognitive–linguistic processing demands
on verbal working memory—a task of story recall—and to compare
findings to the TD control group. It was hypothesized that per-
formance for the verbal school-age children with autism on vary-
ing complexity levels of verbal working memory tasks would be
deficient relative to that of TD children, and that performance on
a story recall task, which is a complex language processing mea-
sure, would be related to performance on complex tasks of working
memory by the group with autism. The following research questions
were asked:
&Do verbal children with autism demonstrate poor performance
and a specific pattern of verbal working memory function on
varying verbal memory tasks relative to TD children?
&Is there a significant difference in performance on a task of
story recall between verbal school-age children with autism
and age-matched TD children?
&Is story recall related to performance on more complex verbal
working memory tasks by children with autism?
METHOD
Participants
The study included 15 children with autism, 13 boys and 2 girls,
between the ages of 5;0 and 7;11, and 10 age-matched, TD children
drawn from a larger, ongoing study of language and cognitive–
linguistic processing ability in verbal children with autism. The in-
clusion in the study of more boys than girls with autism reflects the
finding of increased prevalence of autism in males (Volkmar, Szatmari,
& Sparrow, 1993). More boys than girls were recruited from study
postings and were more likely to meet the entry criteria of the study.
The group with autism was recruited through parent association
and advocacy Web sites, newspaper notices, and word of mouth.
Criteria for entrance in the group with autism were (a) diagnosis
of autism by an expert clinician using at least one of the follow-
ing: the Autism Diagnostic Interview—Revised (ADI–R; Lord,
Rutter, & Le Couteur, 1994), the Autism Diagnostic Observation
Schedule—Generic (ADOS; Lord et al., 2000), or the criteria from
The Diagnostic and Statistical Manual of Mental Disorders, Fourth
Edition (APA, 1994); (b) clinical and educational history of sig-
nificant social, behavioral, and communicative impairment; (c)
current enrollment in special education services; (d) chronological
age between 5;0 and 7;11; (e) verbal ability at the phrase or sentence
level; and (f ) nonverbal intelligence, measured using the Differ-
ential Ability Scale (DAS; Elliott, 1983), ≥70 (standard score).
TD children between 5;0 and 7;11 were recruited from posted notices
Table 1. Theoretical accounts of verbal working memory.
Theory Structure Function Experience
Capacity theory (Just &
Carpenter, 1992)
&Temporary storage space &Constrains comprehension &Not central
&Processing arena &Interacts with stored knowledge
in long-term memory
Phonological loop theory
(Baddeley, 1986)
&Central executive &Central executive provides regulatory
control over a phonological loop
&Not central
&Phonological loop with
rehearsal buffer
&Temporary storage and rehearsal
of phonological sequences
&Supports language processing
and learning
Connectionist theory
(MacDonald &
Christiansen, 2002)
&Multilayer neurobiological
network
&Entire network contributes
to processing
&Central: prior experience
contributes to mapping
and efficiency of the network
&Working memory is the network
itself, not a separate storage area
&Network efficiency and interaction
contribute to language comprehension
Smith Gabig: Working Memory and Story Retelling 501
and word of mouth. None of the TD children had a history of speech-
language delay, hearing loss, or special education services, or a sib-
ling with a diagnosis of autism. Parent/guardian consent, as required
for human subject protection, was granted for all participants.
Parents of both groups of children were asked to complete
the Social Communication Questionnaire (SCQ), Lifetime Form
(Rutter, Bailey, & Lord, 2003), in order to screen the children for
autism and/or confirm the diagnosis of autism before initiation of
further testing. The SCQ focuses on the child’s developmental
history, with particular attention to the ages between 4 and 5 years.
The SCQ provides a total score that is interpreted in relation to a
specific cutoff point that is used to identify individuals who may
have autism. Fourteen of the 15 children with autism scored above
the recommended cutoff score of 15 cited in the SCQ manual;
1 child with autism received a score of 14 but was included in
the study because she had a diagnosis of autism as well as sig-
nificant clinical and educational histories of social, language, and
communication delays and special education services. The group
with autism had a mean SCQ score of 22; the control group had a
mean mean SCQ score of 1.7. See Table 2 for a summary of the
group’s demographic characteristics.
In this study, age-matched TD controls were used for two
reasons. First, the selection of a comparison group in autism re-
search has historically included a language-age-matched control
group, yielding critical information on communication behaviors
that are unique to children with autism ( Tager-Flusberg, 2004).
Although this line of research has proven to be very significant,
Tager-Flusberg recently proposed that future research in autism
embrace the study of within-group variation and performance pro-
files of selected subgroups of children rather than identify and
categorize the language or communication features that distinguish
children with autism from control groups matched on language.
Second, there is an increased focus on the inclusion of children with
autism within the context of the general education classroom in
order to facilitate access to the curriculum and as an essential fea-
ture of intervention and special education (Individuals With Dis-
abilities Education Improvement Act [IDEA], 2004; Simpson,
de Boer-Ott, & Smith-Myles, 2003). Children who are placed in
age-appropriate inclusive classrooms are challenged to acquire
the language of school and negotiate the academic language pro-
ficiency necessary to become literate (Wilkinson & Silliman, 2000).
It is important to investigate how children with autism perform
relative to their same-age peers, especially in the area of verbal
working memory, because this is the peer group of social and
academic interaction for many verbal children with autism.
Nine of the 15 children with autism attended public schools
and were participating in inclusion classrooms for all or some of
the academic day. The other 6 children attended private schools
for children with developmental disabilities, including pervasive
development disorder. Seven of the 10 age-matched controls attended
public schools; the other 3 attended a private religious school.
Participants in both groups received an assessment of nonverbal
intelligence, receptive vocabulary, and word articulation ability.
Nonverbal intelligence was assessed using the DAS, receptive
vocabulary was assessed using the PPVT–III, and single-word
articulation ability was assessed using the Word Articulation (WA)
subtest of the Test of Language Development—Primary, Third
Edition (TOLD–P:3; Newcomer & Hammill, 1997). Group com-
parisons for nonverbal intelligence revealed a significantly lower
mean standard score for children with autism than for TD children.
Children with autism demonstrated an average nonverbal intelli-
gence quotient (NVIQ) of 95 (standard score, range 79–109),
whereas the average NVIQ for the age-matched TD group was
106 (range 91–120), t(23) = 3.01, p< .01, d= 1.28. Although
the group mean score for nonverbal intelligence was significantly
lower for the group with autism, the majority of the children
(n= 13, 87%) received an NVIQ score >85, which is within the
average expected performance range for chronological age. Two
of the children with autism scored <85; 1 received an NVIQ of 83
and the other of 79. These children were included in the study
because the scores, although considered below average, do not fall
below an NVIQ of 70, which is a cutoff score that is considered to
place the individual within the diagnostic category of learning
impaired (American Association on Intellectual and Developmen-
tal Disabilities, 2007). In addition, researchers in autism frequently
make a distinction between low-functioning ( IQ < 70) and high-
functioning (IQ > 70) autism as a basis for subgroups with the autism
diagnostic category (Tsatsanis, 2005). All of the children in this
study would be considered high functioning based on an NVIQ
cutoff of >70 and the presence of intentional speech.
Between-group differences were also seen in the mean stan-
dard score for receptive vocabulary and for word articulation.
The children with autism had a mean standard score for receptive
vocabulary on the PPVT–III of 75 (range 51–94) compared to
Table 2. Participant demographic characteristics.
Variable
Children with autism
(n= 15)
Typically developing children
(n= 10)
MSD MSDtdfp
Age (years;months) 6;6 8.42 months 6;8 10.7 mo. 1.1 23 .479
NVIQ 95 10.6 106 9.5 3.0 23 <.01
PPVT–III 75 14.4 103 10.9 5.1 23 <.01
WA 8.1 1.5 9.9 1.9 2.5 23 <.05
SCQ 22 5.9 1.7 1.5 10.2 23 <.01
Note. NVIQ = nonverbal intelligence score on the Differential Abilities Scales (Elliot, 1983), PPVT = Peabody Picture Vocabulary
Test—III (Dunn & Dunn, 1997), WA = Word Articulation subtest of the Test of Language Development—Primary, Third Edition
(TOLD–P:3; Newcomer & Hammill,1997), SCQ = Social Communication Questionnaire (Rutter, Bailey & Lord, 2003).
502 LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008
103 (range 89–120) for the TD control group, t(23) = 5.1, p< .01,
d= 2.17. Word articulation, as measured by the WA subtest of
the TOLD–P, revealed a mean WA standard score of 8.1 (range
6–11) for children with autism and 9.9 (range 7–12) for TD chil-
dren, t(23) = 2.55, p< .05, d= 1.06. Approximately one half (46%,
n= 7) of the children with autism received a standard score on
the WA subtest within the average range of performance (e.g.,
standard score > 7); the remaining children (53%, n=8)scored
1SD below the expected standard score of 10 (standard score = 7).
Examination of articulation errors made by the children with autism
who scored below average on the WA subtest indicated that the
majority of the errors were substitution of w/r, distortion of vo-
calic –r, substitution of f /th, or cluster reduction on medial context
of –sk blend. The degree of overall intelligibility of the children
with autism was judged by the investigator to be very good, with
speech sound errors noted only occasionally in continuous speech.
Measures of Verbal Working Memory
Three measures of verbal working memory were chosen to reflect
a hierarchy of levels of complexity of the to-be-remembered mate-
rial: nonword repetition (NWR), memory for digits span (MD), and
sentence imitation (SI). Based on the work of Fein et al. (1996) and
Williams et al. (2006), it was hypothesized that children with autism
would find the task of NWR the least difficult, followed by MD as
more difficult and SI as most difficult. NWR requires the individual to
encode the phonological representation of a nonsense word and
immediately repeat it. MD requires the encoding of phonological
and semantic representations as well as more organizational over-
sight in short-term memory because a series of numbers must be
temporarily stored and recalled in sequence. Finally, SI requires
the encoding of phonological and semantic representations of words
as well as the construction of phrase and clause structure to aid in
the organization, temporary storage, and recall of the sentence.
For this study, measures of verbal working memory were chosen
from common standardized assessment batteries used by clini-
cians to assess speech and language ability in school-age children.
NWR and MD were assessed using subtests from the Comprehen-
sive Test of Phonological Processing (CTOPP; Wagner, Torgensen,
Rashotte, 1999); SI was measured using the SI subtest of the
TOLD–P.
Measures of Story Retelling
Story retelling was assessed using The Renfrew Bus Story (Bus
Story; Cowley & Glasgow, 1994), which is a test of story recall
and continuous speech. The story has a prepared story script that is
read by the investigator while he or she shows a series of 12 pictures
depicting events within the story narrative. Children are asked to
retell the story with the aid of the supporting pictures. Bus Story
does not measure the recall of story grammar components (Stein,
1982); rather, it was initially developed by Renfrew (1969) as a
norm-referenced measure of continuous speech via a story retelling
format for children between 3;6 and 6;11. In the examiner’s manual,
standard scores are derived for information and utterance length.
Because children in this study were older than the normative sample
population for Bus Story (e.g., 5; 0–7;11), the use of standard scores
for age was not appropriate; therefore, an index of propositions
recalled (IPR), which is a measure of recall complexity, was devised
to measure the intricacy of information that was recalled by each
child. The IPR is the total number of weighted propositions that
were recalled by the child over the total possible weighted pro-
positional content. For example, the examiner’s manual of Bus
Story lists 32 propositions or ideational units, each weighted either
1 or 2 points, for a total possible score of 53 points. A child’s
IPR would be the total number of weighted propositions that he
or she recalled divided by 53. The second measure of narrative
discourse was longest utterance length (LUL), which is a measure
of continuous oral language ability that is described in the Bus Story
manual. It is calculated as the average of the five longest utter-
ances used by the child during the narrative retelling.
Procedure
Participants were seen individually for one to two sessions in
their home or school in a quiet room away from everyday distrac-
tions. Session 1 focused on the assessment of nonverbal intelli-
gence, receptive vocabulary, and word articulation. During Session 2,
the story retelling task and verbal working memory tasks were
given and were audio-recorded for later transcription and scoring.
Presentation of working memory tasks was counterbalanced to
control for any possible order bias. Presentation and scoring of the
three measures of working memory followed the procedures out-
lined in the examiner’s manuals for the CTOPP and the TOLD–P.
NWRs were scored as incorrect if the responses contained any
syllable or phoneme omission, substitutions, or additions. As per
scoring instructions, misarticulations were disregarded in scoring
for SI and MD, where repetition of word and digit order is the
defining feature in scoring. Tangible reinforcements such as stickers
and frequent breaks between tasks were used with some of the
children with autism and younger TD children in order to facilitate
attention and motivation to complete the testing protocol. The major-
ity of the children easily completed each task presented with little
interruption.
Reliability in Scoring
Interjudge agreement in scoring was calculated for responses
during working memory tasks and for transcription and scoring
of propositional content and length of utterance during the story
recall task. The examiner randomly selected 25% of individual
audio recordings of narrative and response-to-memory tasks from
each group for scoring by a second listener. Percentage of agree-
ment between the two listeners for accuracy in transcription,
scoring for story recall, and scoring for memory task responses
was >90% for each measure.
RESULTS
Working Memory: Group Statistics
Group statistics and t-test results for verbal working memory
tasks and story recall are presented in Table 3. Verbal working
memory was measured using standardized tasks from commercial
tests including NWR, MD, and SI. For each of these tasks, the
expected standard score for age is 10, with an SD of ±3. Children
with autism had significant difficulty with all three measures of
verbal working memory, scoring 1 or more SDs below performance
expectations for chronological age. As anticipated, TD children
Smith Gabig: Working Memory and Story Retelling 503
received standard scores on each of the memory tasks that were
consistent with the expected mean score for age.
For children with autism, the mean standard score for NWR
was 7.2, which was significantly lower than the mean standard
score of 9.6 for age-matched TD children, t(23) = 3.35, p<.01,
d= 1.43. Children with autism had a mean score of 5.5 on MD
compared to a mean score of 9.5 for the TD children, t(23) = 5.4,
p<.01,d= 2.3, and a standard score of 3.0 compared to 10.6 on
the SI, t(23) = 8.79, p< .01, d=3.7.
A pattern of performance across the three measures of verbal
working memory was seen in the group with autism, with some
measures appearing more difficult than others. NWR was below
average for age, but children with autism scored higher on this
measure of verbal working memory than on the other two measures,
MD or SI. Examination of individual performance profiles on the
NWR task for children with autism indicated that 33% (n=5)ofthe
group scored within the average range of performance on this task
(e.g., SD > 8), overlapping with scores achieved by the TD group.
The mean standard score for MD was 5.5, more than 1 SD below the
expected standard score of 10, with 13% (n=2)ofthegroupwith
autism scoring within the average range. The mean standard score for
SI was 3.0, more than 2 SDs below the expected mean score, and
none of the children with autism scored similarly to the TD group.
It appears that the more complex verbal material that was contained in
the processing of digits and sentences proved to be more difficult
for the group with autism. Post hoc paired-samples ttests of verbal
working memory task performance by the group with autism con-
firmed the hypothesis of the complexityof the to-be-remembered ma-
terial. Figure 1 illustrates the finding that for children with autism,
NWR was less difficult than MD, t(23) = 2.66, p< .05. d= 1.2, and
MD was less difficult than SI, t(23) = 2.46, p< .01, d= 1.3. Each
pair-wise comparison was significant, with very large effect sizes
(Cohen, 1988).
Influences of Vocabulary and Articulation
on Verbal Working Memory Task Performance
The possibility that receptive vocabulary and speech articulation
may have influenced the autism group’s performance on the verbal
working memory and story retelling tasks was examined though
correlation analyses. Table 4 contains correlations for receptive
vocabulary and speech articulation with the three verbal working
memory tasks for both groups of children. For the TD group of
children, receptive vocabulary was not related to performance on
any of the verbal working memory tasks. For children with autism,
receptive vocabulary was strongly associated with SI (r= .71,
p< .01), but no relationship was found between receptive vocab-
ulary and NWR or MD. WA ability was not related to any of the
measures of verbal working memory for children with autism, but a
moderate relationship between WA and SI was found for the TD
children (r= .68, p< .05).
Story retelling. Story retelling was measured for the IPR and
LUL. As seen in Table 3, children with autism had an IPR of .26,
meaning that they recalled on average 26% of the story’s proposi-
tional units. Age-matched TD controls had an IPR of .68, mean-
ing that they recalled 68% of the story’s propositional units, which
is a significant between-groups difference, t(23) = 7.08, p< .01,
d= 3.0. Average LUL for children with autism was 5.1 words
compared to 13.5 words for TD children, which is also a significant
difference, t(23) = 8.14, p< .01, d=3.5.Childrenwithautism
recalled significantly fewer propositional units from the story than
did the TD children and they encoded these units in shorter phrases
Table 3. Group statistics: Working memory and story recall.
Variable
Children with autism
(n= 15)
Typically developing children
(n= 10)
MSD M SD t dfp
Working memory
NWR 7.2 .94 9.6 2.5 3.35 23 <.01
MD 5.5 1.68 9.5 1.95 5.4 23 <.01
SI 3.0 2.08 10.6 2.11 8.79 23 <.01
Story recall
IPR .26 .16 .68 .11 7.08 23 <.01
LUL 5.1 2.11 13.5 3.05 8.14 23 <.01
Note. NWR = nonword repetition, MD = memory for digits span, SI = sentence imitation, all of which have standard scores with a
mean of 10. IPR = index of propositions recalled, LUL = longest utterance length.
Figure 1. Means for verbal working memory tasks by group.
504 LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008
and sentences. Age-matched TD controls recalled approximately
two thirds of the story’s propositional content and encoded the
content using longer utterances. Because story retelling has a
significant language processing and verbal memory component,
the effect of vocabulary on the two measures of story recall was
considered for each group. As can be seen in Table 4, receptive
vocabulary was strongly related to both IPR (r= .64, p< .05)
and LUL (r= .67, p< .01) for children with autism, suggesting that
the processing and recall of story propositions and utterance length
was related to stored vocabulary knowledge for the group with
autism. No relationship was found between vocabulary knowledge
and story recall for the TD children.
Verbal working memory and story retelling. Correlations be-
tween the three tasks of verbal working memory and the two mea-
sures of story retelling are presented for both groups of children
in Table 5. None of the measures of verbal working memory were
associated with the measures of storytelling for TD children, which
is an expected finding because essential language competence
and developmentally appropriate working memory function, as
measured by standard score for age, are expected by TD children
on standardized measures of working memory.
For children with autism, two measures of verbal working mem-
ory, MD and SI, were positively related to IPR and to LUL during
the story recall task. MD was moderately related to IPR (r= .53,
p< .05) and to LUL (r= .51, p< .05) during story retelling. SI was
strongly related to IPR (r= .70, p< .01) and moderately related to
LUL (r= .58, p< .05). Because vocabulary scores on the PPVT
were previously reported to be related to SI for children with autism
(r= .71, p< .01), partial correlations between SI and each measure
of story retelling were calculated, controlling for the effects of
vocabulary. When the influence of receptive vocabulary (RVOC)
on SI, IPR, and LUL were controlled, the relationship between SI
and story measures (IPR and LUL) ceased to exist for the children
with autism, suggesting that underlying vocabulary knowledge
and language processing affect performance on both a task of SI and
a task of story retelling for children with autism, r(SI, IPR.RVOC) =
.36, p= .21; r(SI, LUL.RVOC) = .17, p= .54.
DISCUSSION
One purpose of this study was to examine verbal working
memory in verbal school-age children with autism and an age-
matched TD control group because this aspect of cognitive pro-
cessing has been linked to language development and learning
in children. Of particular interest was the level of complexity of
material that was used in the verbal working memory tasks for
children. It was hypothesized that complexity of task in the mea-
surement of verbal working memory would influence performance
by the children with autism but not the TD children, and that
more complex verbal working memory task demands would be
associated with performance on a story recall task. Children with
autism scored significantly below their age-matched cohorts on all
measures of verbal working memory (NWR, MD, and SI), and
significant within-group variability for verbal working memory
task performance was noted. Paired comparisons indicated that a
hierarchy of memory task complexity influenced performance
outcomes by children with autism but not TD children. Although
scoring below TD children, the group with autism clearly per-
formed better on NWR than on the working memory tasks of MD or
SI. A hierarchy of performance on verbal working memory tasks
by children with autism is consistent with previously reported re-
search on verbal working memory function in autism (Fein et al.,
1996; Williams et al., 2006). These results are in agreement with
a pattern of escalating memory deficits with increasing task com-
plexity reported in older children and adults with autism (Minshew
& Goldstein, 2001).
The possibility that underlying linguistic constraints in speech
articulation or receptive vocabulary knowledge were associated
with performance on verbal working memory tasks by children with
autism was considered. Speech articulation, as measured by a WA
test, was not related to any of the verbal working memory tasks
for the group with autism, although a strong positive relationship
was seen for the TD group on WA and SI. The correlation between
WA and SI for TD children may be spurious; replication of this
finding in a larger group of children is needed. For children with
autism, WA did not contribute to performance on any verbal
working memory task.
Receptive vocabulary was not related to either NWR or MD
for either group of children. Unlike in previous studies with younger
TD language learners and children with specific language impair-
ment (Gathercole, Willis, Emslie, & Baddeley, 1992; Gray, 2006),
receptive vocabulary was not related to NWR for children with
autism or TD children between 5;0 and 7;11. However, performance
on longer cognitive–linguistic sequences evident in tasks of SI and
Table 4. Pearson product–moment correlations for receptive vocabulary
as measured by the PPVT and word articulation as measured by the
WA subtest of the TOLD–P:3 with verbal working memory and story
recall.
Variable
Children with autism
(n= 15)
Typically developing
children (n= 10)
PPVT WA PPVT WA
NWR .20 –.05 .34 .60
MD .47 –.20 –.10 .31
SI .71* –.01 .40 .68*
IPR .64* –.11 .55 .42
LUL .67** .01 .57 –.25
*p< 0.05. **p< .01.
Table 5. Pearson product–moment correlations for verbal working
memory and story recall.
Variable
Children with
autism (n= 15)
Typically developing
children (n= 10)
IPR LUL IPR LUL
NWR –.18 –.05 .01 –.26
MD .53* .51* –.09 –.10
SI .70** .58* .13 –.21
*p< 0.05. **p< .01.
Smith Gabig: Working Memory and Story Retelling 505
story retelling was associated with existing vocabulary levels for
the group with autism.
Cognitive–Linguistic Complexity of Verbal
Working Memory Tasks
Why do children with autism perform better on the NWR task
than on the MD or SI task? Children with autism scored 1 SD below
average on NWR, a finding that is consistent with a previously
reported investigation of NWR in children with autism (Kjelgaard
& Tager-Flusberg, 2001). However, post hoc analysis indicated
that performance on NWR was significantly better than on other
measures of verbal working memory for the group with autism.
One possible explanation is that performance on the three verbal
memory tasks was affected by the processing demands of the task
and how well the child could perform the simultaneous processing
of several tacit cognitive–linguistic processes.
The phonological storage capacity hypothesis (Gathercole,
2006; Gathercole & Baddeley, 1990) asserts that phonological
memory is a causal factor in word learning and vocabulary de-
velopment and that deficits in phonological storage capacity can
explain language learning disorders in children with language
impairment. According to this hypothesis, NWR is a particularly
sensitive measure of immediate verbal working memory, the
phonological loop, because it does not involve or it minimizes
the possible contribution of stored lexical representations in long-
term memory, thus manifesting a pure measure of phonological
storage capacity. Because the nonsense word is unfamiliar, the
individual must rely on his or her individual phonological storage
capacity to encode the phonological form and hold it long enough
so that it can be sent to the articulatory control system for speech
output. According to this theoretical framework, deficits in NWR
can be attributed to limitations in storage capacity for phonological
representations and to decay as word length increases. By exten-
sion, increasingly poorer performance in verbal memory should
be seen with tasks requiring the storage of longer linguistic stim-
uli, as was noted in this study, with an obvious declination in
performance relative to increasingly longer and more complex
language contained in the serial recall of digits and sentence
imitation.
Recent challenges to a strict interpretation of a phonological
storage capacity as a quantity-limited storage capacity with no
or minimal interaction with stored lexical representations have
emerged (Bowey, 1997; Chiat, 2006; Ellis Weismer & Edwards,
2006; Metsala & Walley, 1998; Montgomery & Evans, 2006).
There is ample support to the interpretation that NWR is in-
fluenced by the quality of the phonological representations in
working memory, and the quality of these representations is
constrained by underlying phonological processing ability as
well as factors related to working memory capacity (Chiat,
2006; Ellis Weismer & Edwards, 2006; Montgomery & Evans,
2006).
Researchers have noted that there are significant cognitive
factors involved in NWR in addition to phonological storage, in-
cluding accurate input processes (e.g., speech perception, discrim-
ination), phonological processing of the phonological form of the
nonword (e.g., construction of the phonological representation, seg-
mentation of the phonological representation), and output pro-
cesses (e.g., retrieval of articulatory speech production plan, speech
production output) (Bowey, 1997; Chiat, 2006; Montgomery, 1995;
Snowling, Chiat, & Hulme, 1991). An alternative explanation of
deficits in NWR is the phonological sensitivity hypothesis (Ellis
Weismer & Edwards, 2006; Munson, Edwards, & Beckman, 2005).
This theoretical framework asserts that the robust construction of
phonological representations for speech analysis and speech motor
output is necessary in order to repeat nonsense words accurately,
and that the ability to assemble phonological representations de-
velops gradually in children through interaction with existing
vocabulary knowledge (Edwards, Beckman, & Munson, 2004;
Munson et al., 2005). A growing body of literature suggests that
children’s repetition of nonwords is significantly related to the
phonological properties of words in their lexicon, and that the struc-
ture of spoken word representations in memory gradually becomes
increasingly segmental and more robust with increased vocabulary
development, a process referred to as lexical restructuring (Metsala
& Walley, 1998). Evidence for this theoretical framework is seen
in studies that demonstrate that TD children are sensitive to the
phonotactic probability of nonwords (Edwards et al., 2004; Munson
et al., 2005). Phonotactic probability refers to the likelihood that
sublexical sequences of sounds may occur in a lexical item and is
related to stored phonological representations and abstractions of
lexemes in the lexicon. As children’s vocabulary increases, their
stored knowledge of possible phonetic sequences becomes more
robust and defined, facilitating the perceptual parsing of novel words
or nonwords. In this view, vocabulary size would be associated with
NWR. Children with smaller vocabularies, such as younger TD
children or children with language deficits, have fewer phonological
representations that are context sensitive; therefore, repetition of
longer nonwords would be less accurate due to weaker or fewer
phonological representations in the lexicon.
In the present study, the children with autism had reduced re-
ceptive vocabulary size relative to their age-matched cohorts, yet
vocabulary was not related to performance on NWR for the group
with autism. However, children with autism were able to construct
adequate phonological representations for some of the nonword
stimuli in the task. Examination of individual protocols on the
NWR task by the children with autism revealed that all of the chil-
dren were able to repeat nonsense words containing one and two
syllables, but only 37% of the group was able to repeat three- and
four-syllable nonwords without phoneme substitution or syllable
reduction. These findings are similar to what has been reported for
NWR in children with specific language impairment (Ellis Weismer
et al., 1999). All of the TD groups were able to parse, hold, and
repeat three-syllable and some four-syllable nonwords without
phoneme substitution or syllable reduction; only 20% were able
to repeat five-syllable nonword stimuli. This study did not examine
the phonotactic probability of the nonword stimuli used in the
study; therefore, the interaction between vocabulary size and re-
petition of phonotactic probable sequences cannot be determined. In
the present study, the children with autism had reduced receptive
vocabulary knowledge relative to their age-matched cohorts and
were less capable of repeating longer nonwords containing more
syllables and therefore additional phonetic sequences, suggesting
the possibility of fewer or less robust stored phonological represen-
tations for longer phonetic sequences available for processing of
longer nonwords.
Longer sequences of lexically based material in working mem-
ory provide even more of a challenge for children with autism.
Performance on the MD task was poorer relative to performance
on the NWR task and also was independent of stored receptive
506 LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008
vocabulary levels or WA in the children. SI was the most challenging
verbal memory task for the group with autism and was strongly
associated with existing vocabulary knowledge but not WA ability.
The TD children showed the opposite relationship: WA ability was
positively associated with SI but not existing vocabulary levels,
suggesting that the encoding and motor speech execution of pho-
nological representations may assist the TD children in their per-
formance on a task of SI.
MD and SI tasks are both lexically based and require similar
cognitive and linguistic processes, including increased oversight
and organization in working memory and language input/output
processes affecting acoustic phonetic processing, phonological
representation, phonological segmentation, retrieval of semantic
representations, speech motor planning, and speech articulation.
The construction and maintenance of sequential phonological
representations as well as the retrieval of stored phonological and
semantic representations from long-term memory contribute to
the cognitive–linguistic processing demands of each of the tasks.
However, MD differs from SI in that it is a serial recall task, a
measure of sequential processing ability in memory that is thought
to be confounded by attention or regulatory control (Mirsky,
Anthony, Duncan, Ahearn, & Kellam, 1991). In this study, vo-
cabulary knowledge was not significantly associated with MD
for the group with autism, suggesting that retrieval of semantic
representations contributed less to MD task performance than did
the possible influence of a number of other cognitive–linguistic
factors including poor executive regulatory control, incomplete
or weak constructions of phonological representations, deficient
subvocalization control process to refresh verbal material, or
impaired output processes involving motor planning for speech
production. Future research is needed to examine this area more
closely in order to determine the cognitive–linguistic factors that
are associated with performance on an MD task for children with
autism.
SI is a more complex verbal working memory task than MD or
NWR because it involves greater demands on cognitive–linguistic
processing with the encoding of meaningful lexical sequences
within clause structures and the assignment of syntactic and seman-
tic structure to provide an organizational schema to sentences
(Caplan & Waters, 1999). In this study, vocabulary knowledge
was significantly related to SI for children with autism; WA ability
was not related to SI. This result suggests that performance on an
SI task by the group with autism is related to existing vocabulary
knowledge, perhaps due to reduced vocabulary size or to weak
or inadequate phonologic and semantic representations necessary
for the rapid and robust construction of meaningful representations
in verbal working memory.
An additional underlying linguistic factor that could be relevant
to performance on an SI task for children with autism may be gram-
matical understanding; however, sentence comprehension was
not measured in the present study. It is reasonable to assume that
SI requires the simultaneous processing of multiple layers of lin-
guistic representation not germane to the performance of NWR
and MD, and that the language processing demands of the SI
task influenced performance of the working memory capacity
for the group with autism during this task. The evidence from
this research indicates the need to clarify the influence and re-
lative contribution of vocabulary knowledge and grammatical
understanding to performance on an SI task for children with
autism.
Verbal Working Memory Tasks and Story Retelling
A secondary goal of this research was to examine the interaction
of verbal working memory and story retelling in children with
autism with the supposition that more complex verbal working
memory tasks would be associated with story recall. As noted
earlier, story recall places significant demands on verbal working
memory and involves the simultaneous processing of multiple
levels of linguistic knowledge, cognitive organization, and regula-
tory control. As the story unfolds, a child must be able to quickly
retrieve semantic representations for vocabulary within sentence
frames, assign syntactic and semantic organizational structure, and
relate these meanings to previous and subsequent sentences and
ideational units. Story recall was significantly impaired for the
children with autism compared to age-matched TD children for both
IPR and LUL. The performance by children with autism on the
SI task, argued to be the most complex verbal working memory
task, was strongly associated with IPR and moderately related to
LUL during story recall, suggesting a similarity in cognitive–
linguistic factors associated with working memory capacity for
SI and story retelling. MD was moderately related to measures of
story recall for this group of children, also suggesting a similarity
in cognitive factors associated with organization, attention, and
serial recall between the two tasks. However, the relationship be-
tween performance on the tasks of SI and story recall appears to be
strongly influenced by stored receptive vocabulary, suggesting that
both are measures of simultaneous language processing and lin-
guistic competence and not limitations in working memory capacity
per se. The underlying cognitive–linguistic requirements are simi-
lar for both SI and story retelling in that both require the simultaneous
processing and construction of representations of multiple layers
of language, including phonologic, semantic, and syntactic.
Story retelling also requires the ability to construct higher or-
der representations related to the organizational schema of story
grammar as well as the additional conception of story characters
and their mental states—the ability to grasp the intentions, feel-
ings, and beliefs of others, a notion referred to as theory of mind
(Premack & Woodruff, 1978). The development of a theory of
mind is considered to be the result of metarepresentational develop-
ment or the capacity to form second-order representations of the
thoughts and feelings of others (Baron-Cohen, Leslie, & Frith,
1985). Poor performance on a story retelling task by children with
autism may be strongly influenced by the simultaneous construc-
tion and processing of multiple layers of both primary language
representations and organizational story structure, as well as the
concurrent construction and processing of second-order metare-
presentations of the thoughts, beliefs, intentions, and motivations
of the story characters. Theory of mind is linked theoretically to
social interaction and the ability to take another person’s point
of view. Researchers have demonstrated that children with autism
show a specific deficit in social perspective taking and fail to ap-
preciate another person’s point of view—a failure to employ a
theory of mind (Baron-Cohen et al., 1985). Difficulty with story
retelling may be the result of absent or inadequate primary linguis-
tic representations as well as absent or inadequate second-order
representations needed to simultaneously process and recall the
story and not due to limitations in working memory capacity.
Limitations in simultaneous processing have also been offered for
verbal working memory deficits seen in children with specific
language impairment (Marton, 2006; Marton & Schwartz, 2003).
Smith Gabig: Working Memory and Story Retelling 507
It appears that children with autism have significant difficulty
in tasks that require the parallel construction and processing of
complex linguistic and cognitive representations to support their
verbal working memory performance.
Limitations
There are several limitations to the design of this study. One
limitation involves the unequal number of children in the two groups,
with only 10 age-matched TD children in the control group and
15 children in the group with autism. Because the tasks measuring
verbal working memory were chosen from norm-referenced stan-
dardized tests, the normative sample of children used in the stan-
dardized measures also served as an inherent control group for the
comparison of standard scores achieved by children with autism
on the memory tasks. However, the small number of children in the
TD group may have influenced the correlation analyses, as was seen
in the possibly spurious positive relationship between WA and SI
for the TD group of children. Another limitation of the study is the
use of standardized tasks to assess verbal working memory in children.
Although the use of standardized measures allows for examination
of variation in performance between the group with autism and
age-matched TD children via comparison of standard scores, it does
not permit careful examination of performance variability based on
internal item construction, as in experimentally designed stimuli.
Similarly, a limitation is seen in a lack of measures of grammat-
ical understanding and social-pragmatic function that may contrib-
ute to sentence imitation and story retelling, both requiring the
processing of longer linguistic sequences within syntactic frames. It
is likely that the ability to construct an adequate representation of
a sentence during an SI task is causally related to the ability to un-
derstand similar sentence types and clause structures. Likewise, as
noted above, an inherent lack of social perceptive taking and fail-
ure to employ a theory of mind by children with autism may also
have influenced the children’s ability to construct higher order repre-
sentations of the mental states and motivations of the story characters
necessary to gain meaning from the story. Future research should
focus on the effect of existing language knowledge and social-
conceptual perspective taking on variations in verbal working memory
performance by children with autism in order to better understand
their specific contributions to complex language processing.
Prior investigations have noted attention and regulatory control
deficits in children with autism (Ozonoff, 1997; Ozonoff & Jensen,
1999; Tsatsanis, 2005). Such deficits may have influenced the
performance of the children in this study on the verbal memory
tasks and the story recall task. The investigator was sensitive to
attention, motivation, and regulatory control issues for the children
in the study and provided tangible reinforcement for participation
and completion of items and tasks. Frequent breaks from the
experimental protocol were needed with a few of the children with
autism, but they were able to be engaged in each task. Thus, re-
sults were deemed to be a valid indicator of ability in the areas
assessed for all of the children.
Implications for Future Research
and Clinical Practice
Children with autism demonstrate significant deficits on verbal
working memory tasks related to the cognitive–linguistic processing
demands of the task and to existing language knowledge. These
findings can be a starting point in the design of future studies to
examine the relationship between existing cognitive knowledge and
language ability to the construction of adequate representations
needed for processing by children with autism. Studies designed to
determine how children with autism differ on language processing
and memory tasks from children with other developmental language
delays without autism, such as specific language impairment, may
provide critical information needed to understand the nature of
the disorder as well as guide educational and intervention practices.
Do children with autism demonstrate a separate pattern and severity
level of verbal working memory deficits or can a similar hierarchical
pattern of verbal working memory performance be observed in
other diagnostic classifications of children, suggesting an overlap in
populations with similar cognitive–linguistic profiles? For example,
what is the effect of the phonetic structure of nonword stimuli on
performance by children with autism on an NWR task relative to
children with developmental language impairment without autism?
Is there a similar influence of existing linguistic knowledge such
as vocabulary or grammatical understanding on working memory
task performance across diagnostic classifications of children with
language impairment? Do children with autism perform similarly
to children with specific language impairment on memory tasks
that include the interjection of additional processing requirements?
Such research would provide much needed evidence about the lan-
guage competence and processing ability of children with autism.
The impact of deficits and variations in verbal working memory
performance by verbal children with autism has compelling implica-
tions for language processing within the context of the classroom.
Children are faced with increasingly more complex discourse
processing demands when they enter primary school and as each
grade progresses (Cazden, 1988). Children with autism who are
placed in inclusion classrooms may not be at the same ability level
as their classroom peers in the construction of cognitive–linguistic
representations necessary for processing information. Thus, a care-
ful assessment of the language processing demands of the class-
room, the level of existing language knowledge, and the varying
performance on verbal memory tasks by children is needed in order
to provide the necessary support for many high-functioning children
being educated in inclusion contexts.
Conclusion
There is evidence, from this study and previously cited stud-
ies on verbal working memory in children with autism, that the
working memory capacity of children with autism for complex
verbal tasks is inferior to that of TD children, and that a hierarchy
of performance occurs across tasks for the children with autism.
The question is how best to explain the autism group’s variable
verbal memory deficits. The concept of a temporary storage space
or limited capacity resource for verbal information in working
memory has been used to explain deficits in language processing
and comprehension for children with developmental language
delays and adults with neurogenic brain injury (Gathercole &
Alloway, 2006; Gathercole & Baddeley, 1990). However, results
from this study and from previous research suggest a more multi-
causal explanation for poor performance on verbal memory tasks
than just limitations in temporary storage capacity. An alternative
explanation for poor performance on varying tasks of verbal
working memory may be best described within a connectionist
508 LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008
model of language processing (MacDonald & Christiansen, 2002).
MacDonald and Christiansen questioned the existence of an isolated
construct called working memory capacity that can be measured
by various tasks of verbal working memory. Rather, their claim is
that the distinction that is made between verbal working memory
tasks and language processing tasks is an artificial one, and that
essentially both are measuring language processing competence. In
this view, deficits that are seen in performance on verbal memory
tasks can be attributed to variations in exposure to similar language
processing experiences and to neurobiological differences in pro-
cessing accuracy such as variations in the ability to construct robust
phonological, semantic, and syntactic representations.
As children gain more experience with language, their stored
phonological, semantic, and syntactic representations become
more robust and sophisticated and are therefore available to sup-
port processing capacity for varying levels of cognitive–linguistic
processing demands. For the children with autism in this study, a
hierarchy of performance across verbal memory tasks was observed,
with the children performing progressively worse in a stepwise
declination according to the proposed cognitive–linguistic demands
of the task. The observed poorer performance on SI, a memory
task that is considered to be the most cognitively–linguistically
complex, was associated with story recall, which is a complex lan-
guage processing task that also places significant demands on verbal
working memory. However, the relationship between SI and story
recall was mitigated by existing vocabulary knowledge for the
children with autism. It appears that underlying vocabulary knowl-
edge and an ability to construct robust phonologic, semantic, and
syntactic representations during complex language processing
affects performance on both the SI task and the task of story retelling
for children with autism.Variation in performance across the verbal
working memory tasks and the story retelling task is reflective of
subtle language and cognitive differences in the ability to construct
adequate representations for processing. In this account, knowl-
edge and experience play a significant role in that processing efficiency
and performance outcome are enhanced by exposure to and the
successful processing of similar input.
ACKNOWLEDGMENTS
This research was supported in part by Professional Staff Congress, The
City University of New York Faculty Development Award 60159-34-35
to the author. The author wishes to thank the AHA/Asperger Syndrome
and High Functioning Autism Association, Bethpage, NY; the Cody
Center for Autism, Stony Brook University; The School for Language and
Communication Development, Glenn Cove, NY; and the Eden School, Staten
Island, NY for their assistance in recruitment. Thank you to the families
who opened their homes to the author and participated in this research.
REFERENCES
American Association on Intellectual and Developmental Disabilities.
(2007). Definition of mental retardation. Retrieved June 5, 2007, from
www.aamr.org.
American Psychiatric Association. (1994). Diagnostic and statistical
manual of mental disorders (4th ed.). Washington, DC: Author.
Baddeley, A. D. (1986). Working memory. Oxford, England: Oxford
University Press.
Baddeley, A. D. (2003). Working memory and language: An overview.
Journal of Communication Disorders, 36, 189–208.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic
child have a “theory of mind”?Cognition, 21, 37–46.
Bowey, J. A. (1997). What does nonword repetition measure? A reply to
Gathercole and Baddeley. Journal of Experimental Child Psychology,
67, 295–301.
Caplan, D., & Waters, G. S. (1999). Verbal working memory and sentence
comprehension. Behavioral and Brain Sciences, 22, 77–94.
Cazden, C. (1988). Classroom discourse. Portsmouth, NH: Heinemann.
Chiat, S. (2006). The developmental trajectory of nonword repetition.
Applied Psycholinguistics, 27, 552–556.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Cowan, N., Elliott, E. M., Saults, J. S., Morey, C. C., Mattox, S.,
Hismjatullina, A., & Conway, A. R. (2005). On the capacity of
attention: Its estimation and its role in working memory and cognitive
aptitude. Cognitive Psychology, 51, 42–100.
Cowley, J., & Glasgow, C. (1994). The Renfrew Bus Story (American ed.).
Centreville, DE: The Centreville School.
Dunn, L. M., & Dunn, L. M. (1997). Peabody Picture Vocabulary Test—
III. Circle Pines, MN: AGS.
Edwards, J., Beckman, M. E., & Munson, B. (2004). The interaction
between vocabulary size and phonotactic probability effects on children’s
production accuracy and fluency in nonword repetition. Journal of
Speech, Language, and Hearing Research, 47, 421–436.
Elliot, C. D. (1983). Differential Ability Scale. San Antonia, TX: The
Psychological Corporation.
Ellis Weismer, S., & Edwards, J. (2006). The role of phonological storage
deficits in specific language impairment: A reconsideration. Applied
Psycholinguistics, 27, 556–562.
Ellis Weismer, S., Evans, J., & Hesketh, L. J. (1999). An examination of
verbal working memory capacity in children with specific language
impairment. Journal of Speech, Language, and Hearing Research, 42,
1249–1260.
Fein, D., Dunn, M. A., Allen, D. M., Aram, R., Hall, N., Morris, R.,
& Wilson, B. C. (1996). Neuropsychological and language findings.
In I. Rapin (Ed.), Preschool children with inadequate communication:
Developmental language disorder, autism, low IQ (pp. 123–154).
London: MacKeith Press.
Gathercole, S. E. (2006). Nonword repetition and word learning: The
nature of the relationship. [ Keynote]. Applied Psycholinguistics, 27,
513–543.
Gathercole, S. E., & Alloway, T. P. (2006). Practitioner review: Short-
term and working memory impairments in neurodevelopmental disorders:
Diagnosis and remedial support. Journal of Child Psychology and
Psychiatry, 47, 4–15.
Gathercole, S. E., & Baddeley, A. D. (1990). Phonological memory
deficits in language disordered children: Is there a causal connection?
Journal of Memory and Language, 29, 336–360.
Gathercole, S. E., & Baddeley, A. D. (1993). Working memory and
language. East Sussex, UK: Psychology Press, Ltd.
Gathercole, S. E., Willis, C. S., Baddeley, A. D., & Emslie, H. (1994).
The Children’s Test of Nonword Repetition: A test of phonological
working memory. Memory, 2, 103–127.
Gathercole, S. E., Willis, C. S., Emslie, H., & Baddeley, A. D. (1992).
Phonological memory and vocabulary development during the early school
years: A longitudinal study. Developmental Psychology, 28, 887–898.
Smith Gabig: Working Memory and Story Retelling 509
Goldman, R., & Fristoe, M. (1986). Goldman Fristoe Test of Articulation.
Circle Pines, MN: AGS.
Gray, S. (2006). The relationship between phonological memory, receptive
vocabulary, and fast mapping in young children with specific language
impairment. Journal of Speech, Language, and Hearing Research, 49,
955–969.
Individuals with Disabilities Education Improvement Act of 2004, Pub. L.
No. 108-446 § 118 Stat. 2647 (2004).
Just, M., & Carpenter, P. A. (1992). A capacity theory of comprehension:
Individual differences in working memory. Psychological Review, 99(1),
122–149.
Kjelgaard, M. A., & Tager-Flusberg, H. (2001). An investigation of
language profiles in autism: Implications for genetic subgroups. Language
and Cognitive Processes, 16, 287–308.
Korkman, M., Kirk, U., & Kemp, S. (1998). NEPSY. San Antonio, TX:
The Psychological Corporation.
Lord, C., Risi, S., Lambrecht, L., Cook, E., Leventhal, B., DiLavore,
P.,etal.(2000). The Autism Diagnostic Observation Schedule —Generic:
A standard measure of social and communication deficits associated
with the spectrum of autism. Journal of Autism and Developmental
Disorders, 30, 205–223.
Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism Diagnostic
Interview—Revised. Journal of Autism and Developmental Disorders,
24, 659–685.
Losh, M., & Capps, L. (2003). Narrative ability in high-functioning chil-
dren with autism or Asperger’s syndrome. Journal of Autism and Devel-
opmental Disorders, 23, 239–251.
MacDonald, M. C., & Christiansen, M. H. (2002). Reassessing working
memory: Comment on Just and Carpenter (1992) and Waters and Caplan
(1996). Psychological Review, 109, 35–54.
Marton, K. (2006). Do nonword repetition errors in children with specific
language impairment reflect a weakness in an unidentified skill specific to
nonword repetition or a deficit in simultaneous processing? Applied
Psycholinguistics, 27, 569–573.
Marton, K., & Schwartz, R. G. (2003). Working memory capacity and
language processes in children with specific language impairment.
Journal of Speech and Hearing Research, 46, 1138–1153.
Metsala, J. L., & Walley, A. C. (1998). Spoken vocabulary growth and
the segmental restructuring of lexical representations: Precursors to
phoneme awareness and early reading ability. In J. L. Metsala & L. C.
Ehri (Eds.), Word recognition in beginning literacy (pp. 89–120).
Mahwah, NJ: Erlbaum.
Minshew, N. J., & Goldstein, G. (1998). Autism as a disorder of complex
information processing. Mental Retardation and Developmental Dis-
abilities Research Reviews, 4, 129–136.
Minshew, N. J., & Goldstein, G. (2001). The pattern of intact and im-
paired memory functions in autism. Journal of Child Psychology and
Psychiatry, 42, 1095–1101.
Mirsky, A. F., Anthony, B. J., Duncan, C. C., Ahearn, M. B., &
Kellam, S. G. (1991). Analysis of the elements of attention: A neuro-
psychological approach. Neuropsychological Review, 2, 109–145.
Montgomery, J. W. (1995). Examination of phonological working memory
in specifically language-impaired children. Applied Psycholinguistics, 16,
355–378.
Montgomery, J. W. (2000). Verbal working memory and sentence com-
prehension in children with specific language impairment. Journal of
Speech, Language, and Hearing Research, 43, 293–308.
Montgomery, J., & Evans, J. (2006). Commentary on keynote. Applied
Psycholinguistics, 27, 573–577.
Munson, B., Edwards, J., & Beckman, M. (2005). Relationship between
nonword repetition accuracy and other measures of linguistic develop-
ment in children with specific language impairment. Journal of Speech,
Language, and Hearing Research, 48, 61–78.
Newcomer, P. L., & Hammill, D. D. (1997). Test of Language
Development—Primary, Third Edition. Austin, TX: Pro-Ed.
Norbury, C. F., & Bishop, D. V. M. (2003). Narrative skills of children
with communication impairments. International Journal of Language
and Communication Disorders, 38, 287–313.
Ozonoff, S. (1997). Components of executive function in autism and
other disorders. In J. Russell (Ed.), Autism as an executive disorder
(pp. 179–211). New York: Oxford University Press.
Ozonoff, S., & Jensen, J. (1999). Brief report: Specific executive function
profiles in three neurodevelopmental disorders. Journal of Autism and
Developmental Disorders, 29, 171–177.
Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a
“theory of mind”?Behavioral and Brain Sciences, 4, 515–526.
Renfrew, C. (1969). The Bus Story: A test of continuous speech. Oxford,
UK: Collins & Company, Ltd.
Rice, M. L., Warren, S. F., & Betz, S. K. (2005). Language symptoms of
developmental language disorders: An overview of autism, Down syn-
drome, fragile X, specific language impairment, and Williams syndrome.
Applied Psycholinguistics, 26, 7–27.
Russell, J., Jarrold, C., & Henry, L. (1996). Working memory in children
with autism and with moderate learning difficulties. Journal of Child
Psychology and Psychiatry, 37, 673–686.
Rutter, M., Bailey, A., & Lord, C. (2003). The Social Communication
Questionnaire, Lifetime Form. Los Angeles: Western Psychological
Services.
Semel, E., Wiig, E. H., & Secord, W. A. (1995). Clinical Evaluation of
Language Fundamentals, 3rd ed. San Antonio, TX: The Psychological
Corporation, Harcourt, Brace, & Co.
Sheslow, D., & Adams, W. (1990). Wide Range Assessment of Memory
and Learning. Wilmington, DE: Jastak Associates.
Shriberg, L., Paul, R., McSweeny, J., Klin, A., Cohen, D., & Volkmar,
F. (2001). Speech and prosody characteristics of adolescents and adults
with high-functioning autism and Asperger syndrome. Journal of
Speech, Language, and Hearing Research, 44, 1097–1115.
Simpson, R. L., de Boer-Ott, S. R., & Smith-Myles, B. (2003). Inclusion
of learners with autism spectrum disorders in general education settings.
Topics in Language Disorders, 23, 116–133.
Snowling, M., Chiat, S., & Hulme, C. (1991). Words, nonwords, and
phonological processes: Some comments on Gathercole, Willis, Emslie,
and Baddeley. Applied Psycholinguistics, 12, 369–373.
Stein, N. (1982). What’s in a story: Interpreting the interpretations of story
grammars. Discourse Processes, 5, 319–335.
Stone, W. L., & Caro-Martinez, L. M. (1990). Naturalistic observations
of spontaneous communication in autistic children. Journal of Autism and
Developmental Disorders, 20, 437–453.
Tager-Flusberg, H. (1991). Semantic processing in the free recall of
autistic children: Further evidence of a cognitive deficit. British Journal
of Developmental Psychology, 9, 417–430.
Tager-Flusberg, H. (1995). “Once upon a ribit”: Stories narrated by autistic
children. British Journal of Developmental Psychology, 13, 45–59.
Tager-Flusberg, H. (1996). Current theory and research on language and
communication in autism. Journal of Autism and Developmental Dis-
orders, 26, 169–172.
510 LANGUAGE,SPEECH,AND HEARING SERVICES IN SCHOOLS •Vol. 39 •498–511 •October 2008
Tager-Flusberg, H. (2003). Language impairment in children with com-
plex neurodevelopmental disorders: The case of autism. In Y. Levy
& J. Schaeffer ( Eds.), Language competence across populations
(pp. 297–321). Mahwah, NJ: Erlbaum.
Tager-Flusberg, H. (2004). Strategies for conducting research on language
in autism. Journal of Autism and Developmental Disorders, 34, 75–80.
Tager-Flusberg, H., Paul, R., & Lord, C. (2005). Language and commu-
nication in autism. In D. J. Cohen & F. R. Volkmar (Eds.), Handbook
of autism and pervasive developmental disorders (3rd ed., pp. 335–364).
New York: John Wiley and Sons.
Tsatsanis, K. D. (2005). Neuropsychological characteristics in autism and
related conditions. In D. J. Cohen & F. R. Volkmar ( Eds.), Handbook
of autism and pervasive developmental disorders (3rd ed., pp. 365–381).
New York: John Wiley and Sons.
Volkmar, F. R., Szatmari, P., & Sparrow, S. (1993). Sex differences in
pervasive developmental disorders. Journal of Autism and Develop-
mental Disorders, 23, 579–591.
Wagner, R. K., Torgensen, J. K., & Rashotte, C. A. (1999). Compre-
hensive Test of Phonological Processing. Austin, TX: Pro-Ed.
Wiig, E., Secord, W., & Semel, E. (1992). Clinical Evaluation of Lan-
guage Fundamentals—Preschool. San Antonio, TX: The Psychological
Corporation.
Wilkinson, L. C., & Silliman, E. R. (2000). Classroom language and
literacy learning. In M. L. Kamil, P. E. Mosenthal, P. D. Pearson, &
R. Barr (Eds.), Handbook of reading research, Vol. III (pp. 337–360).
Mahwah, NJ: Erlbaum.
Williams, D., Goldstein, G., Carpenter, P. A., & Minshew, N. J. (2005).
Verbal and spatial working memory in autism. Journal of Autism and
Developmental Disorders, 33, 747–756.
Williams, D., Goldstein, G., & Minshew, N. J. (2006). The profile of
memory function in children with autism. Neuropsychology, 20, 21–29.
Williams, K. T. (1997). Expressive Vocabulary Test. Bloomington: MN:
Pearson.
Received March 26, 2007
Revision received August 4, 2007
Accepted February 5, 2008
DOI: 10.1044/0161-1461(2008/07-0023)
Contact author: Cheryl Smith Gabig, Department of Speech-Language-
Hearing Sciences, Lehman College/CUNY, 250 Bedford Park Boulevard
West, Bronx, NY 10468. E-mail: Cheryl.gabig@lehman.cuny.edu.
Smith Gabig: Working Memory and Story Retelling 511
