Professional Documents
Culture Documents
net/publication/293015455
CITATIONS READS
19 9,710
3 authors, including:
Christine Brennan
University of Colorado Boulder
15 PUBLICATIONS 55 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Distinct benefits given large versus small grain orthographic instruction for English-speaking adults learning to read Russian Cyrillic View project
Phonological grain size selectivity in the human auditory cortex View project
All content following this page was uploaded by Christine Brennan on 11 June 2018.
Few studies have explored the comorbidity of autism and articulation/phonological disorders. In this study,
phonological analyses were performed on eight children with autism. Speech elicitation included object
naming and a connected speech sample. Data analyses involved phonetic inventory, phonological
process analyses, and percentage consonants correct. Results showed that all eight children exhibited
both delayed phonological behavior as well as some atypical phonological processes. Findings also
revealed a strong relationship between severity of phonological behavior and severity of language
impairment. Group comparisons showed that children with moderate–severe language impairment
exhibited more typical and atypical phonological processes than children with mild–moderate language
impairment, and overall correlation between phonological severity and language impairment was high.
The findings contrast with some earlier studies but support and extend recent research showing unusual
phonological trends in some children with autism.
Keywords: Phonological disorder, Speech sounds, Phonological process, Autism
children, articulation skills were spared’ ( p. 287). articulation were not clearly reported, the authors
Shriberg et al. (2001) compared 15 males with high suggested that there was no relationship between
functioning autism (HFA) to 15 males with Asperger language and phonological impairment. They con-
Syndrome (AS) and 53 comparison subjects. In cluded that ‘phonology is not a common area of diffi-
addition to measures of prosody and voice, six culty in children with autism’.
indices of phonological behavior were used, including In contrast, several studies using more detailed pho-
percentage of consonants correct, occurrence of errors nological analyses have suggested more significant
involving diphthongs, and an intelligibility index. speech sound impairment in children with ASD. In
They found no significant differences in severity of one of the first papers to use a detailed phonological
sound errors across the three groups. However, the analysis, Wolk and Edwards (1993) reported a case
mean age of participants in this study (19 years in study of an autistic child (B.D., age 8.3 years) with
the AS and HFA groups) was significantly older both delayed receptive and expressive language skills.
than in many other studies of children with ASD. In Phonetic inventory data showed that all stops, front
addition, more in-depth analysis of phonological pro- nasals, glides, and one liquid /l/ were present in his
cesses was not reported. repertoire. Approximately half of all possible frica-
McCleery et al. (2006) compared the consonant tives, both affricates, velar nasal /ŋ/, and liquid /r/
production patterns of 14 severely language delayed were absent in his phonetic inventory. Some later to
children with autism to those of 10 typically develop- develop sounds were present whilst earlier ones were
ing children. Speech elicitation was done primarily by absent. B.D. exhibited both typical (e.g. final conso-
single consonant imitation, and sounds produced were nant deletion) and atypical phonological processes
rated based on the extent to which they were develop- (e.g. glottal replacement and atypical cluster
mentally early sounds versus later sounds. In-depth reduction). The evidence of multiple phonological pro-
phonological investigation, including a phonological cesses, process interactions, use of homonymy, and the
process analysis, was not used. Although the sample use of jargon, resulted in severely reduced intelligibil-
size is somewhat small, results suggested that this ity. Wolk and Edwards concluded that B.D. was
severely impaired group of children exhibited the acquiring his phonological system in a unique way,
same general speech sound patterns typically seen in with a pattern of acquisition rarely seen in normally
normally developing and language learning impaired developing children. This was an important early
children. finding on a single case of childhood autism.
In a recent study to investigate articulation/ In a later case study design, Wolk and Giesen (2000)
phonological impairments and language function in studied four siblings with autism. All four siblings
children with ASD, Rapin et al. (2009) used cluster exhibited a well-structured, rule-governed phonologi-
analysis to evaluate both language and phonology in cal system. Phonetic inventory data showed wide
62 children (mean age = 8.6 years). Findings suggested individual variation. However, each child showed aty-
two main subtypes of language disorders, one with pical speech sound development not commonly
severe impairment in expressive phonology (24%) observed in normally developing children. For
and the other with borderline/normal phonological example, in one child the bilabial nasal /m/, a
skills but impaired comprehension (76%). Thus, one sound very early to develop, was absent from his
in four children with ASD showed significant phono- inventory, whereas other nasals, several fricatives, as
logical impairment. However, because detailed analy- well as affricates were present. Phonological process
sis of sound errors was not reported, this study does analyses across all four children showed both typical
not provide information about the extent to which (e.g. liquid cluster reduction, liquid gliding, final con-
these children presented with normal delayed versus sonant deletion) and atypical (e.g. velarization, frica-
unusual/disordered patterns of phonological behav- tion of stops and liquids, segment coalescence)
ior. However, given the relatively large sample size phonological behavior. Interestingly, the speech of
used, it is an interesting finding. the sibling with the mildest symptoms of autism was
In another study addressing the relationship characterized mainly by typical phonological pro-
between language function and articulation in cesses that would be seen in normal development. In
autism, McCann et al. (2007) investigated prosody contrast, the speech of the more severely impaired sib-
and language skills in 31 children with high-function- lings was characterized by both typical and atypical
ing autism. They found that all children had at least phonological processes. These findings supported
one aspect of prosody affected, and that prosody those of the single case study reported above.
correlated highly with language abilities. However, Few studies have used an in-depth phonological
results suggested little impairment in phonological be- approach to examine speech sound production in
havior in these children. Although data on the larger samples of individuals with autism. Cleland
relationship between language measures and et al. (2010) used a one-word naming task in 69
children with HFA or Asperger syndrome. They was to examine the relationship between phonological
reported that although 41% of children showed some behavior and language measures in children with ASD.
speech sound errors, only 12% of children scored in
the range indicating a phonological disorder. Method
Phonological process analysis showed gliding, cluster Participants
reduction, and final consonant deletion occurred Participants were eight children (five male, three
most frequently. Interestingly, atypical processes or female) with ASD, ages 5.3–15.1 years (M = 7.9).
‘non-developmental error types’ were also evident, Participants were from different schools and different
including phoneme specific nasal emission and initial families except for two who were siblings. Each child
consonant deletion. The authors note that these pro- was diagnosed with ASD between ages 3 and 6
cesses were evident not only in children with severe years. All participants were currently receiving
phonological impairment, but even in some with pho- private speech therapy and ongoing special education
nological behavior in the ‘normal’ range. services at school. All children communicated verb-
The literature review to date, as discussed above, ally, although two used supplemental Augmentive
shows how several of the group studies have failed to and Alternative Communication (AAC) technology
carry out sufficiently detailed phonological analyses. to support language development and communi-
It is important to note that further investigation is cation, primarily during school activities.
needed to evaluate the power of the findings from Participants were recruited from the private speech
larger sample studies against the predominantly and language clinic of the second author. Of 15
smaller case studies which did use detailed analyses clients in this clinic who had a diagnosis of ASD, 14
and found some unusual phonological behavior. were verbal. Three of the 14 were exceptionally high
Furthermore, these studies that used detailed analyses functioning and exhibited no language or speech
did not make comparisons with phonologically disor- impairment, and were not recruited for participation.
dered children without autism. Thus, there is no clear Of the 11 remaining, 8 agreed to participate in the
information to date about how children with autism study. Notably, none of the children who were
may differ from phonologically impaired children recruited for this study had phonological impairment
without autism on the evidence of atypical phonologi- as part of their current treatment program at the
cal features. However, there are data to support evi- clinic or at their respective schools.
dence of a smaller likelihood of atypical features in
normally developing children (Bernthal et al., 2009). Language assessment
In summary, there is growing support to suggest evi- All participants completed standardized language
dence of disordered phonological patterns in some testing. For the two youngest children, the Preschool
children with autism. Further, while some children Language Scale – fourth edition (PLS, 2002) was
with autism may exhibit the same general speech used to obtain receptive and expressive language
sound production patterns as typically developing scores. The PLS was also used for two older children
and language learning impaired children, there is evi- who, due to cognitive and language limitations, were
dence from studies utilizing in-depth phonological unable to participate in tests standardized for their
analyses that some children with autism may also ages. Thus, although age equivalencies were obtained,
exhibit more atypical speech patterns (Cleland et al., standardized scores could not be determined due to
2010; Wolk and Edwards, 1993; Wolk and Geisen, the age limitation of the tests. The Clinical
2000). Existing research on phonological disorders in Evaluation of Language Fundamentals-Preschool
children with autism is still very preliminary and (CELF-P, 2004) was administered to three partici-
more studies including children with more severe pants, and for the oldest child the Clinical Evaluation
deficits and larger sample sizes are needed. of Language Fundamentals-4 (CELF-4, 2003) was
The primary purpose of the current investigation administered. All testing was conducted by the
was to extend previous work on articulatory/ second author, a certified and licensed speech-
phonological behaviors in children with ASD. The language pathologist.
hypothesis was that at least a subgroup of these chil-
dren exhibit not only delayed but atypical phonological Speech assessment
patterns. In-depth phonological analysis of the speech In-depth assessment of speech production was
sound production was conducted in order to better conducted by two trained speech-language pathol-
understand the specific sound errors in eight children ogists. Two methods were used for speech elicitation:
with autism. Second, because research on the relation- (1) object naming, using the Assessment of
ship between cognitive/language measures and phono- Phonological Processes (APP, Hodson, 1986); and (2)
logical behavior has been mixed, a secondary purpose conversational speech sample. For the children who
demonstrated higher language formulation skills,
they were engaged in conversation. For children with Kwiatkowski, 1982). Inter-rater reliabilities for
greater language impairment and mean length of coding participant responses were 100%. The phonetic
utterance less than three, language samples were inventory established a baseline for which sounds each
obtained using a picture description task. During the child was able to produce and were evident in his/her
APP, an attempt was made to elicit 50 words per productive speech output. The criterion used for pres-
child. Because of the inherent difficulties obtaining a ence or absence in the phonetic inventory was more
speech sample from children with ASD, some items than one instance of use. The phonological process
were elicited using delayed imitation. Additionally, analysis determined which phonological processes were
some items were deleted if too unintelligible to exhibited, and frequency and percentage of occurrence
analyze accurately or if it was unclear if the child were calculated. A phonological process was considered
was using the correct word. All speech elicitation present if it occurred more than once. Furthermore,
was video-recorded for later transcription and coding. every time a particular sound, for example fricative,
was present in any position in the target word, it was
Behavioral assessment considered an opportunity for the particular process to
For the purpose of this study, only a perfunctory occur. For example, in the case of final consonant del-
behavioral assessment was believed to be necessary, etion, if a fricative was deleted in word final position,
to enhance the overall understanding of the partici- it could also have had the opportunity for stopping of
pants. Behavioral information was collected for all fricatives and therefore been counted as part of the stop-
participants, including spontaneous communication ping process. Phonological processes were divided into
and sentence production, self-regulation of behavior, common (typical) and uncommon (atypical) phonologi-
comprehension of stories, comprehension of abstract cal processes. Common processes refer to those that are
language, sensory integration issues, and presence of more typically present in normally developing children,
echolalia. Analysis of behavioral information was whereas uncommon processes refer to those processes
predicted to be associated with the severity of language less likely to occur in normally developing children
and/or phonological difficulty. (Bernthal et al., 2009). Percentage occurrence was calcu-
lated based on the number of opportunities for each
Language delay process to be exhibited. For example, if 20 words had
Because different language assessments were used final consonants and a child deleted a final consonant
based on individual age or ability, a measure of in 10 words, the percentage occurrence would be
percent language delay was used to quantify language coded as 50%. The number of phonological processes
impairment and to have a means to compare the along with the percent language delay appears in
children across ages and test measures. All participants Table 1.
exhibited some degree of impairment on expressive
and receptive language measures. Percent was calcu-
lated using age equivalencies on assessment batteries Results
completed. The percentage below age expectations Typical and atypical phonological processes
were calculated for receptive and expressive language. All children showed evidence of some normally devel-
Each participant had comparable delay in receptive oping but delayed phonological processes (Table 2).
and expressive language; therefore, the mean was Across children, 11 typical delayed phonological
used as the index of language delay. processes were observed, as follows: stridency deletion,
The eight participants were then divided into two liquid assimilation, fronting, gliding, stopping, post-
groups using a median split based on severity of vocalic devoicing, reduplication, final consonant
language impairment. The mild language delay deletion, weak syllable deletion, vowelization, and
group (MLD, n = 4) had a language delay less than pre-vocalic voicing. All participants also exhibited
38%; participants in the severe language delay group some atypical phonological processes not often
(SLD, n = 4) had a language delay 42% or higher observed in typical development. Specifically, the
(see Table 1). Mean age for the MLD participants following 13 atypical processes were observed across
was 6.6 years (range 5.25–7.42) and included two children: deaffrication, migration, palatalization, pre-
females and two males. SLD participants were older, vocalic devoicing, post-vocalic voicing, syllable
with a mean age of 9.4 years (range 5.25–15.08) and coalescence, backing, metathesis, epenthesis, initial
included one female and three males. consonant deletion, medial consonant deletion, and
vowel deviations (see Table 2).
Data analysis An analysis of the data using the PCC yielded
Three levels of analyses were conducted: (a) phonetic similar results. Across participants there was a wide
inventory, (b) phonological process analyses, and (c) range of PCC, ranging from a low of 0.27 to a high
percent consonants correct (PCC; see Shriberg and of 0.90. A group comparison using the
Table 1 Percent language delay and total number of phonological processes for participants in the MLD and SLD groups
1 0.17 10 2 0.88 18
3 0.34 7 4 0.42 17
5 0.33 10 7 0.70 18
6 0.38 5 8 0.75 19
Mann–Whitney test was significant, with MLD par- compared to participants in the SLD group (M =
ticipants having a higher PCC than SLD participants 5.75; Z = 2.381, P < 0.05). Similar differences were
(P < 0.05). observed for atypical processes, with MLD children
exhibiting fewer atypical phonological processes
(M = 3.25) than SLD children (M = 7.25; Z = 2.323,
Relationship between language impairment and
P < 0.05). Results are illustrated in Fig. 1.
phonological processes
To examine language delay as a continuous
When divided into MLD and SLD groups, differences
variable, correlation coefficients were computed to
emerged in terms of number and nature of phonologi-
examine the relationship between percent language
cal processes. Because of the small sample size and use
delay and the number of phonological processes exhib-
of percentage data, group comparisons were calculated
ited by all eight participants. There was a strong
using non-parametric T-tests with Bonferroni correc-
significant positive correlation (r = 0.77, P < 0.025),
tion. MLD participants exhibited fewer typical/
showing that the children with the greatest percent
delayed phonological processes (M = 3.00) as
Table 2 Proportions and percentage occurrence of typical processes, atypical processes, and cluster reduction, per child by
group
Participant 1 3 5 6 2 4 7 8
Total number of processes 10 7 10 5 18 17 18 19
No. of typical processes 3 2 4 3 6 5 6 6
No. of atypical processes 5 3 4 1 7 7 6 9
No. of cluster reductions 2 2 2 1 5 5 6 4
Stridency deletion (%) 0 0 0 0 0 0.316 0 0
Liquid assimilation (%) 0 0 0.083 0 0 0 0 0
Fronting (%) 0 0 0 0 0 0.017 0 0
Gliding (%) 0.11 0.167 0 0 0.063 0 0.02 0.133
Stopping (%) 0.031 0.07 0.02 0 0.097 0 0.068 0.019
Post vocalic devoicing (%) 0 0 0 0.012 0 0.017 0.02 0.057
Reduplication (%) 0 0 0 0 0.2 0 0.02 0.038
Final consonant deletion (%) 0.105 0 0.025 0.323 0.708 0.044 0.09 0
Weak syllable deletion (%) 0 0 0.167 0.217 0.161 0 0.4 0.05
Vowelization (%) 0 0 0 0 0 1 0 0.9
Pre-vocalic voicing (%) 0 0 0 0 0.032 0 0 0
Deaffrication (%) 0 0 0 0 0 0.6 0.09 0
Migration (%) 0 0 0 0.012 0.129 0.017 0 0.075
Palatization (%) 0 0 0 0 0 0.017 0 0
Pre-vocalic devoicing (%) 0.038 0 0 0 0 0 0.045 0.057
Post Vocalic Voicing (%) 0 0 0 0 0 0 0 0.019
Syllable coalescence (%) 0 0 0.055 0 0.067 0.158 0.13 0.05
Backing (%) 0.02 0.045 0.02 0 0.194 0.017 0 0.018
Metathesis (%) 0.02 0.045 0.058 0 0.067 0.017 0 0.057
Epenthesis (%) 0 0 0 0 0.032 0 0.16 0.339
Initial consonant deletion (%) 0.04 0 0 0 0.31 0 0.02 0.019
Medial consonant deletion (%) 0 0.143 0 0 0 0 0
Vowel deviations (%) 0 0.136 0 0 0.387 0.1 0.34 0.245
Affrication (%) 0.02 0 0 0 0 0 0 0
Atypical ICR (%) 0 0 0 0 0.643 0.39 0.59 0.25
Typical MCR (%) 0.286 0.2 0.11 0 0 0.143 0.222 0.375
Atypical MCR (%) 0 0 0 0 0.57 0 0.555 0.375
Typical FCR (%) 0.3 0 0.455 0.07 0.25 0.286 0.286 0
Atypical FCR (%) 0 0 0 0 0.625 0.143 0.429 0
Note: The number of processes, typical and atypical processes and cluster reduction are presented in proportions and the specific
processes are presented in percentages.
MLD, mild language disorder; SLD, severe language disorder; ICR, initial cluster reduction; MCR, medial cluster reduction; FCR, final
cluster reduction.
Cluster reduction
Due to the number and importance of cluster
reduction, these data were analyzed separately.
Cluster reduction was categorized into typical and aty-
pical patterns based on the patterns typically observed
in normal development. MLD participants exhibited
fewer cluster reduction processes overall than children
in the SLD group (Z = 2.381, P < 0.05). However,
when broken down by cluster reduction process type,
SLD children exhibited more atypical cluster reduction
processes than MLD children (Z = 2.494, P < 0.025),
while the difference in typical cluster reduction pro-
cesses was non-significant (see Fig. 3).
Discussion
The purpose of this study was to investigate the speech
Figure 1 sound production abilities in eight children with ASD.
Results showed that all eight children exhibited
language delay also exhibited the highest number of delayed phonological behavior, consistent with pre-
phonological processes (Fig. 2). Additional correlation vious findings (e.g. Bartak et al., 1975; Cleland
coefficients were computed to examine the relationship et al., 2010; Paul, 1987; Wolk and Giesen, 2000).
between severity of language impairment and the Further, the current findings support prior research
percent occurrence for both typical and atypical showing that some children with ASD also exhibi
processes. Percent language delay correlated signifi- some atypical phonological processes not commonly
cantly with the percentage occurrence of three atypical observed in normal but delayed development
phonological processes: metathesis (r = 0.739), vowel (Cleland et al., 2010; Wolk and Edwards, 1993;
deviations (r = 0.883), and atypical initial cluster Wolk and Geisen, 2000). Specifically, the following
reduction (r = 0.878). It should be noted, for eight atypical phonological processes were observed:
example that since the highest use of metathesis in initial consonant deletion, epenthesis, metathesis,
any of the eight children was 0.067 ( just under 7%) syllable coalescence, post vocalic voicing, prevocalic
and the total sample from the APP was limited to devoicing, and vowel deviations.
100 words or less, the total opportunities for potential This study reveals that some children with ASD
targets for metathesis may have been only a few words. exhibit unusual phonological trends. This finding cor-
Total number of target words differed for the potential roborates the findings in two earlier studies (Wolk
for each process to occur and this is the nature of this and Edwards, 1993; Wolk and Giesen, 2000). It is par-
type of analysis. Thus, for example atypical cluster ticularly interesting that in the study on multiple sib-
reduction was much higher for these children, but lings, Wolk and Giesen (2000) found that among the
also had a larger number of opportunities for four siblings studied the more severe cases (referring
occurrence. to non-linguistic autistic features) exhibited more
Figure 2 Figure 3
unusual sound patterns whereas the milder ones we observed that two participants exhibited unusual
showed more normal processes typical of delayed voicing patterns. Consistent with the findings from
phonological behavior. It is possible that some pre- Wolk and Geisen (2000), which noted unusual
vious studies may not have shown the same finding voicing patterns in some participants, the current
of unusual phonological trends because they did not investigation also found unusual voicing in two
include more severely impaired children. This issue of participants (one in the MLD and one in the SLD
severity is critical for future studies to examine. When group). Both of these participants exhibited an
phonological processes were evaluated in more detail unusual pattern of devoicing words, with some
by examining cluster reduction separately, particularly words produced completely without voicing, regard-
interesting findings emerged. In children with mild–mo- less of articulation error. For both children, speech
derate language impairment, no atypical cluster was more like a whisper for those words. The partici-
reduction processes were observed. However, children pant with less delayed language exhibited devoiced
with severe language delay exhibited an equal words in 35% of elicited words, while the subject
number/frequency of typical and atypical cluster with a greater language delay exhibited devoiced
reduction processes. It is noteworthy that out of all words in 88.6% of elicited words. Although making
eight children, for all processes examined, there are a comparison of only two subjects results in very
only five instances where a child is using any process limited conclusions, these findings in voicing dysfunc-
more than 50% of the time. Apart from clusters, 92% tion were consistent with other trends observed in this
of the cells show less than 20% occurrence of that study regarding relationship between speech sound
process. This may be partially affected by the relatively production and language delay.
limited verbal output of some of these children. The findings of this study contrast with some earlier
The current investigation also reveals a strong studies, but support and extend the findings of other
relationship between severity of phonological behavior studies showing unusual phonological trends in some
and the severity of language impairment. This is a children with autism. Previous work using in-depth
replication and extension of the classical correlation phonological analyses used primarily single subject
seen in developmentally delayed children. Group com- designs (e.g. Wolk and Edwards, 1993), and few have
parisons revealed that children with mild–moderate examined the relationship between speech sound and
language impairment (MLD) exhibited more phono- language behavior in this population. Studies using
logical processes overall when compared to the more larger sample sizes or comparison group designs
severe language impaired group (SLD). When have typically relied on one-word naming or imitation
broken down by process type, this pattern held true tasks to elicit speech (e.g. Kjelgaard and Tager-
for both typical and atypical processes. Participants Flusberg, 2001; McCleery et al., 2006) or have
with more severe language delay also exhibited a lacked detailed phonological analysis beyond general
higher occurrence of cluster reduction processes with indices of speech sound errors (e.g. Rapin et al., 2009;
more atypical cluster reduction patterns than children Shriberg et al., 2001). Other studies (e.g. McCann
with more mild language impairment. Analysis using et al., 2007) investigated speech errors only among chil-
the PCC as an overall index of phonological impair- dren with high-functioning autism, and thus a restric-
ment yielded consistent results. Future research inves- tion in range may have limited the ability to find a
tigating the relationship between phonological and relationship among these measures. Only one study
language behavior and severity will help to confirm has examined the extent to which phonological impair-
and extend these findings. Future studies would need ments relate to language delay in children with ASD
to look in similar detail at children with language dis- (Rapin et al., 2009) but, again, detailed analysis of pho-
order without autism to see whether similar findings nological behavior was not reported.
on cluster behavior may be evident. Future research employing larger sample sizes
Clinical observation suggested that the MLD group would help to (a) clarify the nature of the speech
exhibited milder behavioral characteristics than the sound difficulties in children with autism; (b) continue
SLD group. These included self-regulation of behav- to explore the differences between mildly and more
ior, comprehension of abstract stories and complex severely language impaired children with respect to
rules/routines, and mild sensory integration difficul- speech sound production skills; (c) determine
ties in the MLD group. Behavioral characteristics in whether detailed phonological investigations may be
the SLD group include poor self-regulation of behav- useful for diagnostic assessments; and (d) determine
ior, poor comprehension of stories, marked difficulty whether phonologically based treatment programs
with abstract language, moderate–severe sensory inte- may be optimal for at least a subset of children with
gration issues, and echolalia in the SLD group. ASD. Many of these clinical implications are appli-
In addition to the primary findings regarding cable to all children with speech disorders.
phonological processes and relationship to language,