Mattingly "AI & Prompt Design: The Basics of Prompt Design"
Prof. Norbury Lecture: 'Overlap between ASD and SLI: diagnostic challenges.
1. overlap between autism
spectrum disorders (ASD)
and specific language
impairment (SLI):
diagnostic challenges
Professor Courtenay Frazier Norbury
Royal Holloway
University of London
1
2. why is there so much variation in
core language skills within ASD?
2
3. outline of the session
• overview of language profile(s) in ASD
• does autism cause language impairment?
• is language impairment a co-morbid (co-
occurring) condition?
• multiple factors that contribute to variation in
language development within ASD. 3
4. Language
Phonology
the sounds of
language
Syntax/morphology
the rules that control
sentence formation and
word endings (plural,
past tense)
Semantics
the meaning of
individual words &
words in context
Pragmatics
the social use of
language in
context & social
exchanges
4
7. ‘normal’
range
= omnibus test of language production and comprehension (grammar)
= receptive and expressive vocabulary (semantics)
= articulation/phonology
= non-sense word repetition (memory and phonology)
language in ASD is hetergeneous
(Tager-Flusberg & Joseph, 2003)
7
8. ‘specific’
language
impairment
“pragmatic” language skills
are extremely variable!
structural language impairments
are universal
deficits in
morphosyntax
and grammar
weaknesses in
semantics and
word learning
limitations in
verbal short-term
and working
memory (poor
non-word
repetition)
8
9. what is the relationship between
autism spectrum disorders and
specific language impairment?
9
10. • autism + Language Impairment
(ALI) is the outcome of autistic
cognition
– poor social interaction results in
reduced language learning
opportunities
– weak central coherence impedes
learning of language in context
• but must explain how some
children have an ALN (autism +
‘language normal’) profile
SLI Autism
ALI
A B
10
12. autistic
symptoms
language impairment
(standard tests of vocabulary and sentence processing)
autism: language ‘normal’ (ALN)
autism: language ‘impaired’ (ALI)
typically developing (TD)
‘specific’ language impairment (LI)
12
13. autistic
symptoms
language impairment
(standard tests of vocabulary and sentence processing)
autism: language ‘normal’ (ALN)
autism: language ‘impaired’ (ALI)
typically developing (TD)
‘specific’ language impairment (LI)
13
14. direct comparison of language
phenotypes…
• allows identification of features that are
‘universal’ and ‘specific’ to ASD
• allows identification of risk factors that
contribute to language impairment (common
across disorders)
• allows identification of protective factors that
facilitate language acquisition, despite ASD
14
16. • set of functions necessary for
flexible, future-oriented
behaviour
– working memory
– inhibitory control
– attentional flexibility
– planning
• poor executive control could
disrupt language learning
– difficulties with joint attention
– problems inhibiting irrelevant
information to the context
– problems learning rules
– problems planning expressive
language
executive function
16
17. implications
• should be a strong relationship between
executive abilities and language competence
– discrepant findings regarding correlations between EF
tasks and measures of verbal ability
• some do: (Pellicano 2007; Liss et al. 2001)
• some don’t: (B&N 2005; Landa & Goldberg 2005; Joseph et
al. 2005)
• different patterns of relationship may depend in part on
which measures are used
• some studies find associations between poor EF
performance and impaired language status, but not
ASD status (e.g. Bishop & Norbury 2005; Liss et al.
2001)
17
18. Kelly, Walker & Norbury (2013) Developmental Science
• explore eye-movement control in relation to
ASD and language impairment
• ALI = autism + language impairment
• ALN = autism with language ‘normal’
• SLI = specific language impairment
• TD = typically developing
18
22. challenges
• direction of causation is not at all clear:
– one alternative explanation is that language important
to encode the arbitrary rules needed to succeed on EF
tasks (Russell et al. 1999, Zelazo 2004)
– another is that EF and language ability are essentially
unrelated, but that individuals with ASD fail to use
language for self-regulation (Joseph et al. 2005)
• relationship between EF and language seen in
non-autistic populations (e.g. SLI; Henry et al.
2012)
• limited investigation of the developmental
relationship between EF and language (especially
in ASD)
22
24. • tendency to focus on
individual, local elements at
the expense of global, holistic
meaning
• explains apparent skill at
solving certain visuospatial
tasks
• could explain disruptions to
language because need to
integrate different pieces of
information in context
– joint attention
– learning from context
weak central
coherence
24
25. implications
• individuals with ASD, of all ages and abilities,
should be impaired at deriving context
dependent meaning (Happe 1999)
– e.g. homograph task (Happe 1997)
• Lucy was cutting onions. In her eye there was a tear.
• Lucy was climbing fences. In her dress there was a tear.
– inferencing, resolution of ambiguous and figurative
expressions
25
26. challenges
• problems with verbal coherence are not
unique to ASD
– seen in virtually every other disorder where
language is impaired
• good visual spatial skill and poor verbal
coherence do not necessarily go together
• series of studies demonstrate that weak CC
aligns with weak verbal ability, independent of
ASD status (Norbury 2004, 2005, Brock et al.
2008)
26
27. weak verbal ability weak CC?
• Snowling & Frith (1986)
– performance on
homograph task
dependent on language
ability in both ASD and
Developmental Delay
groups
0
1
2
3
4
5
6
7
8
9
SLI ALI ALN TD
Group
Faciliatationdifferencescore
27
29. • failure to orient to social
stimuli early in life:
– reduced fixations to faces/eyes
– reduced preference for child
directed speech
• decreased participation in
social interactions
• poor joint attention
– poor gaze following
core social deficit
poor understanding
of social intention /
theory of mind
disruptions to language
learning / pragmatics
29
30. implications
• all individuals with ASD should show pervasive
language impairments (especially in
vocabulary, which relies on use of social cues)
• should be strong relationship between
measures of “socialness” and measures of
language ability
• all individuals with ASD should show poor
processing of social stimuli, from the earliest
ages/stages
30
31. challenges
• all individuals with ASD should show pervasive
language impairments (especially in
vocabulary, which relies on use of social cues)
– most do, though ~25 - 43% of cognitively able
individuals demonstrate ‘normal’ language skills
on standardised tests (Kjelgaard & Tager-Flusberg,
2001; Loucas et al. 2008)
– vocabulary often one of the better aspects of
language in ASD (Mottron 2004)
31
32. challenges
• should be strong correlation between
measures of “socialness” and measures of
language competence
– Kuhl et al. (2005): toddlers with ASD who
preferred to listen to non-speech instead of child
directed speech
• show decreased ability to distinguish different
meaningful phonemes
• poorer scores on expressive language measures
32
33. Norbury et al. (2009)
• Measured symptom severity and social adaptation
scores in adolescents with ASD who did and did not
have additional language impairments
– Social Communication Questionnaire (SCQ)
– Autism Diagnostic Observation Schedule (ADOS-Module 4)
– Vineland Adaptive Behavior Scales (Socialisation)
• Compared scores to adolescents with SLI
33
37. verbal descriptions of
social scenes
• children with ASD will
look more at
background
• children with ASD more
likely to mention
background items in
output
• children with language
impairments (ALI/LI)
will produce fewer
important sentence
elements
37
40. the story so far
• impaired language development is not causally
related to ASD, nor does it account for variation
in ASD symptomatology
• severity of ASD symptomatology does not fully
account for variation in language ability
• no one cognitive theory of ASD can explain
variation in language phenotype
• perhaps language impairments are not caused by
autism?
40
41. so, why is there so much variation in
core language skills within the
autism spectrum?
41
42. • ALI represents the co-morbidity of
SLI and ASD
– separate causal systems that are
themselves correlated.
• ALI overlap is due to causal factors for SLI
being associated with those causing
autism.
– same causal factor yielding different
diseases (Pleiotropy)
• autism and SLI are phenotypic variants of
the same set of genes (Bishop, 2003)
• should expect overlap in both directions
SLI AutismALI
A B
SLI ALI
A B C
Autism
42
43. note: partially overlapping disorders
ASD involves impairment not seen in
SLI (e.g. repetitive
interests/behaviours; regression)
therefore, would not expect identical
patterns of language behaviour, even on
tasks both groups find challenging, or
remarkably similar developmental
trajectories
43
44. the way forward
• a truly developmental approach
– at risk studies so children are assessed over time
• cross disorder and cross cultural comparisons
• unified approach to language assessment and
characterisation of language impairment
• combining behavioural, cognitive,
neurobiological evidence in same cohorts
• intervention studies to test causal hypotheses
44
46. Kuhl (2007) model of language
SOCIALNESS
SPEECH PROCESSING
perception
phonological skill
computational capacity
???????
amongst
many other
things! 46
47. those who develop good language
skills rely on non-social processes
individual differences in non-social
processes could explain variation in
language skill...
47
48. learned 4 new words
(6 exposures to each)
• three measures of
learning:
– recognition
– naming (phonology)
– defining (semantics)
• two time frames:
– immediately after
learning trials
– four weeks later
‘show me the kellow’
48
50. conclusions from this study
• children with ALN very good at learning and retaining
phonological information
– strong prediction that children with ALI will have additional
phonological deficits
– may explain facility with text (non-social avenue to word
learning?)
• unlike peers, children with ALN do not show strong
evidence of consolidating learned information
(see also Henderson, Gaskell, Powell & Norbury, 2014)
– qualitative differences in ASD languagereflect difficulties
integrating new information with existing knowledge
– underlies difficulties with semantic and pragmatic aspects
of meaning?
50
51. ASD (n =13) TD (n = 13) t-value p-value Cohen's d
Age (months) 86.46 85.00 1.02 0.32 0.40
Receptive
Vocabulary
(raw score)
73.54 73.85 0.08 0.93 0.03
Receptive
Vocabulary
(SS)
100.54 102.31 0.50 0.62 0.20
WASI Matrix
Reasoning
51.31 51.77 0.11 0.91 0.05
WASI
Definitions
40.00 51.23 3.05 0.006 1.73
can acquire vast store
of words despite
social limitations
clear qualitative
differences in what
they know about
words 51
52. summary
• single deficit models of ASD cannot explain
the wide variation in language ability that
characterises the disorder
• risk factors for language impairment are likely
to be shared across a number of neuro-
developmental disorders
• disorder specific risk factors also operate,
resulting in rich tapestry of behavioural
outcomes
52