Early indicators of language impairment: precursors of dyslexia and SLI

1
Early indicators of language impairment
precursors of dyslexia and SLI

• Jan de Jong
• University of Amsterdam

2
An overview

• 1. Specific language impairment (SLI) and
  developmental dyslexia
• 2. Linguistic precursors of dyslexia
• 3. Linguistic precursors and later reading status
• 4. SLI and dyslexia revisited
• 5. The natural history of language impairment

3

• Specific language impairment (SLI) and
  developmental dyslexia

4
Specific language impairment (SLI) and dyslexia
exclusion definitions

• children who show a significant limitation in
  language ability, yet the factors usually
  accompanying language-learning problems such as
  hearing impairment, low non-verbal intelligence
  test scores, and neurological damage are not
  evident (Leonard, 1998)

• Dyslexia is a disorder manifested by difficulty
  learning to read despite conventional
  instruction, adequate intelligence, and
  socio-culture opportunity. (Critchley, 1970)

5
Specific language impairment (SLI) and dyslexia
inclusion definitions

• About SLI
• The most common profile in English is a mild
  to moderate deficit in semantic functioning, with
  a more serious problem in morphosyntax.
  Grammatical morphology seems relatively weak.
• Phonology might also be below age level, though
  usually cannot account for the problems in
  grammatical morphology. (Leonard, 2000)

• Dyslexia is a developmental language disorder
  whose defining characteristic is difficulty in
  phonological processing () difficulties include
  problems storing, retrieving and using
  phonological codes in memory as well as deficits
  in phonological awareness and speech production
  (Catts Kamhi, 1999)

6
SLI and dyslexia in education separate routes

• Children with SLI speech therapy, special
  schools for language-impaired children or
  ambulant treatment
• Dyslexic children mainstream schools, remedial
  teaching or treatment focused on dyslexia
• gtgt different routes for children with SLI and
  dyslexic children

7
Separate routes, separate groups?

• Many children with language problems develop
  reading problems later on (estimates between 40
  and 50)
• Many children with reading problems have (had)
  language problems) (according to Leonard, 1998,
  the majority)
• For both groups
• Genetic predisposition (more easily identified in
  dyslexia)
• More boys than girls affected (more clearly so
  for SLI)

8
Overlap between SLI and dyslexia

• Scarborough (1990, 1991) children selected based
  on a family history of dyslexia also fit the
  criteria for SLI. Conversely, scores on
  pre-school syntactic measures correctly predicted
  reading problems in 75 of the cases.
• Mc Arthur et al. (2000) two groups of children
  (with SLI and reading problems) were re-diagnosed
  by blind reviewers. Within both groups, 50
  also fit the other definition.

9
Similarities between SLI and dyslexia for which
variables?

• In a direct group comparison both groups had
  difficulties with
• Phonological awareness (analysis, synthesis
  tasks)
• Word and sentence repetition
• Speech perception
• Grammaticality judgment
• rapid naming
• mental imagery
• (Kamhi Catts, 1986 Kamhi et al., 1988)

10
A preliminary conclusion

• There is overlap between the two groups children
  with dyslexia and children with SLI
• Reading disorders have precursors in the
  understanding, perception and production of
  spoken language.

11
Reading problems more than phonology?

• Bishop (1991), Developmental reading
  disabilities the role of phonological processing
  has been overemphasised
• There are good theoretical reasons why
  syntactic, semantic and pragmatic skills should
  be important in their own right
• Scarborough (1990, 1991)
• diificulties with syntactic and morphological
  aspects of language in the earlier preschool
  years may be more strongly related to later
  reading disabilities than (..) expressive
  phonological deficits

12
2. Linguistic precursors of dyslexia the Utrecht
project (2000-, University of Utrecht)

Petra van Alphen Elise de Bree Ellen Gerrits Jan
de Jong Carien Wilsenach Frank Wijnen
http//www.let.uu.nl/dyslexie/
13
How to study dyslexia in the preschool years?

• Dyslexia can only be assessed once literacy
  education has started (gt8).
• However, 40-70 of children with a dyslexic
  parent have problems with learning to read and
  write (in the general population 5-10).
• ? Assess oral language development of children
  with a familial risk of dyslexia.

14
Basic predictions in the project

• If there is a language problem in the group at
  risk for dyslexia, it will resemble (share
  characteristics with) SLI.
• Since the at risk group includes children who
  will develop dyslexia as well as children who
  will not, the at risk group will perform worse
  than the controls and better than the SLI group

15
Design

• Tasks for speech perception, phonology, word
  recognition, morphosyntax, from 16 to 50.
• Two cohorts
• babies (16 30).
• 70 children at risk for dyslexia
• 40 controls
• toddlers (30 50).
• 70 children at risk for dyslexia
• 40 controls
• 30 children with SLI

15
16
Why compare dyslexia and SLI?

• Identify non-normal language development in
  dyslexia
• SLI as a benchmark

16
17
Subject selection

• At-risk children
• one parent diagnosed with dyslexia
• dyslexia test battery administered to parent
• SLI children
• diagnosis by school speech pathologist

17
18
General procedure

• 4 test sessions, at 6-month intervals
• several experiments / tests at each session
• additional data collected
• N-CDI I.Q. digit span
• educational level mother birth rank

18
19
Tasks to be discussed today

• Grammatical contrasts (babies)
• Phoneme categorisation (toddlers)
• Word recognition (toddlers)
• Phonology (toddlers)
• Grammatical morphology (toddlers)
• Some tasks are phonological. Some not!

20
Sensitivity to grammatical contrasts Carien
Wilsenach, 2006
21
Sensitivity to grammatical contrasts

• Previous research by Santelman Jusczyk (1998)
  on discontinuous dependencies
• Contrast measured is Verbing versus can
  Verb-ing 15 month olds were not sensitive to the
  contrast, 18 month olds were
• Stimuli in this experiment
• 8 grammatical sequences auxiliary heeft (has)
  with past participle (e.g. gelopen)
• 8 ungrammatical sequences modal auxiliary kan
  (can) with past participle

22
Preferential listening task
Green light
Red light
Red light
Video camera
?
Speaker
23
In the Utrecht Babylab the researchers room
24
In the Utrecht Babylab mother and child
25
Sensitivity to grammatical contrasts preference
26
Sensitivity to grammatical contrasts group
comparison
27
Sensitivity to grammatical contrasts conclusion

• Control group significant preference for natural
  passages sensitivity to grammatical dependency
  relation between auxiliary and participle
• At risk group no sensitivity, no discrimination
  between the two stimulus types

28
Phoneme categorisation Ellen Gerrits
29
Phoneme categorisation
30
Categorical speech perception example of normal
pattern
31
Categorical perception and dyslexia what do we
know?

• Children with dyslexia show less consistent
  phoneme identification
• They have a less sharply defined phoneme boundary
  or a shallower categorization function
• more ambiguous speech sounds
• So how is the categorization of younger, at-risk,
  children?

32
Categorical speech perception results
pop
kop
33
Conclusion

• Young children at risk for developing dyslexia
  have subtle speech perception deficits Their
  perception of phonemes with weak acoustic cues is
  less categorical than that of controls.
• Their perceptual performance is highly similar to
  that of young language impaired children and much
  older dyslexic children.

34
Word recognition Petra van
Alphen
35
Word recognition

• A deficit in phonological processing may result
  in less detailed or incorrect phonological
  representations of words in the mental lexicon
• Is word recognition hampered in Dutch at-risk and
  SLI children?

36
Word recognition

• What is the effect of initial phonological
  mismatches on the recognition of words in
  5-year-old at-risk and SLI children?
• Look at phonemes in onset position (e.g. zebra as
  zebra, pebra, or vebra)
• Maximal mismatch three features different (zebra
  as pebra place, voice and manner)
• Minimal mismatch one feature different (zebra as
  vebra place)

37
Procedure phonological mismatch detection

• Familiarisation phase
• Naming pictures of the objects
• Test phase
• Pressing the red button when the word is not
  correct (e.g. zebra as pebra or vebra opposed to
  zebra)

38
Results mismatch detection
39
Phonology Elise de Bree, 2006
40
Phonology at 39

• Participants
• 10 control children (mean age 39) 29 at-risk
  children (mean age 310) 10 children with SLI
  (mean age 41)
• Picture naming task
• lexical factors length, stress, syllable
  structure, phoneme occurrrence controlled for
  Age of Acquisition
• Analysis of phonological processes
• word level truncation (e.g. Banaan -gt naan,
  baan)
• syllable level avoidance of consonant clusters
  (stressed syllable onsets only e.g. Slaap -gt
  saap, laap)
• phoneme level avoidance fricatives (stressed
  syllable onsets only e.g. Fiets -gt tiets)

40
41
Truncation
p lt .001
n.s.
Cluster avoidance
Fricative avoid.
p .022
41
42
Error profiles
42
43
PCC and PMLU
p .002
p .002
43
44
Morphological production Jan de Jong
45
(No Transcript)
46
Morphosyntax at age 33

• Morphosyntax grammar ? morphology
• Verb inflections (agreement with subject)
• Nominal pluralization
• Delays in morphosyntax are typical for SLI (de
  Jong 1999)
• Example using infinitival verb forms instead of
  inflected forms
• Papa koekje eten infinitival construction can
  mean(papa is n koekje aan t eten)(papa gaat
  n koekje eten)(papa eet n koekje)
• Morpho-syntax is a good testing ground for the
  hypothesis that dyslectic children are
  language-delayed

47
Elicitation of plural nouns
This is a bear
...and these are two
48
Elicitation of finite verbs
...and this bear
This bear runs
49
Results production plurals
50
Verb inflection totals (in)correct responses
Correct 3sg, Aux-V Incorrect stem,
infinitive F2,85 5.19, p lt .01
50
51
Verb inflection qualitative
Overall not significant, but stem is
significantly more frequent in at-risks than in
other two groups
51
52
3. Linguistic precursors and later
reading Frank Wijnen, Elise de Bree Jan de
Jong, 2008
53
Reading test outcomes phonology group
53
54
Phonology and Reading Status
54
55
Reading test outcomes morphosyntax group
55
56
Morphosyntax and Reading Status
56
57
Conclusions

• Language development data
• At risk children show an oral language profile
  similar to that of children with SLI, but milder
• Difficult to see qualitative differences between
  dyslexia and at-risk
• Early language later reading
• Trend poor readers have lower performance scores
  on both early phonology and morphosyntax

57
58
4. SLI and dyslexia revisited Esther Parigger
Judith Rispens, 2008
59
Possible relations between dyslexia and SLI

• Dyslexia and SLI may be exactly the same
  disorder, differing only in severity (single
  source)
• Dyslexia and SLI may be separate disorders but
  share similar risk factors (qualitative
  difference)
• Dyslexia and SLI are different and have different
  causes but can (tend to) be co-morbid
  (comorbidity)(Bishop Snowling 2004 Catts et
  al. 2005)

59
60
The role of non-word repetition

• An important finding
• Non-word repetition is weak in dyslexics
• Non-word repetition is weak in SLI and is even
  considered to be a clinical maker of SLI

61
Relation between SLI/dyslexia and non-word
repetition

• NWR is a measure of phonological processing
• Reading problems (RP) are associated with a
  deficit in phonological processing
• Catts et al. (2005)
• presence of RP affects performance on non-word
  repetition task
• only the SLIRP group has a severe NWR deficit
• the SLI RP group has a mild NWR deficit only

62
Relation between SLI/dyslexia and NWR (Catts et
al., 2005)

63
Aim of study

• Data best explained by model 2 or by model 3?
• Prediction of model 2
• SLI characterised by a phonological processing
  deficit, regardless of the presence of a RP
• Prediction of model 3
• SLIRP not characterised by a phonological
  processing deficit.
• SLIRP do show phonological processing deficit

64
Method subjects

• Control children (n15)
• 5 girls 11 boys
• age 83
• non-verbal IQ 115.73
• SLI children (n19)
• 4 girls 15 boys
• age 82
• non-verbal IQ 99.21
• ?assessment of RP with non-word reading
  task (Klepel)

65
Klepel test
66
Method subjects

• SLI children (n19)
• SLI children with RP (n13)
• 1 girl 12 boys
• Age 81
• Non-verbal IQ 95,92
• Klepel 2,46
• SLI children without RP (n6)
• 3 girls 3 boys
• Age 83
• Non-verbal IQ 102,50
• Klepel 8,0

• TD children
• Klepel 9,13

67
Results NWR

• Group effect F(2,31)10,16, plt.001
• Group x Length effect F(6,93)3,34, p.01
• Length SLI RP SLIRP Controls
• 2 syll 89 95 94
• 3 syll 81 93 94
• 4 syll 61 81 82
• 5 syll 47 63 71
• Significant difference between SLIRP and
  SLI-RP/Controls
• Significant difference between SLIRP and
  Controls
• ? no differences between SLI-RP group and
  controls!

68
Results NWR
69
Conclusion

• Deficit in phonological processing as measured
  with the NWR is closely linked to RP but not to
  SLI
• Consistent with Catts (2005) (Model 3) who argues
  that SLI is not characterised by a central
  phonological processing deficit

70
Conclusion

• One third of the sample of SLI children did not
  differ significantly on NWR from typically
  developing children (on all four PPC lengths)
• ? data cast doubt on the usefulness of
  NWR as a diagnostic (clinical) marker for
  SLI

71
The natural history of language impairment (some
slides from Bishop)
72
The problem

• early intervention better than late
• BUT
• the younger the child, the more likely there will
  be spontaneous improvement

73
Decrease in prevalence of language impairment
with age
2 3 4 5 6 7 8
9
74
Can we predict outcome?
75
High likelihood of spontaneous improvement

• in 2-year-old with expressive language delay,
  unless
• comprehension is impaired
• poor use of nonverbal communication
• very restricted expressive vocabulary (8 words or
  less at 2 years)
• family history of SLI/dyslexia

76
Bishop Edmundson, 1987

• longitudinal study of language-impaired children
  from 4 to 5.5 yr
• good outcome in 30/68 with SLI
• 2/19 with general
  delay

77
Predictions versus findings

• Predicted
• worse outcome for uneven than for even profile of
  impairment
• Found
• worse outcome for even than uneven profile of
  impairment
• the more functions impaired, the worse the
  outcome
• hierarchy of vulnerability
• expressive phonology
• expressive syntax
• expressive semantics
• comprehension

78
The iceberg model
impaired
unimpaired
79
The iceberg model
impaired
unimpaired
80
Follow-up at 8 yrs of age(Bishop Adams, 1990)

• Conclusions
• children with early language delay only at risk
  for literacy problems if language difficulties
  persist
• reading comprehension more often a problem than
  grapheme-phoneme decoding

81
Stothard et al, 1998 (same population)
status at age 15 yr
status at age 5 yr
82
A final notice (from Bishop, 2007) What creates
language impairment?
83
Bishop, 2006

• SLI in the majority of cases the disorder is
  caused by the interaction of several genes
  together with environmental risk factors
• Twin studies the sibling of a child with SLI
  more often has a language problem if the twins
  are monozygotic. But the disorder is less
  serious in the sibling that has not been
  diagnosed as SLI.
• Suggests genetic and environmental factors.

84
Bishop, 2006

• Children with SLI must have a disorder in more
  than one domain for it to develop a language
  impairment (i.e. SLI) development is
  compromised (..) because more than one cognitive
  process is disrupted. Conversely if one route
  is blocked, another can usually be found.
• This idea challenges any notion of SLI as a
  single syndrome and also suggests that we may
  need to analyse it in terms of dimensions of
  impairment instead of looking for discrete
  subtypes
• Measuring on one clinical marker (e.g. NWR) does
  not suffice.

85

• Question how do Bishops ideas compare to
  Parigger Rispens findings?
• To finish off what do you think of the model by
  Bishop and Snowling?

86
Two-dimensional model of SLI and dyslexia(Bishop
Snowling, 2004)
87

• J.deJong1_at_uva.nl
• http//home.medewerker.uva.nl/j.dejong1/