Dyslexia and Speech Sound Disorder: How are they related

1
Dyslexia and Speech Sound Disorder How are they
related?

• Bruce F. Pennington
• SSD Team

2
SSD and Dyslexia Projectsupported by NICHD
(HD049027-21A1)
PI Bruce F. Pennington CoPI Richard
Boada Collaborators Consultants Research
Assistants Larry Shriberg Dorothy
Bishop Christa Hutaff Shelley Smith Jan
Keenan Irina Kaminer Rebecca Treiman Dick
Olson Lauren McGrath Maggie Snowling Robin
Peterson Erin Phinney Nancy
Raitano Rachel Tunick
3
Outline

• Background
• A. Diagnostic definitions of dyslexia and
  speech sound disorder (SSD)
• B. Speech and language precursors of dyslexia
• C. Cognitive model
• D. Relations to SLI
• II. Main Hypotheses
• III. Cognitive methods and results
• A. Time 1 studies phonological deficits across
  subtypes of SSD.
• B. Time 2 studies Literacy outcome in SSD.
• IV. Conclusions New Directions

4
Definition of Dyslexia (IDA NICHD,
2002) Dyslexia is a specific learning disability
that is neurobiological in origin. It is
characterized by difficulties with accurate
and/or fluent word recognition and by poor
spelling and decoding abilities. These
difficulties typically result from a deficit in
the phonological component of language that is
often unexpected in relation to other cognitive
abilities and the provision of effective
classroom instruction. Secondary consequences
may include problems in reading comprehension and
reduced reading experience that can impede growth
of vocabulary and background knowledge.
5
WORD ATTACK
ift fit bim bim ut ut rayed rate
kak cake maft mat nen any or
en ab add tash trash/tash wips wippie
s beb bid
6
Definition of Speech Sound Disorder (SSD) A
delay in the acquisition of developmentally
appropriate speech sounds, resulting in reduced
intelligibility. Not due to peripheral problems
(e.g. hearing loss or cleft palate) or a known
syndrome (e.g. autism or mental retardation).
Distinct from stuttering or mutism. Formerly
called articulation disorder and phonological
disorder (DSM-IV-TR).
7
Classic Errors in Speech Sound
Disorder Omissions bathtub ? batub blue
? bue Christmas tree ? Chistmas
tee Substitutions cup ? tup shovel
? sobel bathtub ? baftub Omissions and
substitutions can co-occur sleeping ?
feeping brush ? bus
8
Summary of Definitions Idiopathic (unexplained)
problems with a particular aspect of language
development speech sounds in SSD and printed
word recognition in dyslexia. Phonological
development implicated in both.
9
Early Speech and Language Problems Precede Later
Dyslexia

• Studies of infants and preschoolers born to
  dyslexic parents find problems in speech and
  language development (Gallagher, et al., 2000
  Lyytinen, et al., 2002 Pennington Lefly, 2001
  Scarborough, 1990). About 25 of RDs had SSD.
• Children with SSD have higher rates of later
  dyslexia (e.g. Bishop Adams, 1990 Catts,
  1993), but rates vary widely (8-68).
• So the overlap is far from complete in either
  direction.

10
SSD and RD Share Etiologies

• SSD RD are co-familial (Lewis et al, 1989
• 1990 1992) and coheritable (Tunick
  Pennington, 2002).
• So, the comorbidity between SSD and dyslexia
  is caused by a partly shared genetic etiology.
  Therefore, we predicted some of the risk loci
  already identified for RD would also be risk loci
  for SSD. This genetic overlap may vary by
  subtypes of SSD.
• The linkage results of Stein et al, (2004) and
  Smith et al, (2005) support this prediction.

11
Cognitive Models
12
Phenotypic Overlap Among SD, LI, RD
SD
RD
LI
13
Summary of Oral and Written Language
Constructs Disorders Dyslexia reading and
spelling problem (written language) SSD speech
production problem (oral language) SLI
semantics and syntax problems (oral
language) Some Cognitive Components Phonological
awareness (e.g. Pig Latin) explicit knowledge
of speech sounds (oral) PA Phonological short
term memory (e.g. Nonword Rep.) implicit use of
phonological codes to remember words or nonwords
(oral) PM Rapid Serial Naming phonological
retrieval and processing speed (oral)
RSN Phonological Coding (e.g. nonword reading)
use letter sound relations to read and spell
(written) Orthographic Coding (e.g. spelling
choice) use word-specific letter patterns to
read and spell (written)
14
Motivation for Our Study

• Intersection of two interests.
• A. Testing the phonological hypothesis of RD
• B. Testing explanations of comorbidity
• SSD and RD are comorbid, but there are puzzles
  about their relation.
• A. Although SSD and RD share phonological
  deficits (in PA PM), most children with SSD
  do not become RD. Some even have PA deficits
  without RD.
• ? SSD RD are not same single deficit
  disorder. A PA deficit may not be
  sufficient to cause RD.
• B. The SSD LI subgroup has the highest risk
  for later RD, whereas outcome of SSD only
  varies across studies.
• ? Maybe the real risk factor is LI, not SSD.
• C. Persistent SSD has a higher risk for RD than
  resolved SSD.
• ? Maybe these two subtypes have different
  deficits.

15
Main HypothesesCommon vs. Specific Risk Factors
Severity Synergy
CE
E1
E2
E3
E1
Etiology Cognitive Processes Observed
Behavior
C1
CC
C3
C1
C2
SSDRD
RD
SSD
SSD
RD
SLI
SSDRDSLI
CE Common Etiology, CCCommon Cognitive
Phenotype E1, E2, E3Specific etiologies
C1,C2,C3Specific cognitive phenotypes.
16
Severity Hypothesis
RD
SSDRD
Maturity of Phonological Representations
17
Predictions of Severity Hypothesis
1) RD with and without earlier SSD is mainly
caused by a genetically mediated deficit in
phonological representations, but the
phonological deficit is more severe in SSD RD
than in RD alone. 2) SSD without later RD is a
distinct disorder from SSD RD, both
etiologically and cognitively.
18
SSD and RD Project (e.g. Raitano, et al, 2004)

• SSD (N101), Controls (N 41)
• Similar in age (5-6y), parental education,
  gender
• ethnicity. NVIQ lower in SSD
• Tested at age 5 and age 8 on oral and written
  language

19
SSD and RD Project Initial Results(Raitano, et
al, 2004)
1.) Entire SSD group worse than controls on PA
and letter knowledge (LK), but not RSN.
Normalized LI group worst on RSN. 2.) Within
SSD group, main effects of persistence and LI
status on PA, with NIQ covaried. Main effect of
LI on LK. 3.) Even the normalized SSD, no LI
subgroup worse than controls on PA.
20
SSD and RD Project Speech Perception and Nonword
Repetition
Speech Perception (Nittrouer, 1992) Additive
main effects of persistence and LI status on
training trials. The subset of SSD group that
passed training trials was more immature than
controls on test trials. Therefore, a deficit on
an input phonology task indicating less segmental
phonological representations. Nonword Repetition
(Dollaghan Campbell, 1998) Additive main
effects of persistence and LI status on NW Rep.
Only LI status interacted with syllabic length.
Covarying Digit Span eliminates only the LI main
effect. Therefore, at least two deficits evident
on this task. Summary A phonological deficit
is pervasive in SSD, so we are not finding the
non-phonological subtype of SSD predicted by the
severity hypothesis.
21
Tunick (2003) dissertation Description of SSD and
RD Probands and their Siblings
a Goldman Fristoe Articulation standard score
b Reading composite standard score
22
Group x Task Interaction in the Proband
Comparison,F(2,88)6.80, plt.01
n.s.
p lt.10
p .10
23
Group x Task Interaction in the Sibling
Comparison, F(2,82)3.60, plt.05
plt.01
n.s.
n.s.
24
Conclusions from Time 1 Studies

• Deficits in both explicit and implicit phonology
  are pervasive in SSD, but their severity varies
  as a function of LI persistence.
• Predictions
• a) If phonological deficits are a sufficient
  cause of literacy problems, then the entire SSD
  group, including normalized no LI subgroup, will
  have worse literacy outcome than controls.
• b) Both LI status and persistence will
  contribute additively to literacy outcome.
• Contrary to the severity hypothesis, SSD and RD
  have a similar deficit in PA, but differ on RSN.
  The SSDLI subgroup, however, has a deficit in
  RSN.
• Prediction
• Intact RSN may be a protective factor in SSD
  without LI.

25
Figure 1. Multiple Deficit Model
Non-Independence at each Level Gx E
interaction G-E Correlation Interactive
Development Comorbidity
Level of Analysis Etiologic Risk and
Protective Factors Cognitive
Causes Complex Behavioral Disorders
G1
E1
G2
E2
G3

Phon
Sem
RSN
RD
SSD
SLI
KEY G genetic risk or protective factor, E
environmental risk or protective factor, Phon
Implicit phonological representations, Sem
Semantics, RSN Rapid serial naming, RD
Reading disability, SSD Speech sound disorders,
SLI Specific language impairment

26
Time 2Literacy Outcome in SSD Questions

• Is our sample of children with SSD at increased
  risk for literacy problems?
• Does literacy risk vary by SSD subtype?
• Which hypothesis, severity, synergy, or
  multiple deficit, best accounts for literacy
  outcome in SSD?
• Is RD in SSD similar to RD in general?

27
A) Categorical Risk for RD

• Somewhat elevated Across 11 definitions of RD,
  rate of RD in entire SSD group is 1.6-4.3 times
  higher than rate in controls (median 2.8). But
  risk difference significant for only 4
  definitions.
• Severe subgroup For 2 most severe definitions,
  SSlt85 on all four or three (no RC) literacy
  tests, rates are
• 9-11 in SSD and 0 in controls.
• C) RD Definition for Later Analyses 1.5 SD
  below Control mean on average of 4 literacy
  tests.
• SSD Controls
• 23.8 (N19) 5.9 (N2)
• Fishers exact p 0.03
• This rate of RD in SSD consistent with previous
  studies.

28
A) Mean Literacy Scores (NIQ Covaried)




Group Main Effect F(1,110)9.77, plt.01
pairwise comparisons significant at p lt.01
29
A) Mean Literacy Component Process Scores (Age
and NIQ covaried)

p .06

Ortho Choice NW Spell
NW Read Group Main Effect
F(1,109)10.53, plt.01
30
Literacy Outcome in SSD Answers

• Our sample of SSD children have literacy deficits
  relative to controls across all dimensions of
  literacy word recognition, spelling, fluency,
  comprehension, orthographic and phonological
  coding.
• But the large majority are not RD and their mean
  literacy scores are close to national norms.

31
B) Literacy Risk by LI Subtypes Effect of LI,
but mostly not Persistence
Whole table Pearson Chi-Square 41.38, p lt0.001
Just SSD Pearson Chi-Square 30.134, plt.001,
Fishers Exact Test, plt.001 So about two thirds
(13/19) of SSDRD group have LI In terms of
persistence, no difference on diagnostic reading
variables, but persistent SSD worse at non-word
spelling (plt.05)
32

• Mean Literacy Scores LI Subtypes
• (NIQ covaried)

Effect size comparison for literacy outcome
(Cohens d) SSD SSDLI Lit.
0.18 0.95 Our Study 0.64 1.90
p .08
p .08





Group Main Effect F(1,109)11.15, plt.001
Pairwise comparison significant at p lt.01
33
B) Mean Literacy Component Process scores (Age
NIQ covaried)
a
a
a
a
b
b
c
c (trend)
b
Ortho Choice NW Spell NW
Read Group Main Effect
F(2,108)12.08, plt.001 Trend (p.07) for
interaction Different subscripts denote
significant simple effects
34
B) Subtypes Normalized SSD, no LI (N 41)
p .052
p.08

n.s.
Main effect of literacy measure (F 4.74, p lt
0.01) Trend for a main effect of group (F
2.96, p 0.095) No group literacy interaction
(F lt 1)
35
B) Subtypes Norm, no LI (N41)
n.s.
n.s.

No main effect of literacy measure (F lt 1) Main
effect of group (F 4.10, p lt 0.05) No group
literacy interaction (F lt 1) p lt.05 for
simple effect
36
B) Subtypes Norm, no LI (N41)
Repeated Measures ANOVA with NIQ Covaried No
main effect of predictor (F lt 1) Main effect of
group (F 12.11, p 0.001) Group Predictor
interaction (F 4.91, p 0.001)
Post-hocs PM, Syntax, Semantics
Control gt SSD Norm/no LI PA/Reps
Control gt SSD Norm/no LI (trend) RN
Control SSD Norma/no LI
37
Literacy Outcome in SSD Answers

• Our sample of SSD children have literacy deficits
  relative to controls across all dimensions of
  literacy word recognition, spelling, fluency,
  comprehension, orthographic and phonological
  coding.
• But the large majority are not RD and their mean
  literacy scores are close to national norms.
• B) Literacy outcome varies only by LI subtype,
  not by persistence. LI accounts for most of the
  risk for RD. Despite phonological and language
  deficits, literacy outcome is about average for
  the normalized SSD/no LI subgroup, and RN appears
  to be protective.

38

• Component language processes at Time 1 by RD
• status
• (IQ covaried z-scores relative to control group)

Time 1 profile
Simple effects Controls gt SSD alone gt SSDRD
for all pairwise group comparisons except for RN,
where Controls SSD alone gt SDRD

• Main effect of group, F(1,75) 10.81, p lt .01
• Main effect of language component , F(5, 225)
  2.67, plt.05
• No group by language component interaction

39

• Oral Language Component Processes at Time 2 by RD
  status
• (NIQ covaried z-scores based on control group)

Simple effects Controls gt SSD alone gt SSDRD
for Phon. Mem., PAreps, Semantics, and
Syntax. Controls SSD alone gt SSDRD for
RN Controls SSD alone SSDRD for List.
Comp. (last comparison is trend at p.065)

• Main effect of group, F(1,75) 19.34, p lt .001
• No main effect of language component
• Group by language component interaction, F(5,
  298) 2.33, p.057

40
C) Predicting Literacy at Time 2 From Language
Components at Time 1

• plt.001, plt.01, plt.05, plt.10
• Articulation, Semantics, Syntax do not predict
  unique variance in reading once Language or
• PA/PM have been included in the model.
• Language predicted by PA LTR at time 1
  R-squared .85, Betas significant at plt.001
• PARN and LTRPA interactions were not
  significant.
• Regression analyses were consistent when
  examining SSD probands only.

41
C) Predicting Literacy Outcome at Time 2 From
Language Components at Time 2

• plt.001, plt.01, plt.05,
  plt.10
• PARN interaction was not significant.
• Regression analyses were consistent when
  examining SSD probands only.

42
Literacy Outcome in SSD Answers

• Our sample of SSD children have literacy deficits
  relative to controls across all dimensions of
  literacy word recognition, spelling, fluency,
  comprehension, orthographic and phonological
  coding.
• But the large majority are not RD and their mean
  literacy scores are close to national norms.
• Literacy outcome varies only by LI subtype, not
  by persistence. LI accounts for most of the risk
  for RD. Despite phonological deficit, literacy
  outcome is about average for norm., no LI
  subgroup.
• C) Several results favor multiple deficit over
  severity synergy hypotheses.

43
D) Is RD in SSD like RD in general?
Comparing SSD RD to young twins with RD Group
criteria
44
D) Samples
Note unequal variances for age. Less variance
in twin sample.
45
D) Literacy scores in controls
Comparing control groups directly on diagnostic
literacy standard scores (WIAT or PIAT)
Note unequal variances for all measures. Less
variance in twin sample.
46
D) Literacy scores in RD groups
Comparing clinical groups on diagnostic literacy
measures Looking directly at standard scores on
the diagnostic literacy measures, it appears that
the SSD RD group is more severe than the Twin
RD group
47
D) Prediction variables
Comparing clinical groups on literacy predictor
variables All measures converted into z-scores
based on control group performance.
48
D) Language predictor variables
Main effect of clinical group F 23.65, p lt
0.001 Still a main effect of clinical group if we
remove SSD/no RD group (F 38.1, p lt
0.001) Clinical group predictor variable
interaction F 4.69, p lt 0.001 Without SSD/no
RD group, interaction is even stronger F 8.62,
p lt 0.001 Simple effects PA, PM SSD/no RD gt
Twin RD SSD RD RN SSD/no RD gt Twin RD gt
SSD RD MPS, VC SSD/no RD Twin RD gt SSD RD
49
Literacy Outcome in SSD Answers

• Our sample of SSD children have literacy deficits
  relative to controls across all dimensions of
  literacy word recognition, spelling, fluency,
  comprehension, orthographic and phonological
  coding.
• But the large majority are not RD and their mean
  literacy scores are close to national norms.
• Literacy outcome varies only by LI subtype, not
  by persistence. LI accounts for most of the risk
  for RD. Despite phonological deficit, literacy
  outcome is about average for norm., no LI
  subgroup.
• Several results favor multiple deficit over
  severity synergy hypotheses.
• RD in SSD has a similar PA PM deficit to RD in
  general, but has more severe deficits in RN, MPS,
  language.

50

• New Directions
• 1. Conduct SEM analyses from Time 1 to Time 2.
• Test young RD group on Time 2 measures.
• 3. Expand Time 1 sib pairs sample for genome
  scan.