Introduction to Psycholinguistics

1
Introduction to Psycholinguistics Brian
Byrne University of New England
Genes and environment in early literacy
development Change and continuity from
preschool to Grade 2 Brian Byrne University of
New England with Will Coventry, Richard Olson,
Stefan Samuelsson, Jacqueline Hulslander, Angie
Friend, Sally Wadsworth, Robin Corley, John
DeFries, Erik Willcutt
2
Overview
Text and speech Brief primer on behavioural
genetics Data from International Longitudinal
Twin Study Preschool Kindergarten Grade
2 Continuity and change Exploring the processes
Textual specificity at the word level
3
Text and speech
Variability in processing of written and spoken
language Three reasons for believing its
greater for text Human history Simple view of
reading R D X L Common sense
4
Sources of variance?
Genes, environment the classic nature/nurtue
question. Selection pressures in human history
greater genetic variance for text? (Why do Swedes
read so well? Since the time of Charles XI A
Protestant in order to marry must be able to
read Gods word directly.the priest, with
increasing rigour, made annual tours of every
household, testing the reading and understanding
level of every Swede with respect to Luthers
Little Catechism.)
5
Primer on behavioural genetics The twin method
6
Twin method
So for traits that are highly heritable, MZ twins
will be very similar, DZ twins about half as
similar For traits that are highly influenced by
shared environment, both types of twins will be
equally and highly similar For traits that are
highly influenced by unique environment, twins
will be no more alike that randomly selected
individuals Thus, genetic, shared environment
and unique environment influences on a trait can
thus be estimated from the degree of similarity
between twins as a function of zygosity
7
Types of similarity

• For categorical traits Concordance
• For continuous traits Correlation

8
Concordance
Diagnosis of schizophrenia MZ twins 48 DZ
twins 17
9
Correlation
Intelligence MZ twins .85 DZ twins .60
10
Partitioning variance in a trait
total variance
genetic variance
environmental variance
shared environment
unique environment
11
Twin method Some terminology and analytic tools

• a2 additive genetic variation
• c2 shared environmental variation
• e2 unique environment variation (includes
  measurement error)

12
Univariate data structure for continuous traits

• MZ twins DZ twin
• Twin 1 Twin 2 Twin 1 Twin 2
• X X X X
• X X X X
• X X X X
• . . .
  .
• . . .
  .
• If MZr gt DZr, evidence that Variable X is
  influenced by genes, as in IQ example

13
Multivariate extension

• MZ twins DZ twin
• Twin 1 Twin 2 Twin 1 Twin 2
• X Y X Y
• X Y X Y
• X Y X Y
• If MZr gt DZr, evidence that Variables X and Y are
  influenced by same gene(s)

14
Cholesky decomposition

• A method of analyzing genetic and environmental
  relations among variables
• Similar in principle to hierarchical regression
  in that e.g. genetic influence on a variable is
  examined after the genetic influence on a
  correlated variable is removed.

15
Cholesky decomposition structure

X
a11
A1
Latent variables
Manifest variables
a12
a21
Y
A2
16
Cholesky decomposition structure

X
c11
C1
c12
c21
Y
C2
17
Cholesky decomposition structure

X
e11
E1
e12
e21
Y
E2
18
Cholesky decomposition example
a2 .702 .49

X
.70
A1
.50
.40
Y
A2
a2 .502 .402 .41
19
Cholesky decomposition example
a2 .802 .64

X
.80
A1
.60
.00
Y
A2
a2 .602 .36
20
Cholesky decomposition example
a2 .802 .64

X
.80
A1
.00
.60
Y
A2
a2 .602 .36
21
Issues concerning twin method

• Equal environments assumption
• Gene-environment correlation
• Assortative mating
• The process of twinning
• Prematurity
• eg earlier exposure to high frequency sounds
• MZ twins more vulnerable
• longer, more variable NICU stays greater
  birth-weight differences, etc

22
ILTS

• Sites in Australia, USA, Norway, Sweden
• To date approximately 2000 children (1000 sets of
  twins) assessed in seven successive cohorts
  starting at age 4-5
• Virtually 100 retention rate from year to year

23
ILTS Twins seen in homes or pre-schools prior to
kindergarten
Note Each child assessed by a different tester
24
They are tested after kindergarten year
25
And again after first grade
26
and after second grade
27
ILTSPreschool
Verbal memory
Phonological awareness
Verbal fluency
28
ILTSPreschool (cont.)
Print knowledge.
Vocabulary
Syntax/morphology
29
ILTSKindergarten
Word reading
Spelling
Phonological awareness
30
ILTSGrade 1
Word reading
Reading comprehension

Spelling
Rapid naming
31
ILTSGrade 2
Word reading
Reading comprehension

Spelling
Vocabulary
32
TEDSAge 7 Reading
Boys

Girls
33
Gene-environment interactionsKindergarten word
reading
U.S.
Scandinavia
Australia
34
but in Grade 1
Scandinavia
U.S.
Australia
35
Longitudinal analysisPreschool - Kindergarten
print knowledge
A1
.22
.37
preschool
A2
phon. aw.
.25
.17
verb. fluency
.52
.34
A3
kindergarten
.17
A4
word reading
36
Preschool Kindergarten (cont.)
print knowledge
C1
.72
.27
preschool
C2
phon. aw.
.08
.06
verb. fluency
.10
X
word reading
kindergarten
37
Longitudinal analysisPreschool Grade 2
print knowledge
A1
.21
.19
preschool
A2
phon. aw.
.30
.09
.27
verb. fluency
.40
A3
.10
Grade 2
.39
A4
word reading
38
Longitudinal analysisPreschool Grade 2
print knowledge
A1
.21
.19
preschool
A2
phon. aw.
.30
.08
.27
verb. fluency
.42
A3
Grade 2
.42
A4
spelling
39
VocabularyPreschool and Grade 2

Preschool
Grade 2
40
Exploring processes

Genes Cognition
41
Dyslexia Loci
42
Chromosome 15 EKN1
43
Chromosome 6 KIAA0319
44
Chromosome 6 DCDC2
45
Congitive processes

• Genes that influence current achievement levels
  are also (in part) genes that influence new
  learning

46
Grade 2 spelling, orthographic learning and
decoding

• Spelling achievement measure
• WRAT-R Spelling
• Orthographic learning measure (after Share)
• Children read texts that include a novel word
  with ambiguous spelling possibilities, like
• The new word is Feap. There are lots of rocks on
  the planet Feap. Feap is very far away. To get
  to Feap you need a rocket
• Decoding
• Errors in text reading

47
Data 321 MZ, 314 DZ pairs
Orthographic learning
Decoding
rg .95
Spelling
rg .85
rg .97
48
Decoding and orthographic learningTwo models
A
C
E
A
C
E
D
OL
D
OL
A
C
E
A
C
E
Cholesky
Direction of causation
49
Our preferred model
current achievement (LK, D, Sp)
genes for learning
new learning (OL)
50
summary

• Text processing (reading and spelling) are highly
  heritable in early schooling
• Genetic influence is in evidence prior to formal
  instruction
• Some of these early genetic sources continue to
  exert influence as chidlren develop in literacy
• Other genes come on stream during development
• There is evidence of gene-environment interaction
• We propose a genetically-influenced learning
  parameter as common to several literacy processes

51
acknowledgements

• Australian Research Council
• Australian Twin Registry
• National Institutes for Health (NICHHD)
• Norwegian Research Council
• Swedish Research Council
• Universities of New England, Colorado, Stavanger
  and Linköping
• Our many testers and coordinators
• Our many twins and their families