Transcript Titel van de presentatie

Behavioural Science Institute
Nijmegen, the Netherlands
Working memory, long-term memory, and reading:
The case of catastrophe theory versus regression analysis
Anna M. T. Bosman
Fred Hasselman
Ralf Cox
EWOMS-2006
W4 = (Who*What*Why*Where)
Who: Scientists & Practitioners
What: Reading and reading difficulties
Why:??????
Where: are we now????
EWOMS-2006
Memory & Reading
STRAND
S
/ s/
T
/ t/
R
/ ɪə/
A
/e/
N
/ n/
D
/ d/
Likely reading errors: /stand/, /sand/, /trend/, /spend/, /rand/
DEAR
/ ɪə /
PEAR
/ eə /
DEAD
/e/
BREAK
/ eɪ /
EWOMS-2006
Memory & Reading
Beware of heard, a dreadful word
That looks like beard and sounds like bird,
And dead: it's said like bed, not bead For goodness sake don't call it deed!
EWOMS-2006
Tests
Working memory : Digit Recall
Backward Digit Recall
Block Recall
Long Term Memory:
12-Words Test
Reading level decoding:
DMT: Score = Ncorrect words / minute
EWOMS-2006
Experiment 0
99 Dutch, Grade-1 students (mean age 80 months)
46 without and 53 with reading delays
Test
without RD
with RD
Significance
WM: Digit recall
22.5
22.2
F<1
WM: Block recall
22.4
21.6
p > .30
WM: Backward recall
8.7
7.2
p < .005
LTM: capacity
7.4
6.3
p < .01
EWOMS-2006
Working Memory and remediation
Digit recall: RemediationSuccessful > RemediationUnsuccessful p <
.01
Backward recall: RemediationSuccessful =RemediationUnsuccessful
Block recall: RemediationSuccessful = RemediationUnsuccessful
EWOMS-2006
Long-term memory and remediation
Build-up
significant linear and quadratic trends
Capacity
RemediationSuccessful > RemediationUnsuccessful p < .05
EWOMS-2006
Multiple linear regression model
Y = b0 + b1 X1 + b2 X2
EWOMS-2006
Change
 Is at the heart of psychology (everything!)
 What we want to achieve in RD
So why not study it in terms of a dynamics?
EWOMS-2006
Catastrophe models
 Describe dynamical systems in terms of
mathematics
 Enable us to understand discontinuities in
behaviour (i.e., change over time)
 With the help of so-called control
parameters
EWOMS-2006
Un po’ di matematica
dx
 f (x )
dt
dV ( x )
f ( x)  
dx
 x is the psychological variable of interest (i.e.,
reading success)
 V is a potential function describing the
possible states in which x might eventually
occur
EWOMS-2006
Potential function of the Cusp-catastrophe model
V ( x; α, β)  x  βx  αx.
1
4
4
1
2
2
 α en β are control parameters determining the exact
shape of the function.
x = ‘order parameter’
α = ‘asymmetry parameter’
β = ‘bifurcation parameter’
EWOMS-2006
Non-linear or Cusp-catastrophe model
V ( x; α, β)  x  βx  αx.
1
4
4
1
2
2
EWOMS-2006
Ancora un po’ di matematica
Dynamic systems tend to seek particular end states,
called attractors (the variable x does not change
anymore)
In terms of mathematics, we need to establish when
dx
 0.
dt
or
dV ( x)
f ( x)  
 0.
dx
EWOMS-2006
Canonical cusp-surface equation
x  βx  α  0.
3
Bifurcation parameter: LTM
Asymmetry parameter: WM
EWOMS-2006
Experiment 1
• 47 Dutch, Grade-1 students with reading problems
– 25 boys
– 22 girls
• Mean age = 80 months (SD = 5); at memory assessment
• Assessment
– Memory: October/November 2003
– Reading level 1: January/February 2004
– Reading level 2: June/July 2004
EWOMS-2006
Results: Linear difference model
dx = b1LTM + b2WM + b3
R2
Model
LTM
WM: Digit recall
.05
n.s
LTM
WM: Backward recall
.04
n.s
LTM
WM: Block recall
.02
n.s
Factors
EWOMS-2006
Results: Linear interaction model
dx = b1LTM + b2WM + b3LTM*WM + b4
R2
Model
LTM
WM: Digit recall
.05
n.s
LTM
WM: Backward recall
.05
n.s
LTM
WM: Block recall
.08
n.s
Factors
EWOMS-2006
Results: Linear pre-post model
x2 = b1LTM + b2WM + b3x1 + b4
R2
Model
LTM + Digit recall
Reading ***
.56
p < .0001
ß = .74
LTM + Backward recall
Reading ***
.57
p < .0001
ß = .73
LTM +Block recall
Reading ***
.57
p < .0001
ß = .74
Factors
EWOMS-2006
Results: Non-linear Cusp-catastrophe model
dx = b1x13 + b2x12 + b3LTMx12 + b4WM + b5
LTM
2
Z
3
Z
ß = .23
ß = .81
ns
p < .0001
ß = -2.2
p < .04
ß = 2.2
p < .04
ß = .17
ß = .80
ns
p < .0001
ß = -1.9
p < .01
ß = 1.6
p < .02
ß = .20
ß = .80
ns
p < .0001
ß = -2.1
p < .02
ß = 2.2
p < .03
All models
p < .0001
R2
WM
Digit recall
.61
Backward
recall
.62
Block recall
.62
EWOMS-2006
What did we learn?
• Scientifically
LTM, WM, and Reading are dynamically related. Thus,
the search for independent components as causal
mechanisms seem futile
• Practically
impossible to predict reading-remediation success
based on LTM and WM levels.Thus:
EACH CHILD DESERVES THE EXTRA HELP!
EWOMS-2006
Many thanks to
Tom Braams, MA for keeping us in touch with daily practice
Marion IJntema-de Kok, MA for running the Experiment
Braams & Partners, Instituut voor Dyslexie
Deventer, the Netherlands