Models of speech errors a developmental study

1
Models of speech errors - a developmental study
Mary-Jane Budd University of Essex Supervised
by Prof. Rick Hanley Dr. Yvonne Griffiths
2
Overview

• introduction to the Dell model of speech
  production
• findings from adult aphasic data
• childrens picture naming errors
• theoretical issues
• relationship between picture naming and reading

3
Models of speech production in picture naming
General consensus
semantic level - conceptual meaning from visual
input
lexical level - nonphonological representation of
the word
phonological level - phonological encoding of the
word
4

Dell model of picture naming - cat (eg. Dell,
1997 Foygel Dell, 2000)
semantic layer
rat
mat
lexical layer
dog
cat
fog
phonological layer
f
k
m
ae
o
t
g
r
d
cat
output
5
Errors on picture naming - semantic error
semantic layer
rat
mat
lexical layer
dog
cat
fog
phonological layer
f
k
m
ae
o
t
g
r
d
dog
output
6
Errors on picture naming - formal error
semantic layer
rat
mat
lexical layer
dog
cat
fog
phonological layer
f
k
m
ae
o
t
g
r
d
mat
output
7
Errors on picture naming - mixed error
semantic layer
rat
mat
lexical layer
dog
cat
fog
phonological layer
f
k
m
ae
o
t
g
r
d
rat
output
8
Errors on picture naming - unrelated error
semantic layer
rat
mat
lexical layer
dog
cat
fog
phonological layer
f
k
m
ae
o
t
g
r
d
fog
output
9
Errors on picture naming - nonword error
semantic layer
rat
mat
lexical layer
dog
cat
fog
phonological layer
f
k
m
ae
o
t
g
r
d
cag
output
10
Models of picture naming - cat
Foygel Dell, 2000
Dell et al, 1997
semantic layer
semantic parameter s
weight parameter w
rat
mat
lexical layer
dog
cat
fog
decay parameter d
phonological parameter f
phonological layer
f
k
m
ae
o
t
g
r
d
cat
output
11
Models of picture naming - cat
Superior model Schwartz et al, 2006 - case-series
test of 94 patients. 76/94 with adequate fit.
Dell et al, 1997 vs Foygel Dell,
2001 weight/decay model semantic/phonological
model
12
Errors in picture namingusing the Foygel Dell
(2001) model

• Semantic-lexical level deficit will lead to

• Lexical-phonological level deficit will lead to
• nonword

• semantic
• formal
• mixed

real word errors nonword errors
13
Adult aphasic picture naming examples - data
taken from Hanley et al (2004)
PS makes more nonword errors
phonological parameter deficit MF makes more real
word errors semantic parameter deficit
14
Childrens picture naming and the Dell model

• 43 children 15 reception age (m55) 14 year 3
  (m84) 14 year 6 (m112) from a mainstream
  primary school in Ipswich
• 56 item picture set using the same items Nation,
  Marshall, Snowling (2001) - 1/2 low/1/2 high
  frequency, 1/2 short/1/2 long names, matched for
  imageability, AoA, name agreement
• picture naming task and picture-word matching
  task (56 item set were each matched with a
  semantic foil, a phonological foil, and a visual
  foil.

15
Picture naming task
Participant has to name picture
beep

1000 ms
Responses are recorded using sound studio
0 ms
16
Picture-word match example
17
Childrens picture naming and the Dell model-
results
Correctness Using only items correct in
picture-word matching task Reception 0.66 (range
.55 - .76) Year 3 0.82 (range.73 - .89) Year
6 0.82 (range.67 - .91) main effect of age
(F(2, 40) 35.6 plt.001)
18
Childrens picture naming and the Dell model -
results

19
Childrens picture naming and the Dell model -
results
main effect of model (F(1, 40) 35.6 p lt .001)
Reception 4/15 Year 3 14/14 Year 6 12/14
reach the criteria used in the aphasic literature
for adequate model fit to the actual data (using
the s/p model of Foygel Dell, 2001)
20
Childrens picture naming and the Dell model-
results
Does the strength of the semantic and
phonological parameters increase with age?
main effect of age (F(2, 40) 20.9 p lt .001),
21
Summary of Findings
The models appear to struggle with the very young
childrens error data but are successful at
simulating the picture naming performance of
children from year 3 onwards.

• Model Fit - the Foygel Dell (2000) model
  outperformed the Dell et al (1997) model at all
  age groups but the model fit was least good on
  the reception data.
• JF pattern (Foygel Dell, 2000) was observed
  in the reception group data
• -gt under prediction of semantic and mixed errors
• -gt over prediction of formal and unrelated
  errors

22
Other theoretical issues

• can we predict the auditory repetition
  performance of the children from their picture
  naming responses?
• relationship between reading and picture naming

23
Reading and picture naming
Shared processes written word/ picture
semantic system
lexicon
phonological representation
output
24
Literature on the relationship between reading
picture naming

• Examples
• Katz, 1986
• Wolf Obregon, 1992
• Swan Goswami, 1997
• Nation, Marshall, Snowling, 2001

25
The Next Step
Is there a relationship between the strength of
the semantic and/or phonological parameters and
reading? Reading measures on reception and year
3 children Reception - floor Year 3 -
ceiling Repeat study using year 2 children
(age 7)
26
Thank you
mbudd_at_essex.ac.uk