Transcript Kein Folientitel - uni

A usage-based approach to grammatical
development
Holger Diessel
University of Jena
[email protected]
http://holger-diessel.de/
Language learning
• Connectionism (Rumelhart & McClelland 1986;
Elman et al. 1996; Lewis and Elman 2001).
• Corpus studies of the ambient language (Redington
et al. 1998; Mintz et al. 2002; Monaghan et al. 2005).
• Experimental studies with infants (Saffran et al. 1996;
Saffran 2002; Newport and Aslin 2004).
Construction grammar
Meaning
Meaning
Form
Form
Meaning
Form
Meaning
Form
Meaning
Form
One-word utterances / holophrases
Daddy.
[Adam 1;4]
Mommy.
[Adam 1;4]
Doggy.
[Adam 1;5]
Milk.
[Adam 1;5]
Allgone.
[Adam 1;6]
Lexically-specific constructions
Take this key off .
Take this paper off.
More corn.
More cookies.
Block get-it.
Take that off .
More mail.
Mama get-it.
Take this dress off.
More popsicle.
Bottle get-it.
Take that belt off me.
More jump.
Towel get-it.
Take it off.
More Peter water.
Books get-it.
Phone get-it.
Emergence of schematic constructions
VERB __
Get __
Get doggy
Get milk
Get him
Get Billy
Hypotheses
• Relative clauses form a network of related
constructions that children acquire in a piecemeal,
bottom-up fashion.
• The development originates from relative constructions
that are only little different from simple sentences.
• The development can be seen as an example of
‘abductive constructivist learning’.
Study 1 (Diessel 2004)
Adam
Sarah
Nina
Peter
Naomi
Age range
Finite
2;3-4;10
2;3-5;1
1;11-3;4
1;9-3;2
1;8-3;3
1;9-5;1
178
32
62
25
8
305
Nonfinite
120
36
71
44
16
287
Head of the relative clause
(1) The man who we saw was reading a book.
SUBJ
(2) He noticed the man who was reading a book.
OBJ
(3) He saw to the man who was reading a book.
OBL
(4) The man who was reading a book.
NP
(5) That’s the man who was reading a book.
PN
Head of relative clause (total)
60
50
48,5
proportions
40
30
23,8
21,5
20
10
5,6
0,7
0
PN
NP
OBJ
OBL
SUBJ
Head of relative clause (earliest)
90
80
80
70
proportion
60
50
40
30
20
7,5
10
10
2,5
0
0
PN
NP
OBJ
OBL
SUBJ
Head of relative clause (development)
90
80
70
proportion
60
50
PN
OBJ
PN
NP
OBJ
NP
OBL
OBL
40
30
20
10
OBL
SUBJ
0
3;0
4;0
age
5;0
Input frequency
PN-relatives are among the most frequent relative
clauses in the ambient language, but they are not as
frequent in the ambient language as in the children’s
data.
Semantic complexity
(1)
Here’s the tiger that’s gonna scare him.
> The tiger is gonna scare him.
(2)
This is the sugar that goes in there.
> The sugar goes in there.
(3)
It’s something that you eat.
> You eat something.
Semantic complexity
(1)
You left this toy I’m playing with.
> You left this toy. + I’m playing with the toy.
Information structure
The information structure of PN-relative constructions
is similar to the information structure of simple
sentences, i.e. they do not include presupposed
information.
Pragmatic function
PN-relatives are pragmaticlly very useful in parent-
child speech: They occur in constructions that focus
the hearer’s attention on elements in the surrounding
situation.
Conclusion
PN-relatives are the earliest relative clauses that
children learn because:
(1)
they suit the communicative needs of
young children
(2)
they are semantically similar to simple
sentences.
Syntactic amalgams
(1)
That’s doggy turn around.
[Nina 1;11]
(2)
That’s a turtle swim.
[Nina 2;2]
(3)
Here’s a mouse go sleep.
[Nina 2;3]
(4)
That’s the roof go on that home. [Nina 2;4]
(5)
That’s the rabbit fall off.
[Nina 2;4]
Relativizsed syntactic role
(1) The man who met the woman.
subj
(2) The man who the woman met.
obj
(3) The man who the woman talked to.
obl
(4) The man who the girl gave the book to.
io
(5) The man whose dog bit the woman.
gen
Relativized syntactic role (total)
70
60
57,3
proportion
50
37
40
30
20
10
5,7
0
subj
obj
obl
0
0
io
gen
Relativized syntactic role (development)
90
80
70
proportion
60
obj
50
subj
40
30
20
10
obl
0
>3;0
3;0-4;0
age
4;0-5;0
Diessel & Tomasello (2005)
This is the girl who saw Peter on the bus this morning.
This is the girl who the boy teased at school yesterday.
This is the girl who Peter borrowed a football from.
This is the girl who Peter played with in the garden.
This is teh girl whose horse Peter heard on the farm.
subj
do
io
obl
gen
Das ist der Mann, der mich gestern gesehen hat.
Das ist der Mann, den ich gestern gesehen habe.
Das ist der Mann, dem ich das Buch gegeben habe.
Das ist der Mann, mit dem ich gesprochen habe.
Das ist der Mann, dessen Hund mich gebissen hat.
subj
do
io
obl
gen
Results
80
70
70
71,1
60
60
50
50
40
40,5
40
31
30
32,8
30
31,5
21,4
20
20
12
10
10
0
68,5
2,4
subj
do
io
obl
gen
0
0,5
subj
do
English
io
obl
German
subj vs. do
p =. 001
subj vs. do
p =. 001
do vs. io
p = .173
do vs. io
p = .061
Do vs. obl
p = .169
io vs. obl
p = .001
gen
Subj-relatives
Do-, io-, and obl-relatives were often converted to subjrelatives.
English
ITEM:
CHILD:
This is the girl who the boy teased at school.
This is the girl that teased … the boy … at school.
German
ITEM:
Da ist der Mann, den das Mädchen im Stall gesehen hat.
CHILD: Da ist der Mann, der das Mädchen im Stall gesehen hat.
Subj-relatives
However, children were not consistent in their
performance.
In addition, they often repaired their conversion errors
before they reached the end of the sentence:
(1) This is the girl who bor/ Peter borrowed a football from.
(2) Da ist der Junge, der/ dem Paul … die Mütze weggenommen hat.
Hypothesis
The conversion errors are due to the fact that subj-
relatives are more easily activated than other types of
relative clauses.
Frequency and ease of activation
The more frequently a grammatical construction
occurs, the more deeply entrenched it is in mental
grammar, and the easier it is to activate in language
use.
Input frequency
60
53.8
50
40
35.6
30
20
7.7
10
0
0
IO
GEN
0
SUBJ
OBJ
OBL
(Diessel 2004)
Subj-relatives and simple sentences
Children’s good performance on subject relatives can be
explained in terms of the similarity between subject
relatives and simple sentences.
AGENT
VERB PATIENT.
PRO is AGENT
rel VERB PATIENT.
PRO is PATIENT rel AGENT VERB.
Simple clause
subj
do / io / obl
Word order in English relative clauses
NP [V …]
subj
NP [NP V …]
do
NP [NP V …]
io
NP [NP V …]
obl
NP [[GEN N] V …]
gen
Relative pronouns in German relative
clauses
Der Mann, der …
subj
Der Mann, den …
do
Der Mann, dem …
io
Der Mann, mit/von dem …
obl
Der Mann, dessen N
gen
Gen- and io-relatives
Both gen- and io-relatives are basically absent from the
ambient language.
Io-relatives caused fewer errors than gen-relatives
because they are similar to do-relatives.
Summary
Important is the similarity between constructions:
• Subj-relatives caused few problems because they are
similar to simple sentences.
• English do-, io-, and obl-relatives caused basically the
same amount of problems because they have the same
word order.
• Io-relatives caused relatively few problems because they
are similar to direct do-relatives.
• Gen-relatives and German obl-relatives caused great
problems because they are dissimilar to other relative
clauses.
Why does similarity matter?
Relative clauses are constructions (i.e. form-function
pairings) that are related to each other in a network like
lexical expressions.
Children acquire this network in a piecemeal, bottom-up
fashion by relating new relative clause constructions to
constructions they already know.
A network of relative constructions
… [gen-relative]
…-relatives
…-relatives
…-relatives
That is N [subj-relative]
Simple Sentences
References
Diessel, Holger. 2004. The Acquisition of Complex Sentences.
Cambridge: Cambridge University Press.
Diessel, Holger & Tomasello, Michael. (2005). A new look at
the acquisition of relative clauses. Language.