Acquisition de la syntaxe.

1
Acquisition de la syntaxe.

• Anne Christophe
• Laboratoire de Sciences Cognitives et
  Psycholinguistique,
• CNRS-EHESS-ENS, Paris
• http//www.lscp.net
• http//www.lscp.net/persons/anne
• enseignement cours Acquisition de la syntaxe

2
Créolisation

• Aujourdhui deux articles.
• Singleton Newport (2004). When learners surpass
  their models The acquisition of American Sign
  Language from inconsistent input. Cognitive
  Psychology, 49, 370-407. Présenté par Perrine
  Brusini
• Carla Hudson Kam Elissa Newport (2005),
  Language Learning and Development, 1, 151-195.

3
When learners surpass their models The
acquisition of American Sign Language from
inconsistent input. Singleton Newport (2004).
Cognitive Psychology, 49, 370-407.
4
Regularizing unpredictable variation The roles
of adult and child learners in language formation
and change.

• Carla Hudson Kam Elissa Newport (2005),
  Language Learning and Development, 1, 151-195.

5
Language description

• 51 words 36 nouns, 7 intransitive verbs, 5
  transitive verbs, 2 determiners corresponding to
  2 noun classes gender (arbitrary) and mass/count
  (semantic)
• S -gt (NEG) V NPsubj (NPobj)NP -gt N
  DET1 (if NC1/mass)NP -gt N DET2 (if
  NC2/count)
• apparition de DET 45, 60, 75, 100

6
Experimental design

• Exposure auditorily presented sentences (with
  English prosody and phonology) together with
  matching videos
• ex flim rongmout po blergenfal po hit
  bowling-ball det2 bowling-pin det2 The bowling
  ball hits the bowling pin
• exposure set 230 sentences (out of 13200
  possible ones, even taking into account semantic
  constraints)
• 6 exposure sessions (25mn each), 1 test session
  (on 7 different days)
• 40 Subjects in 8 groups (4 consistency conditions
  x 2 groupings of nouns, 5 ss per group)

7
4 tests

• vocabulary (provide name for object on monitor)
• sentence completion task watch video, hear verb
  (ex gernfall), have to complete sentence (ex
  gern kowalt ka fall cotton det). 24 test
  sentences (12 trans, 12 intrans), 36 NPs, 36
  opportunities to produce a det 12 different
  nouns.
• Grammaticality judgment task listen to 123
  sentences, judge if they like or dislike them
  (1 to 4)
• gern ferluka po (correct det follows noun)
• gern po ferluka (incorrect det precedes
  noun)
• gern ferluka (incorrect no det)
• Forced choice grammar test 16 sentence pairs
  (verb subcategorization presence of verb).
  Results 15/16 correct

8
Exp.1 adults, sentence completion task
Effect of consistency, no interaction with noun
classes (arbitrary vs count/mass) group data
could reflect average on individuals who formed
different rules? NO. (NB were incorrect
determiners produced?)
9
Exp. 1 adults, production task
10
Exp. 1 adults, sentence completion task
Subjects produce more determiners the second time
they utter a given noun no comment could be
performance constraint?
11
Exp.1 adults, grammaticality judgment task


Interaction between condition and consistency
all participants liked correct sentences and
disliked sentences with determiner in wrong
position whether they accepted sentences without
determiners depended on the presence of
determiner-less sentences in the input they
received.
12
Conclusion of Exp. 1

• In the face of inconsistent input, adults do not
  tend to regularize the language they produce
  inconsistent output (and this is very
  convincingly demonstrated).
• What do children do?

13
Design of Exp. 2

• 15 6-year-old children 8 adults
• 17 words 4 verbs, 12 nouns, 1 determiner(pilot
  testing language from Exp. 1 too large)
• live presentation (pilot video does not work)
• taught vocabulary directly
• if you want to say hit in SillySpeak you say
  flim
• 6 exposure sessions, 1 test
• 2 determiner conditions 100 (consistent) 60
  (inconsistent)
• 3 tests sentence completion determiner
  judgment general grammar test.

14
Exp.2 production task
Main effect of input consistency (more determiner
produced in 100 condition) No effect of age, no
significant interaction between age and
consistency
15
Exp.2 production task
Percentage of participants in each production
systematicity category
16
Exp. 2 sentence completion data
4 always 2 never 1 other 1 variable
4 always
10 always
1 always 4 never 2 variable
1 always 1 never 7 variable
2 never 2 variable
2 never 6 variable
0 never 10 variable
Khi2 between children and adults in Exp. 2 not
sig. Khi2 between children Exp2 and adults Exp.
1 sig. there is a large difference between
adults and children in their overall tendency to
use determiners systematically, particularly in
the 60 condition(!!!)Ccl children are more
systematic than adults! Alternative
interpretation children have a bias towards
producing less determiners? (see 2 never in
100!) (maybe not 1 other in 100 1 always
in 60)
17
Exp2 Grammaticality judgment task







Claim diff. between No Det and Wrong Position
sig. only for adults inconsistent input
interpretation children and adults exposed to
consistent input judged both types of sentences
to be equally bad children 60 did not prefer
sentences with missing determiners to a sentence
type they had never heard.
18
Conclusion

• Possible to compare children and adults in
  artificial language learning experiments indeed
  no age effect on general grammar task (where
  performance is rather bad!)
• children learn unpredictable variation
  differently than adults they have a stronger
  tendency to impose systematicity on inconsistent
  input also in judgment task, where children
  exposed to inconsistent input preferred sentences
  with determiners in their most frequent location
  but did not distinguish between sentences missing
  determiners and those with the determiner in the
  wrong location. In this task they performed like
  children exposed to consistent input rather than
  like the adults exposed to inconsistent
  input.Children, it seems, prefer regularity in
  language and sometimes perceive or produce such
  regularity even when it is not present in their
  input (!!)