A fundamental issue in language acquisition, especially from a usage-based

perspective, is to explain how children are able to generalise from item-based

knowledge to a more general pattern-based form of knowledge. Casenhiser and

Goldberg (2005) demonstrated experimentally how children came to learn mappings

between novel phrasal forms and novel meanings. They showed that children could

learn meaning associated with a novel word order on the basis of a minimal input,

even without the help of a stable morphological cue.

The goal of the current work is to try to reproduce the experiment of Casenhiser

and Goldberg in another language, i.e. French. Moreover, French has some notable

differences from English. Especially, although it follows the same verb order as

English, word order is not as constrained, changing the word order for emphasis and

other pragmatic reason is not unusual. So changing word order may not provide a

sufficient cue for children to learn new syntactic patterns (Kail and Charvillat,

1986) .

The experiment will reproduce exactly Casenhiser and Goldberg's. A new, non-

canonical word order (NNV) will be associated to an original meaning (apparition of

an object). Children have only 16 examples to learn this new association between

grammatical construction and meaning. The test procedure is a forced choice between

new uses of the NNV constructions (all test examples were not used during the

training phase) and distracters.

If Casenhiser and Golberg results are confirmed, this would mean that young

children are indeed able to learn new mappings ? between grammatical structure and

meaning very quickly, which would make a huge step in explaining how children learn

to generalise grammatical knowledge so quickly. This would stress out the fact that

huge repetition effects are not necessary to learn a language using simple

association principles. It remains to see if a similar experiment could be done in

language production and not only in language comprehension.