fonolog

Anne-Michelle Tessier's PhD thesis Biases and Stages in Phonological Acquisition presents a novel approach of OT acquisition, based on the Biased Constraint Demotion algorithm (BCD).

The BCD is a fairly abstract, competence-based view of language acquisition, but Tessier aims to develop a more psychologically realistic version of it, which she calls 'Error-Selective Learning' (ESL). The principal extension of the theory is with a 'cache', where learners store their 'errors' before they are crucially used to change the constraint ranking. Only when there are enough 'errors' in the cache, this will be used to rerank constraints.

The idea is interesting — as is the comparison with the main alternative view of constraint ranking acquisition, the Gradual Learning Algorithm — but for me it is hard to see why the language learning device would use a Cache, and the notion of 'enough errors'. Why is one error not enough to decide that a constraint ranking is wrong? It may be true that adding such a Cache to the BCD provides us with a more satisfying model of what is going on, but we do not yet know why this happens. Intuitively, it seems to me that this should have something to do with the fact that the child can not be sure (i) whether she was really making an error (for instance because she is not sure about the underlying form yet), and (ii) whether the error is really in the relevant constraints, for instance because she is not sure about what the Tessier calls the 'hidden structure' of the word yet (morphology, syllable structure) and a different hidden structure might lead to a different analysis and constraint violation profile. However, I do not see that these ideas are encoded in any way in the model provided by ELS.

89 Redtube red tube porno 89

Redtube red tube porno 89

porno Redtube red tube porno 89