Premium
The ‘Shell Game’: Why Children Never Lose
Author(s) -
Satterfield Teresa
Publication year - 1999
Publication title -
syntax
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 24
eISSN - 1467-9612
pISSN - 1368-0005
DOI - 10.1111/1467-9612.00013
Subject(s) - learnability , rule based machine translation , computer science , grammar , selection (genetic algorithm) , linguistic universal , task (project management) , natural language processing , artificial intelligence , linguistics , production (economics) , representation (politics) , philosophy , management , politics , political science , law , economics , macroeconomics
This paper articulates a formal solution to the puzzle of child language learnability within the Principles and Parameters–based framework. The language learning (parameter‐setting) task requires, in principle, that the selection of syntactic knowledge be sufficiently constrained for the child to arrive at the appropriate target grammar, expending a minimum of computational effort and time. Since previous L1 analyses impose very strict requirements on the learner, solutions are achieved only at a very high cost. Further, not only do the standard accounts frequently contradict fundamental empirical facts of child linguistic development, such as the degree of variability observed in the production of early grammatical structures regardless of input; the accounts also fail to consider a crucial aspect which impacts selection: the young child's innate potential to efficiently acquire multiple languages simultaneously. The primary aim of the current paper is to provide a computational model that demonstrates a “bilingual universals” (in the spirit of Roeper 1996) stage of development based on real world data. The proposed model actually reflects a more precise UG‐based representation within early monolingual grammars, as well as plausibly accounting for variability found in child L1 grammars.