What can be learned from positive data? Insights from an ‘ideal learner’. Commentary on ‘A Multiple process solution to the logical problem of language acquisition’ by Brian MacWhinney

Abstract

MacWhinney's stimulating discussion suggests that there are many lines of argument that may address concerns raised by theorists who are concerned that there is a logical problem of language acquisition. This commentary argues: (1) that if the ‘logical problem’ applied to language, it would apply, with curious consequences to any learning by experience; (2) that the logical problem does not apply – given sufficient positive data from any reasonable language, language can be learned, in a probabilistic sense, by an ‘ideal learner’ using a simplicity principle; and (3) that a simplicity, or minimum description length, principle may provide a useful methodology for assessing claims concerning learnability of particular linguistic structures.