Published online by Cambridge University Press: 28 November 2008
It is commonly thought that (1) the description of English auxiliary inversions requires a feature with a default value, (2) that non-default values must be stipulated and learned as exceptions and (3) that when languages exhibit different values for a feature in different contexts, learning theory requires grammars to stipulate a default value. Distinguishing two perniciously confused uses of the term DEFAULT enables a demonstration that the first and third assumptions are incorrect. Conseqently, any argument that depends on them is invalid, and the absence in a theory of a mechanism for default-value declarations is not a deficiency. It is then shown that a comprehensive account of inverted structures has to encompass considerably more diversity of structural types than is generally recognized, but is entirely possible in a constraint-based grammar with monotonic multiple-inheritance and no overridable default specifications.