Syntactic category Information and the semantics of derivational morphological rules

Abstract

In standard generative approaches, word-formation rules contain, among other things, information on the semantics of the suffix and the syntactic category (or word-class) of possible bases. Based on the general assumption that word-class specification of the input is a crucial ingredient of derivational morphology, far-reaching claims have been made. For example, the unitary base hypothesis (Aronoff 1976) states that word-formation rules can take only bases as input that share syntactic category information. This generative position is a reflection of the well-established traditional descriptive practice of talking about affixes as being, for example, 'deverbal', 'deadjectival' or 'denominal'. Thus, across the different theoretical camps, the syntactic category of the base is, implicitly or explicitly, assumed to be central to any account of affixal behavior. In the present paper this view is challenged and it is claimed that derivational morphology does not necessarily make reference to syntactic category information in the input. It will be shown that the word-class specification of the input of word-formation rules creates numerous empirical and theoretical problems that have been largely ignored in the literature. It is argued that there are productive affixes where the syntactic category of potential base words is only a by-product of the semantics of the process. The data even strongly suggest that input word-class specification should be generally abandoned. Based on the analysis of a large number of English affixes and an in-depth analysis of a subset of these affixes it will be shown that the inclusion of syntactic category information in the word-formation rules under investigation makes wrong predictions for significant portions of the data and fails to account for the phonological and semantic patterns observable in these morphological categories. In contrast, a semantic, output-oriented approach is much more successful in coping with these problems. Th

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 72 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.