Retour vers liste

Détail de la contribution

Auteur: Melanie J. BELL

Titre:
Morphosyntax and relative frequency: a usage-based account of the distinction between words and phrases in English.


Abstract/Résumé: Phrases and words are sometimes said to be formed by different components of grammar (e.g. Pullum & Huddleston 2002:4) and even by different mental mechanisms (e.g. Giegerich 2005), yet in an inflectionally impoverished language like English the distinction between phrases and compound words is difficult to delineate. Criteria are usually based on the principle of lexical integrity, and one such criterion is whether the constituent words can be modified independently of one another. For example, a construction like 'London colleges' is classed as a phrase, since both nouns can be separately modified to produce e.g. 'south London colleges' or 'London theological colleges' (Payne & Huddleston 2002:449). But it has been unclear what properties of a construction determine whether such modification is possible. The present study is the first large-scale empirical investigation of this question, and is based on a large database of [AN]N and N[AN] constructions randomly extracted from the British National Corpus. It is shown that the possibility of modifying the first noun (N1) depends on there being a combination of adjective plus N1 that is lexicalised, institutionalised, or at least more frequent than the corresponding noun-noun combination. For N2, the possibility of modification depends largely on the nature of N1. Following a suggestion by Plag (2003:160), it is shown that modification of N2 is associated with certain types of noun in N1 position: proper nouns, nouns with incorporated numerals e.g. 'one way' or material nouns such as 'silk'. These types are typically used attributively. However, it is not only classes of noun that produce this effect but also individual lexical items that occur frequently in attributive position compared to their occurrence elsewhere. Overall, the results suggest that, for a given noun-noun construction, the probability of either constituent being modified independently of the other depends largely on the relative frequencies with which the two nouns occur together and in combination with other lexical items. Rather than differentiating between the products of two mental modules, the modification test therefore appears to reflect usage patterns of the constituent nouns. The results are compatible with a view of language in which memory is central: frequency effects can only arise if speakers keep track of, i.e. remember, their linguistic experiences. Furthermore, if we accept that such modification distinguishes between objects usually viewed as compounds and those usually viewed as phrases, then a possible conclusion is that the distinction between morphological and syntactic objects is itself based on relative frequencies: as such, it is gradient and usage-based, and the lack of a clear boundary is expected.