Up and Down the Cline: The Nature of Grammaticalization (review)

Within the field of diachronic linguistics, grammaticalization has become an increasingly important term, and grammaticalization theory has been launched and developed as a theory of linguistic change. Gradually, grammaticalization theory has achieved the status of a theoretical framework in its own right, and grammaticalization theorists have established themselves as an important linguistic subcommunity. The work on grammaticalization has spawned a number of books—monographs and collections of articles—as well as journal articles, and even a series of conferences, ‘New reflections on grammaticalization’, which has now become a permanent event in historical linguistics. This volume contains seventeen of the papers from the second conference, held in Amsterdam in 2002. In their introduction, the editors place the theme of the conference and the contributions to the volume in the context of recent developments in grammaticalization theory, before briefly summarizing each contribution.

Grammaticalization theory was originally developed by functionalists and typologists, that is, by people outside the generative field. Nowadays, grammaticalization has also become an important concept in diachronic generative syntax (e.g. Roberts & Roussou 1999, 2003, van Gelderen 2004a,b). The two camps do not yet seem to be on speaking terms, however, and the ‘New reflections on grammaticalization’ series, as well as the present volume, is mainly nongenerative (with perhaps one or two exceptions). The main difference between a functional/typological approach and a generative approach to grammaticalization lies in the concept of the ontology of language and therefore in the type of questions to be asked about change. One question not asked by any of the contributors to this volume is where the grammaticalization cline actually comes from, although everybody, including the most ardent antigenerativists, will probably accept a weak version of mentalism and agree that changes are caused by speakers.

One central issue in grammaticalization theory is the principle of unidirectionality. No one will deny the fact that changes in one direction along the grammaticalization cline are much more common crosslinguistically than changes in the opposite direction. The question, however, is what theoretical status this observation should have. Grammaticalization theorists will promote it to a principle. This raises two issues, which concern several of the contributors—as well as grammaticalization theorists in general. (1) Does the principle of unidirectionality tolerate exceptions? (2) What constitutes an exception? Several of the papers discuss the principle of unidirectionality and the problem of counterexamples, so-called degrammaticalization, at length. There are basically two prevalent attitudes toward this question. One is that unidirectionality follows from the definition of grammaticalization, and is therefore tautologous. Evident instances of degrammaticalization are something else, and therefore not counterexamples to grammaticalization. The other view admits certain counterexamples, but these are so few that they do not prevent us from stating the principle of unidirectionality as an interesting generalization.

In the first chapter of the book, Martin Haspelmath defends aversion of the latter view, promoting the generalization to an explanatory theory. He operates with universals of language change, which he calls general laws. Those laws are then parts of a theory that explains language change. Then, according to Haspelmath, the counterexamples are not a problem, as long as they are not too many. He makes an important distinction between real counterexamples, which he calls antigrammaticalization and which he claims to be very few (in fact exactly eight!), and all the claims about degrammaticalization found in the literature, which he argues are not counterexamples to unidirectionality, but different kinds of phenomena. It is of course a moot point whether a theory consisting of general laws can tolerate eight exceptions (in addition to all the other hitherto unknown and undescribed exceptions in the world’s living and extinct languages). Even if a general law can tolerate certain exceptions, the problem still remains that Haspelmath implicitly assumes (as do most of the other contributors to this volume) that language has...