This paper concerns some interactions between syntax, pragmatics, and semanticsexemplified by the development of the Discourse Markers indeed, in fact , besides , and what rolethey might play in a theory of grammaticalization, especially of the unidirectionality ofgrammaticalization.Nominal clines (nominal adposition > case) and verbal clines (main verb > tense, aspect,mood marker) are staples of grammaticalization theory. I will argue that a further cline: Clause-internal Adverbial > Sentence Adverbial > Discourse Particle (of which Discourse Markers are asubtype) should be added to the inventory. In some languages like English this cline involvesincreased syntactic freedom and scope, and therefore violates the principles of bonding andreduced scope frequently associated with grammaticalization. It nevertheless illustrates a clusterof other long-attested structural characteristics of early grammaticalization, specificallydecategorialization, phonological reduction, and generalization; it also illustrates a number ofmore recently recognized characteristics, especially pragmatic strengthening and subjectification. Iwill suggest that these characteristics should be considered salient to grammaticalization, butdecrease in syntactic freedom and scope should notFurthermore, I will argue that grammaticalization is the process whereby lexical materialin highly constrained pragmatic and morphosyntactic contexts becomes grammatical, in otherwords that lexical material in specifiable syntactic functions comes to participate in the structuraltexture of the language, especially its morphosyntactic constructions. The mechanism that bringsabout particular instances of grammaticalization is local reanalysis. How we should think of thislocal reanalysis and its relation to analogy and extension has been the topic of much recentresearch (e.g. Hopper and Traugott 1993: Chap. 3, Roberts 1993, Tabor 1994, Harris andCampbell 1995: Chap. 4, Vincent 1997). My interest here is not in mechanisms but primarily inthe processes--the interactions of syntax with semantics, of structure with use--that sow the seedsfor reanalysis and analogy in the first place.Grammaticalization is the process whereby lexical material in highly constrainedpragmatic and morphosyntactic contexts becomes is a unidirectional statement, in sofar as it predicts that grammatical material will not become lexical. So are other formulations indifferent traditions from the functional one adopted here. For example, in his study of the historyof the habere auxiliary in Romance within a GB framework, Roberts subsumes grammaticalizationunder optimization of grammars:Diachronic Reanalysis [of a lexical head as a functional head] is driven by theoperation of the Least Effort Strategy in acquisition. There is a real sense thenin which Diachronic Reanalysis leads to a local simplification in a grammar(Roberts 1993:251)).Indeed, unidirectionality in various forms has been one of the strongest hypotheses associatedwith grammaticalization theory. Like any good hypothesis, unidirectionality has been used as agatekeeper to assess whether some change X is or is not a case of grammaticalization. Some
Stored only when both papers are in the corpus — 476 citations exist upstream.