2.3. Diachrony

Grammaticalization, simply put, is "the shift of an independent word to the status of a grammatical element" (McMahon 1994, p.160). A lexical element (noun, verb) becomes a grammatical element (preposition, auxiliary), or a grammatical element becomes even more grammatical (affix). This type of category change has repercussions in the morphology, phonology, and semantics of the elements involved.

According to Lehmann (1985), there are several parameters of grammaticalization: integrity, scope, paradigmaticity, and bondedness. Only the first two will be of concern to us here. Integrity is the phonological or semantic "weight" of an element; loss of integrity is attrition. Scope has to do with the complexity of constructions which the element appears in. Reduction in scope is condensation. When an element grammaticalizes, its meaning becomes "bleached" (it loses certain senses), although often some shade of its earlier meaning may persist, affecting its behavior or meaning after it has become fairly functional. Usually, some sort of phonological reduction (loss of stress, shortening) is experienced. Further, an element may lose the morphological or syntactic trappings that identify it as a member of a given lexical category. Condensation entails a reduction in the complexity of the potential constituents with which the element can combine. One final property of grammaticalization is layering, or the co-existence of both the old lexical forms, and the new grammatical ones.

Traugott (1989) views grammaticalization in terms of semantic-pragmatic shifts, and suggests that inferences become conventionalized, and meaning in general moves from less to more subjective.

One assumption of lexical diffusion (Chen & Wang, 1975) is that a change (be it phonological or syntactic) will originate in a small subset of items (or even a single item). Other elements may change directly while still others fluctuate between the old and new forms for a while. Gradually, the change may take effect in these items, and then continue to others, so that it can diffuse throughout the lexicon, potentially affecting all eligible items.

A certain change does not diffuse at a constant rate; rather, it can be represented as an S-curve as in Figure 2.1 (McMahon, p.53). At first only a few items change, producing the initial shallow incline. Then the rate of change accelerates rapidly, so more items change in a shorter period of time. Finally, the rate of change slows as most items have already been affected, and eventually stops, possibly leaving some virgin elements. One aspect of this model is that it can account for apparently exceptional or irregular old forms which persist, because it does not require that a change be completed.

Reanalysis occurs when the underlying structure of a syntactic pattern changes without altering the surface form. Because of this, the only way to tell when a reanalysis has occurred is by observing further changes (often by analogy or extension of some rule). This process is possible because language learners abduce their native grammars, that is, they observe their linguistic environment (output of the previous generation of speakers) and then guess as to what rules might underlie what they hear. This allows for a measure of error, especially where ambiguous constructions are concerned.