Reanalysis, as understood in traditional historical linguistics, lacks explanatory force as a mechanism of language change, because the assumption that reanalysis works through ambiguity is logically flawed, and because reanalysis postulates a shift from an old to a new representation without specifying the source of the new representation. Therefore, two cases of syntactic reanalysis are examined to produce a more convincing picture of how reanalysis operates. One is the development of the English adjectives worth and worthwhile when used with gerund clauses. The other is the emergence of the English for. . .to-infinitive. On the basis of these case-studies it is shown that reanalysis can be decomposed into more basic mechanisms of change. These mechanisms involve 'category-internal change' resulting from semantic change, 'categorial incursion' through analogy, and 'automation'. Each of these underlying mechanisms obtain additional plausibility from the fact that they are firmly based in synchronic language use as understood in current usage-based models. One consequence is that reanalysis itself becomes to some extent epiphenomenal to the more basic mechanisms. #
The problem and its ramificationsAs a mechanism of linguistic change, reanalysis has an important share in explaining the syntactic changes that take place in language history. Along with analogy, reanalysis is commonly considered one of the two principle languageinternal mechanisms of syntactic change (e.g. Harris and Campbell, 1995;Hopper and Traugott, 2003). Essentially, while analogy works across syntagms, involving the extension of a form from one syntactic environment to another, reanalysis occurs within syntagms and causes the assignment of new syntactic representations to existing surface forms.The following more precise description of reanalysis (conforming to the definitions in Harris, 2003;Harris and Campbell, 1995;Langacker, 1977;Timberlake, 1977) is more or less commonly accepted. As stated, reanalysis takes place on the syntagmatic level of language, causing a single surface sequence of linguistic elements to receive a new syntactic and semantic interpretation. This happens as an alternative analysis is assigned to an existing surface sequence in ambiguous environments. The immediate result is a split between an old representation and a new one for the same surface sequence; later, the newly established representation may manifest itself in new surface sequences