Analyzing V2 triggers in Historical English

Introduction It is well established that pronouns can intervene between a topicalized constituent and a finite verb in Old English (OE); this has been analyzed as evidence that evidence that the verb moves only as high as T in OE matrix clauses, and that a landing site between T and Spec,CP is available to subject pronouns (cf. Pintzuk 1991). This paper considers a well known exception to this pattern, involving a class of clause-initial, “discourse-anaphoric” adverbs such as þa ‘then’, swa ‘so’ and nu ‘now’, which trigger subject/verb inversion even with pronominal subjects. Since Pintzuk (1991), the literature has generally assumed that these adverbs are a class of operators in Spec,CP, which trigger verb movement to C. However, Trips and Fuß (2009) (henceforth T&F) have proposed an alternative analysis, suggesting that personal pronouns and þa-type adverbs compete to fill Spec,TP in an OE clause. Under this analysis, subject/verb inversion occurs when þa fills Spec,TP, forcing the subject to remain lower in the clause. This is argued to be a result of the “discourse configurational” characteristic of OE; Spec,TP is not filled to satisfy an EPP in this account, but in order to license discourseanaphoric elements. T&F argue that for both pronouns and anaphoric adverbs, interpretation “involves a variable that must be bound by (or identified with) a topical element/referent in the given discourse” (184). In this paper, we will consider these alternative analyses of þatype adverbs, and show with quantitative and empirical evidence that the data does not support T&F’s analysis. Embedded contexts It is unclear what predictions T&F’s theory makes about the behavior of þa in embedded contexts. They concede that their theory cannot be as restrictive as it is for main clauses, saying that “embedded tense is dependent on the temporal anchoring of the matrix clause [...] Hence, þa/þonne are not required to occupy SpecTP.” (186) Thus, sentences such as (1) would be analyzed by T&F as involving adjunction of þa,1 because unlike in matrix clauses it does not need to occupy Spec,TP. Crucially, what T&F do not predict is the lack of the order þa Vfin Subjpro in embedded contexts, as this should be permitted given the possibility of verb-first subordinate clauses plus high adjunction of þa. In contrast, this prediction is made by the operator analysis, because inversion triggered by þa would be blocked if C were occupied. This is in fact what we find. We considered “strict” V2 orders, where only a þa-type adverb precedes the finite verb, in a subordinate clause with an overt complementizer. In the OE corpus we found only one example with this word order, and in Middle English only three; each case involved a copular clause, and was clearly an example of a specificational sentence as in (3) – thus irrelevant to the present issue. Thus, we conclude that inversion triggered by þa does not occur in embedded contexts, which supports the op-