The development of the declarative complementizer in German

It is standardly assumed that the German declarative dass-complementizer evolved from the demonstrative pronoun (Old High German thaz) used cataphorically. On this analysis, the source structure would be a paratactic sequence of two sentences in which thaz occurred in the final position of the first clause and pointed forward to the content of the second clause. Out of this structure, thaz developed into a subordinate conjunction/complementizer via a shift of the clause boundary (as in Mary knows that: Peter is lying → Mary knows that Peter is lying). This article takes issue with the standard assumption and puts forward an alternative account in which the declarative complementizer developed from a correlative construction (as in Mary knows that, that Peter is lying). The correlative construction (arguably also with an optionally silent correlative) is robustly attested in the old Germanic and old Indo-European languages. The source structure was thus not a sequence of syntactically independent clauses, but a hypotactic structure with an explicative relative clause associated with a silent correlative element in the main clause. In line with the hypothesis currently under discussion—that apparent noun-complement clauses are in fact relative clauses—it is argued that the explicative clause in the correlative structure was a relative clause, and thus the declarative complementizer developed from the relative complementizer thaz. The syntactic reanalyses involved in this new scenario are far less radical than in the traditional one. The new scenario is furthermore supported by crosslinguistic evidence: the development of relativizers into complementizers is a frequently attested grammaticalization path.