Discourse relations and information structure : Evidence from German discourse particles in embedded domains

Abstract

This dissertation explores non-standard uses of German discourse particles, especially 'ja' (roughly: ‘as we know / uncontroversially’). After discussing arguments for the focus sensitivity of discourse particles, ample structural evidence in favor of this assumption is presented by natural instances of 'ja' in syntactically integrated and semantically embedded structures like restrictive relative and central, i.e. proposition-modifying, adverbial clauses. The acceptability of 'ja' in such environments and e.g. the nominal domain, i.e. mainly in adjectival phrases, depends, first, on the fulfillment of the felicity conditions of the particle. Second, it depends on information-structural properties of the syntactic complement of the discourse particle, which must contain or at least comprise a focus, or, more precisely, an element for which alternatives are evoked, even if the latter is not the primary focus of the containing sentence. My analysis of the information structure of complex sentences is based on a theory of discourse structure in terms of Questions under Discussion. This allows for an account of ordering alternations between discourse particles and clausal constituents, which sheds a new light on scrambling and comparable adverbial movement in German. I show that qualified claims on the grammatical requirements of discourse particles, particularly in non-root structures, cannot be made without due regard to the discourse context. I distinguish between unmarked uses of 'ja' and uses that are only acceptable in the right context, and infelicitous in isolation or the wrong context, but not unavailable, contrary to previous claims. Besides 'ja' and other exemplary German discourse particles, I consider comparable expressions in other languages and observe interaction with information structure as well as syntactic association with subsentential phrases. In German, debate of such small particle phrases is limited to focus and additive particles, and discourse particles in wh-phrases. I argue that discourse particles behave more like focus and additive particles than hitherto acknowledged and show that e.g. 'ja' occurs with topical arguments and adverbial phrases in particle-final small particle phrases in the prefield of verb-second clauses. Consequently, small particle phrases in the middle field of clauses are a theoretical possibility, empirically supported by 'ja' in the background of wh-questions, occurrences of the same discourse particle in two positions in the middle field of one clause, open particle combinations, and combinations of discourse particles violating the otherwise rather strict order of discourse particles. Particles evidently associate with subparts of wider primary foci, too, i.e. nested foci that can consist of more than one constituent. 'ja' in any embedded position is interpreted differently from 'ja' at the level of the matrix clause (unless embedded 'ja' operates on an embedded narrow primary focus), i.e. matrix 'ja' relates the propositional at-issue meaning of the containing sentence to another utterance in the context, but embedded 'ja' generally relates an embedded proposition to the containing utterance. The same should hold – and be explored in more detail – for other discourse particles, assuming they are focus sensitive, if they occur in different positions in the same assertions and questions, as discussed. Further, we suggest that 'ja' can modify non-at-issue meaning. Besides occurring inside conventionally implicated structures like non-restrictive relative clauses or appositions, the particle occurs in small particle phrases with epistemic and evaluative adverbs and in rhetorical wh-questions, but only when followed by the discourse particle 'schon' (which disambiguates toward a rhetorical reading of wh-questions) or sometimes 'auch' (literally ‘too/also’). Beyond explaining why 'ja' appears in these structures, the assumption that 'ja' can modify non-at-issue meaning conveyed by non-at-issue expressions, besides meaning conveyed by conventionally implicated structures, can explain why 'ja' is consistently perceived as gaining acceptability in some non-standard uses when other particles and phrases are present: according to my approach, many structures are ambiguous, and ambiguity alleviates processing.