The paper examines syntactic features of non-canonical relativization in spoken Polish that loosen the structural integration of two types of relative clauses – one introduced by the complementizer co, the other by the wh-pronoun który. The resulting unintegration holds between the head NP and the co/który clause and contrasts with the integrated structure of canonical relatives. I discuss the range of unintegration features observed for both types in corpus data and indicate the distinct quantitative extents to which the two types are unintegrated. Although the nature of spontaneous conversation is such that it imposes some loosening of structural cohesion in both types, co clauses (especially non-subject relative clauses) are far more frequently unintegrated than który clauses. Also, co clauses depart functionally from the canonical relative structure in that the complementizer co serves functions other than that of a straightforward relativizer, namely it has conjunction-like uses (temporal, spatial, and general conjunction), indicating an expansion of the categorial status of co. The observed unintegration of Polish conversational relatives is in line with previous analyses of the syntax of unplanned speech (e.g. Miller and Weinert 1998)

The present study aims at differentiating between semantically-coded and pragmatically-conditioned meaning components of Polish and German sentence adverbs whose meaning is conventionally associated with hearsay (≈ Eng. allegedly, reportedly, supposedly). In the first part, we present a systematic corpus study of hearsay adverbs in Polish and German providing the empirical basis for our analysis and conclusions. In the second part, we provide reasons why our objective should be reached on the basis of Generalized Conversational Implicatures (GCIs), and we show which particular communicative principles distinguished in Neo-Gricean frameworks can sensibly be considered as triggers of GCIs that evoke ‘epistemic overtones’ in the use of hearsay adverbs. We differentiate between GCIs which work for all relevant adverbs and implicatures which only apply to more individual properties of hearsay adverbs on more specific levels of their meaning structure. In accordance with this more descriptive task, we discuss general issues concerning presumable hierarchies of factors that influence (trigger or cancel) epistemic implicatures in the usage of lexical markers of information source. We argue that many discourse properties on the semantics-pragmatics interface which are characteristic of grammatical evidentials also hold true for lexical markers of information source