Many languages (e.g. Spanish, German, and Swedish) have more than one morpheme corresponding to the English “but”. It has been argued by Anscombre & Ducrot (1977) and by Horn (1989) that two of these morphemes covary with two different functions of natural language negation. Anscombre's and Ducrot's descriptions of these BUTs are examined, and it is shown that there is no such link. There is, however, an interesting connection between “but” and negation. One of the two morphemes imposes a restriction on the use of negation in the clause that follows it, which suggests that natural language negation does not correspond to a truth function.

Anaphors and the grammatical roles of their antecedents are examined in Finnish language data in order to establish a correlation between two universal hierarchies: the Keenan & Comrie (1977) NP Accessibility Hierarchy and the hierarchy of accessibility of referential expressions (Accessibility Marking Scale) (Ariel 1985, 1988, 1990). A more or less clear correlation pattern between the type of anaphoric NPs and the grammatical roles of their antecedents arises in Finnish intuition data (Section 2), and this pattern is corroborated by the data from prose text counts (Section 3). Even though a one-to-one mapping between the two hierarchies remains an idealization, it is clear that grammatical relations do bear on the type of anaphoric expression employed.

A multiple methods approach was applied to the study of morphology on the processing of lexical items in Swedish. Data from slips-of-the-tongue, agrammatic speech production, agrammatic oral reading, and lexical decision experiments were used. The results indicate that whole word processing as well as morphological processing takes place in the different types of tasks. The type of processing seems to vary along a continuum, with whole word processing as the most commonly applied type in automatized and relatively simple processing (such as lexical decision for common Swedish words), whereas signs of morpheme-based processing appear less often, and perhaps in less automatized tasks (such as agrammatic speech production).