This paper examines the non-situational (i.e., non-exophoric) pragmatic functions of the three adnominal demonstratives, óyo , wâná , and yangó in the Bantu language Lingala. An examination of natural language corpora reveals that, although native-speaker intuitions sanction the use of óyo as an anaphor in demonstrative NPs, this demonstrative is hardly ever used in that role. It also reveals that wâná , which has both situational and discourse-referential capacities, is used more frequently than the exclusively anaphoric demonstrative yangó. It is explained that wâná appears in a wide range of non-coreferential expression types, in coreferential expression types involving low-salience referents, and in coreferential expression types that both involve highly salient referents and include the speaker’s desire to signal a shift in the mental representation of the referent towards a pejorative reading. The use of yangó , on the other hand, is only licensed in cases of coreferentiality involving highly salient referents and implying continuation of the same mental representation of the referent. A specific section is devoted to charting the possible grammaticalization paths followed by the demonstratives. Conclusions are drawn for pragmatic theory formation in terms of the relation between form ( yangó vs. wâná ) and function (coreferentiality vs. non-coreferentiality).