An anaphor is said to be associative when it refers to a referent (i) which has not been explicitly mentioned in the prior context, and (ii) which can nevertheless be identified on the basis of information provided by this context. Numerous linguists consider that associative anaphora can only be introduced by definite NPs and claim that demonstrative NPs cannot take on the role of associative anaphora. In this paper, we defend the idea that demonstrative NPs are not incompatible with associative anaphora and are in fact current in French. We examine the various explanations and interpretations provided by the authors who claim that demonstrative NPs cannot take on the role of associative anaphora (polyphony effect, memory deixis, empathy, etc.). We show that, in numerous cases, these interpretations are not satisfactory. On the contrary, it is possible to describe non-coreferential demonstrative NPs by applying a dynamic model for the production and reception of discourse reference. It is based on the idea that interactors may have diverging strategic interests, a different representation of the universe of discourse, and that the selection of referring expressions may be a reflection of these diverging interests. It can also be shown that demonstrative associative NPs have various discursive and communicative functions or effects, such as avoiding a generic interpretation, signalling a problematic presupposition, signalling that the identification of the intended referent will be somewhat difficult, etc. A distinction is proposed between positive and negative function. Finally, we contend that an investigation of anaphoric phenomena should be conducted along three lines of description, to avoid oversimplification .