The principles that underlie pronoun interpretation have been extensively studied in both computational and psycholinguistics, yet little consensus has emerged. In this talk, I revisit Hobbs's (1979) hypothesis that coreference is simply a by-product of establishing discourse coherence, in light of counterevidence that has motivated attentional-state theories such as Centering (Grosz et al, 1995 [1986], Brennan et al. 1987, inter alia). I will argue that the seemingly inconsistent collection of data found in the literature can only be explained by modeling how attentional state is influenced by both (i) syntactic form (per attentional-state theories) and (ii) the inference processes that underlie the establishment of different types of coherence relations. As such, while the analysis I present incorporates several tenets of attentional-state approaches, it rejects the idea that attentional state can be adequately modeled using a clause-by-clause update mechanism and without reference to inference and world knowledge. I will present a (rather simplistic) illustration of how the integration of the two factors cited above might be achieved; ultimately it may be the behavior of pronominal reference itself that provides the best view into some of these deeper complexities of discourse processing.