The ultimate goal of most events of sentence comprehension is to correctly update a ‘discourse representation’ on the basis of the intended message of the speaker/writer and relevant aspects of how they conveyed that message. However, cognitive models of the discourse representation are still underspecified and incomplete, and research in education and in populations with disordered language comprehension still struggle to define sensitive measures of these representations in reading and speech comprehension. In this talk, I will discuss our recent efforts to develop better behavioral and neural approaches for investigating referential processing and representation, one key component of the discourse model. Our results from eyetracking and the N400 component in ERP provide some of the best evidence to date for rapid resolution of pronoun interpretation in reading, and indicate that readers further benefit from rich expectations about whether a prior referent is likely to be mentioned next. Critically, these mechanisms are likely to strongly depend on depth of processing and to vary across populations, and I will describe how these measures might be used in future work to more precisely quantify the impact of these factors on the quality of language comprehension.