Following two individually mentioned characters in a text it is possible to successfully refer to either the individuals, or the set of two. Various factors, syntactic and pragmatic, have been found to affect the ease with which these types of reference can be made, however. This is therefore an interesting puzzle for those attempting to work out how linguistic and pragmatic factors interact in determining the representation we build in the process of understanding text. We build on previous ideas of "equivalence" and a "common association base" to propose that the Scenario mapping and focus model may explain at least some of the factors which affect plural pronominal reference. Two experiments are described which together suggest that the roles assigned to the two individual characters, and the likelihood of these roles being associated with joint activity in subsequent text are crucial in determining the likelihood of plural pronominal reference. Our findings broadly support the view that our representation of characters is heavily dependent on the scenario-based roles which we assign to those characters as we read. (Contains 2 figures, 4 tables and 2 footnotes.)