We conducted five experiments to examine readers' access to background
ed causal information in the absence of a local coherence break. Parti
cipants read surprise-ending stories in which some story actions were
explained by information in the preceding context, as well as informat
ion about the protagonist's ulterior motives disclosed in the story's
conclusion. Some participants received information about the story's e
nding before reading the story, whereas other participants received no
prior information. We found that participants who received prior know
ledge about the story exhibited longer reading times for actions relev
ant to this knowledge, used this knowledge to explain the actions, and
recalled more ideas from the story than did participants with no prio
r knowledge. We discuss the implications of these results for evaluati
ng the constructionist model of inferential processing.