Studies of adult sentence processing have established that the referential context in which sentences are presented plays an immediate role in their interpretation, such that referential features of the context mitigate, and even eliminate, so-called ‘garden-path’ effects. Perceivers experience garden path effects almost exclusively when they are attempting to parse locally ambiguous linguistic structures in the absence of context, or in infelicitous contexts. The finding that the referential context ordinarily obviates garden path effects is compelling evidence for the Referential Theory of parsing, advanced originally by Crain and Steedman, (1985) and extended in Altmann and Steedman (1988). Recent work by Trueswell, Sekerina, Hill and Logrip (1999) suggests, however, that children may not be as sensitive as adults to contextual factors in resolving structural ambiguities. This conclusion is not anticipated by the Referential Theory and, worse still, it runs counter to the Continuity Assumption, which supposes that children and adults access the same cognitive mechanisms in processing language (Crain 1991; 2002; Crain and Wexler 2000; Pinker 1984). The purpose of this paper is to reexamine the observations that have led researchers to conclude that children, unlike adults, may lack sensitivity to referential contexts in comprehension in certain circumstances. Adopting a research strategy which we call the method of subtraction, we present evidence that the performance systems of children and adult differ only minimally; that children are sensitive to the referential context; and that children make use of contextual factors to resolve structural ambiguities in on-line sentence interpretation.