Questions concerning relatively immediate determinants of syntactic analysis can be tackled by comparing the reading times within crucial regions of subject- and object-relative clauses. Using such a comparison, (Altmann, Garnham, and Henstra (1994) have presented data which they interpreted as evidence for a discourse-driven account of parsing. This article reexamines this evidence and highlights a number of potential artifacts in the data. An experiment was conducted to demonstrate that at least some of these artifacts may have introduced distortions which were large enough to be of practical concern. It is concluded that the findings of the Altmann et al. experiment failed to provide unambiguous support for the discourse-driven model of parsing and that, overall, structure-driven models still offer a better account of the data