The aim of this study was to determine whether a general or specific context facilitation mechanism should be incorporated into information-processing models of reading. General facilitation models claim that a context can facilitate recognition of any word that is related to it. Specific facilitation models claim that a context will facilitate recognition of only those words that are highly expected on the basis of the context--that is, a subset of words predicted by general models. Three experimental procedures were used. The first required a lexical decision (word or nonword) about a test item sometimes preceded by a context, the second used a successive lexical decision procedure with two words, and the third used a sentence acceptability judgment task. Overall, the experiments provided some evidence for a general facilitation mechanism but did not eliminate the possibility that a specific facilitation mechanism was also operating. To determine whether these results should influence the construction of models of reading, tentative criteria were proposed for deciding whether an experimental effect needs to be accounted for by a model. (Author/AA)