The focus of this proposal is the investigation of the mechanisms by which sentence contexts influence lexical access (word recognition) processes during reading. Prior research has established that lexical access processes for a given word are affected by the sentence context in which the word appears. The mechanisms by which context exerts its influence, however, remain to be clearly specified. For example, is lexical access affected by access processes for preceding words? Is access affected by the message of the sentence as a whole, apart from the identities of the individual words? The answers to these questions have implications for current models of language comprehension. Under one model, the lexical access process is assumed to be immune to the influence of many aspects of sentence context, including the representation of the whole sentence message. Under an alternative model, lexical access is influenced by the representation of the whole sentence message. Within this proposal a third model is developed, based upon representational assumptions that allow one to specify which aspects of the representations of the whole sentence message should influence lexical access and which should not. A series of experiments is proposed to test this model. In all of these experiments, subjects first see a sentence context displayed on a CRT screen. The context is then followed by a single target word that the subject must respond to. Response time is measured as an index of access time. Contexts are varied to test specific predictions of the model. The goal of this research is an understanding of a fundamental process within language comprehension, the process by which the comprehension system recognizes written words.