Two experiments are presented to clarify the relationships among different information codes available to a fluent reader and the utilization of syntactic and semantic knowledge to facilitate word identification. Use of context is measured with a word boundary task on passages with either coherent or random organization. Orthographic pattern and prior knowledge are also manipulated. The results indicate a reduction in use of context with unfamiliar orthography but an independence of information code levels. Results are discussed in terms both of alternate models of the reading process and of the development of fluency in reading. (Author/AA)