Word recognition and reading failure are examined in this report using an interactive framework of the reading process based on the premise that reading is both a top-down and a bottom-up process, both conceptually guided and graphically based. Experiments are discussed that show that less-skilled readers are affected by anomalous contexts and that they do make accurate predictions. Experiments in which print is degraded (letter segments are deleted) are also described, indicating that both the accuracy and the identification times of less-skilled readers were helped more by context. The conclusion notes that middle grade children of normal intelligence who are poor readers can use context very well in word identification and that they show signs of being dependent on context in compensation for inefficient word coding abilities. (MKM)