This study sought to add to the evidence on predictable texts by asking two questions: (1) What type of reading do predictable texts produce? and (2) How do predictable texts influence sight vocabulary acquisition and decoding? Subjects, six black male students selected from a first-grade classroom in an urban setting, had minimal or no sight vocabulary. Subjects (in two groups of three) were pulled out of their regular classroom reading lesson twice a week from mid-October to mid-April for a total of 30 sessions. Sections were 20-25 minutes long and consisted of each subject reading orally the same complete predictable story, followed by word recognition activities. Data included scores on word recognition, pseudoword decoding, and standardized tests. Results indicated that: (1) an extremely high proportion of the miscues were contextually appropriate; (2) using a 90% criterion, three subjects failed by a substantial amount to learn to recognize the words presented in the texts; and (3) decoding as measured by a pseudoword test showed a similar pattern--three of the students showed an almost total lack of decoding skills. Findings suggest the existence of large individual differences in decoding and word recognition which seem to be related to sensitivity to the code. (One table of data is included; 20 references are attached.) (RS)