A study examined how word recognition automaticity develops and its relationship to the acquisition of comprehension skill. Two different methods for teaching reading were used: (1) a global method using the Houghton Mifflin basal reading program, and (2) a code method using the New Reading System, which emphasizes word decoding skills along with comprehension skills. Core data were collected with children over a three-year period from first through third grade. Measures included reaction times for oral reading of individual words and for judgments of word meanings, comprehension of sentences and passages, coding of errors for sensitivity to context and fidelity to the phonemic code, and subtest scores from achievement tests. Results indicated a relationship between word recognition efficiency early in learning and reading comprehension performance later on. The two different teaching approaches resulted in substantial qualitative differences in acquisition of word recognition ability. These differences in oral reading disappeared by the end of third grade. The findings do not provide a basis for choosing between code and global methods, but they do suggest that neither method sufficiently develops word recognition efficiency. Eight figures and four tables of data supplement the text. (SRT)