The use of words in isolation, words with pictures, and words in sentence context in teaching word recognition to beginning readers was studied in a sample of 54 boys and 54 girls. These 108 pupils were randomly assigned to the three treatment groups; each child was individually taught the target words. Learning was assessed on five dependent variables: trials to criterion, word in isolation, word with picture, word in sentence context, and word in story context. Results indicated significant differences for methods of instruction but not for sex or for sex by method interaction. On trials to criterion, children in the sentence-context group required fewest trials, and those in the word in isolation group required the most trials. Preplanned contrasts suggested that a teach/test bias existed on the word in isolation and word in sentence context dependent variables, suggesting that previous research may be biased in favor of one of the teaching conditions. Results were interpreted as supporting Goodman's contextual hypothesis rather than Samuels' focal-attention hypothesis in reading. (Author/AA)