Abstract Five hypotheses were proposed and tested to account for Reicher's (1968) finding that recognition of letters is more accurate in the context of a meaningful word than alone, even with redundancy controlled by a forced-choice design. All five hypotheses were rejected on the basis of the experimental results. Performance on the forced-choice letter detection task averaged 10% better when the stimuli were four-letter English words than when the stimuli were single letters appearing alone in the visual field. Three classes of models were proposed to account for the experimental results. All three are based on analysis of the task in terms of the extraction of features from the stimuli.