ABSTRACT
This paper addresses the question of whether readers of alphabetic English ever bypass letter identification by utilizing features that span single letters. The first section outlines the problem and briefly reviews evidence from several tasks that have not previously been considered together. The conclusions that have been drawn from the results of these different tasks conflict. The second section reconciles these disparate conclusions by making a distinction between two kinds of recognition: one in which the response uniquely specifies a stimulus, and one in which figurai familiarity is assessed. The former requires preliminary letter recognition, the latter operates on transletter features. The third section considers the hypothesis that certain classes of words are identified without recourse to preliminary letter identification; none of the evidence supports such a possibility. The fourth section extends the discussion to the performance of patients with certain acquired reading disorders. The final section summarizes the conclusions and notes a caveat.
