Carrasco, M., Kinchla, R.A., & Figueuroa, J.G. (1988). Visual letter-matching and the time course of visual and acoustic codes. Acta Psychologica, 69: 1-17.

Abstract
Two experiments on visual letter-matching have been widely cited in the literature as indicating a role of acoustic codes in such tasks. Posner and Mitchell (1967) concluded that a visual code is sufficient to compare two identical letters (e.g., AA), whereas a slower forming acoustic code is required to compare nonidentical letters, whether with the same name (e.g., Aa) or with different names (e.g, AB or aB). Thorson, Hochhaus, and Stanners (1976) asked subjects to compare sequentailly presented letter pairs and found that visual confusability increased latencies at short delays (ISLs), whereas acoustic confusability increased latencies at longer delays. Our original objective was to use these paradigms to reveal differences in processing by English and Spanish speakers (Spanish has a much more direct grapheme-phoneme correspondence). Extensive testing of both native English and Spanish speakers reveals a highly consistent pattern of results that clearly differs from those reported in the aforementioned papers. The new data fail to indicate any role of acoustic codes in such tasks.