Six experiments used an illusory words paradigm to demonstrate that repetition blindness (RB) in orthographically similar words only affects the wordsÕ shared letters. RSVP streams of words and word fragments allowed the unique letters of the second critical word to combine with a subsequent fragment to create a word, as in rock shock ell. The illusory word shell was reported 2-3 times as frequently in RB conditions as in control conditions. Further experiments ruled out letter migration, contour summation, and differences in processing load as explanations for our results. These findings are inconsistent with current proposals that orthographic RB represents similarity inhibition or lexical competition, or that it reflects problems with word-level token individuation.