The occurrence of implicit reading in brain-damaged patients with letter-by-letter dyslexia suggests a process of covert lexical activation, whereby lexical access occurs on the basis of parallel letter encoding. The extent and limitations of this process w ere studied by examining masked orthographic and phonological w ord priming as well as orthographic neighbourhood size effects in letter-by-letter reader IH. In Exp. 1, masked repetition priming occurred with primes displayed in a case-alternate format that were show n for 100 m sec (a duration that does not reliably support overt word identification in IH). Under sim ilar exposure conditions, however, prim es that are homophones to the target failed to affect performance, in contrast to neurologically intact observers (Exp. 2). Exp. 3 showed that IH's naming latencies are reduced for words with m any (vs. few) orthographic neighbours. This result suggests that overt word recognition in the patient is not strictly mediated by sequential letter recognition, but rather that it is conjointly affected by covert lexical activation. Relative to neurologically intact subjects, however, the pattern of the neighbourhood size effect shown by IH as a function of word frequency is abnormal and suggests that lexical activation based on the parallel processing of letters is w eakened in the patient compared to normal readers. Overall, results from IH point to a w eak form of activation of abstract orthographic lexical representations on the basis of parallel letter encoding, but no significant degree of phonological access. This account is discussed in relation to other similar proposals seeking an explanation of letter-by-letter dyslexia and of the covert lexical activation phenomena that accompany the disorder.