The present study examined whether semantic activation for words occurs by encoding whole word shape in addition to individual letters. We used LEET stimuli, where digits were used as parts of words, such as “R34DING” instead of “READING”. Previous studies have suggested that LEET stimuli are encoded in a letter-like fashion and can automatically activate lexical information because of their physical similarity (e.g., “3” vs. “E”). We measured the N400 effect elicited by words, a measure of whether participants detected a mismatch between the word and the current semantic context (e.g., belonging to a category or not). The N400 effect can occur only if a word has been identified and processed up to the semantic level. Participants performed a categorization task – determining whether a word or LEET stimulus was related or unrelated to a given category name. We found that LEET stimuli produced a similar N400 effect as regular words, suggesting that word shape beyond individual letters is sufficient to access word meaning.