Using event-related potential (ERP) measures, we examined the time course of Chinese compound word processing in 15 dyslexic and 10 normal children in a lexical decision task with three conditions including real words (e.g.,[Formula: see text] (house)), reversed nonwords (e.g.,[Formula: see text][Formula: see text] can be transposed to a real word [Formula: see text](ocean)) and random nonwords (e.g.,[Formula: see text] is not a real word when transposing). Behavioral results showed that dyslexic children performed slower and less accurately than normal children did across conditions. ERP data revealed that normal children exhibited significant N400 effects across conditions. The dyslexics did not show any difference on N400, however, suggesting a possible weakness of morphological processing in dyslexic children.