A large number of Chinese characters are made up by pairing a semantic radical and a phonetic radical. The phonetic radical usually gives a phonological clue for the pronunciation of the whole character but does not contribute to its meaning. Using an event-related potential (ERP) measurement, the present study was able to trace the very intricate interplay between phonological and semantic information embedded in the phonetic radical. It was found that, within the first 50 to 100 ms of perceiving the character, the semantic information of the phonetic radical was better preserved when the constituent phonetic radical provided a valid phonological clue (e.g., regular phonogram) than when it contained an invalid phonetic cue (e.g., irregular phonogram). However, no trace of semantic information from the phonetic radical was preserved after 300 ms. These ERP results detail the neuropsychological steps in reading Chinese and support the framework of the lexical constituent model.