It is often assumed that oral reading of Chinese script proceeds from print to phonological output via semantic representations (e.g. Wang WS-Y. The Chinese language. Scientific American 1973; 228: 50-60). However, it is possible (at least in principle) that Chinese characters can be read aloud without access to the mappings between semantic representations and phonological output that are presumed to underlie normal spoken word production. We report the confrontation naming and oral reading performance of a Chinese anemic patient, YQS, whose word comprehension, word repetition and oral reading skills are intact. When her retrieval of names from pictures is compared with her retrieval of the same names from print, there is a highly significant advantage of oral reading over picture naming (P < 0.0001). Indeed, her oral reading of hundreds of Chinese characters is flawless. We argue that the data from YQS show that oral reading in Chinese does not require access to the mappings between semantic representations and phonological output, but instead can proceed via a non-semantic reading pathway that maps orthographic units, such as radicals and characters, directly onto phonological output.