Aim: To explore the category-specific semantic deficits in two Chinese-speaking patients, so as to provide new evidence for the psychological reality of the deficits, which is also helpful for clinical diagnose and rehabilitation. Methods: Two Chinese-speaking mal e patients aged 75 and 35 years, respectively, with semantic impairment were selected between October 1999 and January 2005. They kept normal language abilities before the onset of cerebrovascular disease, which caused deficits in comprehension and production of language. Two groups of typical cognitive tests (semantic test and category-specific semantic deficits test) were used to investigate their semantic system and identify whether the deficits existed or not. The semantic tests involved five tasks: 1 auditory word/ picture matching; 2 auditory sentence/picture matching; 3 visual word/ picture matching; 4 visual sentence/picture matching; 5 oral picture naming. The first two tasks were used to examine the patients' ability of auditory lexical comprehension, the following two tasks to examine their visual lexical comprehension, and the last one to inspect their oral lexical production. The oral picture-naming task was still used for the semantic category test, which increased to 232 items, to deeply analyze the patients' naming ability in different semantic categories. To examine whether there were disparities in the naming of living and nonliving items of the patients, three analyses were carried out: 1 Comparison among all items: There were 72 living things and 160 nonliving things of the 232 items, to compare whether there were significant differences in the correct rates of naming of the two groups items. 2 Comparison after removed some items: Body parts and musical instrument were excluded, and left 70 living things and 130 nonliving things. 3 comparison after matching: 45 living items and 45 nonliving items were selected to conduct critically matching of confounding factors including word frequency, familiarity, age of acquisition, naming agreement, imaging agreement and so on. Results: The two subject s completed all tests. 1 They performed badly in the tests. The correct rates of case one and case two in the auditory word/picture matching, auditory sentence/picture matching, visual word/ picture matching, visual sentence/picture matching and oral picture naming, were 88% (44/50), 84% (42/50); 50% (10/20), 45% (9/20); 72% (36/ 50), 94% (47/50); 55% (11/20), 55% (11/20) and 44% (36/82), 43% (35/82), respectively. Meanwhile, they both made a lot of semantic errors (lion named tiger, mouth named ear). 2 Their semantic system had impaired, which presented different deficits in different semantic categories. The case one had worse deficits in living things than nonliving things, which was opposite to the case two. The two cases formed a clear pattern of double separation. Conclusion: Both two cases cannot complete the se mantic tasks normally because of their impaired semantic system. Furthermore, the two cases present different impairment patterns in living and nonliving things: the case one appears worse deficits for living things, but case two for nonliving things. The category-specific semantic deficits are objective phenomena, but not artifacts from some confounding factors.