The purposes of this project were twofold: first, to explore the memory pattern in patients with schizophrenia using the Wechsler Memory Scale-Third Edition (WMS-III); second, to compare five competing models of the underlying factor structure of the WMS-III. Convenience sampling was used to recruit 119 patients with schizophrenia from several sources, including psychiatric clinic, day care center and community rehabilitation center of the Kaohsiung Medical University Hospital. Patients who met all of the following criteria were eligible for this study: (1) diagnosis of schizophrenia or schizoaffective disorder (Diagnostic and Statistical Manual of Mental Disorders ; DSM-IV), (2) symptomatically stable outpatients, (3) age between 18 and 65, (4) Mini-Mental State Examination(MMSE)total score ≧24, (5) ability to follow test instructions, and (6) willingness to participate in the study. Patients were excluded if they had a past history of stroke, traumatic brain injury, Parkinson’s disease, or dementia, had underwent WMS-III testing within three months of the commencement of this study, had a history of substance abuse or alcoholism, and were unable to comply with the testing schedule. Demographic and clinical information of each patient were gathered from clinical interview and medical file. The WMS-III and MMSE were administered on an individual basis to the participants. To facilitate analyses, all raw scores of the WMS-III were first converted to standard scores using the publisher-provided norms. Next, a one-sample Z test was computed to assess whether patients performed differently from standardization samples on WMS-III measures. Since group mean may hide the distribution of test scores within a group, the proportion of patients whose score on an individual subtest fell more than 1.5 standard deviations below the normative mean was calculated. Finally, the factor structure of the WMS-III was investigated using the confirmatory factor analysis module of the Analysis of Moment Structures 5.0 for Windows(AMOS 5.0). Several important findings emerged from our study. First, performance on seven out of the eight WMS-III index scores was classified as below the medium, whereas working memory index score fell within the medium category. Secondly, results of z tests revealed significant impairment on all test measures in patients with schizophrenia, with the worst performance on the general memory (delayed memory) index. Thirdly, significant difference was found between immediate and delayed auditory memory ( p < .0001 ). Fourthly, approximately 70% of the patients scored in the designated impaired range. In particular, 53 ( 44.2% ) of patients scored in the impaired rage on the family picture II recognition subtest, whereas only 10.8% of patients scored in the impaired range on the face memory II recognition subtest. Fifthly, the WMS-III was best represented by a three-dimensional model comprising auditory memory, visual memory, and working memory. Our results provided a reference for clinical assessment and explanation for patients with schizophrenia. Future research should aim at systematic replication of this three-factor model across different subgroups of patients with schizophrenia.