This study investigates the validity of a newly developed computerized cognitive assessment system (CCAS) that is equipped with rich multimedia to generate simulated testing situations and considers both test item difficulty and the test taker's ability. It is also hypothesized that better predictive validity of the CCAS in self-care of persons with stroke can be obtained. The CCAS has been constructed and validated by an expert review panel. The system was pilot-tested with 14 patients who had suffered a stroke and fulfilled specific selection criteria. In addition, data on patients' performances in the Mini-Mental State Examination (MMSE) and Neurobehavioral Cognitive State Examination (NCSE) or Cognistat, Modified Barthel Index, Functional Test for the Hemiplegic Upper Extremity were collected. Their demographic characteristics were also collected. Correlation coefficients of the scores among the CCAS, MMSE, and NCSE were used to show the concurrent validity of the former. With regard to the content validity, the intraclass correlation coefficient [model 2, ICC(2,k)] among the panel members was 0.972 with a "P" value of less than 0.01. The scores of the CCAS correlate with MMSE (r = 0.676, P = 0.011) and with the four subtests of the NCSE (repetition, naming, construction, and calculation). The content and concurrent validity as well as the predictive abilities in self-care function of the CCAS were initially established. The degree of the usefulness and accuracy of the new computerized system in measuring the cognitive performance in persons with stroke remains to be determined.