Computer aided tests replace traditional written answers on paper sheets with electronic records. For subjective tests, computers are not able to do all grading jobs, due to limited comprehensive ability of computers. Subjective answers need be reviewed by different graders in order to improve justice. We propose a distributed computational model for grading electronic subjective answers. Answer data are divided into numerous independent data cells. Computational resources are automatically allocated by the system to do grading jobs and independently report grading results to the system. Since the grading results may vary slightly or magnificently, we assign computational factors to data cells and computational resources to minimize the variation. In addition, rules are defined to obtain a final result. We successfully apply the proposed model in our grading system in computer aided spoken English tests.