As it is known, the use of the Item Response Theo ry (IRT) with individuals must rest on a demonstration that the test responses of individuals are in accordance with the testing model hypothesized. However, sometimes it obviously cannot give satisfaction. In such a case, an extention of the model becomes compulsory. Torgerson proposes a general normal ogive model postulating that both an item and a subject are represented by frequency distributions of points on a psychological continuum. That is, different from IRT model in the generalized IRT (we call it GIRT) model, each subject, not only by his level of ability but also by the degree of its dispersion are here described. The present study compares the two models in a case of japanese and foreigners taking a japanese vocabulary test, and substantiates that by introducing person parameters we ought to get satisfying results. It was found that the specific feature of the foreigners were well described by the values of person parameters in the GIRT.