In the present paper, we propose an equating method of two scales based upon two separate tests using common subjects' item response patterns for both tests, and to examine the validity of such new equating method. We assume that there are two latent trait ability scales, based upon test X and Y respectively. In this method, each subjects' item response patterns for various tests are obtained, and then maximum likelihood estimates of the equating coefficient k, l are extracted from these data. In this paper, test X and Y are formed with the items in the word meaning comprehension test (Shiba, 1978). By actual subjects' and computer simulational item response data for these tests, the validity of this method is examined in various situations, such as changing the difficulty levels of two tests, the distribution of subjects and the number of subjects. As a consequence, this new equating method proves to be valid in situations where (i) the two tests are not so different in difficulty levels, and (ii) the subjects' ability distributes for a wide range.