Utility of multivariate analysis in evaluating the values of laboratory tests on the basis of intra-individual variations (IV) was studied. Periodic health examination on workers has many clinical laboratory tests and these values are generally evaluated in comparison with "normal ranges." Recently it has been ascertained that the individual normal ranges were narrower than the populational normal ranges. Although the values of an individual should be estimated with normal ranges of the individual, it is almost impossible to determine the normal ranges from only one or two health examinations conducted per annum at the place of work. The author in his study could easily determine IV with a method using Mahalanobis' distance on a basis of populational normal range. In this method, if many values of a person in the tests vary in several years, the distance becomes larger, and if a value in a test falls outside the normal limit, it also becomes larger. This method makes it possible to detect not only an "abnormal" value out of the normal range but also a large value out of IV. To examine the practical validity of the method, the data of 24 employees (males, 45 to 65 years old) from 3 periodic health examinations and those of 20 persons (males, 48 to 71 years old) from 3 to 6 medical checkups were analyzed with regard to 11 items. It was found that 33 out of 44 persons were abnormal in one or more items and that 10 abnormal persons had large IV. They would be missed in the judgement with the use of only "normal ranges." As the distances were computed by a portable computer with a basic program, the method can be easily used in a small place of work. It is considered that this new method is effective and valuable not only as the screening procedure but also in the evaluation of IV.