1. Experiments are carried out in the experimental paddy field of the Kyushu Imperial University, to find out the variation of yield of rice plants under different combination of plats, in arranging them to bear different shape or different area. The result of the experiment is summerised and interpreted here to sanction the most adequate recommendation for laying o ut variety test field plats. 2. The error decreases when the area increases, but the mode of decrease depends upon the shape of the plats. With increase of area by in ceasing the width of plats of equal length, the error does not decrease to reach the theoritical figure calculated by the principle that the error is inversely proportional to the radical of the area. By increasing the length of the plats of equal width, the error decreases just as much as the theoritical calculation . 3. In general, the error is smaller when the plats are taken relatively long and narrow , than they are taken short and broad. 4. In the experiments on rice plants managed under the intensitive Japanese method, the error is much smaller compared with the case on wheat experiments in the Western countries. It can be expected therefore to obtain fairly accurate data from experiments layed out in a small area, under prevailing Japanese condition. 5. To adjust the yield of test plats by including check plats in the ordinary method is somewhat efficient to reduce the error, but the mode of the reduction depends upon the shape and the size of plats. Also in this case, the magnitude of the error itself is smaller in the relatively long and narrow plat than in the short and broad one, but the reduction of the error compared with the magnitude calculated directly without adjusting by including check plats is less conspicuous in the former. With increase of area, the error calculated in this method tends to decrease, but not so marked. Within certain limit, extensive replication of check plats does not relate to the appreciable reduction of the error. 6. The method of adjusting the yield by means of applying the percentage of the average yield of grouped plats is more effective to reduce the error, than calculating the yield by including the check plats, provided the grouped plats are not too numberous. When the plats are relatively long and narrow, this is particularly true, and the error decreases with increase of area by increasing the length of plats. With increase of area by increasing the width of the plats, the error rather tends to increase.