This study investigates the impacts of R&E activities on output bias as well as input biases of technical change in postwar Japanese agriculture. In an attempt to explain the rapid change in the relative price between crop and livestock products from the supply side, we calculate incremental or marginal cost elasticities of producing each product. We construct the required data set from four classes of farm size by adopting the Caves-Christensen-Diewert method and we then differentiate the pattern and magnitude of the impacts of R&E among these classes. To accomplish these objectives, we employ the framework of the restricted translog cost function which consists of two-output and four-variable and one-fixed input. To examine whether or not the multiproduct framework is preferable to the single product framework, weak separability of outputs and input non jointness are tested.