Recently, an implantable force transducer (IFT) has been introduced [Xu et al. (1992, J. Biomech. Engng 114, 170-177)] which can be used in tight spaces where force recordings with established transducers, such as the buckle-type transducers, are not possible because of impingement artifacts. The IFT is easily implanted in chronic animal preparations; however, calibration of the IFT in terminal experiments has produced unreliable results. The problems of IFT calibration are that minute movements of the transducer within the tendon, slight misalignments of the tendon, or slight errors in the line of pull cause dramatic changes in the IFT voltage output for a given applied calibration load. Here, we propose a method that eliminates the above calibration problems primarily because the target tendon is left in situ, the calibration loads are applied by the muscles which insert into the target tendon, and the transducer is implanted into the target tendon about two weeks prior to calibration. The theoretical and experimental approaches are demonstrated for the cat patellar tendon, but in principle can be performed with any tendon. The results are repeatable, lie within expected values, and reproduce some of the basic properties which have been observed in prior IFT testing.