When a relaxed joint is subjected to a small sinusoidal torque, the amplitude of the steady-state displacement response is increased up to severalfold by a transient larger perturbation. The original state, in which the relaxed joint is unexpectedly stiff, is restored by several seconds of inactivity. This thixotropic phenomenon has previously been observed in a variety of human joints. We have now investigated the mechanism of thixotropic behavior at relaxed joints in rats anesthetized with pentobarbital sodium, by using a series of preparations including the intact ankle joint, a blood-perfused soleus muscle preparation, an isolated soleus muscle, and ankle joint isolated by severing all muscular attachments. Thixotropic behavior was observed in all intact, isolated muscle, and isolated joint preparations. The contribution of the joint to thixotropic behavior was comparable to, and at times exceeded, the contribution of muscle. We also analyzed the short-range stiffness properties of relaxed, blood-perfused soleus muscles and found them to be similar to thixotropy with respect to range of action (0.2-0.3% of muscle length), elastic modulus (approximately 4 kg/cm2), and time course for redevelopment (time constant = 2.5 s at 34 degrees C). Thus thixotropic behavior at a relaxed joint may be attributed both to the joint structures and to short-range stiffness of muscles acting at the joint.