The inversion-angle replication error of the ankle joint was measured in subjects with a functionally stable or unstable ankle. Testing was performed as a passive set-active replication design. The absolute replication error on the unstable side of 23 unilaterally, functionally, and mechanically unstable subjects was found to be significantly greater (2.5 degrees) than the error on the healthy contralateral side (2.0 degrees ; P < 0.05) and the error in a group of 40 ankle healthy controls (1.7 degrees; P < 0.01). In 10 ankle-healthy high-level cross-country runners the replication error was reduced by 38% after warming up (P = 0.04).