The characteristics of vibration-induced illusory joint movements were studied in healthy human subjects. Unseen by the subject, constant frequency vibration trains applied to the distal tendon of the Triceps or Biceps induced an almost constant velocity illusory movement of the elbow whose direction corresponded to that of a joint rotation stretching the vibrated muscle. Vibration trains of the same duration and frequency applied alternatively to the Biceps and Triceps evoked alternating flexion-extension illusory movements. During successive application of vibration trains at frequencies from 10 to 120 Hz, the perceived velocity of the illustory movements increased progressively from 10 to 70-80 Hz, then decreased from 80 to 120 Hz. The maximal perceived velocity was three times higher during alternating vibration of the Biceps and Triceps than during single muscle stimulation. Unit activity from 15 muscle spindle primary endings and five secondary endings located in Tibialis anterior and Extensor digitorum longus muscles were recorded using microneurography in order to study their responses to tendon vibration and passive and active movements of the ankle. Primary endings were all activated by low amplitude tendon vibration (0.2-0.5 mm) previously used to induce illusory movements of the elbow. The discharge of some was phase-locked with the vibration cycle up to 120 Hz, while others responded one-to-one to the vibration cycle up to 30-50 Hz, then fired in a sub-harmonic manner at higher frequencies. Secondary endings were much less sensitive to low amplitude tendon vibration. Primary and secondary ending responses to ramp and sinusoidal movements of the ankle joint were compared. During the movement, the primary ending discharge frequency was almost constant, while the secondary ending activity progressively increased. During ankle movements the primary ending discharge appeared mainly related to velocity, while some secondary activities seemed related to both movement velocity and joint angle position. Muscle spindle sensory ending responses to active and passive ankle movements stretching the receptor-bearing muscle (plantar flexion) were qualitatively and quantitatively similar. During passive reverse movements (dorsiflexion) most of the sensory endings stopped firing when their muscle shortened. Active muscle shortening (isotonic contraction) modulated differently the muscle spindle sensory ending discharge, which could stop completely, decrease or some times increase during active ankle dorsiflexion. During isometric contraction most of the muscle spindle sensory endings were activated. The characteristics of the vibration-induced illusory movements and the muscle spindle responses to tendon vibration and to active and passive joint movements strengthened the possibility of the contribution of primary endings to kinesthesia, as suggested by several previous works...