Junctional microfibrils, which span the lamina lucida of the vertebrate myotendinous junction, are thought to function in force transmission at the junction. This hypothesis has been tested by disrupting junctional microfibrils through elimination of extracellular divalent cations, and determining the effects of this treatment on the ultrastructure and mechanics of whole frog skeletal muscles passively stretched to failure. Muscles incubated in divalent cation-free solution failed exclusively in the lamina lucida of the myotendinous junction, while control muscles all failed within the muscle fibres, several millimetres away from the junction. Failure sites from divalent cation-free muscles incubated with antibodies against collagen type IV, laminin, and tenascin showed no labelling of the avulsed ends of the muscle fibres, indicating that remnants of junctional microfibrils observed on the cell surface are not composed of any of these extracellular proteins. All three proteins were present on the tendon side of the failure site, confirming that the lamina densa remains attached to the tendon. Breaking stress for control muscles was 3.47 x 10(5) N m-2, and for divalent cation-free muscles, 1.84 x 10(5) N m-2, or approximately half the control value. Breaking strain averaged 1.17 for divalent cation-free muscles and 1.39 for controls, although the difference was not significant. We conclude that junctional microfibrils are components of a divalent cation-dependent adhesion mechanism at the myotendinous junction. In addition, ultrastructural analysis of divalent cation-free fibres stretched just short of failure suggests that a second, divalent cation-independent mechanism persists along the non-junctional cell surface, and can transmit substantial passive tension from myofibrils laterally to the extracellular matrix, bypassing the failed myotendinous junction.