Transient peaks or "spikes" frequently appear in the initial segments of torque curves recorded with the Cybex isokinetic dynamometer. The purpose of this investigation was to determine whether these spikes represent artifact or transient initial surges of true muscular force output. Cinematographic analysis using both inert weights and a human subject as the source of torque input to the Cybex revealed that the dynamometer's input lever initially exceeded the pre-set angular velocity by an amount ranging from 11% (inert weight; pre-set Cybex velocity = 180 degrees/s) to 200% (human hip abduction; pre-set velocity = 30 degrees/s). The majority of this "overspeeding" occurred in the latter part of the free acceleration period, prior to the engagement of the dynamometer's resistance mechanism. The remainder occurred in the initial part of the elastic loading phase, just after resistance had engaged. A sharp deceleration of the overspeeding lever and the affixed weight or limb then followed in response to the dynamometer's continuing build-up of resistance. Simultaneous with this deceleration, a prominent torque spike was recorded that superseded the correct (mechanical equilibrium) torque value. Within our error of measurement, the deceleration observed in the film quantitatively accounted for all of the "overshoot" torque, i.e., that amount of the spike that exceeded the correct value. It was concluded that such prominent, initial torque spikes represent inertial forces and should not be confused with true muscular tension development.