To capture prey, chameleons ballistically project their tongues as far as 1.5 body lengths with accelerations of up to 500 m s⊃–2. At the core of a chameleon's tongue is a cylindrical tongue skeleton surrounded by the accelerator muscle. Previously, the cylindrical accelerator muscle was assumed to power tongue projection directly during the actual fast projection of the tongue. However, high–speed recordings of Chamaeleo melleri and C. pardalis reveal that peak powers of 3000 W kg⊃–1 are necessary to generate the observed accelerations, which exceed the accelerator muscle's capacity by at least five– to 10–fold. Extrinsic structures might power projection via the tongue skeleton. High–speed fluoroscopy suggests that they contribute less than 10% of the required peak instantaneous power. Thus, the projection power must be generated predominantly within the tongue, and an energy–storage–and–release mechanism must be at work. The key structure in the projection mechanism is probably a cylindrical connective–tissue layer, which surrounds the entoglossal process and was previously suggested to act as lubricating tissue. This tissue layer comprises at least 10 sheaths that envelop the entoglossal process. The outer portion connects anteriorly to the accelerator muscle and the inner portion to the retractor structures. The sheaths contain helical arrays of collagen fibres. Prior to projection, the sheaths are longitudinally loaded by the combined radial contraction and hydrostatic lengthening of the accelerator muscle, at an estimated mean power of 144 W kg⊃–1 in C. melleri. Tongue projection is triggered as the accelerator muscle and the loaded portions of the sheaths start to slide over the tip of the entoglossal process. The springs relax radially while pushing off the rounded tip of the entoglossal process, making the elastic energy stored in the helical fibres available for a simultaneous forward acceleration of the tongue pad, accelerator muscle and retractor structures. The energy release continues as the multilayered spring slides over the tip of the smooth and lubricated entoglossal process. This sliding–spring theory predicts that the sheaths deliver most of the instantaneous power required for tongue projection. The release power of the sliding tubular springs exceeds the work rate of the accelerator muscle by at least a factor of 10 because the elastic–energy release occurs much faster than the loading process. Thus, we have identified a unique catapult mechanism that is very different from standard engineering designs. Our morphological and kinematic observations, as well as the available literature data, are consistent with the proposed mechanism of tongue projection, although experimental tests of the sheath strain and the lubrication of the entoglossal process are currently beyond our technical scope.