Many previous models for how SecA drives translocation have been proposed, designed to accommodate the results of structural and functional studies. So far, however, such models make various assumptions, e.g. they postulate the existence of conformational changes that lack direct experimental evidence. In general these can be divided into three types: models involving a power stroke within SecA, those that invoke quaternary interactions between multiple SecA molecules or those which act through biased diffusion. (A) An example power-stroke mechanism, whereby conformational changes within SecA during the ATPase cycle physically push polypeptides through the channel. In the model shown [36], the 2HF binds to the pre-protein substrate, pushes it into the channel, then releases it and returns to its resting position. (B) The observation that SecA can exist both as a monomer and in several different dimer forms has led to the proposal of multiple models in which quaternary interactions drive transport. In the example shown, one SecA protomer holds the pre-protein substrate in the channel whereas the other binds to downstream regions. ATP binding alters the SecA dimer interface, pushing the substrate through the channel, whereas ATP hydrolysis releases SecA, allowing it to rebind downstream. (C) Rather than physically pushing the substrate through the channel, directional movement can be achieved by selectively allowing diffusion in one direction, while preventing it in the other. Such a ‘Brownian ratchet’ would act by using ATP to somehow prevent backsliding. In the version shown, SecA senses backsliding and constricts to halt movement; however this is entirely speculative, as an illustration of the core concept.