We summarize some evidence in favor of the view that the transmembrane region of many, perhaps most integral membrane proteins is made up of an aggregate of hydrophobic alpha-helices, each of which behaves as an autonomous folding domain. Folding of these proteins is seen as a two-stage process during which individual transmembrane helices first form in response to local interactions between the polypeptide and the aqueous and lipid phases, and then pack without extensive rearrangement to yield the three-dimensional structure. This two-stage model is supported by examination of those few structures that are known to a sufficient resolution, by experiments in which functional integral membrane proteins are "microassembled" from separately folded fragments, and by the existence in the inner membranes of organelles of a large number of very small integral subunits, often barely longer than a single transmembrane alpha-helix. We describe application of microassembly to establishing the path of the polypeptide in the tertiary structure of bacteriorhodopsin by neutron diffraction and we briefly discuss its possible role in the biosynthesis of organelle inner membrane complexes and its implications for model building from sequence data.