The kinetics of an individual helix of bacteriorhodopsin have been monitored during folding of the protein into lipid bilayer vesicles. A fluorescence probe was introduced at individual sites throughout helix D of bacteriorhodopsin and the changes in the fluorescence of the label were time-resolved. Partially denatured, labelled bacteriorhodopsin in SDS was folded directly into phosphatidylcholine lipid vesicles. Stopped-flow mixing of the reactants allowed the folding kinetics to be monitored with millisecond time resolution by time-resolving changes in the label fluorescence, intrinsic protein fluorescence as well as in the absorption of the retinal chromophore. Monitoring specific positions on helix D showed that two kinetic phases were altered compared to those determined by monitoring the average protein behaviour. These two phases, of 6.7 s−1 and 0.33 s−1, were previously assigned to formation of a key apoprotein intermediate during bacteriorhodopsin folding. The faster 6.7 s−1 phase was missing when time-resolving fluorescence changes of labels attached to the middle of helix D. The amplitude of the 0.33 s−1 phase increased along the helix, as single labels were attached in turn from the cytoplasmic to the extracellular side. An interpretation of these results is that the 6.7 s−1 phase involves partitioning of helix D within the lipid headgroups of the bilayer vesicle, while the 0.33 s−1 phase could reflect transmembrane insertion of this helix. In addition, a single site on helix G was monitored during folding. The results indicate that, unlike helix D, the insertion of helix G cannot be differentiated from the average protein behaviour. The data show that, while folding of bacteriorhodopsin from SDS into lipids is a co-operative process, it is nevertheless possible to obtain information on specific regions of a membrane protein during folding in vitro.