The main-chain conformations of 80 proteins were analysed to identify helical structures that commonly occur but do not fall into the known classes of alpha-helix, 3(10)-helix and beta-sheet. The analysis yielded 96 occurrences of four or more sequential residues forming the threefold left-handed poly-L-proline II (PPII) helix. This contradicts the previously held opinion that left-handed helices are rare in globular proteins. The main-chain dihedral angles of these helices form a cluster in phi,psi space that has a maximum at -75 degrees, 145 degrees, corresponding to conformations with the number of residues per turn (n) = -3.0. We show that 51% of PPII-helices lie within the range of n = -3.0(+/- 0.2). It is shown that the PPII segments are distinct from the conformation typical of beta-pleated sheets. Although proline residues commonly occurred in PPII-helices, this side-chain is not obligatory, as 28 of these helices did not contain proline. In addition, we found 120 segments with three C alpha atoms forming a PPII-helix. PPII-helices tend to occur on the surface of the protein and, having few main-chain hydrogen bonds with the rest of the protein, tend to be the more mobile segments of the molecule. The geometry of PPII helix allows the polypeptide chain to progress immediately from this conformation to right-handed alpha-helix and 3(10) helix, as well as to beta-sheet or reverse turn. We conclude that PPII-helices should be considered as a regular conformation and should be added to beta-sheets, alpha-helices and 3(10)-helices in databases of protein structures, in secondary structure prediction and in tertiary model-building.