As determined by freeze fracture electron microscopy, increasing levels of bovine brain galactosylceramide (GalCer) altered the surface structure of 1-palmitoyl-2-oleoyl-phosphatidylcholine (POPC) bilayers by inducing a striking "macro-ripple" phase in the larger, multilamellar lipid vesicles at GalCer mole fractions between 0.4 and 0.8. The term "macro-ripple" phase was used to distinguish it from the P beta' ripple phase observed in saturated, symmetric-chain length phosphatidylcholines. Whereas the P beta' ripple phase displays two types of corrugations, one with a wavelength of 12-15 nm and the other with a wavelength of 25-35 nm, the macro-ripple phase occurring in GalCer/POPC dispersions was of one type with a wavelength of 100-110 nm. Also, in contrast to the extended linear arrays of adjacent ripples observed in the P beta' ripple phase, the macro-ripple phase of GalCer/POPC dispersions was interrupted frequently by packing defects resulting from double dislocations and various disclinations and, thus, appeared to be continuously twisting and turning. Control experiments verified that the macro-ripple phase was not an artifact of incomplete lipid mixing or demixing during preparation. Three different methods of lipid mixing were compared: a spray method of rapid solvent evaporation, a sublimation method of solvent removal, and solvent removal using a rotary evaporation apparatus. Control experiments also revealed that the macro-ripple phase was observed regardless of whether lipid specimens were prepared by either ultra-rapid or manual plunge freezing methods as well as either in the presence or absence of the cryo-protectant glycerol. The macro-ripple phase was always observed in mixtures that were fully annealed by incubation above the main thermal transition of both POPC and bovine brain GalCer before rapid freezing. If the GalCer mixed with POPC contained only nonhydroxy acyl chains or only 2-hydroxy acyl chains, then the occurrence of macro-ripple phase decreased dramatically.