Magmatic, tectonic, and sedimentary processes at slow spreading ridges are known to create complex distributions of rock types at the surface that do not reflect the classical model for oceanic crust with horizontally stratified lithologic layers comprising extrusives, sheeted dikes, and gabbro overlying mantle rocks. A database of seafloor lithology observed in >100 submersible dives is analyzed in order to characterize this complexity. As a way of summarizing the relations between units exposed on rock slopes, the transitions between different lithologies are counted to create transition probability tables, which represent the relative likelihood of passing between different rock types. They reveal that extrusives exposed on Atlantic rock slopes crop out above dikes, serpentinite, and gabbro, in order of decreasing, though with similar, likelihood. Tables derived from the occurrences of the different rock types in dredge hauls are similar to the tables derived from dive observations, so the dredges record a similar heterogeneity. It is suggested that slope failure of gabbroic and ultramafic escarpments, implied by sedimentary breccias in off-axis drill sites, may contribute to the heterogeneity, in addition to the magmatic and tectonic processes that have been discussed extensively by previous workers. Further analysis of the tables reveals that the net effect of tectonic, magmatic, and sedimentary processes is to produce a distribution of rock types that is indistinguishable from a random distribution. This result is derived from the immediate transitions between units and therefore applies to the fine-scale structure. If occanic crust is also heterogeneous on a larger scale, however, it could have more general implications for resolving different crustal models. For example, the presence of a sheeted dike layer is critical evidence for continuous seafloor spreading, but its presence or absence in any one location may be an artifact of the heterogeneity or discontinuous exposure. Methods will therefore be needed to characterize large-scale heterogeneity so that the significance of these large-scale observations can be assessed. ¿ 2001 American Geophysical Union