Long-lived, large scale segmentation of the mid-ocean ridge, as defined by variations in subsidence rate of the ocean crust, zero-age depth, and geoid rate, has been documented along three mid-ocean ridge systems: the South Atlantic, the Pacific-Antarctic, and the Southeast Indian. Many of the observations in these three areas are inconsistent with standard thermal models for the generation of oceanic crust and lithosphere. Three models for subsidence rate and geoid rate and two models for zero-age depth are examined. Sensitivity tests are done to determine how much the thermal and physical ''constants'' in these models must vary in order to explain the observations and whether or not such values are plausible. Only the zero-age depth variations can be adequately explained using existing models. An internally consistent, qualitative model is presented which can explain the data; it involves regional and local temperature changes in the mantle both along and across strike of the mid-ocean ridge. ¿ American Geophysical Union 1994