Models of ridge segmentation, mantle flow, and melt focusing predict how the chemical compositions of mantle melts should vary along a mid-ocean ridge axis. The compositions of basaltic lavas can be compared to these predictions to test the models. Such tests have been carried out using basalts from the neovolcanic zone south of the Kane fracture zone (the MARK area), where there are both a large transform and nontransform offsets. Before evaluating mantle models, the effects of differentiation must be accounted for. Fractional crystallization at low pressures (constrained by new melting experiments on these samples) does not account for the data. High pressure or in situ crystallization better account for the differentiation trends; however, these two processes imply different relationships between magmatic differentiation and position within a segment. Irrespective of the differentiation model, significant differences exist among parental magmas. Magmas near the transform have much lower levels of highly incompatible trace elements but higher levels of moderately incompatible trace elements, suggesting both lower extents of melting and a more depleted source. These two characteristics may be natural consequences of the truncation of a melting regime by a large-offset transform: depleted mantle from across the transform may contribute to the melting regime, while the cooler thermal environment produces less melt. Quantitative modeling of these geochemical characteristics produces thin crust near the transform, consistent with seismic and gravity studies. In contrast, thin crust adjacent to nontransform offsets is associated with no reduction in extent of mantle melting. These results, along with data from other regions, suggest that nontransform offsets overlie a continuous melting regime, and melt focusing creates the variations in crustal thickness. Focused flow may also lead to incompatible element enrichment at segment centers, and relative depletion at segment margins. Only offsets that truncate the melting regime, such as large transforms, are associated with diminished extents of melting within the mantle. Petrological evidence obtained thus far is not consistent with active upwelling to explain crustal thickness variations along nontransform offset bounded segments.¿ 1997 American Geophysical Union