Abyssal peridotites have been interpreted to be residues of mantle mel
ting beneath ocean ridges. Recent experimental data and models of mant
le melting allow quantitative tests of this hypothesis. The tests show
that abyssal peridotites are not purely melting residues. Their modal
proportions and whole-rock compositions have far more olivine than wo
uld be predicted from melting models. Nonetheless, the correlations be
tween modal proportions of olivine and residual mineral chemistry, and
the relationship between associated basalt and peridotite composition
s, require an important role for melting. We suggest that abyssal peri
dotite compositions result from a combination of melting and crystalli
zation processes that are both a natural response to ascent of solid a
nd melt beneath an ocean ridge. Different extents of melting create a
range of residual peridotite and mantle melt compositions. The buoyant
melts migrate upwards, where they encounter the surface thermal bound
ary layer and crystallize olivine. The greater the ambient extent of m
elting of the mantle, the higher the normative olivine contents of the
melt, and the more melt is produced. Hence greater extents of melting
lead to more olivine crystallization at shallow levels. This correlat
ion between melting and crystallization within the mantle preserves th
e observed relationships between peridotite modes and mineral composit
ions. Significant implications of these results are: (1) the bulk comp
osition of the oceanic crust differs from the primary melt composition
s produced by partial melting of the mantle because of olivine crystal
lization at the thermal boundary layer; (2) the actual thickness of ig
neous crust may be variably thinner than would be calculated assuming
total melt extraction; and (3) peridotite modes can be used to infer p
olybaric mantle melting reactions only if the accumulated olivine is r
emoved appropriately. (C) 1997 Elsevier Science B.V.