Subduction related magmas are characterised by distinctive minor and t
race element ratios which are widely attributed to the introduction of
a hydrous component from the subducted crust. In general, high LIL/HF
SE ratios are best developed in low HFSE rocks, and the variation in L
ILE is less than that in HFSE. Moreover, the size of the contributions
from subducted material estimated on the basis of minor and trace ele
ment variations are consistently greater than those inferred from radi
ogenic isotopes. The Nd isotope ratios of island arc rocks are lower t
han those of N-MORB, and there is a broad negative correlation between
Nd-143/Nd-144 and Pb-208/Pb-206* consistent with a small contributio
n (2-4%) from subducted sediment in arc magmas. However, the isotope a
nd trace element compositions of IAB cannot be modelled by two compone
nt mixing of sediment and N-MORB from the mantle wedge, and they requi
re an additional high LIL/HFSE component which is typically attributed
to the release of hydrous fluids from the subducted slab. A number of
models are considered for the effects of fluid percolation in the man
tle wedge. D(wedge/fluid) values are not well constrained but, if loca
l equilibrium occurs, the available data on arc rocks are consistent w
ith D(wedge/fluid) approximately 10(-2) for the LILE. Several authors
have proposed that some of the trace elements in the subduction compon
ent were scavenged from the mantle wedge, in order to reconcile the di
fferent contributions from subducted materials inferred from radiogeni
c isotope and trace element data. In the models outlined here that sug
gests that D(slab/fluid) for the LILE is > 0.1. Within low Ce/Yb arc r
ocks there is a general increase in average Ce/Sm, but not K/Ce, with
increasing crustal thickness. The observed range in Ce/Sm in the low C
e/Yb rocks is consistent with 3-18% melting of slightly LREE depleted
source rocks, and it appears that the degree of melting varies with cr
ustal thickness.