We have discovered annual variations in the U-238/Ca-40 ratio of Porit
es coral skeletons. Measurements were made using thermal ionization ma
ss spectrometric techniques, yielding precisions of +/-2 parts per tho
usand (2 sigma) for U-238/Ca-40 and +/-1 parts per thousand (2 sigma)
for Sr-88/Ca-40. Coralline aragonite subsamples weighed similar to 2 m
g, enabling submillimeter sampling resolution corresponding to similar
to monthly temporal resolution. The annual nature of the cycles was c
onfirmed by comparison with annual banding observable in X-radiographs
. For two modern and one fossil sample, the amplitude of the U/Ca vari
ation ranges from 6 to 23%, well outside of analytical error. As annua
l U/Ca cycles appear to be a general feature of primary coralline arag
onite, the preservation of such features will be important in identify
ing unaltered coral for U-series dating studies. U/Ca variations mimic
and are in phase with annual variations in Sr-88/Ca-40. For given fra
ctional shift in Sr/Ca, the fractional shift in U/Ca is about 6 times
larger. For the two modem corals, U-238/Ca-40 is strongly anticorrelat
ed with measured temperature, suggesting that U-238/Ca-40 has potentia
l as a paleothermometer. If temperature is the only significant contro
l on coralline U-238/Ca-40, We reach the following conclusions from an
alyses of fossil samples: a Vanuatu sample, which grew halfway through
the last deglaciation, gives U/Ca temperatures 4 to 5 degrees C below
modern values. U/Ca thermometry applied to published data for Barbado
s corals indicates that (1) temperature generally correlates with sea
level, (2) glacial temperatures (stages 2 and 4) were 4 to 6 degrees C
lower than interglacial temperatures (stages 7a, 5e, and 1), and (3)
temperatures rose from glacial to interglacial values early in the las
t deglaciation. Thermometry applied to Papua New Guinea corals indicat
es that (1) temperatures were 5 to 6 degrees C lower than interglacial
temperatures from similar to 13 to similar to 10 Ky BP, then rose to
present values similar to 9 Ky BP, (2) the temperature depression betw
een similar to 13 and similar to 10 Ky BP is consistent with low tempe
ratures observed in Vanuatu during the same time interval, and (3) the
Papua New Guinea deglacial temperature history differs from that of B
arbados. The results generally support estimates of tropical temperatu
res obtained from Sr/Ca thermometry and snow line elevation data, but
disagree with those based on foram transfer functions. A thermodynamic
model suggests that coralline U-238/Ca-40 may also be sensitive to ma
rine carbonate ion concentration, raising the possibility that some of
the observed glacial-interglacial U-238/Ca-40 variation may result fr
om glacial-interglacial carbonate ion changes. However, the key experi
ments that might establish a coralline U-238/Ca-40-carbonate ion relat
ionship have yet to be performed.