The oxygen isotopic composition of coexisting carbonate and phosphate from belemnite rostra was measured according to well established techniques in 42 samples of Early and Middle Jurassic age and in five samples of oyster shells. Most of the samples come from various locations in the Western Carpathians of Slovakia and Ukraine, and from central Poland. Three samples come from the Isle of Skye. The phosphate content of belemnite rostra, though variable, is systematically very low: consistently lower than about 0.3%. However, this phosphate concentration is close to that found in shells of modern marine organisms including pelecypods, gastropods and Sepia cuttlebones which, in some way, could be considered the modern belemnite counterpart. The measured oxygen isotopic composition of carbonate is within the normal range of values obtained from these fossils ranging from about −1.3 to about +0.6‰ (PDB-1) with the exception of three samples; the δ 13C values range from about −0.8 to about +2.8‰ (PDB-1). With the single exception of one sample from the Isle of Skye, the oxygen isotopic composition of phosphate from belemnite rostra ranges from +19.8 to +24.9‰ (V-SMOW), 22 of the samples measured showing δ 18O values equal to or heavier than +23.0‰. In contrast, the oyster values are considerably lighter, in the case of both carbonate and phosphate. 18O-enriched values can hardly be related to diagenetic processes that normally cause an oxygen isotope shift towards light values. If deposition temperatures are calculated from the heavily enriched values by means of the equation of Longinelli and Nuti [Earth Planet. Sci. Lett. 19 (1973) 373–376] and assuming the δ 18O of Jurassic ocean water to be equal to −1‰ taking into account the lack of ice caps during the Jurassic, the obtained temperatures range from about 8°C to about zero. These temperatures are obviously unreliable when Mesozoic palaeoceanographic conditions and palaeoclimate are taken into account. Two different hypotheses are suggested to explain these results, other hypotheses being rejected as unreliable. (1) Phosphate derived from the decaying organic matter of belemnites might have been introduced into belemnite rostra by early diagenetic fluids. If the phosphate of belemnite organic matter was isotopically heavy as happens nowadays in the flesh of molluscs, the inflow of this phosphate into the rostra could be responsible for the very positive δ 18O values shown by many belemnite rostra (this hypothesis is suggested by H.W.); (2) previous oxygen isotope measurements on Upper Cretaceous belemnites yielded δ 18O values very close to the most positive values obtained from Lower Tertiary pelecypods and fish teeth which are known to precipitate their phosphate under isotopic equilibrium conditions with seawater. These data suggest the possibility that the phosphate in belemnite rostra is primary phosphate so that the very positive data reported here can be considered the result of good preservation of the pristine isotopic composition of primary phosphate. Consequently, the only way to explain the very positive δ 18O values is to consider the oxygen isotopic composition of Jurassic ocean water to be more positive than nowadays by at least 3‰. This hypothesis is suggested by A.L. and A.D.M.