Measurements of opal preservation in deep sea sediment cores have been presented in three ways: the opal concentration as a fraction of total dry weight (%opaltot), the opal concentration normalized to calcite-free dry weight (%opalcalcite-free), and the opal accumulation rate (opal MAR). It is tempting to interpret changes in these indices as indicators of rates of biological production in past oceans. Based on theoretical constraints, we argue that in typical tropical and subtropical sediments, both %Opalcalcite-free and opal MAR reflect a significant artifact of dilution by other phases. Thus the band of high %Opalcalcite-free in the equatorial Pacific appears to be caused in large part by the high %Calcite in that region, rather than by high opal productivity. The best candidate for a reliable paleoproductivity proxy appears to be %Opaltot. Unfortunately, present-day %Opaltot data from tropical and subtropical regions show little or no systematic trend with the rain rate of opal. Pore water silica concentration data reveal that the apparent pore water opal solubility is not constant but correlates regionally with the rain rate of opal to the seafloor. A model that treats opal as a single homogeneous phase with a single well-defined solubility product predicts a strong dependence of opal concentration on rain rate (in stark contrast to the data), and a constant asymptotic pore water Si. Two models representing opal as multiple heterogeneous phases with different solubilities are able to reproduce the observed asymptotic pore water Si/rain rate relationship, but not the lack of rain rate trend in the opal concentration data. Only by assuming a systematic trend in the quality of opal (i.e., the solubility) as a function of opal production, can we reproduce the observed pattern of opal preservation. The implication of this study is that changes in opal preservation in the geologic record cannot simply be interpreted in terms of changes in surface ocean productivity until our understanding of opal diagenesis can be improved. ¿ American Geophysical Union 1993