On average, 50--60% of diatomaceous opal produced in the euphotic zone redissolves within the upper 100 m of the water column. High specific silica dissolution rates in the surface ocean contrast with order-of-magnitude lower values in deep-sea sediments. The lack of a mechanistic understanding of these large variations in dissolution kinetics is a major source of uncertainty in models of nutrient silicon cycling in the oceans. Here we show that the observed variability in biogenic silica reactivity is consistent with results of experimental dissolution studies. Differences in aluminum content and specific surface area of the diatom skeletons, plus differences in temperature and degree of undersaturation, all contribute to lowering silica dissolution rates in deep-sea sediments relative to the euphotic zone. When variations in these material and environmental properties are accounted for, the predicted specific opal dissolution rate in average surface ocean water is over 2 orders of magnitude faster than the rate measured in recently deposited biosiliceous oozes of the Southern Ocean. The predicted specific dissolution rate, however, is approximately 5 times higher than the value obtained from global estimates of the depth-integrated dissolution rate and biogenic silica concentration in the upper water column. This discrepancy likely reflects the inhibitory effect of protective organic coatings on fresh diatom shells in surface waters. The experimental data further imply that aluminum incorporation in the silica matrix, during biomineralization or after death of the organisms, enhances the preservation efficiency of biogenic opal in marine sediments.