Diatom preservation can be a major taph- onomic issue in many lakes but is often unrecognised and its impacts on qualitative and quantitative infer- ences (such as productivity and biodiversity estimates) from sedimentary archives are seldom explored. Here two palaeolimnological case studies of 20th-century anthropogenic eutrophication of freshwater lakes in Northern Ireland (Lough Neagh and Lough Augher) are re-visited and new data presented on diatom preservation. Assessing problems of taphonomy challenges previous interpretations of silica dynamics and diatom productivity at these sites. Diatom preservation was assessed in both sediment trap material and sediment cores from Lough Neagh, and in sediment cores from Lough Augher. Preservation data, combined with geochemical analysis (Si, Fe), provide an insight into silica cycling and diatom accumulation over a range of temporal scales from these lakes. Diatom preservation was generally good for the Lough Neagh material, although differential (better) preservation of the smaller Aulacoseira subarctica compared to the larger Stephanodiscus neoastraea sensu lato valves was clear, especially in sediments. Porewater silica showed a complex sea- sonal pattern in the upper sediment, against expecta- tions of steady-state. The Lough Augher material was generally poorly preserved, although preservation (dissolution) was significantly (and positively) correlated to bulk sedimentation rate, and was found to be a major control on (net) diatom accumulation rate across the basin. Past seasonal and severe anoxia at Lough Augher did not improve diatom preservation, contrary to some previous studies, which may be due to extreme changes in sedimentary redox conditions. Finally, using published experimental relationships between dissolution and diatom valve loss, correction factors were applied to previously published profiles of diatom accumulation over the last approx. 150 years (biovolume from Lough Neagh and frustule accumulation rate from Lough Augher), which suggest that diatom productivity estimates from sedimentary records are underestimated by a factor of 2–4 due to dissolution effects alone. The results clearly have implications for the reliability and accuracy of diatom-based inferences made from sediment records, both qualitative and quantitative, especially for those that employ diatoms as direct measures of productivity or biodiversity.