Comparisons between death assemblages and their source living communities are among the most common actualistic methods of evaluating the preservation of compositional and environmental information in fossil assemblages. Although live-dead studies have commonly focused on marine mollusks, the potential of diatoms to preserve ecological information in continental settings has been overlooked. Thus, little is known about the nature and magnitude of the taphonomic biases affecting live-dead agreement of diatom assemblages, despite their extensive application as modern and fossil bioindicators in paleoecological and paleoenvironmental reconstructions. In this study, I analyzed three live-dead data sets in order to evaluate the compositional and environmental fidelity exhibited by diatom death assemblages in shallow lakes. I find that diatom death assemblages (DAs) do differ significantly in their taxonomic composition from living assemblages (LAs), mainly as a consequence of (1) differences in the temporal resolution between time-averaged DAs and non-averaged LAs, and (2) differential preservation of diatom taxa related to the intrinsic properties of their valves. Despite compositional dissimilarities, DAs were able to capture the same environmental gradients as LAs, with high significance. This decoupling between live-dead agreement in community composition and community response to gradients can be related to the existence of at least two mutually exclusive subsets of species that significantly captured compositional dissimilarities based on the full set of the species in the three lakes. This functional redundancy implies that the between-sample relationships of living assemblages can be significantly preserved by DAs even if some taxa are removed by taphonomic processes. The preservation of environmental gradients thus does not require good preservation of all living taxa. Structural redundancy compensates for the loss of compositional fidelity caused by postmortem processes in the diatom data set.