Fossil nannoplankton assemblages are controlled by both initial plankton production and dissolution throughout the water column and on the seafloor. It is usually assumed that dissolution can be identified by subjective visual assessment of preservation state. We have tested this by a quantitative comparison of mid-Pliocene nannofossil assemblages in two closely located ODP sites, Ceara Rise Sites 926 and 929, both with apparently good preservation but located respectively above and below the lysocline. We demonstrate that, despite an absence of obvious qualitative preservation differences, dissolution has removed perhaps 80% of the assemblage in the sub-lysocline Site 929. This has produced consistent distortions in the assemblages and artificially enhanced abundance variations patterns of individual species. The results provide strong cautions for palaeoceanographic interpretation of nannofossil data and allow us to uniquely extract a dissolution variation record and provide a detailed ranking of dissolution susceptibility of Pliocene nannofossils.