Pollen analysis is one of the most important methods to reconstruct past climate change and to understand prehistoric and early historic human-environment interactions. Every study based on fossil pollen assemblages from sedimentary archives starts with the preparation of collected sample material. The most widely employed protocols to concentrate pollen involve the use of several chemicals including hydrofluoric acid (HF), which is extremely hazardous to human health. As an alternative to HF, we have tested the reliability of dense media separation using two non-toxic substances, sodium polytungstate (SPT) and lithium heteropolytungstate (LST). Our test, which is based on five different samples representing different palaeoenvironmental archives partly revealed statistical disagreement between HF-treated samples and those treated by SPT and LST. In most cases, the observed differences in taxa proportions of the SPT and LST samples are unidirectional. In general, they do not appear to be pollen-taxa-specific but sample-specific thus probably linked to properties associated with the respective study material. However, sample comparison indicates that SPT-based dense media separation produces pollen concentrates that are statistically more comparable to those obtained by protocols based on HF-treatment. Discrepancies between both methods were also recognised for pollen concentrations and generally support the sample-specific character of dense media separation performance as suggested by the pollen proportion comparison. To verify the observed significant differences in pollen proportions and concentrations and to understand the factors that control them, further studies based on a larger number of test samples are required. In addition, we evaluated the effect of ultrasonic-aided fine sieving to bi-saccate pollen types. Our results indicate that this commonly used method to remove clays may lead to fragmentation of bi-saccate pollen into corpora and sacci, thus making identification more complicated. Although more time-consuming, we recommend to use less destructive differential centrifugation as an alternative, if indicated by preliminary tests.