Get reliable science

In quantitative phytolith analysis, chance error associated with insufficient counts can affect the robustness of the interpretation, whether it is vegetation reconstruction or taxonomic differentiation. It is therefore vital to choose a count size that will ensure statistically reliable results, while minimizing the time expended. Numerical statistical methods (bootstrapping) that have become available over the past few decades have made it possible to model even complex phytolith assemblages with relative ease. This study used bootstrapping as well as analytic statistical formulas to evaluate the influence of count size on vegetation reconstruction by means of two commonly used indices, D/P (tree cover index) and Iph (aridity index). The analysis indicates that the count size needed to ensure statistical precision depends on the question as well as the observed assemblage composition. Importantly, it is the count of specimens relevant to a specific ratio or other index ('index-specific' count) that matters, whereas the total count is less important. Based on these results, some general guidelines for choice of count size and for the use of statistics in phytolith analysis are suggested.