It is not easy to draw sensible conclusions from a pollen diagram. A pollen diagram registers vegetation, nothing else; all conclusions ref. other problems must be inferred from the vegetational changes by botanical judgment. Because of different pollen production in different spp. the relative percentages are of different significance according to which spp. form the rest of the vegetation. Selective destruction of pollen grains and local over-representation can easily bo detected and compensated for. Longdistance transport is generally of little significance as a scource of error, except outside the forest area. Pollen-analytic zones must be so clearly defined that isolated spectra can be referred to the proper zone. Correlations between diagrams based upon small-scale variations of curves are not reliable. Statistic errors have not been taken sufficiently into consideration. Formulas are given for the preliminary calculation of statistic errors. The number of pollen grains counted for the different spectra should always bo stated.