Abstract. Diatoms play a key role in the development of quantitative methods for environmental reconstruction in lake ecosystems. Diatom-based calibration datasets developed during the last decades allow the inference of past limnological variables such as TP, pH or conductivity and provide information on the autecology and distribution of diatom taxa. However, little is known about the relationships between diatoms and climatic or geographic factors. The response of surface sediment diatom assemblages to abiotic factors is usually examined using canonical correspondence analysis (CCA) and subsequent forward selection of variables based on Monte Carlo permutation tests that show the set of predictors best explaining the distributions of diatom species. The results reported in 40 previous studies using this methodology in different regions of the world are re-analyzed in this paper. Bi- and multivariate statistics (canonical correlation and two-block partial least-squares) were used to explore the correspondence between physical, chemical and physiographical factors and the variables that explain most of the variance in the diatom datasets. Results show that diatom communities respond mainly to chemical variables (pH, nutrients) with lake depth being the most important physiographical factor. However, the relative importance of certain parameters varied along latitudinal and trophic gradients. Canonical analyses demonstrated a strong concordance with regard to the predictor variables and the amount of variance they captured, suggesting that, on a broad scale, lake diatoms give a robust indication of past and present environmental conditions.