Collocation extraction has gained much attention in natural language processing, its results are important in various areas of applied linguistics. The research is a comparison of fifteen association measures based on a subset of Russian “Taiga” corpus. The paper studies automatically extracted Verb-Noun collocations. The aim of experiments is two-fold. First, to examine the difference between statistical measures and second to find out which measure is most efficient for Russian data. The former assumes calculation of Spearman’s rank correlation coefficient, whereas the latter implies evaluation of extracted lists against a Russian-language dictionary, i.e. to identify automatically extracted and manually collected collocations. The results are far not straightforward; we can distinguish groups of measures that demonstrate relative interchangeability. More so, produced bigrams can be of interest to lexicographers and may therefore enrich dictionaries.