The present study aims at an empirical evaluation of lexical association measures (AMs) for collocation extraction from Korean spoken corpus. Compared was the effectiveness of five widely used AMs: cooccurrence frequency, mutual information, chi-square test, t-test and log-likelihood ratio. The results revealed that for the entire reference data set, cooccurrence frequency outperformed the rest of the AMs. For candidate pairs of nouns and non-hada verbs, however, log-likelihood ratio turned out to be better than or at least comparable to the other four AMs. In comparison with that, cooccurrence frequency showed the worst performance among the five AMs. Finally, the present study has implications for development of a methodology for evaluation of collocation extraction methods in the field of Korean linguistics by making its procedures and findings reproducible.