The separation of anomalies from backgrounds is a critical step in geochemical prospecting. However, the determination of stream sediment geochemical anomalies is always affected by lithologic backgrounds in areas with variable lithologies. If this influence is not eliminated prior to anomaly determination, some serious errors may occur. In fact, a problem of lithologic background is essentially a problem of multiple population, which can be effectively solved by the clustering method based on the Expectation-Maximization (EM) algorithm. In this study, the authors applied the EM clustering method to a geochemical data set from a 1∶200,000 scale stream sediment survey, and then discussed the influence of separating multiple populations on anomaly determination. A practical example demonstrates that geochemical anomalies of stream sediments in lithologically complex regions can be determined in a more reasonable way by using the EM clustering method. This is mainly reflected in two aspects: on the one hand, strong but false anomalies can be eliminated, and on the other hand, weak but important anomalies can be uncovered.