AbstractWe study the detectability of multiple stellar populations in small kinematic data sets. We investigate the expected significance of detection in several cases, with different parameters for the populations, i.e. number of stars, the separation between the velocity dispersions, the errors in the velocities and fraction of stars in the subpopulations. We find that for data sets smaller than 100 stars, it is hard to get detections at the 3σ confidence level using only kinematic data and conclude that additional information must be taken into account for the robust identification of distinct populations. (© 2008 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)