In an attempt to improve the information obtained after computer-assisted sperm analysis (CASA), data from five stallions (three ejaculates from each) were analysed before (fresh, extended semen) and after cryopreservation using traditional statistics as well as a cluster analysis. The data matrix consisted of 13 987 observations of individual spermatozoa for fresh, extended semen, and 8305 for frozen-thawed samples. As expected, freezing and thawing resulted in a marked decrease of CASA-derived variables of sperm kinematics. All sperm velocities were significantly lower in frozen-thawed samples than in samples before cooling. Using sperm velocities, six sperm subpopulations were identified in fresh semen (S1-S6). As such, subpopulations S1 and S2 were characterized by low sperm velocities, subpopulations S3 and S4 corresponded to spermatozoa depicting medium speed values, and finally, subpopulations S5 and S6 were those depicting the highest velocities. After freezing and thawing, four sperm subpopulations were identified, listed as nr FT1 to FT4. While subpopulations FT1-FT3 were characterized by low sperm velocities, and thus corresponded speed-wise to those listed as S1-S4 for fresh, extended semen, the one called number FT4 in frozen semen was characterized by high velocities, of the same range as that of the subpopulations S5 and S6 for fresh spermatozoa. The sperm subpopulation structure varied among stallions, but the cluster analysis hereby assayed was able to provide valuable information about the freezability of the samples that the customary statistics did not reveal.