Between 2004 and 2006, 145 African horse sickness viruses (AHSV) were isolated from blood and organ samples submitted from South Africa to the Faculty of Veterinary Science, University of Pretoria. All nine serotypes were represented, with a range of 3-60 isolates per serotype. The RNA small segment 10 (S10) nucleotide sequences of these isolates were determined and the phylogeny investigated. AHSV, bluetongue virus (BTV) and equine encephalosis virus (EEV) all formed monophyletic groups and BTV was genetically closer to AHSV than EEV. This study confirmed the presence of three distinct S10 phylogenetic clades (a, b and c). Some serotypes (6, 8 and 9 in a; 3 and 7 in b; 2 in c) were restricted to a single clade, while other serotypes (1, 4 and 5) clustered into both the a and c clades. Strong purifying selection was evident and a constant molecular clock was inappropriate. The S10 gene is the second most variable gene of the AHSV genome and the use of S10 in molecular epidemiology was illustrated by an AHS outbreak in the Western Cape in 2004. It was shown that two separate AHSV were circulating in the area, even though AHSV serotype 1 was the only isolate from the outbreak. The small size of the gene (755-764 bp) and conserved terminal regions facilitate easy and quick sequencing. The establishment of an S10 sequence database is important for characterizing outbreaks of AHS. It will be an essential resource for elucidating the epidemiology of AHS.