Non-invasive pregnancy diagnosis in mares by measuring faecal oestrogens has been performed over years with great accuracy. However, results have indicated breed-related differences in the amount of excreted steroids during late pregnancy. Therefore faecal samples were collected during the last 4 months of pregnancy of Thoroughbred (n = 10), New Forest pony (n = 9), Shetland pony (n = 10) and Iceland pony mares (n = 11). Concentrations of oestrogens, 20alpha-hydroxy- and 20-oxopregnanes were measured using enzyme immunoassays. Breed differences concerning both levels (though significant only in case of oestrogens) and time course of measured steroids were observed. There was a highly significant time effect (p < 0.00001) and an interaction between time and breeds (p < 0.02) for all steroids measured, suggesting that the time effect differs for different breeds. Oestrogen concentrations showed a decrease towards parturition, whereas in 20alpha-hydroxy- and 20-oxopregnane levels a pronounced increase was found 2 and I months, respectively, before parturition. A breed effect was only significant (p = 0.001) when comparing oestrogen concentrations and was mainly due to Iceland ponies, which had the lowest concentrations especially during the last 2 months of pregnancy. An almost significant (p = 0.06) breed effect was found for 20-oxopregnanes. In Iceland mares an additional increase in faecal pregnane content was already observed earlier, reaching maximum levels before the 60th day ante-partum (a.p.), followed by a decrease until the 30th day a.p. The ratio of 20-oxopregnanes to oestrogens in the samples was significantly higher (p < 0.006) in Iceland ponies in comparison with any other breed throughout all months before parturition. The breed differences observed in the amounts of oestrogens and/or progestagens present during late pregnancy may demonstrate micro-evolutionary changes in the endocrine system of a species.