Twelve sows each farrowed in an experimental pen designed especially for this study or in a crate. Viable counts of enterobacteriaceae were performed in samples taken from the laying area and from the teats. Secretion from every mammary complex was examined repeatedly for bacteria and for somatic cells. The sows in the experimental pen did not lay down in their own faeces. The viable counts in samples from the laying area and the teats were much lower than with the sows kept in farrowing crates. Infection with E. coli was observed in 3 mammary complexes of the sows in the experimental pen as compared to 27 complexes of the sows in the crate. More than half of the infections was detected in the samples taken before farrowing began. In the average the bacteria persisted for 1.3 days. On the first 4 days of life piglets sucking teats with cytologically defined mastitis had an average daily gain of 105 g as compared to 125 g with piglets sucking healthy teats. In conclusion puerperal mastitis is a consequence of faecal contamination of the mammary gland. Soiling of the laying area with faeces and urine can be reduced by improvements in the farrowing accommodations.