Data on 16406 cows (6025 recorded) from the area of the East & Central Al Society were analysed by least squares procedures. Ketosis incidence was determined from the milk by a commercial keto test reagent by AI technicians in connection with the ordinary first insemination of a particular cow. Ketosis incidence was on average 11.5 %, increased up to about the 4th - 5th parity and was significantly higher in larger herds. Breed did not significantly affect ketosis incidence. There was a tendency for higher incidence with increasing usage of commercial feed mixtures. Cows with ketosis milked less than unaffected ones and cows inseminated earlier than 60 days after calving had significantly higher incidence than cows inseminated later. The complete statistical models accounted for only 5 % (all herds) and 9 % (recorded herds) of variation in ketosis incidence, indicating the need for better measures of the herd environment. Heritabilities for ketosis incidence were not significantly different from zero. Cows with ketosis had significantly poorer non-return rates than unaffected ones, indicating the need for prophylactic measures.