The study evaluated the use of the Mastazyme ELISA for quantification of Neospora caninum (N. caninum) specific IgG in bovine milk and examined the relationship between serum and milk antibodies in two dairy herds. The serum and milk antibodies both had bimodal distributions in each herd. This was mainly due to between cow variation: in both herds, approximately two thirds of cows were either clearly and consistently seropositive or seronegative for N. caninum with one third consistently near the threshold. Milk and serum N. caninum IgG were strongly related. This relationship was modelled using a linear mixed model including a polynomial term for serum, the effect of herd, and between and within cow variance components. The latter gave a significantly better fit to the data than a model that allowed for a different relationship for the positive and negative (according to the serum test) groups of observations. The sensitivity and specificity (based on serum percentage positivity (pp)) of the milk antibody was determined for different milk pp thresholds. In spite of the differences between the relationship of milk to serum seen for the two herds, for those estimates with sufficient precision, sensitivity and specificity greater than 0.73 for both herds were obtained using single thresholds of 14 and 15.5 for milk pp in both herds based on, as our gold standard, serum antibody pp thresholds of 22.5 and 25, respectively. If milk antibody is to be used for detecting persistently infected cows, the higher threshold of 15.5 may be suitable while for epidemiological screening 14 would be preferable. Further validation in a greater number of herds is required, but our results suggest that this test may prove to be a useful adjunct to serum N. caninum IgG assays in the monitoring of N. caninum infection as part of herd health programmes and epidemiological studies.