Cows with subclinical intramammary infections were identified by milk bacteriology. The mastitis pathogens included Staphylococcus aureus (n=9), Streptococcus uberis (n=10) and coagulase-negative staphylococci (n=10). Samples of first fore milk, main flow milk and strippings milk fractions were collected from each quarter and laboratory measurements were made of electrical conductivity, milk fat concentration and somatic cell count. Conductivity measurements were corrected for milk fat concentration and within-cow inter-quarter conductivity ratios calculated. Repeatability estimates of all measurements between days were calculated. In the case of infected quarters, all conductivity values decreased markedly (P<0·05) from first fore milk to main flow milk fractions. Conductivity differences between quarters of infected cows were substantially lower during the main milk flow phase. For quarters infected with Staph. aureus an increase in conductivity was observed (P<0·05) from main flow to strippings fractions. For uninfected quarters, conductivity declined as milk fat concentration increased with successive milk fractions. Variation, both within and between milk fractions, was greater for somatic cell count than for conductivity. Differences in conductivity between milk fractions from individual infected quarters were not accounted for by changes in fat concentration and may result from the mixing of milk from infected and uninfected regions of the gland. Localized infection may produce a decrease in conductivity between fore milk and mid-flow fractions while differential drainage from an infection site in the secretory tissue may additionally produce an increase in conductivity from mid-flow to strippings fractions. Such changes may thus provide information on the location and magnitude of an infection. The results clearly demonstrate the importance of the milk fraction when using conductivity as a diagnostic of intramammary infection, the highest diagnostic sensitivity being achieved by using first fore milk samples.