Half-udder comparisons were made using 72 and 59 cows in 2 experiments of 8 months duration to assess the incidence of pulsation failure, and its effects on teat condition and mastitis, as a consequence of milking with shortened teatcup liners. Diagonally opposite quarters of all cows were milked with control liners of 148 mm effective length or with the same liners shortened to an effective length of 110–130 mm by inserting a rigid sleeve in the lower barrel of the liner. Milking with shortened teatcup liners resulted in a high incidence of pulsation failure (defined for this experiment as failure of the liner to collapse fully beneath the teat in each pulsation cycle throughout milking). More than 50% of hind teats and 90% of front teats in the research herd penetrated deeply enough into the teatcup to prevent cyclic collapse of liners shortened to 130 mm. Such failure occurred with all teats in liners shortened to 125 mm for front teats or 110 mm for hind teats. The combined results for the 2 experiments showed that almost twice as many new mastitis infections occurred in quarters milked with the shortened liners (47 υ. 25, P < 0·01). This effect was confined to front quarters for which there was a 3-fold difference in the number of new infections (32 for the short liners υ. 11 for the controls, P < 0·001). In both lactations, significantly more teats had petechial haemorrhages when milked with the shortened liners and the electrical conductivity of milk from uninfected front quarters was higher. These results might reflect an interaction between pulsation failure and differences in the degree of overmilking that usually occur between front and hind quarters.