Deflector shields were inserted in the teatcup to prevent ‘impacts’ against the teat end, and their effect on the rate of new udder infections was studied on 15 commercial dairy farms in Britain and 16 in Australia. Small metal shields were fitted between the liner barrel and short milk tube of 2 teatcups in each cluster in a milking installation so that all cows in a herd always had the same 2 quarters shielded. Half-udder comparisons were made using 1039 cows in the British herds and 1268 in the Australian herds. Herds were quarter-sampled for cytological and bacteriological examination when shields were fitted at the start of the experiment and again after 6 months. The shield effects were similar in the 2 experiments. The combined results from the 31 farms showed that 10·5% fewer quarters became infected when protected by shields (P < 0·05). In the 1287 cows that had no infected quarters at the start of the trial, 15% fewer quarters became infected when protected by shields (P < 0·01). The shield effect was mainly confined to hind quarters in which 23% fewer shielded quarters became infected (P < 0·001). The small reduction of 2% in forequarters was not significant. In the 16 herds in which post-milking disinfection of teats was practised, shielded quarters had 23·5% fewer new infections than unshielded quarters (P < 0·01). In 6 British herds milked in installations classed as non-standard, shielded quarters had almost 50% fewer new infections than the unshielded quarters (0·05 < P < 0·10). The results provide evidence of a milking machine effect, at varying levels of intensity, to increase new infection rates on all or most farms. Control of this machine effect could lead to substantial reduction in new infections in some herds and a modest reduction in most.