In order to assess the potential value of milk conductivity sensors for the detection of subclinical mastitis, measurements have been made manually on quarter foremilk samples for 10–74 days from 44 cows in two herds. Some measurements were also made of quarter cell counts and yields but the true status of the cows was established by bacteriological examination near the end of each run.
For uninfected cows, conductivities of the four quarters tended to move in parallel from day to day whereas for infected cows there were usually substantial departures from parallelism. A statistic which is a measure of departure from parallelism is the quarter × day interaction mean square (I.M.S.), the magnitude of which was used to discriminate between infected and uninfected cows. Using conductivity data obtained at the time of the bacteriological examination, I.M.S. correctly identified all 19 un-infected cows and 24 out of 25 infected cows, the single misclassified cow being only mildly infected. The procedure was further tested on data obtained from one of the herds a year later. In this case, using the I.M.S. limits established the year before, 14 out of 18 infected cows were correctly identified, the four failures being only mildly infected. With nine uninfected cows, there were three false positives but these cows were all pregnant and had been lactating for 44 weeks. Although special limits may have to be set for cows in late lactation, two of these animals had been infected during the lactation, so that conductivity changes may reflect past damage to the udder.
The same data used to set I.M.S. limits were also examined in simpler ways that required little or no calculation. These were ‘differential conductivity’ (the quarter with the highest conductivity divided by the value for the lowest), ‘absolute conductivity’, ‘mixed conductivity’ (mixed foremilk from all four glands) and ‘out-of-balance conductivity’ (the product of the conductivities of the fore quarters divided by the product of the hind quarters or vice versa). These methods of analysis were all less sensitive than the I.M.S. test for parallelism. Differential conductivity, absolute conductivity and mixed conductivity all detected severe cases but missed more mild cases. Out-of-balance conductivity was much less satisfactory because it missed some severe cases.