fond 
New York AGRICULTURAL EXPERIMENT STATION, 87 
been rinsed free of brine a small amount of water commonly 
remained in the filter cups and this was hard to remove. This 
tended to wet the cotton, causing it to mat down and not filter 
the air properly. Also the bulk of the cotton in the filter was. 
so small that on screwing on the filter cover the cotton was 
easily entangled and rolled to one side practically destroying 
the filtering action. By careful attention to details these fac- 
tors were reduced to a minimum in our work but in practical 
use they would have been important. On machine No. 2716 
the cylindrical filter cups were so much larger that neither of 
these factors was so troublesome. 
Comparison of later forms of relief filters——On account of 
the important modifications we thought it best to to test two 
new machines, Nos. 3157 and 3158. These differed from the 
preceding form in that they did not have a relief filter on 
the dome of the milking machine and the relief filters at the 
teat cup connectors had correspondingly larger openings for the 
entrance of air. In connection with these machines there were 
furnished two different forms of relief filters which are shown 
in Plate VI, figs. 1 and 4. It will be noted from the wooden 
‘blocks which indicate the capacity of the filter cups that these 
were of practically the same size and shape. The difference in 
the filters lies in the extension of the air chamber in the case 
of filter 1 in the form of a cone which admits the air into the 
teat cup connector in front of the openings coming from the 
forward pair of teats. The filtering cotton was prevented from 
entering this cone by a perforated sheet of metal. The object 
of this change was to equalize the milking action on the four 
teats. The test was intended to determine whether this change 
in form influenced the filter action when the capacity of the 
filter remained constant. In this test machines 3157 and 3158 
were used on four cows, the conditions being kept as constant 
and the machines handled as well as possible. Each cow was 
milked four times with each form of relicf filter. The results 
of these 32 tests are given in Table IX. 
