554 Scientific Proceeding?, Royal Dublin Society. 



In three eases out of thirty-seven four-day tests the average is outside 

 the error limit (* on table). On applying the same test to cow 128 it is 

 found that in no case does the result diverge seriously from the average of a 

 period extending equally on both sides of the period of the test. The same 

 thing holds for all the other cows under observation. Even in the case of 

 the hyper-sensitive cow 97, the three exceptional results above (* on table) 

 are easily explained. It is known that cows are affected differently as 

 regards milk and butter-fat yields by the return of the oestrum period ; 

 in some cases the fat rises, in others falls, and in others still remains 

 unchanged. Cow 97 belongs to the group which is seriously affected. The 

 oestrum recurred on the night of June 15th ; and the result is a rapid rise in 

 fat (see diagram III) : the increase is not sufficient, however, to upset the 

 accuracy of the four-day test. On July 2nd and July 3rd, and again on 

 July 2oth, it was noticed that other cows in the same field were in 

 oestrum ; and it is a fair inference that the excitation of cow 97, due to 

 the disturbing effect of those others, influenced her fat yield in a fashion 

 similar to that shown after July 15th. Because of these disturbances 

 the four-day test for June 30th-July 3rd, for July 4th-7th, and again 

 for July 25th-28th are outside the usual limits. But even the appearance 

 of these exceptions in the case of a hyper-sensitive cow does not 

 invalidate the four- day test, because the careful sampler always makes 

 observations regarding oestrum, violent weather changes, and so on ; and if 

 one of these disturbing elements should intrude into the period of sampling, 

 the number of samples may be extended to six; and a six-consecutive-day 

 test is in all cases, without exception, reliable. While it is necessary to 

 make provision for serious disturbing factors by an extension of the test to 

 six days, a four-consecutive-day period is in normal circumstances sufficiently 

 long to give a result which comes within the 10 per cent, limit of error, and 

 in the majority of cases within a 5 per cent, limit, frequently approximating 

 even to the correct average for the period concerned. 



The average of a two-day test would be unreliable, thus — 



Per cent. Per cent. Per cent. 



July 15, . . 4-06 July 16, . . 5-16 July 18, . 3-42 



„ 16, . . 5-16 „ 17, . . 3-69 „ 19, . 3-54 



Total, . . 9-22 8-85 6-96 



Average, . 4-61 4-42 3-48 



The average calculated in this way is, in one instance, 20 per cent, outside 

 the true average for the period. 



A successive three-day test frequently gives a better result than a two- 



