Refinement of Feeding Experiments for Milk Production 237 



as these records constituted only six per cent of the total, even assuming 

 that they were subject to greater variation than the records of normal 

 individuals, the effect of their inclusion would be small. However, it 

 was believed that under the conditions of this selection no greater varia- 

 tion would result. A study was made on this point. Groups of six cows 

 were made up as follows: (i) all aborting cows; (2) all normal cows; 

 (3) half aborting cows and half normal cows. In order to get enough 

 aborting cows to make the results more trustworthy, the number of records 

 of such cows used in these studies was increased by all such records in 

 the herdbooks which satisfied the conditions of selection. The results 

 are shown in table 14: 



TABLE 14. Effect of Inclusion of Aborting Cows on the Variation of Groups 



ry -i- r i^uiiiuci Coefficient of 



Composition of group r • ur* 



f >= ^ or groups variability 



All aborting cows 5 3 . 30 ± o. 70 



All normal 7 5 . 61 ± i . 01 



Half aborting cows 7 5 . 20 ± o. 94 



It is shown in table 14 that under the conditions of this selection 

 abortihg cows were apparently less variable than normal cows. Altho 

 the nimiber of groups used is too small to make the results, absolutely 

 conclusive, they indicate that aborting cows are not more variable than 

 normal cows. The variability of mixed groups, even, shows no increase 

 above that of groups of all normal cows, when selected on the basis used. 



Hills (1896) found that abortion causes a decrease of one- third in milk 

 flow from the production of the preceding lactation period. When 

 aborting cows are selected on the basis of their production after abortion., 

 however, their variation is apparently less than the variation of normal 

 cows selected in the same way. 



This discussion is included to indicate the soundness of the procedure 

 followed. It should not, however, be concluded that animals which have 

 aborted should be included in an experimental group. At all times in 

 experimental work the animals used should be free from disease and from 

 all residual effects of disease. While aborting cows are apparently no 

 more variable when selected on the basis of production after abortion, 

 the residual effects of the disease on the health of the animal and on her 

 production are not known. Such an animal is diseased, is abnormal, 

 and hence should not be included. It is of .the greatest importance to 

 use none but healthy cows in experimental work. 



347 



