Sept. IS, I9I9 Variation in Milk of Ayrshire Cows 287 



seemed good reason for the belief that one was likely to get better — that 

 is, more nearly physiologically normal — values for the two characters 

 here studied — mean fat percentage and mean weekly yield — if one con- 

 sidered only lactations eight months or more long. Furthermore it is 

 clear that no error of any consequence can be introduced by leaving out 

 of account short lactations, since Vigor {28) has shown that there is no 

 significant net correlation between duration of lactation and either 

 percentage of fat or average weekly yield of milk, the two characters 

 studied in this investigation. 



The third restriction obviously needs no argument in its justification. 

 Under it were excluded cases of abortion, "off -food" at particular tests, 

 diseases, and accidents of various sorts. Undoubtedly some records 

 were excluded which might fairly have been regarded as normal; but it 

 was thought best, where one was working entirely from records and could 

 not see the cow itself, to err, if at all, on the side of too great rather than 

 too little strictness. 



The two characters dealt with in this paper are (a) average milk yield 

 per week in gallons, and (6) average fat percentage. The values for the 

 former were obtained by dividing the total yields as given in the reports 

 by the weeks in milk. The fat percentage figures were taken directly 

 from the reports. The ages were taken as centering at the mid-point of 

 each year. For example, all cows recorded as 3 years or more in age but 

 less than 4 years were put in the 3-year class in the tables of the present 

 paper. Hence a 3-year-old is to be taken as including anything between 

 3 and 4 years. 



The biometric methods used were the ordinary ones. All of the dis- 

 tributions containing enough individuals to make the results significant 

 were fitted with Pearson's {18, 20) skew frequency curves, following in 

 the computations some simplifications of method. 



FREQUENCY DISTRIBUTIONS 



The frequency distributions, showing the variation in the two char- 

 acters studied, are exhibited in Tables I and II, in both absolute and per- 

 centage figures. 



122502°— 19 4 



