13S Prohohle Error in PUj-feedlmj Trials 



animals and that "the requisite precision in feeding trials can onlj' be 

 obtained by increase of numbers, or, if that is impossible, repetition of 

 the experiment." This conclusion is criticised by Mitchell and Grindley 

 as being "not based upon sound and sufficient evidence" and not in 

 harmony with the results of their own analysis of more comprehensive 

 data. They quote, in particular, the results of three Canadian experi- 

 ments with swine, comprising 10 lots of 5 to 10 pigs each, in which 

 the probable errors arc remarkably, and, with one exception (13-5 per 

 cent.), uniformly low, averaging only T-l per cent. The probable 

 errors of the separate lots, excluding the highest, ranged from 2-5 to 

 10-4 per cent. "The reports of these experiments are too meagre to 

 enable one to tell what feature or features of experimental control are 

 responsible for this low variability." 



The case of the pig is clearly one for special consideration in this 

 connection, since its more prolific breeding qualities enable a much 

 closer equalisation of experimental lots w-ith regard to breed and age 

 than is possible with cattle and sheep. This is notably of advantage 

 as regards breed, since this factor undoubtedly contributes more than 

 any other to the disturbing complex of "individuality." This is 

 illustrated by Mitchell and Grindley from the records of 18 lots of pigs, 

 each lot consisting of a single litter, ranging from 5 to 10 pigs. In only 

 3 lots is the probable error greater than tiie average (\\\ per cent.) 

 of all American results as ({noted above. 



In further illustiation of this point the records of a pig-feeding 

 experiment carried out at the Manor Farm, Garforth ( Kxperiniental 

 Farm of the University of Leeds), in 1913 are summarised below. 

 A more detailed account of the objects and results of the experiment 

 has been given elsewhere^. Ten Large White pigs, about 8 weeks old 

 at the outset, were subjected for 24 weeks (June 30-December 15) 

 to identical conditions of feeding and general treatment. The pigs 

 were drawn from two litters and for convenience were treated as two 

 groups, each drawn from one litter only. With one exception (No. 9) 

 all were castrated males, but owing to the unsatisfactory progress 

 made by No. 7 it was replaced at an early stage of the experiment by 

 a female from the same litter. The data given in the tables for No. 7 

 refer to this latter animal. 



Special feeding pens were used in order to make it possible for each 

 animal to be fed separately and the amount of food supphed was carefully 

 adjusted to the appetite of the least voracious animal. In this way it 



' Crowtlier and Ruston, Journal Bd. Agric. xxi. 789. 



