INHERITANCE OF WINTER EGG PRODUCTION 97 



years, a good fit between observed and expected ratios is ob- 

 tained, although such a result is hardly to be expected. More- 

 over, the data for 1914-15, if combined with that of whatever 

 other year may be involved, give satisfactory ratios except for 

 one male, viz., male no. 271. The ratios for this male are, on 

 Pearl's theory, 3:32:12 observed, to 0: 35.50: 11.50 expected. 

 On the alternative theory the ratios are 3 : 44 observed, to 

 1 1 . 50 : 35 . 50 expected. 



The various years are not wholly comparable to one another. 

 Certain changes in management, described in later sections, 

 have been forced upon us. Selective matings have also been 

 made in various years. Some were made for high production, 

 but others were made primarily for low production, late maturity, 

 hatching quality of eggs, vigor, broodiness, and size. Thus, a 

 portion of the matings each year are made at random as far as 

 egg production is concerned. 



In compiling these tables the progeny of each pair was first 

 distributed into the three (respectively two) groups of pro- 

 ducers required by each theory and the expected ratio that 

 agreed most closely with the observed ratio determined by 

 inspection (compare table 3). It was of course necessary to 

 select a gametic constitution for the male that would give 

 suitable ratios for all the females with which he was mated. 



An exact fit for all females cannot be expected, but in only two 

 instances has the deviation been greater than three and one- 

 half individuals, i.e., a change of three individuals and one-half 

 from one class to another makes the observed and expected 

 ratios agree perfectly. Changes of three or three and one-half 

 individuals are rare. The sum of the observed and expected 

 ratios of the various mates of each male constitute the data 

 given in table 5. 



Within the limits designated, we have not hesitated to choose 

 that theoretical ratio which gives the closest fit in the total. In 

 some instances it is possible to assign more than one theoretical 

 ratio, but we have not done so. 



By accepting the discrepancies mentioned between observed 

 and expected ratios, we avoid classifying an individual breeder 



