26 



AN ANALYSIS OF THE EFFECT OF SELECTION. 



60 



50 



40 



30 



20 



10 



bristle number of the two classes of Dichset offspring and the differ- 

 ence between these two means. The sign of the difference is given as 

 positive when the not-specks had a higher mean than the specks (as 

 in all these cases); negative, as in other results (see below), when 

 the specks were higher. In the col- 



umn headed p^r is given the quotient 



of the difference between the two means 



divided by its own probable error, a 



measure of the probable significance of 



that difference. The last two columns 



give the x 2 and P values for the two 



(speck and not speck) distributions, 



considered as wholes. 

 These data make it certain that the 



864 plus line and crossbred plus line 



both contained second chromosomes 



with one or more plus modifiers domi- 

 nant to minus modifiers in the second 



chromosome of the speck stock. The 



1002 plus line had similar modifiers, 



and also had the same relation to the 



1331 minus line. It is probable, from 



the results obtained with the 1002 line, 



that the speck stock and the 1331 line 



had some minus modifiers in common, 



but that the second chromosome of the 



1331 line was more strongly minus. 



Both these latter results would have 



been expected, since the second chro- 

 mosome of the 1331 line came, in part 

 at least, from that of the speck stock, 



but has been minus selected, while the speck stock has not been 

 selected at all for bristle number. 



The experiments just discussed show that the second chromosome 

 contains one or more modifiers, but give us no information regarding 

 the loci of such modifiers. It is possible, on the basis of this data 

 alone, that speck itself is the minus modifier. If, however, a heterozy- 

 gous female is tested by mating to a speck not-Dichaet male, there 

 will be a possibility of crossing over between speck and any modifiers 

 in the second chromosome. The result would be that the speck and 

 not-speck offspring differ less than when an Fi male is tested. There 

 is, of course, also an opportunity for crossing over in the third chromo- 

 somes of such females, so that the Dichset offspring will not be all 

 alike with respect to their third chromosomes, as they were when the 

 male was tested; but the same crossover classes should occur among 



Fig. 14. — Not-speck (solid line) and 

 speck (broken line) offspring from 

 male back-cross tests of 864 inbred 

 plus line against speck stock. Curves, 

 based on 153 not-speck and 106 speck 

 flies, are both reduced to the percent- 

 age basis. See table 19 for statistical 

 treatment. 



