No. 595] THE MECHANISM OF CROSSING-OVER 431 



ozygous condition, twenty-two factors have been contin- 

 ually outcrossed, in each successive generation, to their 

 allelomorphs. Yet after about seventy-five generations of 

 outcrossing, these characters do not show the slightest 

 contamination. The experiment therefore forms an ex- 

 tensive test and verification of the "purity of Mendolian 

 segregation." Castle has, however, raised the point that 

 in determining whether characters change, we should not 

 be content with casual inspection. One of the characters 

 in the above experiment— dachs legs— lends itself readily 

 to quantitative work, since one of its main features is a 

 shortening of the tarsus and metatarsus. Measurements 

 o* the legs of about a dozen of these dachs flies, derived 

 from the stock which had been subjected to continual out- 

 crossing, were therefore made, as well as measurements of 

 the legs of some dachs flies derived from a stock which had 

 been kept pure; the values for normal flies were de- 

 termined also. At the same time the thorax length of the 

 flies was observed, in order that any difference in leg 

 length due merely to variation in the size of the "whole 

 animal might be allowed for. The results for each individ- 

 ual are shown in the following table. Measurements are 

 given in eyepiece micrometer divisions, each of which 

 represented .026 mm. 



In order to discover whether the character had become 

 more variable as a result of outcrossing, the standard de- 

 viation of the ratios of foot to thorax, in the two stocks of 

 dachs, was calculated from the above data. In the out- 

 crossed stock the standard deviation was found to be .036, 

 and in the original stock .035; that is, so far as these re- 

 sults can show, the variability of dachs after outcrossing 

 has remained just the same. However this may be, the 

 fact remains that the character, after being subjected to 

 such long-continued outcrossing, had not approached one 

 whit nearer to the type of its allelomorph. The slight dif- 

 ference in the other direction observed between it and the 

 original mutant stock is of no significance, since just about 

 as great differences in thorax length occurred between the 



