142 



allelomorphs even when two recessive 

 allelomorphs and one dominant are 

 present. Unpublished data collected by 

 Bridges (in part verified in an inde- 

 pendently arisen series of triploids in 

 my own experiments) show that this 

 relation is a general one for allelo- 

 morphs that do not produce an ob- 

 viously intermediate diploid heterozy- 

 gote. But Bridges (Bridges and Mor- 

 gan 1923) has shown that a different 

 relation may occur, even for some of 

 the same genes that show the former 

 relation in triploids. In the example re- 

 ferred to, a portion of the second 

 chromosome, carrying the normal al- 

 lelomorph of plexus (among other 

 genes) has become attached to a 

 chromosome III. It is possible, there- 

 fore, to obtain individuals with two 

 complete second chromosomes, each 

 of these carrying the plexus gene, 

 while a normal allelomorph of plexus 

 is present in the section attached to a 

 chromosome III. Such individuals are 

 plexus in appearance,— not as extreme 

 as those in a pure plexus stock, but far 

 more like such a stock than like the 

 ordinary triploids carrying two plexus 

 and one not-plexus allelomorphs. 

 While it is possible that the difference 

 here is due to a different "balance" of 

 modifying genes in the extra section of 

 chromosome II, it seems likely that the 

 effective agent is a difference in posi- 

 tion. It is to be remembered that in 

 Drosophila the homologous chromo- 

 somes lie closely apposed in somatic 

 divisions (see Metz 1916), so that there 

 is probably a real difference in relative 

 positions in the two cases. 



Besides the comparison of round 

 (ultimately from vermilion stock) 

 with round obtained by reversion from 

 bar and from infrabar, two other 

 derived types have been compared (as 

 to facet number) with the correspond- 

 ing original types. 



STURTEVANT 



An infrabar from bar-infrabar over 

 round (table 13) was introduced into 

 the inbred strain by six successive 

 backcrosses, and was then compared 

 with the old inbred infrabar, both 

 types being made heterozygous for 

 double-bar. The control (old infra- 

 bar) value is rather lower than the 

 value given in table 23, probably be- 

 cause the temperature ran slightly 

 higher. 



Since the difference between the 

 two means is about 4.5 times its own 

 probable error, it is probably signifi- 

 cant, but more extensive data will be 

 required to establish this point. 



A female of the inbred series that 

 was double-bar over round, mated to 

 a round male, gave rise to one bar 

 male by mutation. This male was 

 mated to double-bar-over-round fe- 

 males, and the resulting double-bar- 

 over-bar daughters were compared 

 with double-bar over the old inbred 

 bar, derived from cultures made up 

 at the same time and put side by side. 

 Here again there is a slight difference 

 from the value of table 23, perhaps 

 due to a temperature difference. 



The difference between the means is 

 slightly over 3 times its probable error, 

 and may be considered as doubtfully 

 significant. Here again, more data are 

 needed. 



It may be pointed out that in tables 

 25 and 26 the derived type is presum- 

 ably the larger in both cases. It is pos- 

 sible that this result is to be correlated 

 with that recorded in table 24, namely, 

 that round by reversion is perhaps 

 smaller than wild-type. Both relations 

 are consistent with the view that there 

 exists a normal allelomorph of bar that 

 has an effect on facet number opposite 

 to, but much weaker than, that of bar; 

 for both of the derived single types 

 tested were from double type over 

 round, so that these derived forms 



