136 



zygous. It therefore serves to complete 



the demonstration of the relation of 



crossing over (between forked and 



fused) to mutation in the bar locus. 



The experiment of table 19 was, 



however, planned for another purpose. 



It will be seen that in the mother, 



... BB' , 



which was . „ , eqjial crossmg over 



might give rise to new types, namely, 

 double-bar and double-infrabar. The 

 first could not be distinguished, in 

 somatic appearance, from the unmu- 

 tated double types (BB' and B'B); but 

 the double-infrabar should be readily 

 detected. Such an individual would be 

 forked. It may accordingly be con- 

 cluded that none of the 35 forked 

 (not-fused) offspring represented 

 equal crossing over between the halves 

 of the rvvo double-type bar allelo- 

 morphs present. It therefore seems 

 probable that crossing over of this 

 kind is not much, if any, more fre- 

 quent than is that between the two 

 elements of a double-type allelomorph 

 when the other chromosome carries 

 round (tables 4, 5, 12, 13, 15, 16). 



Several of the above tables agree 

 with a small series of infrabar over 

 round, heterozygous for forked and 

 for fused, in showing that infrabar lies 

 between forked and fused. It must 

 clearly be either an allelomorph of bar, 

 or bar plus a modifier that lies near bar. 



The experiments with bar-infrabar 

 and with infrabar-bar show that these 

 two types both contain infrabar as a 

 unit distinct from bar. 



Since bar-infrabar was produced by 

 an unequal crossover that occurred 

 very close to the left of infrabar, it 

 becomes unlikely that a modifier can 

 lie on that side of the bar locus; infra- 

 bar-bar furnishes similar evidence that 

 there is no modifier to the right. All 

 the evidence thus indicates clearly that 

 the infrabar gene is really a modifica- 

 tion of the bar gene itself. 



STURTEVANT 

 FREQUENCY OF BAR MUTATIONS 



The data presented in tables 1 to 19 

 have been examined in an attempt to 

 formulate some general statements as 

 to the relative frequency of the vari- 

 ous types of mutation in the bar locus. 

 It is probable that homozygous double 

 types, and double over single show the 

 lowest frequencies of mutation, and 

 that double type over round shows the 

 highest. Both these results might have 

 been expected. There is, however, so 

 much variability among crosses of the 

 same general nature that these conclu- 

 sions must be accepted with caution. 

 For example, the two largest series are 

 those from homozygous bar (20,438 

 offspring) and from homozygous in- 

 frabar (16,918 offspring). The me- 

 chanical conditions should be alike in 

 the two cases, since both represent 

 homozygous single types. Yet from 

 the first there appeared 0.03 percent 

 of reversions, or 1 in each 2920 off- 

 spring; from the second there were 

 0.11 percent, or 1 in 940 offspring. In 

 view of such unexplained differences 

 as this, and in view of the statistical 

 difficulty of determining probable 

 errors for such small percentages, it 

 does not seem profitable to discuss 

 further this aspect of the data, except 

 to note that mutation frequency does 

 not appear to be correlated with fre- 

 quency of forked-fused crossing over. 



THE CROSSOVER VALUES FOR FORKED, 

 BAR AND FUSED 



The experiments recorded in tables 

 1 to 19 include by far the largest series 

 of data yet accumulated for the cross- 

 over values of the three loci, forked, 

 bar and fused. These are summarized 

 in tables 20 and 21. In these tables all 

 the mutant individuals have been 

 omitted. Their inclusion would not 

 have affected any of the values ap- 

 preciably. In table 22 the data already 



