No. 636] GAMETIC RATIOS IN DBOSOPIIILA 57 



improvement of culture conditions has been that of the char- 

 acter and methods of use of the food for the larva. There are 

 specific viability differences among the larvae of the different 

 mutant types and combinations. Such viability differences 

 must depend upon differences in the characters of the larvae, 

 and these, because of the intervening metamorphosis, have little 

 direct relation to the characters of the adult, but are products 

 of the action of the same mutant gene. The high correlation 

 observed between extensive change in adult characters and high 

 degree of inviability must, then, mean that such genes generally 

 cause changes which interfere directly with the success of the 

 larvje. 



Three larval characters are known— the tumor responsible for 

 the death of Iethal-7 larvje, the much shortened larvae of 

 the mutation "chubby," and a marking on the posterior end, 

 viz., "barette." It is supposed that a high proportion of tlie 

 larval characters that lead to inviability are differences in in- 

 ternal structures, but some of these might be detected. How- 

 ever, no systematic search for larval characters has been made 

 even in the case of the inviable mutants where such differences 

 are probably present. 



As we have stated, the distortion in ratios that arises from 

 inviability, especially inviability originating in the larval stage, 

 can be very materially reduced by improvements in culture 

 media and in methods. Many poorly viable mutants can be 

 made quite generally usable, as, for example, dachs. But when 

 a mutant such as dachs is to be used in linkage determinations, 

 the experiments should be so planned as not to include more 

 than one of these characters. The presence of a single 

 poorly viable character in an experiment does not prevent the 

 calculation of correct crossover values. The complementary 

 classes that do not include the inviable character should be in 

 the same proportions as in the gametic series. Even in cases of 

 mutants completely lethal, the linkage relations of the lethal 

 gene can be calculated accurately from the ratios shown by the 

 other characters of the cross. The classes that include the 

 inviable character are often also usable, but with less certainty 

 that the values are correct. Such values are correct when the 

 presence of the mutant decreases by the same percentage the 

 size of every class in which it occurs. The fewer the mutants 

 involved in an experiment, the greater the likelihood that this 



