1919] Frost: Mutation in MatthioJa 139 



probably unlike genetically. In the 1916 cultures, on the other hand, 

 with better development, this type seemed substantially as uniform 

 as the others. 



If we ignore these possible genetic differences and attempt to 

 apply the scheme worked out for crenate, difficulties appear at once. 

 First, the scarcity of Snowflake singles would indicate much closer 

 linkage than with crenate, while the relative abundance of slender 

 doubles apparently contradicts this supposition. Second, the in- 

 adequate results from crossing with Snowflake (table 36) suggest 

 that the sperms carry the supposedly crossover slender factor at least 

 as often as do the eggs. While crenate as pollen parent gives results 

 agreeing tolerably with the hypothesis, slender gives results differing 

 from these in the wrong direction. 



No doubt, however, the disagreements can be over emphasized. Both 

 crenate and slender as seed parent seem to give the expected relations 

 between singles and doubles, and series 23 also does this with the 

 Snowflake progeny. Obviously the functional sperms and eggs of these 

 mutant-type parents exhibit different ratios between types, and the 

 peculiar results in other respects with slender may be related to the 

 added complication suggested above. The astonishing feature of the 

 data, of course, is the great excess of single slender over double slender 

 in series 23 — an excess which suggests an actual significant excess of 

 singles in the totals of all types given by this cross — while with selfed 

 slender there is a great total deficiency of singles. We may at least 

 feel confident that the modifications of the single-double ratio, with 

 this type and with crenate, are due to lethal action which also affects 

 the proportions of viable slender and crenate gametes or zygotes. 



If differential viability before germination is an important factor 

 with these types, very probably it differs according as Snowflake or 

 the mutant type is the seed parent, and according to the parental 

 environment. In other words, partial selective elimination during 

 seed formation may vary with the environment of the embryos, accord- 

 ing as this environment is affected by either the genetic constitution 

 or the external environment of the seed parent. Until such uncer- 

 tainties are eliminated, we are hardly justified in ruling out, for 

 the types discussed, the probability that regular segregation and (in 

 the last two cases) true linkage are concerned in these phenomena. In 

 fact, the definite differencas in ratios between reciprocal crosses and 

 between at least one of the crosses and selfing encourage further 

 attempts at satisfactory factorial analysis. 



