336 J. A DETLEFSEN AND E. ROBERTS 
chosen on the basis of external characters which can be deter- 
mined by direct observation. To select for high and low cross- 
over values is rather more tedious and time exacting because 
the individual cannot be classified directly with respect to its 
crossover potentiality. Its character is disclosed only after 
obtaining a reasonably large progeny. Characters which could 
be recognized easily and classified rapidly and accurately were 
indispensable. The two allelomorphic pairs, white eye vs. 
red eye, and miniature wing vs. long wing seemed to fulfill these 
conditions, and they have the added advantage of giving a large, 
initial, normal percentage of crossovers (about 33 per cent), 
which means that variations are thus more readily detected. 
The procedure followed in the low-selection series A is typical 
of all the series and can be taken as a sample. A single white 
miniature female mated to a red long male gave Fi white minia- 
ture males, w m, and long red females ^ ~, the latter 
being double heterozygotes. The Fi sibs were mated in pairs 
in 8-drachm homeopathic vials, and the pairs were removed to 
new vials about every three days. The culture methods were 
those commonly used with Drosophila. The F2 offspring from 
each vial were classified daily until a fair sample of each Fi 
female's 'crossover capacity' was obtained. As expected, the 
offspring were of four kinds: the parental types, red long, and 
white minature and the crossovers red miniature and white 
long. It was impossible to anticipate which Fi female was going 
to be selected because of her low crossover ratio determined by a 
reasonably large progeny, and it was likewise virtually impos- 
sible to continue mating in pairs the sibs from each Fi female 
until we could find out which line was going to be used to con- 
tinue the selection. Thus in Fi series A, low selection (table 1), 
there were twenty-eight pairs of Fi individuals, several of which 
appeared to be promising material, but we eventually chose 
pair 15, which gave 21 : 98 = 21.43 per cent.'^ By the time we 
' In giving crossover values we shall put the data in the following order 
throughout this paper: crossovers: total-per cent of crossing over. The classes 
are always the same and repetition can thus be avoided. 
