Vol. 6, 1920 GENETICS: MORGAN, STURTEVANT, BRIDGES 163 
experiments, however extensive, in establishing the order of the loci, 
or in testing the validity of the hypothesis of linear arrangement. For 
such purposes, as we have already pointed out, it is essential to use data 
in which all the loci are followed at once. 
The purpose of the maps is twofold: first, to give the sequence of the 
loci, and secondly, to indicate, by the relative spacing of the loci, the cross- 
over values most likely to coincide with the results of future experiments. 
For the latter purpose it is evident that mean values are needed. These 
can best be obtained by using all the data. And this, as stated, has been 
the method used for determining the distances on the maps published. 
Castle has used all the data for criticizing the sequence of the loci on our 
maps, and has accused us of using only the three-point data for determining 
the map-distances. Obviously this is an inversion of the correct relationship. 
Even when all the available data are used in constructing the map, 
these data should themselves be subjected to all the aontrols that it is 
possible to apply to them, as everyone familiar with the treatment of 
rough data will understand. For example: It is well known that under 
certain conditions, such as crowding, some mutant classes run behind 
their expected ratios. Under favorable conditions of culture, normal 
ratios are obtained. Clearly, data of the first kind unless corrected are 
unsuited for determining distances on the chromosome map; and the 
larger the amount of such data, the larger would be the discrepancies in 
a map based upon such data used with normal data. We have shown 
how the aberrant ratios due to the inviability of given classes can be 
balanced by the use of converse crosses. Further corrections and weighings 
are also desirable in order that no one class of data shall unduly 
prejudice the result. The methods for making such corrections have 
been briefly outlined in Carnegie Publication No. 237, 1916, and given 
more fully in Carnegie Publication No. 278, 1919. 
Besides environmental disturbances of the kind just described there are 
/actors that are known to modify crossing-over. Obviously such data 
should be eliminated from material from which a normal chromosome map 
is to be made. That Castle himself admits the validity of elimination 
of such data is shown by the fact that he deliberately rejected the data 
involving lethal 2 which we had shown to cause aberrant linkage rela- 
tions. If he had used these data, all of his long wires would have been 
bent. Castle has set up the claim that one of the advantages of his model 
is that such a bending of a wire in the case of white forked revealed the 
fact that the experimental value of this interval was too great. The 
same fact would have been "revealed" on any system by a comparison of 
the white forked value with the values for white rudimentary and white 
bar, both of which were based on larger numbers than white forked itself. 
Castle's discussion of interference in connection with yellow white bifid 
(p. 503) shows his failure to understand interference, and hence his criti- 
