Vol. 6, 1920 
GENETICS: C. W. METZ 
of this stock makes it impossible to test the latter two of his predictions. 
The other two predictions are also such that they cannot be tested at 
present; the first because of the loss of hairy stock through sterility, and 
the second because of the sterility of rugose-glazed hybrids.^ Had it not 
been for these practical obstacles the values would not have been left un- 
determined in my original paper. 
The matter need not be dropped here, however. It should be noted 
that, in any event, tests of the predictions would have been futile, for the 
two hypotheses may be compared just as well by means of known values 
as by testing unknown ones. In addition, the very nature of the pre- 
dictions themselves prevents their being used for critical tests, because 
they are so constructed that, on either view, they must be fulfilled if my 
data give sufficiently accurate ratios.^ There is no alternative. This 
being the case it becomes a question of ascertaining the accuracy of my 
data^ or the validity of the assumptions underlying Castle's predictions. 
An illustration may be used to amplify this point. Castle's last two 
predictions apply to the gene for frayed. My data on frayed (pp. 112 
and 126) as tabulated by Castle would locate the gene 1.3 units from yellow 
and 18.6 units from vesiculated. Castle's own calculation of the value 
yellow-vesticulated is 17.4. This would put the three genes in almost 
exactly a straight line (1.3 + 17.4 = 18.7, as compared with 18.6). These 
are the only data available for the direct determination of the location of 
frayed, but so far as they go they conform to the linear hypothesis. Castle's 
predictions relate to the cross-over values that should be given by frayed 
and forked, and by frayed and glazed. The former should be "between 
39 and 41" and the latter "between 43 and 46." The predicted values 
are, of course, to be calculated entirely on the basis of single cross-overs, 
as all values are determined on his system. 
These predictions are evidently based on the cross-over values given 
b}^ yellow and forked, and yellow and glazed, respectively. By leaving 
out of account the double cross-overs, Castle calculated the latter values 
from my data as 40 and 44.5. Since frayed is approximately one unit 
from yellow it should give within approximately one unit of the same 
cross-over values as yellow. This principle would apply on either hy- 
pothesis. It i^ evident, then, that by calculating the value frayed-forked 
in the same manner that yellow-forked was calculated — namely, by using 
only single cross-overs — the result would have to be within approximately 
one unit of 40, if crossing over is a consistent process. The same principle 
applies to the frayed-glazed relation. Both predictions would have to be 
fulfilled (within the limits of experimental error), under these conditions. 
But this does not in any way substantiate the three dimensional hypothesis. 
Predictions in such cases as these can be made and fulfilled on either hy- 
pothesis, providing the calculations are made in accordance with the 
hypothesis (the one including, the other excluding, double cross-overs). 
