EVIDENCE OF GERM CELL SELECTION 405 



large and the difference between the two percentages is sufficiently 

 great to warrant the conclusion that treating parents which 

 are heterozygous for brachydactyly with alcohol vapor re- 

 sults in an increase in the number of brachydactyl offspring 

 produced. 



It is perhaps significant in this connection that in the first 

 three experiments, those in which the males were treated, the 

 increase in brachydactyly is roughly proportional to the aver- 

 age daily dosage, the greater the dosage (table 1) the higher 

 the percentage produced. In experiment 4, while the actual 

 dosage was less, it was very apparent that the general physio- 

 logical effects were much greater, and this fact seems to be 

 mirrored in the more pronounced increase in brachydactyly 

 in 4- A. 



It is also of interest that in the three experiments in which 

 the. males are involved there is a correlation between the magni- 

 tudes of the C's and A's. In other words, if the percentage of 

 brachydactyly is relatively high in one part of the experiment 

 it is also relatively high in the other part, and vice versa. For 

 example, in 3-C the percentage is several points higher than for 

 all of the C's combined, and in 3-A a similar condition obtains 

 in reference to the total A percentage. In 2, A and C are both 

 below their respective averages while 1-A is intermediate be- 

 tween 2-A and 3-A, and 1-C likewise intermediate between 2-C 

 and 3-C. This point will be reverted to in a later paragraph. 



The conclusion in reference to brachydactyly that seems 

 justified is that by treating a heterozygous parent with alcohol 

 vapor of sufficient strength the proportion of brachydactyl to normal 

 offspring can be increased. 



Polydactyly. Three experiments are available for the study 

 of Polydactyly. The results of these experiments are not uni- 

 form, and when the percentages for all the C's and for all the 

 A's are computed there is found to be an exact coincidence. 

 Such a group of data might well serve to illustrate fluctuations 

 of percentile values in individual samples, and the tendency of 

 these, in a sense provisional, values to approximate the true 

 values as the magnitude or number of samples increases. Such 



