508 TRANSMISSION 



All the rest are mixed in color, no matter what their appear- 

 ance. Of these the 4 &R and the 4 B*IV will most likely 

 appear black, as will also, in all probability, the 12 EPRW, 

 because the B elements are clearly in the majority. Similarly, 

 equal numbers will appear red, and other equal numbers will 

 seem to be white, unless both colors appear, as in roans and 

 piebalds. 



There are three sets of six each (6 B*R*, 6 B*W 2 , and 

 6R*W*) in which but two color elements are present, but in 

 which the appearance will probably be fixed by the color that is 

 most pronounced and which, therefore, tends to dominate the 

 other ; thus the 6 B^lP will appear as black or very dark red. 



Thus it is that appearances are often deceiving, and that 

 which looks like a heterogeneous jumble is, after all, an orderly 

 collection of mathematically exact combinations. A scheme like 

 the above serves to show the exceedingly complicated, yet 

 orderly, systems that necessarily arise in bisexual reproduction, 

 whatever the characters involved, complexity that increases 

 rapidly, indeed almost inconceivably, as generations multiply. 



Because of these facts reproduction would be reduced to a 

 problem in probabilities, and we should have all possible combi- 

 nations presented, were it not for the fact that selection is 

 always at work to eliminate certain unfavored forms, and that 

 differences in fertility serve to give certain combinations still 

 further advantage over others. However, we are not to over- 

 look the fact that, even though certain values be withdrawn 

 from such a distribution, the laws of probability continue to ap- 

 ply to the remaining values, whose combinations will take place 

 as before, and in the end give rise to a distribution not very 

 different in form from that which would have arisen if no 

 values had been removed. 



An ultimate confirmation of this statement is found in the 

 fact that most frequency distributions are fairly symmetrical, 

 and that one large enough to be fairly "smooth" whatever the 

 number of its terms or the size of its frequencies, can be closely 

 reproduced by expanding a binomial. If the distribution be 

 symmetrical, the terms of the binomial should be numerically 

 equal (B + R, or ^ -f |) ; but if its mode is not near the middle 



