Growth anil variation in maize. 



161 



20. 



bution of all the plants iu each series which have their ineau quintil« 



classes indicated. 



tile classes 



3-4 



4-2— 50 



Standard Devi- 

 ation 



Mean 



Standard Devi- 

 ation 



Mean 



Standard Devi- 

 ation 



1-0840 ±-0357 I 3-9026 ±-0550 

 1-0506 ±-0404 I 3-6703 ±-0481 

 1-1182 ±-0346 : 3 -8442 ±-0554 



1-0114 ±-0389 

 0-9614 ±-0340 

 1-0201 ±-0392 



4-5000 ±-0446 

 4-5071 ±-0412 

 4 -5804 ±-0230 



0-7427 ±-0316 

 0- 7221 ±-0291 

 0-3607 ±-0163 



1-0843 ±-0211 3-8057±0295 



0-9676 ±-0209 ' 4-5292 ±-0211 0-6085 ±-0149 



r3198±-0047; 3-7106 ±-0(i82 I l-2502±-0058 4-4073±-0285 



0-9932 ±-0202 



-0-2355 ±0217 i -H 0-0951 ±-0306 \ - 0-2826 ±-0217 ! + 0-1219 ±-0355 i - 0-3847 ±0251 



As was pointed out above the most interesting question in con- 

 nection with the standard deviations is whether the end classes are 

 less variable than the middle classes when account is taken of the 

 differences in the size of the means. An examination of the above 

 table will make it clear that there is a marked difference in this 

 respect. Thus in the small plants ha\'ing- their mean quintile positions 

 between 1"0 and 1-8, the average observed standard deviation is 0-35 

 less than the standard deviation of the theoretical distribution. Tliis 

 difference is some fourteen times its probable error. In the plants whose 

 means fall in the middle class, i. e., between 2'6 and 3-4, the obser- 

 ved deviation is only 0*24 less than the theoretical. This difference 

 is more than ten times its probable error. Thus the extremely small 

 plants have a standard deviation nearly fifty percent lower than the 

 medium sized plants after all allowances are made for the differences 

 in the size of the means. A similar but even more marked difference 

 is shown between the very large and the medium sized plants. 



On the assumption which we made that the manner of growth 

 was determined by Meudelian factors this is exactly the result which 

 would be expected. The extreme classes would tend to be more nearly 

 homozygous for all factors and hence would be less variable than their 

 heterozygous neighbors. On this assumption the two end classes would 

 tend to be equally homozygous and ought to show about the same 



Induktive Abstammangs- und Vererbungslehre. XIV. \l 



