No. 546] INFLUENCE OF STARVATION 329 



lots, and consider the Navy D and Navy H comparable, 

 we get: 



DD-HH DD-HHH 



DDD-HHH DDD-HH 



HD-DH HD-DHH 



HDD-DHH HDD-DH 



2. Statistical Formulae Employed 

 Methods ample for all the needs of this study are fur- 

 nished by the simplest of the Pearsonian statistical for- 

 mulae. The comparisons in the main are restricted to 

 those based on the mean, standard deviation and coeffi- 

 cient of variation. 



These do not fully describe a population, but they fur- 

 nish more information concerning it than do any other 

 three simple constants, and are sufficient for our pur- 

 poses. The methods of calculation are now familiar 

 or readily accessible to all biologists. The original data 

 are available for any other comparison, e. g., that based 

 on skewness. 



The chief possibility of untrustworthiness in the sta- 

 tistical constants seems to me to lie in a possible biolog- 

 ical source of error introduced by growing the compar- 

 ison series in rows instead of mixing all the individually 

 labeled seeds together and scattering them quite at ran- 

 dom over the entire field. 14 If because of the irregular- 

 ity of the fields, some of the rows were subjected to 

 slightly better and some to slightly poorer conditions 

 than the average, and if the rows of an individual series 

 were not distributed over the field in a perfectly random 

 manner, a slight source of differentiation quite undetect- 

 ible by the statistician's simple probable error would be 

 introduced. I suspect this to be the case, and conse- 

 quently our probable errors are perhaps too low as cri- 

 teria of the existence of differentiation due to the treat- 

 ment of the ancestry. 



Fortunately we are not limited to a single comparison, 



