iz8 THE COTTON PLANT IN EGYPT CHAP. 



can be condemned unless it shows incredible abnormalities 

 in several characters. Such treatment does not lend 

 itself to precision. 



A way out of the difficulty which has been employed 

 several times has been to dispense with netting in F 2 , 

 growing the F 3 from natural seed, leaving the F 2 plants 

 in the ground as rattoons, and netting any of them which 

 appear to be important. The natural F 3 is thus used for 

 indicative, while the selfed F 3 is used for critical confir- 

 mation in the following year. The method involves waste 

 of a year, waste of labour, and waste of land, but it seems 

 to be the only plan which gives trustworthy data. 



The few crosses which have been made and studied were 

 easily effected, taking the usual precautions. 



The Fj plant was invariably netted after 1906, and the 

 experimental error of the F 2 data is therefore very low. 



The F 3 results, on the other hand, which are of supreme 

 importance in disentangling the complex F 2 data, are 

 subject to the errors which we have just detailed. The 

 same holds good for F 4 and F 5 which is the highest 

 generation we have cultivated though by this time we 

 usually know what plants to net, so that the data are 

 more likely to be trustworthy ; on the other hand 

 many such families have been grown but not studied 

 critically, through insufficient opportunity. 



The elucidation of inheritance in cotton is thus no light 

 task, and should not be undertaken without a residential 

 laboratory, ample skilled assistance, and financial 

 resources. 



The difficulty of obtaining accurate data beyond F 2 is so 

 much the more regrettable in that the factorial analysis of 

 cotton hybrids often requires the highest precision. We 

 shall see that the composition of an F 2 is commonly very 

 similar to a Gaussian curve of error ; this similarity is 

 usually probably always fictitious, and is due to the 

 appression of several true modes, which blend into one 



