46 STUDIES IN IXDIAN TOBACCOS. 



which can be accompHshed. The descriptions of the various 

 types of tobacco ah'eady pubhshed show that the different forms 

 of N. fabacum can be arranged as regards each character in a 

 series, each member differing very shghth^ from the next (see 

 also Tables VII, XII, XX). This would lead one to expect a 

 large number of factors in connection with each character, and 

 in the ¥, generation a long series of intermediates, with the very 

 rare occurrence of the parent forms. The chances of directly 

 analysing such an F2 generation are small. A very large 

 number of individuals would have to be raised in the Fo genera- 

 tion to obtain any indication of the percentage of plants 

 resembling the parents, and, since the range of variation in the 

 parents due to external influences is considerable, it would be 

 quite impossible to recognise such forms with certainty. This 

 being the case, there seems very little point in undertaking 

 the labour of raising many thousands of plants in the Fo gener- 

 ation. The plan adopted has been to grow not less than 1,000 

 plants in the F2 generation in order to obtain a fair idea of the 

 extent of variation by eye, and then measure as many as possible 

 of these, generally 300 to 600. Self-pollinated seed from plants, 

 which together cover more or less completely the range of 

 variation, was taken and these cultures raised in the F3 generation. 

 In the F3 cultures about 200 plants w ere grown from each parent, 

 and 100 to 200 measured. No choice was exercised in connection 

 with the plants measured. Two or three complete lines of the 

 culture were taken. The limits of variation in the F^ generation 

 are much smaller. In some cases the cultures apparently bred 

 true in some one character. A similar procedure was adopted 

 in the F4 generation. 



The number of plants used is small and the investigation 

 cannot pretend to be statistic, but thelarger the number of plants 

 the smaller must be the number of cultures. The number 

 employed is sufficient to determine whether the culture is 

 uniform or is splitting within narrow limits. The object in 

 adopting this procedure is to obtain as many as possible 



