226 Studies in Indian Cotton 



for the parental leaf factors are not those of the actual plant but 

 the average of the type as given by the offspring (produced by self- 

 fertilisation) in the two subsequent generations. In two cases only is 

 the variation from the parental mean at all marked and in both these 

 this difference is not shown by the reciprocal. 



In the F„ generation a continuous series of forms is produced in 

 which every value of leaf factor between the parental limits is obtained. 

 Diagram 1 illustrates one such case and is derived from the series given 

 in Table XIII (b). It is here noticeable that, while the series appears 

 continuous, in that every value of leaf factor (within the limits imposed 

 by the experimental error) occur, the number of individuals is by no 

 means regularly distributed throughout the series — in other words, the 

 frequency of each class exhibits marked variation. The curve is, in 

 fact, multimodal (Davenport (6)) and possesses three modes. The posi- 

 tion and value of these modes are instructive. While the values of 

 the outer modes differ but slightly from the values of the two p;ireutal 

 leaf factors, the value of the intermediate mode shows a fair degree of 

 approximation to the value of the mean between the leaf factors of the 

 two parental types. The proportion between the number of individuals 

 grouping themselves about these three points is 1 : 22 : 1. The 

 curve retains its trimodal nature, if for the actual values obtained by 

 direct measurement of the leaves of individual plants — the values 

 here given — the mean value of the leaf factor of the F3 offspring be 

 substituted. 



A similar curve has with one exception been obtained in every case 

 submitted to a critical examination. In this instance, the cross between 

 type 2 and type 3, there is no trace of a multimodal curve and the 

 ratio between the number of individuals in each group (Table XIII («)) 

 diverges markedly from that obtained in the instance given above. 

 Lack of opportunity and the difficulty of handling a cross between two 

 monopodial types have rendered it impossible to continue investigation 

 into the behaviour of this cross and for the present it must remain 

 undecided whether, on further examination, this too will fall into line 

 with the example more fully investigated or whether a different series 

 of phenomena is here instanced. 



So far the results have been described in outline only, and as 

 a close examinatioQ of the tables will show, are only approximate. 

 Complete agreement is, perhaps, hardly to be expected in dealing with 

 a character which, as has been already shown, cannot be measured 

 with absolute accuracy. It will be observed that the modal values of 



