C. A. BARBER. 199 



Roughly, in the Saretha series, the widest part is half way up the leaf, 

 and two-fifths of the way up in the Sunnabile. This emphasizes the shortness 

 and broadness of the leaves in the latter series. 



(c) A fuither expression of this feature may be obtained by an estimation 

 of the rate of widening in the leaf, per unit of length. This upward gradient 

 can be put in figures as follows : — • 



Upward gradient = g''"^^"^ width-ieaat width -^^ ^j^^ distance from widest 

 to narrowest part. The narrowest part may be safely taken as 1" above 

 the base in each case, so that the distance referred to in the formula will 

 be the distance above the base of the widest part, less 1". Working out the 

 figures by the formula, we obtain the following for 1917. The actual total 

 length of leaves was not taken in 1916. 



1917. Katha section Gradient of 1 in 100 



Mesangan section 1 in 109 



Saretha group 1 in 102 



Sunnabile group 1 in 55 



Thus the upward gradient of widening in the leaf is practically twice as 

 steep in the Sunnabile series as it is in the Saretha. 



The downward gradient, from the widest point to the tip of the leaf, may 

 be obtained in a similar manner, the formula being "-''?I5|^^'l-*° in the 

 difference between total length and distance of widest point from base of leaf. 

 The 1917 figures give us the following : — 



1917. Katha section 1 in 48 

 Mesangan section 1 in 40 

 Saretha group 1 in 44 

 Sunnabile group 1 in 33 



The downward gradient is considerably steeper than the upward, but 

 there is less difference in the two groups, that of the Saretha leaves being three- 

 fourths that of the Sunnabile, instead of nearly one-half. 



From the above study of leaf measurements we have thus succeeded 

 ill extracting new and stable differences in the form of the average leaf 

 in the two groups, and this gives important support to the reality of the 

 grouping of the varieties concerned. We are in a position to draw the outline 

 of a typical leaf in each group, but the length is so enormously greater than 

 the width, that some sort of compromise is necessary. I have accordingly 

 drawn the two leaves to scale, multiplying the width by ten, and the 



