HOiMOTlTOSIS IN THE VEGETABLE KINGDOM. 
293 
All the possible pairs were now taken, i.e., |^(26 X 25) = 325 in number, and entered 
on a correlation table in the usual manner, the two variables being the number of 
veins in the first leaf and the number of veins in the second leaf. But as either 
member of the pair might be a “ first ” leaf, the table so formed was rendered sym¬ 
metrical by starting with either leaf in the pair as first or second. Thus a single tree 
led to 650 entries in the correlation table, or with 100 trees there were 65,000 entries. 
This large number must not lead the reader to overweight the importance of the 
constants calculated upon it. There were only 100 trees leading to 32,500 pairs of 
leaves, each pair coming from the same tree. I should have much preferred a 
thousand trees, but the great labour of collecting, counting, and calculating precluded 
any such number. In many cases also it would have been practically impossible to 
have obtained I 00 individuals growing under fairly like environment. For example, 
I only succeeded in approximating to 100 Spanish chestnuts from one district and 
near one age. When I took a second hundred, gathered from several districts, partly 
old trees and partly others of a pollard growth of fifteen to twenty years, I found the 
heterogeneity at once increased the correlation (see below, p. 301). In many cases, of 
course, it was impossible to obtain twenty-six undifferentiated like organs from the 
same individual. In these cases the pairs were formed in the same manner, but in 
some series the total number due to each individual varied very considerably, and 
accordingly the work of verifying the tables was much increased. When but few 
pairs could be obtained from each individual, we have sometimes Increased the 
number of individuals dealt with up to a couple of hundred. But the labour 
of dealing even with a hundred individuals is often—for exam 2 )le, in the case of 
mushrooms and onions—very serious. Had it not been for the generous help of a 
number not only of willing but of competent collectors and calculators, the material 
here dealt with would have taken not eighteen months but years of my own 
unaided efforts. 
The calculation of the means, standard deviations, probable errors, and correlations 
of each table was carried out in the manner sufficiently discussed in earlier papers of 
this series."^ The symmetry of the tables leads to slight simplifications in calculating 
the product moment which will readily suggest themselves to the reader, and of 
course only a single mean and standard deviation is required for each table. Tests 
for the accuracy of the last two constants are at once provided in the case in whlcli 
the same number of organs are taken from each individual, for their values must be 
identical with those obtained for the whole series of organs entered only once and not 
for each possible pair. 
(5.) The quantitative measurement of the degree of resemblance between undif- 
ferentiated-like organs being, so far as I am aware, a quite novel branch of investiga¬ 
tion, I venture, with some hesitation, to introduce certain terms to describe oft- 
* See especially Memoir III., ‘ Phil. Trans.,’ A, vol. 187, pp. 2.53-.318, and Memoir IV., ‘ Phil. Trans.,’ 
A, vol. 191, pp. 229-311. 
