J. A. Harris 
313 
It appears, therefore, that there is a very substantial interdependence between 
the length of the peduncle and the length of the fruit. It will be borne in mind, 
however, that the fruit in this series had not quite reached its maximum size, and 
that owing to difficulties inherent in the species, constants involving peduncle 
length and fruit length will always be open to some question. 
Blakeman* has shown that, if ^ = rj" — a good test for linearity of regression 
is given by using the formula 
^ ^ JN 1 1 
0-67449 ■ 2 ^ y 1 + ( 1 - r,''f - ( 1 - r-f ' 
How large the value of ^jE^ shall be to be considered significant must be 
decided by each writer for himself, but when we find ^jE^ > 2 5 we may begin 
to be suspicious of the linearity of our regressions. From our data we find 
Length of Fruit. 
2S 29 30 31 32 33 31 35 36 37 38 30 iO 41 42 4 3 4 4 4 5 4 6 4 7 48 4 9 
DiAOKAM I. Regression of fruit length on peduncle length. (Means depending on a few 
isolated observations not joined up.) 
* Blakeman, J., Iliometrika, Vol. iv. pp. 335—3-50, 1905. 
