J. A. Harris 
TABLE XVII. 
471 
Parental 
Ascendant-Descendant 
Avuncular 
Correlation 
Correlation 
Correlation 
Weight of 
Offspring 
Seed 
Total Weight 
Weighted 
Total Weight 
Weighted 
Total Weight 
Actual 
of Weighted 
Frequency of 
of Weighted 
Frequency of 
of Weighted 
Frequency 
Parental 
Ascendant 
Ascendant 
Avuncular 
Avuncular 

Seeds 
Seeds 
Seeds 
Seeds 
Seeds 
o 
3 
30 
127 
1213 
124 
1183 
4 
7 
67 
263 
2514 
256 
2447 
41 
370 
1473 
13393 
1432 
13023 
6 
111 
1012 
3901 
35562 
3790 
34550 
7 
270 
2437 
9672 
88297 
9402 
85860 
8 
426 
3892 
15421 
142308 
14995 
138416 
9 
416 
3841 
15190 
139924 
14774 
136083 
10 
251 
2366 
9333 
88150 
9082 
85784 
11 
98 
939 
3642 
34725 
3544 
33786 
12 
30 
293 
1098 
10830 
1068 
10537 
13 
12 
134 
511 
5757 
499 
5623 
H 
2 
22 
82 
891 
80 
869 
15 
1 
13 
43 
502 
42 
489 
Totals 
1668 
15416 
60756 
564066 
59088 
548650 
III. Recapitulation. 
The term intra-class is suggested for the correlation describing the degree of 
resemblance of the individuals of the series of sub-groups or sub-classes into which 
a general population may be divided for some logical {e.g., biological, sociological) 
reason. The term inter-class is applied to those correlations which describe the 
degree of interdependence of variables belonging to the logically related sub-classes 
of different populations. Such coefficients have a very wide but hitherto little 
realized range of usefulness. 
Generally, their determination requires the comparison of each individual with 
the array of individuals constituting its own or its associated class. Heretofore 
the use of such coefficients has been limited, indeed almost precluded, by the 
excessive labour of forming the weighted tables. Methods for the rapid formation 
of the tables by the aid of mechanical calculators have been described, and 
formulae are now given by which such correlations can be determined without 
the formation of tables at all. 
The labour of calculation may thus be reduced to a small fraction of that 
formerly required. The formulae present numerous opportunities for checking 
the arithmetic, but of course certain errors may be more easily detected when 
tables are used. Furthermore the tables, full or condensed, should be formed in 
60—2 
