CORRELATION AND LOCATION OF PARTS. 



25 



From this table it is seen at once that the results with homologous 

 joints are quite different from those with non-homologous. Considering 

 only the pairs of homologous joints at the upper end of the table, there is 

 only one case out of the possible six in which the correlation is higher 

 between the non-contiguous pair of legs. In that case the difference 

 between the two coefficients is altogether insignificant in comparison with 

 its probable error (difference ==0.0021 0.0036). The "rule of neigh- 

 bourhood" thus evidently holds for the correlation between homologous 

 joints of the different legs. The result may be stated in the following 

 way: The correlation betiveen the homologous segments of two legs is 

 higher when these two legs belong to contiguous metameres than when 

 they are separated by an intervening metamere. 



There are twelve chances of comparison when the correlations of the 

 non-homologous segments are considered. A reference to the last column 

 of the table shows that in six of these cases the correlation is higher in 

 the non-contiguous pair of legs; while in five cases it is higher in the 

 contiguous pairs. In one case there is practically exact equality between 

 the coefficients, the difference being only 0.0001. The indication clearly 

 is that, so far as the non-homologous joint correlations are concerned, it is 

 about an even chance whether the excess will be in favor of the contiguous 

 or the non-contiguous pairs of legs. Before definitely accepting this con- 

 clusion, however, it will be well to determine whether the plus ( + ) differ- 

 ences are in the aggregate larger than the minus ( ) , or vice versa. 

 Further, it is clear that it will be better to compare the proportions of 

 the differences to their probable errors rather than to the absolute values. 

 Accordingly, in table 10 is given for each plus ( + ) and minus ( ) differ- 

 ence in the lower half of table 9, the value of the ratio of the difference to 

 its probable error (i. e., Diff. /P. E. 



TABLE 10. Correlation differences, to show the relative significance of the 

 plus ( + ) and minus ( ) entries in the last column of table 9. 



1 Mean. 



