CERTAIN TAXONOMIC CHARACTERS OF GRYLLUS. 



11 



table 55, form two distinct curves, each fluctuating according to the 

 law of error about its own mean. It is not surprising to find that the 

 greatest difference is in the length of the tegmina. It is 16 times the 

 difference which might have been expected in random samples. It is 

 least in the case of the body and the posterior femora, being in them 

 barely significant. The tegminal polygons are shown in fig. 2. Both 

 series are plotted on a per centum basis. 



In comparing the variability of corresponding organs in the two 

 groups of this dimorphic population our conclusions will differ accord- 

 ing to our views concerning the relative value of the ' 'standard devia- 

 tion" and the "coefficient of variation." The coefficient of variation is 

 that per centum which the standard deviation is of the mean. Being 

 abstract numbers, all such coefficients are immediately comparable. It 

 seems to me more reasonable to use the coefficient of variation than the 

 standard deviation when the series compared have different means, 

 even if the variates are measured in the same units. The standard 

 deviation is usually, although not necessarily, larger for series having 

 large means than for those having small means. For example, the 

 standard deviation of the weight of men is about 16.5 pounds. The 

 average weight of new-born boys is 7.3 pounds. It would be absurd to 

 think of a standard deviation of 16.5 in the latter case. As a matter 

 of fact, it is about 1.1 pounds. However, considering the coefficient of 

 variation, we reach comparable results and find that infant weights are 

 nearly half again as variable as adult weights instead of being much 

 less. Turning to table 4, we see that the coefficients of variation of 

 the long-winged group are less throughout than those of the short- 

 winged one. The difference is greatest in the case of the wings. 



TABLE 5. Differences in correlation between short-winged and long-winged 



groups from Gotha, Florida. 



Table 5 gives the differences between the respective coefficients of 

 correlation in the two groups, the signs having the same significance 

 as in table 4. In only one case -the tegminal-wing correlation is the 

 difference equal to three times the difference which might have been 

 expected from random samples, and hence it alone can be considered 

 as having any significance. Even it is not very great. The table as a 

 whole shows the correlations in the two groups to be practically equal. 



