THE AMERICAN NATURALIST [Vol. XL VII 



These relationships are made clear by Diagrams 3 and 



4. The first of these shows the differences in standard 

 deviations expressed as percentages of the population 



5. D. The second shows the ratio of the differences to 

 their probable errors. 



The distribution of the differences which are at least 

 2.5 times their probable errors may be summarized : 



-Mi r. r, ... . \', 1^69 ^2L64 



All differences. 20 -.343 -10.78 



Thus of the 17 positive differences, 12 or about 71 per 

 cent, are statistically untrustworthy (i. e., < 2.52?) while 



