6 Carl C. Engberg 



vs and yu's is as probable as the previous one. Thus we can, by 

 very slight manipulations of our data, obtain a new set of con- 

 stants which differ considerably from the old set, but which must 

 give rise to a curve equally as good as the first one. To carry 

 out the computations to six or more places, when an exceedingly 

 slight accidental change in the data will make a proportionately 

 large change in the constants, is a pronounced case of "saving at 

 the spigot and spending at the bung." 



In fig. I are drawn the curves obtained, using six and three 

 decimal figures respectively. A comparison of the two curves 

 with the frequency polygon shows the one to be as good a fit as 

 the other. 



Ill 



The distribution of .dorsal teeth on the rostrum of pij specimens 

 of Palaemonctes Varians from Saltram Park, Plymouth. 

 Professor Karl Pearson, The Mathematical Theory of 

 Evolution. Phil. Trans., A., vol. i86, pp. 403-4. 



The centroid vertical lies .314 of a tooth beyond 4, i. e., at 4.314 

 teeth. The moments about the vertical through 4 ai'e : 



»'i=-3i37. i'2=.8426, 7/3=.4973, V4=i.9705. 



These give to the moments about the centroid vertical the 

 following values : 



/i.2 = .9i09> /^3= — .234, )u,i = 2.6259; 

 whence 



/3i = .o724, /52 = 3-i647, 



92 



