1918] Waynick: A Statistical Study of Nitrification in Soil 259 



In the incubated samples, the probable error of the difference be- 

 tween the samples is 0.50 ± .17 milligrams, for the samples to which 

 ammonium sulfate was added 1.90 ± .31 milligrams, and for dried blood 

 4.00 ± .97 milligrams. It is worthy of note that the probable error of 

 the difference's between eighty-one samples, to which ammonium sulfate 

 was added, and the ten samples given above is no less than to l.v 

 The coefficient of variability of the ten samples is greatly increased. 



being 48.4 ± 3.9 per cent. On the other hand, it happens that the 

 incubated blanks vary even less than the eighty-one samples, the co- 

 efficients of variability being 17.9 ± 2.7 per cent and 31.2 ±1.7 per 

 cent respectively. It is seen from these results that very wide varia- 

 tions may be found when only ten samples are used in making a 

 composite sample and that that number of determinations is by no 

 means enough to be an accurate measure of the actual nitrate nitrogen 

 content when the variations are of the magnitudes of those noted above. 

 Turning for a moment to the results given in table 10. we find that 



t For table of odds to aid in estimating the significance of differences between 

 two results, see Batchelor and Reed, Jour. Ayr. Res., vol. 12, pp. 265-266, 1918. 



