512 
Journal of Agricultural Research 
Vol. XXXI, No. 6 
The range of the fluctuations during the first half of the tests was 
between 25 per cent above and 15 per cent below the general average. 
During the second half there was a falling off in the numbers, with a 
fluctuation between 8 per cent above and 30 per cent below. In 
only two cases was the fluctuation from one determination to the 
next more than 25 per cent, even though one or more days had passed. 
The fourth graph in Figure 2, showing the sample deviation on soil- 
extract agar, was made by taking the plate counts of the odd-num¬ 
bered squares on each day as 100, and computing the percentage 
deviation of the counts of the even squares. Comparison of this 
graph with the preceding one shows that the variation between odd 
Fig. 2—Graphic representation of plus and minus percentage fluctuations from general average of plate 
counts, counted on soil-extract agar 
and even samples is greater than the fluctuations in the counts from 
one day to the next. But the averages of the odd and even squares 
for the entire series are practically identical, 27.9 and 28.2, respect¬ 
ively, showing that the experimental plot as a whole was very uniform 
and that the irregularities in the counts were most probably due 
to variation in the individual samples or unavoidable analytical errors. 
The uniformity of the soil throughout the experimental plot can be 
further demonstrated by combining and averaging the counts from 
every second square across the plot, irrespective of the date of sam¬ 
pling (Table VII). 
