30 BULLETIN 1193, U. S. DEPARTMENT OF AGRICULTURE. 
alteration in adsorptive capacity of the colloidal material produced 
by extraction. Each of these sources of error affects the dye, water, 
and ammonia determinations oppositely or to slightly different 
extents. 
The error caused by inability to secure a sample of colloidal material 
which is exactly representative of all in the soil apparently is not 
large in the case of most soils. Data given in Table 3 indicate that 
a small sample of colloidal material is usually only 5 to 10 per cent 
more adsorptive for dye, water, or ammonia than all the colloidal 
material which is extractable b}^ our methods. However, a small 
or large sample of the extractable colloid as a rule does not appear 
to be closely representative in adsorptive capacity of the unextractable 
colloid. Data given in Table 6 indicate that the unextractable 
colloid in one soil was only one-fourth as adsorptive as the extracted 
colloid for one substance. In other soils examined, however, the 
unextracted colloid was 60 to 100 per cent as adsorptive for all three 
substances — dye, water, and ammonia — as the extracted material. 
Errors induced by the different adsorptive capacities of the extract- 
able and unextractable colloids are reduced by the fact that the 
unextractable colloid usually constitutes only about one-half of the 
total colloidal material (Fry, 13). The unextracted colloid as a 
rule is more nearly like the extracted colloid in its adsorptive capacity 
for ammonia than in its adsorptive capacity for dye or water. The 
error in sampling the colloidal material would, therefore, make 
the ratio 
adsorption per gram of soil 
adsorption per gram of colloid 
too low, but the ammonia ratio should, on the whole, be least affected 
by this error. 
The alteration which apparently takes place in the adsorptive 
capacity of the colloidal material on extraction may be the chief 
source of error in the proposed method of determining the colloidal 
content of soils, at least when the ratios are based on the quantities 
of ammonia adsorbed. Extraction seems to render the colloidal 
material as a rule slightly more adsorptive of dye, slightly less ad- 
sorptive of water and considerably less adsorptive of ammonia. 
This would tend to make the adsorption ratio low bv dye, and high 
by water and ammonia, in proportion to the degree of alteration. 
Determinations of the colloidal content of soils based on the dye 
adsorption ratio, when inaccurate, should be too low as a rule, since 
the sampling and alteration errors are additive in this direction. 
In the case of the water and ammonia adsorption ratios the sampling 
and alteration errors are in opposite directions and, therefore, tend 
to neutralize each other. The relative accuracies of the water and 
ammonia adsorption ratios would, therefore, depend on the relative 
magnitudes of the two possible errors, the ammonia ratio being less 
affected by the sampling and the water ratio being less affected by 
the alteration. 
Fortunately it was found possible to determine the total quantity 
of colloidal material in the soil by a method entirely independent of 
adsorption determinations (Fry, 13). The method seems to be 
accurate within a few per cent; hence a direct comparison can be 
made of the relative accuracies with which the dye, water, and 
