36 BULLETIN 1311, U. S. DEPARTMENT OF AGRICULTURE 
errors. The fact that calculations show excesses of silica, alumina, 
and iron oxide does not necessarily disprove the existence of the com- 
pounds assumed in the calculations. The colloidal matter may be a 
mixture of the compounds assumed above and free silica, alumina, 
and iron oxide. If the calculated excesses of silica, alumina, and iron 
oxide could be demonstrated, the existence of such a mixture would 
seem probable. 
Direct proof of the presence of free silica and alumina is want- 
ing, but the oxalic acid and hydrohloric acid treatments, together 
with the color of the colloids, have indicated the presence of free 
iron oxide in certain colloids. The colors of the colloids are in gen- 
eral agreement with the stoichiometrical calculations. Most of the 
soil colloids showing the larger excesses of ferric oxide in the calcula- 
tions are distinctly red or yellow in color and most of the colloids 
showing a small or no excess of ferric oxide are not distinctly red or 
yellow. The stoichiometrical calculations on 15 colloids varying 
from light gray to black in color show no excesses of Fe 2 3 , and, 
with the exception of No. 29, these 15 colloids show an excess of 
silica varying from 7.6 to 23.6 per cent. Of 27 colloids which are 
brown to yellow or red in color, 23 show from 2.7 to 14.2 per cent 
of excess Fe 2 3 . There are seven colloids which are yellow to brown 
in color which show no excess of ferric oxide; in these, however, 
the excess of silica is generally small, though it reaches 19.1 per cent 
in the Yolo clay colloid. Only a general relation between the color 
and excess of ferric oxide is to be expected on account of the varia- 
tion in color of pure hydrated and anhydrous ferric oxides and the 
presence of organic matter in the soil colloids. 
There is in general a fair agreement between the combined water 
(that driven off above 110° C.) and the water of constitution in the 
calculated quantities of kaolinite and nontronite. This general 
agreement might be interpreted as a substantiation of the correctness 
of the assumptions made in the stoichiometrical calculations, since 
the method for determining combined water is largely empirical. In 
the calculations, however, no allowance is made for water combined 
with the excesses of Si0 2 , A10 3 , and Fe 2 3 . Even without such 
allowance, therfc are 14 colloids in which the calculated water ex- 
ceeds the combined water by more than 2 per cent, which is probably 
more than the error of the " combined " water determination in most 
cases. In these colloids, Nbs. 10, 12, 15, 17, 18, 20, 25, 26, 29, 30, 36, 
40, 43, and 44, it is certain that there can not be the quantity of the 
hydrated silicates assumed in the calculations. In the extreme case, 
that of the Ontario surface colloid, there can not be more than one- 
third the quantity of kaolinite and nontronite calculated. Of course, 
it is not certain from the combined water determination whether all 
or any of the so-called " combined " water is definitely water of con- 
stitution. All of the combined water may possibly be present in the 
same condition as in artificial inorganic gels. 
The question naturally arises whether there are other compounds 
of silica, alumina, and water which would better fit the conditions. 
The assumptions of other common hydrated silicates in stoichiomet- 
rical calculations lead to no more probable conclusions than the as- 
sumption of kaolinite has done. For instance, halloysite (A1 2 3 . 
2Si0 2 .2H 2 0+aq.) contains the same ratio of silica to alumina as 
