14 
BULLETIN 551, U. S. DEPARTMENT OF AGRICULTURE. 
of Rustonfme sandy loam differ markedly in iron, alumina, and sulphur. 
Still further, on comparison of soils of different types, it is possible 
to choose two or more that agree quite as closely as any two of the 
same type. 
This is brought out in Table 
Table 
Variation in the chemical composition of soils of the same texture but of differ- 
ent series. 
Type. : Si0 2 . 
Fe 2 3 . 
AI2O3. CaO. MgO. 
K2O. 
Xa 2 0. P2O5. 
SO3. 
Norfolk fine sandy loam 
Huston fine sandy loam 
Tif ton fine sandy loam 
Portsmouth fine sand y loam . 
Per ct. 
95. 54 
95.51 
94. 15 
94.85 
80. IS 
79. 35 
71.38 
70.99 
Perct. 
0.62 
.68 
.94 
.41 
3.05 
4.44 
3.63 
4.23 
Per ct. 
1.70 
1.70 
1.67 
1.37 
8.48 
8.S9 
12.29 
11.39 
Perct. 
0.06 
.12 
.05 
Trace. 
.27 
.63 
1-09 
.93 
Perct. 
Trace. 
Trace. 
Trace. 
Trace. 
0.45 
.39 
.36 
1.0s 
Per ct. 
0.C6 
.16 
.10 
.06 
1.84 
.67 
2.28 
2.71 
Perct. 
0.18 
.04 
Trace. 
.06 
.72 
.24 
1.14 
.82 
Perct. 
0.05 
.04 
.04 
.05 
.10 
.18 
.10 
.19 
Per ct. 
0.03 
.23 
.06 
.03 
• 03 
.13 
IS 
.39 
Probably no one has ever seriously contemplated classifying soils 
on the basis of their chemical composition alone, but it seems probable 
that soils that are alike in color, texture, relation of soil to subsoil, 
and formed by the same agencies, in other words, having such similar 
characteristics that they would be given the same type name by 
field observers, should have some chemical resemblance. The soils 
just compared, the analyses of which are more or. less alike, are soils 
that because of characteristics other than texture have been given 
distinct type names; but how wide the variation of a single type may 
be in chemical composition, or whether some types should be sepa- 
rated into two or more because of chemical differences, or two or more 
types amalgamated because of chemical resemblances, is a matter for 
future investigation. 
LIMIT OF ERROR IN ANALYTICAL WORK 
Earlier in the paper the authors stated what they considered the 
limit of error in the analytical work involved in the analyses here 
presented and discussed. It was stated that what was meant by 
limit of error was the allowable variation in results obtained from the 
same sample by two analysts familiar with the method used. 
Several factors may contribute to this error: Lack of uniformity 
of sample, impure reagents, contamination from glassware or other 
utensils, and, finally, the error incident to the method, which may be 
of both a personal and chemical nature. All except the last may be 
nearly eliminated by care and blank determinations. 
All analytical results are a compromise, usually arrived at by 
taking advantage of the solubility of compounds to be removed 
and the relative insolubility of the compound to be recovered and 
determined. Xo compound is absolutely insoluble, and when rela- 
