over low and high ground alike, smoothing over irregularities of surface, and often more than 1000 feet in thickness.  It has no stratification, but tends to cleave vertically, and is traversed in every direction by sudden crevices, almost glacier-like, narrow, with vertical walls of great depth, and infinite ramification.  Smooth as the loess basin looks in a bird’s-eye view, it is thus one of the most impracticable countries conceivable for military movements, and secures extraordinary value to fortresses in well-chosen sites, such as that of Tung-kwan mentioned in Note 2 to chap. xli.

Agriculture may be said in N. China to be confined to the alluvial plains and the loess; as in S. China to the alluvial plains and the terraced hill-sides.  The loess has some peculiar quality which renders its productive power self-renewing without manure (unless it be in the form of a surface coat of fresh loess), and unfailing in returns if there be sufficient rain.  This singular formation is supposed by Baron Richthofen, who has studied it more extensively than any one, to be no subaqueous deposit, but to be the accumulated residue of countless generations of herbaceous plants combined with a large amount of material spread over the face of the ground by the winds and surface waters.

[I do not agree with the theory of Baron von Richthofen, of the almost exclusive Eolian formation of loess; water has something to do with it as well as wind, and I think it is more exact to say that loess in China is due to a double action, Neptunian as well as Eolian.  The climate was different in former ages from what it is now, and rain was plentiful and to its great quantity was due the fertility of this yellow soil. (Cf. A. de Lapparent, Lecons de Geographie Physique, 2’e ed. 1898, p. 566.)—­H.C.]

Though we do not expect to find Polo taking note of geological features, we are surprised to find no mention of a characteristic of Shan-si and the adjoining districts, which is due to the loess; viz. the practice of forming cave dwellings in it; these in fact form the habitations of a majority of the people in the loess country.  Polo has noticed a similar usage in Badakhshan (I. p. 161), and it will be curious if a better acquaintance with that region should disclose a surface formation analogous to the loess. (Richthofen’s Letters, VII. 13 et passim.)

NOTE 2.—­Taianfu is, as Magaillans pointed out, T’AI-YUAN FU, the capital of the Province of Shan-si, and Shan-si is the “Kingdom.”  The city was, however, the capital of the great T’ang Dynasty for a time in the 8th century, and is probably the Tajah or Taiyunah of old Arab writers.  Mr. Williamson speaks of it as a very pleasant city at the north end of a most fertile and beautiful plain, between two noble ranges of mountains.  It was a residence, he says, also of the Ming princes, and is laid out in Peking fashion, even to mimicking the