202 EXPLORATIONS IN TURKESTAN. 
The number of terraces ranges from none to nine, but neither extreme repre- 
sents the true state of affairs. V.'here terraces are absent it is either because the 
slope of the streams is so gentle that there is no erosion or because the slope is so 
steep and the country- rock so resistant that the streams lia\-e as yet been able to cut 
only narrow gorges. Where the number is over five the material is usualh- uncon- 
solidated gravel, and some of the teiTaces are usually .small and seem to be mere 
stages of larger ones. Ordinarily there are from three to fi\-e terraces. The num- 
ber of valleys for which the writer has a record is fortj'-three, and the number of 
terraces in these vallej's is shown in Table IV. The number of valleys with only 
one or two terraces was really larger than appears from the table; for after the 
widespread distribution of the terraces had been noticed, valleys where only one or 
two occurred were not recorded. 
Table IV. — Terraces. 
Valleys with — 
One terrace 3 
Two terraces 3 
Three terraces 12 
Four terraces 8 
Five terraces 9 
Valleys with — 
Six terraces 4 
Seven terraces 3 
Eight terraces o 
Nine terraces i 
The terraces are sometimes cut in gravel (fig. 140), and sometimes in rock 
(fig. 141), but in the latter case there is always a cover of gravel lying over the rock. 
It may happen in a single valley that the upstream portions of the terraces are 
almost wholly cut in rock, while the downstream portions are entirely in gravel, 
as, for instance, along the Kuzzil Su, at the southeast end of Issik Kid. In vallej-s 
such as those of the Ispairan Su, flowing to Marghilan, and of the Ak Bura, flowing 
to Osh, it often happens that the terraces are cut for a certain distance in soft strata, 
or in gravel that fills a basin where soft strata have been excavated, although farther 
down the stream flows through a narrow canyon in hard strata, without a trace of 
terracing; l)Ut when the hard strata end and soft ones begin once more, the ter- 
races are resumed as though the)' had ne\er been interrupted. 
Another and perhaps the most characteristic feature of the terraces is the per- 
sistence with which the different members of a series preserv^e the same relati\-e 
height and width. In terraces due merely to the swinging of the stream from side 
to side as it cuts steadily downward, one terrace is here or there cut off either at the 
upper or lower end by another terrace of later date, and a pattern of cusj^s and bays 
is thus fonned along the valley side. In such cases a single terrace can only be 
traced a short distance, and the number of terraces is continually changing. In the 
mountain valleys of Central Turkestan, on the other hand, although it sometimes 
liapjiens that one terrace truncates another and thus fonns a cusp, this is far from 
being a prevalent condition ; each le\-el is, as a rule, distinct and does not interfere 
with its neighbors. Several terraces often nin for many miles side by side without 
interfering with one another, each one preser\-ing an almost uniform width with 
remarkable persistency. As a rule, too, the uppermost terrace possesses not only 
the greatest width, but the greatest height. Such a regular diminution in size can 
mean only that the cause of the terracing was of steadih- decreasing efficiency. 
