ber the ground in marked contrast to the smaller fragments of the eastern 
margin of the typical sinkhole limits. It is possible that some of the cir- 
cular depressions may have been exposed by the roof of a cave falling, 
but if such was the case there is no evidence of it left. The roof must 
have been composed of limestone as their rims are several feet below the 
geologic horizon of the Lost River chert. If limestone fragments of the 
roof were ever present they have disappeared; and the more probable 
theory seems to be that the depression is the result of erosive forces act- 
ing equally upon all the sides. The rocks exposed in place where the 
sinkholes are common in Lawrence, Orange, Washington and Harrison 
counties are always members of the upper or middle portion of the Mitch- 
ell limestone, and the angular chert masses and fragments scattered 
over the surface and mixed with the red residual clay come from the same 
strata or from the Lost River chert stratum. 
The sinkhole area, as a rule, has no surface creeks and branches, and 
such as reach its limits from without soon find an opening and disappear 
wholly or in part, except Blue River and Buck Creek. Occasionally the 
creek or branch is replaced by a dry-bed channel. The dry-beds only 
come into use after heavy rains or when the subterranean passages are 
burdened beyond their capacity. Lost River through a part of its surface 
course is a typical dry bed. When it reaches the eastern edge of the sink- 
hole region it finds a number of underground channels that take in all 
the water of the perennia] stream east of the Orleans and Paoli road. 
If the first openings are overtaxed, the overplus of water passes through 
a dry-bed channel farther west into other sinks, but after an excessive 
rainfall all the sinks fail, and water runs on the surface through the 
whole extent of the dry-bed system and again becomes a part of the per- 
ennial stream a short distance below the Orangeville “rise.” Indian Creek 
for a part of the year runs underground, but, unlike Lost River, the 
greater part of its water passes over a surface channel and a dry bed is 
only exposed during the summer months. It sinks two miles southwest 
of Corydon and “rises” again five below on an air line, and twice that 
distance following the meanderings of the creek bed. There is ample 
evidence that Lost River, like Indian Creek, at some period in the past 
was wholly, or for the greater part of the year, a surface stream over its 
dry-bed channel. 
Contrary to what might be expected, the subterranean channels do not 
greatly increase in capacity as they unite and pass under the Kaskaskia 
