the valley. West of Mill Plain station the valley is
encumbered with ridges of stratified drift, interspersed with heavy
accumulations of till. Near Andrew Pond the true width of the
valley--one-eighth mile--is shown by rock outcrops on both the north
and south slopes. The valley at this point gives no indication of
narrowing toward the headwaters; in fact, it becomes broader toward
the west.

Between Andrew Pond and Haines' Pond is the divide which separates the
waters of the Still system from those of the Croton. It consists of a
jumbled mass of morainal hills, seemingly of boulder clay, that rise
from 50 to 60 feet above the level of the ponds. The divide is thus
merely a local obstruction in what was formerly a through drainage
channel.

THE STILL-CROTON VALLEY

It is evident that before the advent of the glacier a stream must have
flowed through the Still-Croton valley past the present divide in
order to have excavated the rock valley there found. The Housatonic
could not have flowed west through this valley if it was as narrow and
shallow as is indicated by known rock outcrops; the river could have
flowed through it only in a deep narrow gorge which was later buried
under drift, but the evidence at hand does not support this view.

[Illustration: ~State Geol. Nat. Hist. Survey Bull. 30. Plate III.~
Limestone Plain southwest of Danbury, in which are situated the
Danbury Fair Grounds and Lake Kanosha.]

It is most probable that this valley was made by the preglacial
Croton River. This explanation demands no change in the direction of
Still and Croton Rivers but calls for a divide at some point east of
the present one. From a divide between the Fair Grounds and Danbury, a
small stream may be supposed to have flowed toward the east, joining
the larger northern branch of the Still at a point near the middle of
the city of Danbury. The stream flowing westward from this divide
formed the headwaters of one branch of the Croton system.

The presence of till in a ravine can be used as a criterion for
locating the site of a former divide, for where till is present in
the bed of a stream the channel is of preglacial date. Where the river
crosses a divide it should be cutting through rock, though till may be
present on the valley slopes. Judged by this test, the old divide was
situated either just east of the Fair Grounds plain or at the east end
of the ravine described in the preceding topic. Of these tw