The present profile of Rocky River and the singular manner in which
the lower course of the river is doubled back on the upper course are
believed to represent changes wrought by glaciation. Before the advent
of the glacier, Rocky River probably flowed southward through the
"Neversink-Danbury Valley," to be described later, and joined the
Still at Danbury, as shown in fig. 4. The profile of the stream at
this stage in its history is shown in fig. 3, B.
At Sherman a low col separates Rocky River basin from that of the
small northward flowing stream which enters the Housatonic about a
mile below Gaylordsville. Streams by headward erosion at both ends of
the belt of limestone and schist on which they are situated have
reduced this divide to an almost imperceptible swell. The rock
outcrops in the channel show that the glacier did not produce any
change in the divide by damming, though it may have lowered it by
scouring. Assume that at one time a divide also existed on the eastern
fork of Rocky River, for example near Jerusalem. According to this
hypothesis there was, north of this latter divide, a short northward
flowing branch of the Housatonic located on a belt of weak rock,
similar to the small stream which now flows northward from Sherman, and
very like any of the half-dozen parallel streams in the rock mass
south and southwest of Danbury, all of which are subsequent streams
flowing along the strike. While these stream valleys were growing, the
southern ends of the same weak belts of rock were held by
southward-flowing streams which united in the broad limestone area now
occupied by the city of Danbury.
[Illustration: ~Fig. 4.~ Preglacial course of Rocky-Still River.
Dotted lines show present courses of the two rivers.]
The southward-flowing streams whose heads were, respectively, above
Sherman and near Jerusalem joined at the southern end of the long
ridge which includes Towner Hill and Green Mountain. Thence the stream
flowed southward along the valley now occupied by Wood Creek and
reached Still River by way of the valley which extends southward from
Neversink Pond (fig. 4).
The preglacial course of Rocky River, as above outlined, is subject to
possible modification in one minor feature, namely, the point where
the east and west forks joined. The junction may have been where
Neversink Pond is now situated, or three miles farther south than the
indicated junction near the mouth of Wood Creek. A low ridge of till
is the only barrier that at present prevents the western branch from
flowing into the head of Barses Pond and thence into Neversink Pond
(fig. 1).
As thus reconstructed the greater part of Rocky River formerly
belonged to the Still-Umpog system and formed a normal tributary in
that distant period when the Still joined the Saugatuck on its way to
the Sound (fig. 9). However, the normal condition was not lasting, for
the reversal of Still River, as later described, brought about a
complex arrangement of barbed streams (fig. 4) which remained until
modified by glacial action.
In a large stream system which has been reversed, considerable
evidence may be gathered from the angle at which tributary streams
enter. As the original direction of Rocky River in its last 2-1/2
miles is unchanged, normal tributaries should be expected; whereas
between Jerusalem and the head of the stream entering Neversink Pond
from the south, in accordance with the hypothesis that this portion of
the stream was reversed, tributaries pointing upstream might be
expected. Such little gullies as join Rocky River near its mouth are
normal in direction; between Jerusalem and the mouth of Wood Creek, a
distance of 4-1/2 miles, there are no distinct tributaries. South of
the mouth of Wood Creek are four tributaries: (1) the brook which
enters the valley from the west about one mile south of Neversink
Pond, (2) Balls Brook, which empties into Neversink Pond, and (3) two
streams on the east side--Mountain Brook and one other unnamed (fig.
1). All these, except Mountain Brook, are normal to the reconstructed
drainage. The evidence of the tributaries, though not decisive, is
thus favorable to the hypothesis of reversal.