Barriers to Dispersion of River Fishes [21 
of the lower Roanoke. In this case it is likely that we have 
to consider the results of local erosion. Probably the divide has 
been so shifted that some small stream with its fishes has been 
cut off from the Holston and transferred to the Roanoke. 
The passage of species from stream to stream along the 
Atlantic slope deserves a moment’s notice. It is under present 
conditions impossible for any mountain or upland fish, as the 
trout or the miller’s thumb,* to cross from the Potomac River 
to the James, or from the Neuse to the Santee, by descending 
to the lower courses of the rivers, and thence passing along 
either through the swamps or by way of the sea. The lower 
courses of these streams, warm and muddy, are uninhabitable 
by such fishes. Such transfers are, however, possible farther 
north. From the rivers of Canada and from many rivers. of 
New England the trout does descend to the sea and into the 
sea, and farther north the whitefish does this also. Thus these 
fishes readily pass from one river basin to another. As this is 
the case now everywhere in the North, it may have been the 
case farther south in the time of the glacial cold. We may, I 
think, imagine a condition of things in which the snow-fields 
of the Alleghany chain might have played some part in aiding 
the diffusion of cold-loving fishes. A permanent snow-field on 
the Blue Ridge in western North Carolina might render almost 
any stream in the Carolinas suitable for trout, from its source 
to its mouth. An increased volume of colder water might carry 
the trout of the head streams of the Catawba and the Savannah 
as far down as the sea. We can even imagine that the trout 
reached these streams in the first place through such agencies, 
though of this there is no positive evidence. For the presence 
of trout in the upper Chattahoochee we must account in some 
other way. 
It is noteworthy that the upland fishes are nearly the same 
in all these streams until we reach the southern limit of possible 
glacial influence. South of western North Carolina the faunz 
of the different river basins appear to be more distinct from 
one another. Certain ripple-loving types are represented by 
closely related but unquestionably different species in each 
* Cottus ictalops Rafinesque. 
