Scotia), because of the extended history of stocking Maine hatcheries with fish from Canadian streams.
During the early European colonization of North America, Atlantic salmon were found as far south as the Connecticut and Hudson rivers, but continuous attempts to reintroduce the species to the Connecticut have failed to establish self-sustaining wild populations, and the southern limit of wild Atlantic salmon is now the Sheepscot River in Maine (Figure 1). The “at least eight [Maine] rivers” listed by NMFS and FWS (DOI and DOC 2000) as containing wild salmon populations are (from west to east) Sheepscot, Ducktrap, Penobscot, Narraguagus, Pleasant, Machias, East Machias, and Dennys. The Saint John River, whose mouth is in New Brunswick, drains areas in Maine’s northeastern highlands, but its salmon are considered to be among New Brunswick’s populations. The Penobscot and its tributaries harbor the largest populations in Maine, and those from the Narraguagus, Machias, and Dennys are substantial. Those from the other watersheds are smaller (Maine 2000). Occasionally, Atlantic salmon have been seen in the Androscoggin, Kennebec, Union, and several smaller rivers, but they probably include strays or aquaculture escapees. Those additional rivers might well figure in recovery plans, but with the exception of the lower Kennebec and its tributaries, they are not thought to support wild populations.
Much of the evidence on genetic distinctiveness of Atlantic salmon populations is based on laboratory analyses of the variations in gene products (i.e., proteins) taken from samples in the field. In some cases, the genetic material itself (DNA) is analyzed for variation. In either case, the variation observed is compared among populations, and a variety of tests are performed to decide whether the populations are statistically distinct from each other and how accurately an individual salmon can be identified correctly with its source population. Although there might be some relationship between the genetic variation detected in these analyses and adaptively significant differences among the various populations, none can be inferred from these analyses. The differences could be due to random processes (sampling effect or genetic drift), and the markers themselves are thought to be adaptively neutral. In other words, most of the laboratory analyses of genetic variation discussed below cannot provide information on the degree to which different populations have adapted to different local conditions. They can provide information on the degree of isolation of populations—and isolation of populations is a prerequisite for the development of genetically based adaptive differences in