Evolutionary biologists have long researched the dynamics of hybrid zones to determine how reproductive barriers between species originate and become established. It is unknown, however, to what extent hybrid zones persist through time. Do hybrid zones eventually disappear with perfection of isolating mechanisms or by fusion of the taxa, for example, or do they continue to exist, perhaps stretching and shrinking with environmental fluctuation (Wallace, 1889; Dobzhansky, 1951; Grant, 1963; Mayr, 1963; Lewis, 1966; Remington, 1968; Harrison, 1983)? Clearly, historical information is necessary for an adequate understanding of the evolution of isolating mechanisms, but such information has only been available on a scale of decades. Here, fossil evidence is used to propose that two freshwater bivalve species that now hybridize readily did not do so about 200,000 years ago. Isolating barriers are hypothesized to have been perfected in sympatric populations in the past, but then to have been lost when these adapted populations were eliminated by advancing glaciers. Reproductive isolation between related species could therefore have been a transient phenomenon. North America contains one of the richest freshwater clam faunas (family Unionidae) in the world (Burch, 1975; Davis and Fuller, 1981; Kat, 1984), but fossils necessary for interpretation of the origins and species diversity of this fauna through time are scarce (Simpson, 1895; Baker, 1920; Henderson, 1935; LaRoque, 1966; Kat, 1983a). Unionacean fossils are uncommon in the Interior Basin (IB), a faunal region that essentially covers the entire Mississippi River drainage, and very rare on the Atlantic Coastal Plain (ACP). For example, of the 122 Pleistocene freshwater molluscan faunas described by LaRoque (1966) from seven IB states, only six faunas contained unionid fossils. One exception to this trend occurs in the Fish House fauna from Camden, New Jersey (Fig. 1). This ACP fauna contains bivalves morphologically resembling ten extant species, at least three of which are presently restricted to the IB (Kat, 1983a). The Fish House fauna was first described by Lea (1868) and Whitfield (1885) and appears not to have been examined since the turn of the century. Lea and Whitfield both assumed the fossils to be of Cretaceous age, but more recently the Fish House clays have been correlated with the Accomac beds on the Delmarva Peninsula and with beds near the mouth of the Rappahannock River in Virginia as deposits 187,000 ? 20,000 years old (J. P. Owens, pers. comm.; see Cronin et al., 1981; McCartan et al., 1982). No radiometric dates are available directly from the Fish House clays. The climate at that time, as inferred from pollen analyses and plant remains, was warm temperate and suggestive of interglacial conditions (Woolman, 1896; see also Cronin et al., 1981). A sewage disposal facility at the Fish House site prevents further collection from this locally exposed deposit, but a number of Lea's type specimens as well as specimens from subsequent collections are housed at the Academy of Natural Sciences of Philadelphia (ANSP). In the Fish House fauna, fossils phenotypically similar to recent IB Anodonta grandis and ACP A. cataracta (Fig. 2) occur in the same stratum (Whitfield, 1885). Collections at the ANSP consist of 22 specimens preserved as articulated valves or interior casts, and while there is no information about orientation of the specimens in the sediment (i.e., if they were found in life position), the presence of articulated specimens suggests that postmortem transport, if any, was minimal. Ofthe 22 specimens, 7 resemble A. cataracta and 15 resemble A. grandis. Shell shapes of unionids can be quantified by analyses of curvature along the shell outline. Shell outlines are digitized, and the coordinates of 300400 points stored in a computer. Circles are then fit to sets of points along the outline of the shell using the best-fit criterion of Bookstein (1978), involving a multiple linear regression of X2 + y2 as the dependent variable against x and y as independent variables (Kat, 1983a). Centroids of point clusters along the outline were determined by the points of intersection of lines, rotated through predetermined numbers of degrees from the umbo, and the shell edge. The results of such analyses indicate significant differences in curvature at several points along the shell edge when specimens of Recent Anodonta cataracta and A. grandis are compared (Fig. 3A). Previous analyses unequivocally classified all fossil specimens as either A. cataracta or A. grandis (Kat, 1983a). Use of soft-part, shell microstructural, and electrophoretic data has clarified the taxonomy of Recent ACP and IB anodontines (Kat, 1983b, 1983c, 1985), and the cataracta and grandis phenotypes cannot be confused with other extant species. Today, Anodonta cataracta and A. grandis hybridize extensively where their geographic ranges, newly expanded after the retreat of Wisconsinan