The existing warm (Larrea) deserts of the Southwest are Holocene expansions replacing late—Pleistocene, evergreen woodlands of low—statured junipers, pinyon pines, and live oaks; these woodlands have been isolated by complementary contraction to the slopes of higher mountains that rise like islands from the modern desert sea. Because pinyon—juniper woodland is now so widespread on the similar fault—block mountains of the Great Basin, even as far north as southern Idaho, it would seem reasonable to suppose that the modern "cold" (Artemisia, Atriplex) deserts were similarly wooded during the last glacial. However, conclusive Neotoma macrofossil evidence (45 14C—dated assemblages are reported here) documents major latitudinal displacement of vegetation that precludes pinyon—juniper woodland in the northern and central Great Basin at that time. On the other hand, the entire Mohave Desert sector (south of °37°N) served as an extensive Pleistocene refugium for pinyon—juniper woodland, as documented by an additional 48 dated Neotoma deposits. During the Wisconsinan glacial in the southeastern corner of Oregon, a 42°27'N, there was a subarctic landscape of hyperboreal, prostrate shrublet—junipers (Juniperus horizontalis and J. communis) and widespread patterned ground, even at the near—basal elevation of 1460 m. The pleniglacial vegetation of the central Great Basin at 39°N in eastern Nevada and western Utah, was dominated by a regional subalpine forest of bristlecone pine (Pinus longaeva), associated with minor but consistent boreal juniper (J. communis) down to 1660 m, close to the base level imposed by pluvial Lake Bonneville. Spruce has not been recorded below 1900 m during the last glacial. At a lower range of elevation (1350—1525 m), available south of the southeastern rim of the Bonneville basin at 37°30'N. Pinus longaeva was replaced by limber pine (P. flexilis), Douglas—fir (Pseudotsuga), and montane red cedar (J. scopulorum); existing woodland juniper (J. osteosperma) was lacking, but the subalpine J. communis was present at this local base level. Theory of island biogeography, as applied to ecological islands atop the high mountains of the Great Basin, is reexamined in the light of the drastic vegetational displacements documented in the detailed Quaternary macrofossil record. Species/area plots of montane—subalpine conifers presently distributed on 54 Great Basin mountaintops show an overall insular pattern that is especially well developed on the subset of 38 ecological islands east of 116°W; the slope of z = 0.26 is close to the theoretical value for islands in equilibrium. All 11 taxa of montane—subalpine conifers that penetrate the Great Basin deeply have their main distributions in the Rocky Mountains; only three wide—ranging species occur also in the Sierra Nevada. A long sundering trough in the western Great Basin parallels and isolates the Cascade—Sierran uplift with low—elevation barriers that impede migration, but in the eastern Great Basin there are high connecting divides to the western Rockies, especially via an axial route southeast of the Bonneville basin. There is an east—west pattern of declining species richness of11 montane conifers in the Great Basin that correlates with distance from the rocky Mountain pool of 12 coniferous species. The pleniglacial subalpine forests in the lowlands of the central Great Basin had only one to three species of conifers (e.g., Pinus longaeva, Picea engelmannii, Juniperus communis). During the great late—glacial/Holocene (12 000—8000 yr BP) warming of climate, these shifted upward in elevation and were augmented in the east (but not in the west) by as many as five additional species of montane conifers. Macrofossil evidence indicted that the later Holocene arrivals dispersed across barriers of woodland and desert that by then isolated the shrunken montane islands. Moderately long—range transport of seeds by birds is deduced as follows: a northward latitudinal shift of 500—640 km during the Holocene is documented for several species of relatively thermophilous conifers, including the heavy—seeded, late—maturing pinyon pine. A 640—km migration in 8000 yr (80 m/yr) is indicated for pinyon, but the most generous estimate of its dispersal rate via the wind/gravity mode is a plodding 0.4 m/yr (3.2 km/8000 yr), orders of magnitude too slow. Seed dispersal by Clark's Nutcrackers and Pinyon Jays, however, is both the prevalent mode and amply swift enough to fit the known migrational history. Hence, the islands—in—equilibrium pattern indicated by the typically insular slope (z = 0.26) for montane conifers on the eastern set of mountaintops in the Great Basin is a reflection of Holocene immigration via sweepstakes dispersal being offset by extinction on the smaller islands. Both extinction and immigration of conifers are documented in the late—glacial/early—Holocene Neotoma record from the small Confusion Range in the east—central Great Basin of western Utah.