A 14-m core of lake sediments from St. Paul Island yields a long environmental history of the south coast of the Bering land bridge. Tritium assay demonstrates that sands in the bottom 8 m of deposit are injected with modern water, suggesting that a radiocarbon anomaly is the result of modern carbon introduced in groundwater. The remaining radiocarbon chronology, coupled with a time-stratigraphic pollen-zone boundary suggests that the record penetrates to the mid-Wisconsin interstadial. Pollen percentage data, Picea pollen influx, and pollen species lists allow reconstruction of the land bridge vegetation, which was tundra, without shrubs or trees, with bare ground, and comparable to Bering land bridge tundras found further north. There was no coastal mild or wet strip. Plant associations comparable to those of the modern Aleutians or Pribilofs probably did not exist along the land bridge coast and the region was probably not suitable for breeding by fur seals and other marine mammals. A cold, dry, continental air-mass system reached to the coast itself. The south land bridge coast did not offer an environment to aboriginal human populations that was significantly milder than that of the land bridge plains to the north. At about 11,000 yr B.P. the Wisconsin dry climate was replaced by a regimen comparable to that at the modern tree line of the interior, and this climate in turn was replaced with the modern system at about 9500 yr B.P. Climatic change was independent of fluctuating sea level.

A long pollen record is derived from sediments of a lake dammed behind a low moraine of the last glaciation at 3°S latitude in the Ecuadorian Andes and is compared with a glacial age pollen record from the Amazon rainforest immediately below. Lake Surucucho (Llaviucu) lies at 3180 m on the Amazonian flank of the Andes and above the glacial age pollen record from San Juan Bosco at 970 m. The Surucucho pollen record is interpreted as showing treeless vegetation in glacial times, advance of treeline in late-glacial time, and Holocene development of modern Andean forests. Combining the Surucucho and San Juan Bosco records shows that Andean vegetation was affected by glacial cooling at all elevations. Vegetation did not move up and down slope as belts. Rather, plant associations were reformed as temperature-sensitive species found different centers of distribution with changing temperature.

Paleoecological (pollen, phytolith, and wood) analyses of sediments, radiocarbon dated 33,000 to 26,000 yr B.P., from two sites in Ecuadorian Amazonia provide data that suggest a cooling of ca. 7.5°C below present in equatorial lowlands from 33,000 to 30,000 yr B.P. A period of warning followed in which novel species assemblages, a blend of montane and lowland floral components, persisted for at least 4000 years. These data of forest community change, from sites lying within the postulated glacial rain forest Napo refugium, provide the strongest paleoecological refutation of the refugial hypothesis yet obtained. The large temperature depression at ca. 30,000 yr B.P. allows the possibility that if maximum cooling at the equator was synchronous with the last glacial maximum (LGM) of the northern hemisphere, freezing temperatures would have been experienced in parts of lowland Amazonia between 25,000 and 18,000 B.P.

The first pollen and phytolith data covering the entire Pleistocene/Holocene transition from the lowlands of the Central American isthmus indicate that the forests of late-glacial Panama at an altitude of 650 m resembled those currently found at ca. 1500–1800 m. A temperature depression of ca. 5°C and reduced precipitation/evaporation ratios in the late-glacial period are suggested. Forest composition from ca. 14,000 to 10,500 yr B.P., although primarily montane in character, contained a low biomass of species today segregated in lowland forests and, hence, a floral assemblage with no modern analog. The sudden appearance of carbon and burnt, weedy plant material at ca. 11,000 yr B.P. is attributed to the earliest human impact yet recorded from tropical America and may perhaps have been associated with the first human occupation of the region.

Sediment cores from three lakes in the Upper Kolyma region, northeast Russia, provide the first well-dated continuous record of late Quaternary vegetation change from far southwestern Beringia. The oldest pollen zone, tentatively assigned to the Karginsk (mid-Wisconsinan) Interstade, indicates an Artemisia shrub tundra with Pinus pumila, Betula, and Alnus at mid- to low elevations. With the onset of the Sartan (late Wisconsinan) Stade, Pinus disappeared, probably indicating severely cold, dry winters and cool summers. As conditions deteriorated further, an Artemisia -Gramineae tundra developed. Selaginella rupestris and minor herb taxa indicate the presence of poor soils and disturbed ground. This herb tundra was replaced by a short-lived (< 1000 yr) Betula-Alnus shrub tundra followed by the rapid establishment of a Larix dahurica forest with a Betula exilis-ericales-lichen understory. Populus suaveolens and Chosenia may have formed limited hardwood gallery forests at this time. Modern vegetation associations probably developed during the early Holocene with the arrival of Pinus pumila ca. 9000 yr B.P. This shrub became important in the forest understory and, with B. exilis, formed a belt of shrub tundra beyond altitudinal treeline. Comparison of the Upper Kolyma and Alaskan pollen records indicates that important differences in vegetation types and timing of vegetation change occurred across Beringia during the late Quaternary.

The influence of gamma- (γ) and beta- (β) diversity on modern pollen rain is assessed using data from three Amazonian forests. Pollen rain of 79 forest locations was collected in modified Oldfield pollen traps between 1991 and 1993. Pollen diversity in the traps was high with > 280 palynomorph types recognized. Gamma diversity was assessed by comparing lowland terra firme forests in Cuyabeno, Ecuador, with two terra firme forests near Manaus, Brazil. The influence of β-diversity on local pollen rain was investigated using samples collected from neighbouring terra firme forests, seasonally flooded forests, and Mauritia-rich forests at Cuyabeno, Ecuador. Multivariate analyses revealed that γ-diversity produces a stronger signal in the pollen rain than β-diversity. However, β-diversity is accurately reflected in the pollen rain when the diversity is an expression of strong environmental gradients.