The vegetation history of an important conservation area, the Flow Country in northern Scotland, is described using the results of pollen analysis from two closely spaced cores from the Cross Lochs, eastern Sutherland. Buried lake sediments and shallower surface peats yield a continuous record covering the Devensian Late-glacial and the Holocene. The earliest Late-glacial pollen spectra recovered suggest a transition from open habitat to more stable communities, dominated by *Empetrum heath. This Windermere Interstadial vegetation shows some variability, with a phase of increased growth of juniper and birch, followed by deteriorating climate and less stable soil conditions in the transition to the Stadial, with Huperzia selago and Salix. There are poorly defined Loch Lomond Stadial and Stadial-Holocene transition periods, characterized by Rumex/Oxyria, after which much warmer, Holocene conditions occur, with a succession of birch-juniper and birch-hazel woodland. Birch persists in small amounts locally, but the presence of other arboreal pollen is thought to be mostly due to long distance pollen transport. From c. 5500 BP, the landscape is largely treeless, apart from a brief phase of local pine forest c 4500–4000 BP. These changes are compared with other pollen records for northern Scotland. The trends in the Late-glacial vegetation are broadly similar to those displayed by other sites, although many of the changes appear to be less extreme than for areas to the south and west. The lack of Artemisia during the Loch Lomond Stadial in the east Sutherland and Caithness area is particularly noteworthy. This may imply more extensive snow cover. The early Holocene is similar to the rest of northern Scotland, but invasion of Betula was more rapid and the later growth of this genus and hazel was not so dense. The lack of pine in the early Holocene allies the site with eastern Caithness and the Northern Isles, rather than north-west Scotland and the Grampians. It is suggested that the effects of early human populations on such marginal forest areas may have been underestimated in the past. The paradoxical existence of pine macrofossils with no pollen evidence in several areas of Scotland is explained, at least in part, by the lack of temporal precision of pollen diagrams rather than by low pollen productivity, extremely localized occurrences of the species or strong prevailing winds.