Hebridean sites of the coastal sand cliffs and associated machair, or sandy plain have been known for many years. Artefacts and ecofacts of various types have long been collected from archaeological sites in the eroding sand-cliffs of the machairs of the Outer Hebrides. Early in 1983, personnel of the then Central Excavation Unit of Historic Scotland's predecessor revisited very nearly all of the coastal archaeological sites then known in the Long Isle, with the specific task of identifying those at immediate threat from coastal erosion and of assessing the feasibility of their excavation or preservation. Some 32 sites were seen to be undergoing active erosion; at nine of them preservation was not being pursued and excavation was feasible. These sites were of two morphotypes: sites exposed in roughly vertical sand-cliffs and sites exposed over relatively large horzontal areas of sand deflation. It was decided to examine one sand-cliff site along its exposed face. The site selected was Balelone in North Uist, its excavation designed to explore both the problems associated with the excavation of deep midden sites with complex stratigraphy and the not-inconsiderable problems of excavation in sand. In the light of the Balelone trial excavation, a new approach was called for. A structured approach aimed firstly at establishing the three-dimensional extent of the sites to be examined. Four sites were then sampled (the sand-cliff sites of Baleshare, on the island of the same name off the west coast of North Uist and Hornish Point, South Uist and the deflation sites of South Glendale, South Uist and Newtonferry, North Uist) within a rigorously-defined research framework.The machair sites were formed by sand accretion, facilitated by human activities ranging from construction to refuse disposal and cultivation. Their formation was intermittent and they underwent episodes of major erosion, isolating the sites from the landscape mass of the machair sands. Despite their apparent wealth of suitable material, the dating of Hebridean coastal sites presents special problems. The strategy here was to provide a dating framework for the sequences on each site, from which the dates of archaeological significant structures and events could then be arrived at by extrapolation. Preliminary dates from the earliest and latest strata at Balelone spanned such a small period that a First Millennium BC date-range could be assigned. At Baleshare, the deposits investigated were chiefly later Bronze Age; following abandonment (roughly 200 radiocarbon years) of the Period I cultivated soil Period II represented extensive, manured, cultivated fields in the vicinity of a settlement now lost to the sea. As Period II went on. the settlement seems to have moved closer to the excavated area. After another hiatus of a minimum of 350 radiocarbon years, there were further cultivated plots and associated settlement of Iron Age date (Period III). By contrast, the site at Hornish Point (including successive wheelhouses and associated cultivation areas) is considered to be all of one - dynamic, Iron Age - period, lasting some 300 radiocarbon years (with potentially earlier structures unexcavated). A post-medieval blackhouse of characteristic Lewisian form had been cut into the settlement mound. The three dates from Newtonferry suggest that some Early Medieval activity took place at the site, while the bulk of the deposits date from the thirteenth-fourteenth centuries AD. At South Glendale, the radiocarbon dates indicate occupation sometime between the thirteenth and fifteenth centuries AD; stratigraphically lower, fragmented and truncated remains were prehistoric, probably early Bronze Age.