Stonehenge on Salisbury Plain is one of the most impressive British prehistoric(c.3000–1500 BC) monuments. It is dominated by large upright sarsen stones, some of which are joined by lintels. While these stones are of relatively local derivation, some of the stone settings, termed bluestones, are composed of igneous and minor sedimentary rocks which are foreign to the solid geology of Salisbury Plain and must have been transported to their present location. Following the proposal of an origin in south-west Wales, debate has focused on hypotheses of natural transport by glacial processes, or transport by human agency. This paper reports the results of a programme of sampling and chemical analysis of Stonehenge bluestones and proposed source outcrops in Wales.Analysis by X-ray-fluorescence of fifteen monolith samples and twenty-two excavated fragments from Stonehenge indicate that the dolerites originated at three sources in a small area in the eastern Preseli Hills, and that the rhyolite monoliths derive from four sources including northern Preseli and other (unidentified) locations in Pembrokeshire, perhaps on the north Pembrokeshire coast. Rhyolite fragments derive from four outcrops (including only one of the monolith sources) over a distance of at least 10 km within Preseli. The Altar Stone and a sandstone fragment (excavated at Stonehenge) are from two sources within the Palaeozoic of south-west Wales. This variety of source suggests that the monoliths were taken from a glacially-mixed deposit, not carefully selected from anin situsource. We then consider whether prehistoric man collected the bluestones from such a deposit in south Wales or whether glacial action could have transported bluestone boulders onto Salisbury Plain. Glacial erratics deposited in south Dyfed (dolerites chemically identical to Stonehenge dolerite monoliths), near Cardiff, on Flatholm and near Bristol indicate glacial action at least as far as the Avon area. There is an apparent absence of erratics east of here, with the possible exception of the Boles Barrow boulder, which may predate the Stonehenge bluestones by as much as 1000 years, and which derived from the same Preseli source as two of the Stonehenge monoliths. However, 18th-century geological accounts describe intensive agricultural clearance of glacial boulders, including igneous rocks, on Salisbury Plain, and contemporary practice was of burial of such boulders in pits. Such erratics could have been transported as ‘free boulders’ from ‘nunataks’ on the top of an extensive, perhaps Anglian or earlier, glacier some 400,000 years ago or more, leaving no trace of fine glacial material in present river gravels. Erratics may be deposited at the margins of ice-sheets in small groups at irregular intervals and with gaps of several kilometres between individual boulders.‘Bluestone’ fragments are frequently reported on and near Salisbury Plain in archaeological literature, and include a wide range of rock types from monuments of widely differing types and dates, and pieces not directly associated with archaeological structures. Examination of prehistoric stone monuments in south Wales shows no preference for bluestones in this area. The monoliths at Stonehenge include some structurally poor rock types, now completely eroded above ground. We conclude that the builders of the bluestone structures at Stonehenge utilized a heterogeneous deposit of glacial boulders readily available on Salisbury Plain. Remaining erratics are now seen as small fragments sometimes incorporated in a variety of archaeological sites, while others were destroyed and removed in the 18th century. The bluestones were transported to Salisbury Plain from varied sources in south Wales by a glacier rather than human activity.