Sm–Nd isotope ratios of 1.9–1.8 Ga granitoids delineate the Archaean–Proterozoic boundary in northern Sweden, an important feature in the Fennoscandian Shield. The boundary strikes approximately WNW–ESE and is defined as a c. 20 km wide zone with juvenile Palaeoproterozoic rocks to the SSW and Archaean and Proterozoic rocks, derived to a large extent from Archaean sources, to the NNE. It therefore constitutes the strongly reworked margin of the old Archaean craton. Extrapolation of the boundary offshore into the Bothnian Bay and correlation with the marine reflection seismic BABEL Lines 2 and 3/4 indicates that the boundary dips to the south-southwest, consistent with interpretation of the Sm–Nd data. In order to tie the BABEL results with onshore surface geology and obtain detailed images of the uppermost crust a short (30 km of subsurface coverage) pilot profile was acquired in the Luleå area of northern Sweden during August 1999. The profile consisted of a high-resolution shallow component (1 kg shots) and a lower-resolution deep component (12 kg shots). Both components image most of the reflective crust, with the deep component providing a better image below 10 s. Comparison of signal penetration curves with data acquired over the Trans-Scandinavian Igneous Belt (a large batholith) indicate the transparent nature of the crust there to be caused by geological factors, not acquisition parameters. Lower crustal reflectivity patterns on the Luleå test profile are similar to those observed on the BABEL lines, suggesting the same lower crust onshore as offshore. Interpreted Archaean reflective upper crust in the NE extends below more transparent Proterozoic crust in the SW. This transparent crust contains a number of high-amplitude reflectors that may represent shear zones and/or mafic rock within granite intrusions. A marked boundary in the magnetic field in the SW has been interpreted as being the result of a gently west-dipping contact zone between meta-sediments and felsic volcanic rocks, however, the seismic data indicate a near-vertical structure in this area. By correlating the onshore and offshore seismic data we have better defined the location of the Archaean–Proterozoic boundary on the BABEL profiles. Our new interpretation of the crustal structure along the northern part of the BABEL Line 2 shows a more bi-vergent geometry than previous interpretations. Comparison of the re-interpreted crustal structure in northern Sweden with that found in the Middle Urals shows several similarities, in particular the accretion of a series of arcs to a stable craton. Based on this similarity and geological data, we deduce that a continental arc accreted to the southwestern margin of the Archaean craton at c. 1.87 Ga. Shortly thereafter, the Skellefte island arc underthrust the continental arc owing to a collision further to the southwest resulting in the bi-vergent crustal structure observed today.