From July 1993 to September 1994, seasonal variations in the sources of SO42- aerosols in the Arctic lower atmosphere at Alert, Canada, (82 degrees 30' N, 62 degrees 20'W) were investigated using the sulphur isotope abundance of as little as 10 mu g of sulphur analyzed by combustion-flow isotope-ratio mass spectrometry. In conjunction with air mass trajectories and in parallel with measurements of aerosol composition, the sulphur isotope composition was used to discern sources of aerosol SO42-. Total SO42- is composed of seasalt SO42-, marine biogenic, and nonmarine SO42-. From June through September the fraction of biogenic SO42- in the non-sea-salt (nss) component ranged from 0.09 to 0.40 with an average of 0.31 +/- 0.11. Summertime nonmarine SO42- is likely anthropogenic in origin since it is isotopically indistinguishable from SO42- in the polluted winter/spring period of arctic haze (delta(34)S = +5 parts per thousand). In summer there was no significant difference in isotope composition of aerosol sulphate between air which recently traversed Eurasia and the Arctic Ocean and air arriving from North America. In contrast to summer and late winter/spring, delta(34)S values for nonmarine SO42- in fall and early winter were often less than +5 parts per thousand. These isotopically light samples were divisible into two groups: (1) those associated with air mass trajectories potentially affected by North American soils and/or smelters and (2) three weekly samples between October and December which could be attributed to fractionated sea-salt aerosol formed on refrozen Arctic Ocean leads. For the latter the ratio of SO42-/Na was estimated to be a factor of 3.6 lower than in bulk seawater. From November to May, nonmarine aerosol SO42- was apportioned into 10 aerosol components using positive matrix factor analysis of 18 aerosol ions and trace elements [Sirois ann Barrie, this issue]. In turn, a multiple linear regression of delta(34)S values against the scores of the components was used to predict the isotope composition of six components. It was concluded that the main mass of anthropogenic SO42- had a delta(34)S value near +5 parts per thousand and that biogenic SO42- had a delta(34)S Of +16 +/- 3.9 parts per thousand. Reasonable agreement between model results and sulphur isotope measurements at Alert show that SO42- apportionment using positive matrix factor analysis is a reasonable approach which gives realistic results.