The statistics of union density in Britain are thoroughly documented and there is little disagreement about the overall pattern across British industry between the turn of the century and the present day. However, discussion of their significance has largely been dominated by the two apparently incompatible approaches of the 'Marxian' and 'Industrial Relations' schools. In this article, it is argued that an interpretation derived from the theory of social selection can reconcile what is valid in these two approaches and thereby suggest a more satisfactory explanation than either of the reported pattern of variation, both over time and across different sectors of the economy. In conclusion, some wider implications are discussed which bear on recent debates over the class position of white-collar employees.