The changing relationship between class and political party in Britain has been the
subject of considerable debate among social scientists, much of it about the measures
used to identify patterns of change. The paper updates previous work by extending
it to the most recent British election, by comparing different procedures for allocating
respondents to classes, and by using a recently developed log-multiplicative (`UNIDIFF')
model which is especially appropriate for testing whether or not classes have converged
in their voting behaviour. The analysis confirms and strengthens previous interpretations
which have argued that the class basis of partisanship is not in continuing decline.