The concept of a neighbourhood effect within British voting patterns has largely been discarded, because no data have been available for testing it at the appropriate spatial scales. To undertake such tests, bespoke neighbourhoods have been created around the home of each respondent to the 1997 British Election Study survey in England and Wales, and small-area census data have been assembled for these to depict the socio-economic characteristics of voters' local contexts. Analyses of voting in these small areas, divided into five equal-sized status areas, provides very strong evidence that members of each social class were much more likely to vote Labour than Conservative in the low-status than in the high-status areas. This is entirely consistent with the concept of the neighbourhood effect, but alternative explanations are feasible. The data provide very strong evidence of micro-geographical variations in voting patterns, for which further research is necessary to identify the processes involved.