A simulation study of the effects of linked selection on a neutral locus is described. Of six models of selection examined, four could be called variation-reducing as they lower the level of molecular variation at the neutral locus. Unfortunately, all four distort the neutral frequency spectrum in similar ways, making the problem of inferring the appropriate model quite difficult. Similarly, an examination of the effects of linked selection on the non-random association of mutations at two linked neutral loci did not suggest that this property would be useful for inferring the nature of linked selection either. Linked selection may have a profound effect on the 'effective population size' of the neutral locus. For two of the variation-reducing models, the effective size at the neutral locus is almost independent of the actual population size.