Selection on one or more genes inevitably perturbs other genes, even when those genes have no direct effect on fitness. This article reviews the theory of such genetic hitchhiking, concentrating on effects on neutral loci. Maynard Smith and Haigh introduced the classical case where the perturbation is due to a single favourable mutation. This is contrasted with the apparently distinct effects of inherited variation in fitness due to loosely linked loci. A model of fluctuating selection is analysed which bridges these alternative treatments. When alleles sweep between extreme frequencies at a rate lambda, the rate of drift is increased by a factor (1 + E[1/pq]lambda/(2(2lambda + r))), where the recombination rate r is much smaller than the strength of selection. In spatially structured populations, the effects of any one substitution are weaker, and only cause a local increase in the frequency of a neutral allele. This increase depends primarily on the rate of recombination relative to selection (r/s), and more weakly, on the neighbourhood size, Nb = 4(pi rho sigma)2. Spatial subdivision may allow local selective sweeps to occur more frequently than is indicated by the overall rate of molecular evolution. However, it seems unlikely that such sweeps can be sufficiently frequent to increase significantly the drift of neutral alleles.