Determination of the pattern of variation in population abundance among spatial scales offers much insight into the potential regulating factors. Here we offer a method of quantifying spatial variance on a range of scales derived by sampling of irregularly spaced sites along complex coastlines. We use it to determine whether the nature of spatial variance depends on the trophic level or the mode of dispersal of the species involved and the role of the complexity of the underlying habitat. A least-cost distance model was used to determine distances by sea between all pairs of sites. Ordination of this distance matrix using multidimensional scaling allowed estimation of variance components with hierarchical ANOVA at nested spatial scales using spatial windows. By repeatedly moving these spatial windows and using a second set of spatial scales, average variance scale functions were derived for 50+ species in the UK rocky intertidal. Variance spectra for most species were well described by the inverse power law (1/fbeta) for noise spectra, with values for the exponent ranging from 0 to 1.1. At higher trophic levels (herbivores and carnivores), those species with planktonic dispersal had significantly higher beta values, indicating greater large- than small-scale variability, as did those on simpler coastlines (southwestern England and Wales vs. western Scotland). Average abundance and proportional incidence of species had the strongest influence on p values, with those of intermediate abundance and incidence having much greater large-scale variance (beta approximately 0.5) than rare or ubiquitous species (beta approximately 0).