The conservation of lotic ecosystems has historically focused on maintaining the structural properties and processes of river systems, considered as surrogates for the protection of biological diversity. However, the geological structure of the catchment unit and the hierarchical, longitudinal nature of the rivers that drain them impose a number of potential barriers to dispersal. This creates a mosaic of aquatic islands within a terrestrial landscape. As such the protection of biodiversity and biological processes requires considerations that extend beyond the catchment unit. Understanding the extent to which barriers limit the movement of individuals is important in developing an integrated approach toward conservation of river systems. It is also important in understanding the role of dispersal in the process of species formation and population structure. While catchment units represent the logical social, economic and often political scale upon which to manage water resources they are increasingly being defined as the appropriate functional unit for the conservation and management of freshwater ecosystems. The aim was to determine the extent to which catchment units represent the appropriate scale for the conservation of lotic biodiversity. This was done by examining the effect of catchment units on the distribution of genetic variation and population structure in four aquatic taxa among streams in the south-western Cape, South Africa. All four taxa are part of the ancient paleoendemic Gondwanaland fauna characteristic of the Cape region and reflect relative differences in dispersal. The taxa were the freshwater fish, Galaxias zebratus (Teleostei: Galaxiidae); the net-winged midge Elporia barnardi (Diptera: Blephariceridae), which given its specialised morphology and specific habitat requirements has a very limited potential for dispersal; the stonefly Aphanicerca capensis (Plecoptera: Notonemouridae), a species with intermediate dispersal; and the widely distributed dragonfly Aeshna subpupillata (Odonata: Aeshnidae), with the potential for wide dispersal. Allozyme electrophoresis and direct sequencing of a fragment of the cytochrome oxidase subunit 1 (COI) region of the mitochondrial DNA were used to examine genetic structuring within and among streams in two discontinuous mountain ranges. F statistics were calculated from allele frequencies derived from allozyme electrophoresis as a measure of population subdivision and population trees constructed. Nucleotide diversity and levels of divergence were calculated among mtDNA sequence data. Genetic distance, and the relationship among haplotypes, was examined using neighbour-joining trees and an analysis of molecular variance in order to determine the effect of catchment units on dispersal, the distribution of genetic variation and population structure. Low levels of allozyme variability were observed in all four taxa, with no variable loci resolved for the stonefly A. capensis. Significant population structure among all sites in the two ranges in G. zebratus, E. barnardi and A. subpupillata highlights the effect of discontinuous habitat (~0.70, 0.39±0.10 and 0.03±0.01 respectively), while FST values among streams on Table Mountain reflect differences in the dispersal potential of the three species (~0.70, 0.23±0.04 and 0 respectively). Population trees for the Cape galaxiid G. zebratus and the net-winged midge E. barnardi revealed two highly divergent groups (Genetic Identity = 0.41 and 0.73 respectively). Both reflect poor dispersal potential, with the pattern among G.zebratus reflecting a pattern of connectivity between ancient drainages during periods of lower sea levels. Mitochondrial DNA data obtained from the COI region similarly revealed two highly divergent clades in populations of the Cape galaxiid G. zebratus (~7%), the netwinged midge E. barnardi (~5%) and the stonefly A. capensis (~7%). Additional data derived from the cytochrome b region for G. zebratus revealed five highly divergent clades from across the species range (from 7 to 17%). Congruence between monophyletic clades and catchment units in G. zebratus and E. barnardi, along with an analysis of the distribution of genetic variation, suggest movement is confined to within the catchment. In contrast, the distribution of haplotypes and genetic variation in A. capensis and A. subpupillata suggests movement beyond the catchment boundary. Similarities in the degree of divergence in A. capensis and E. barnardi indicate a vicariant event around 3-4 MYBP, coinciding with the erosion of the land bridge between Table Mountain and the Hottentot’s Holland. Divergence among G. zebratus, A. capensis and E. barnardi suggests the presence of more than a single species in all three taxa.