Following a long-term herbivore-exclusion study (2003-2008) in the mid-eulittoral zone at Kalk Bay, Cape Town, South Africa, the post-herbivore-exclusion effect (12 years later) of the limpet Cymbula oculus, the dominant herbivore in the community, was examined after it had recolonised the area. Despite the limpet having recruited into all plots (previous exclusion plots and control plots), the species richness, percentage cover of sessile organisms, invertebrate densities and invertebrate biomass were all significantly higher in previous exclusion plots compared with in control plots. Cymbula oculus density was higher in previous exclusion plots (11.11 [SD 12.54] ind. m−2) than in control plots (9.26 [SD 7.97] ind. m−2) (p = 0.038); however, its biomass was similar between the previous exclusion plots and control plots (366.39 [SD 421.63] vs 367.96 [SD 285.44] g m−2, p = 0.179), indicating that limpets outside exclusion plots were generally larger. This was likely because of younger C. oculus individuals, heightened interspecific competition, or insufficient space in previous exclusion plots. The observed differences might be attributed to a combination of factors, including the persistence of species that have achieved a size refuge from grazing and bulldozing by C. oculus, as well as the establishment of invasive mussels Mytilus galloprovincialis which offer secondary substrates and microhabitats for the establishment of macroalgae and other invertebrates. The recruitment of M. galloprovincialis into the previous exclusion plots is likely to be the cause of the alternative community state observed. Unless some extreme disturbance event occurs, the community from which C. oculus was previously excluded is unlikely to revert to its pre-C. oculus exclusion state.