Mounds built in pastures by the subterranean ant Lasius flavus often have a vegetation which differs markedly from their surroundings. I set out to compare the relative abundances of plant species on and off ant-hills in some grasslands in southern Britain, and to define those characteristics of species which mainly determine their success or failure on the mounds. Large ant-hills and the surrounding grassland ware sampled in detail at five acid grassland sites in the Gower peninsula, Glamorgan, and thirteen chalk grassland and chalk heath sites in southern Britain. Mounds were often dominated by perennials which could grow up through heaped soil, like Polytrichum piliferum, P. juniperinum on the Gower and Thymus drucei and Helianthemum chamaeciatus on the chalk. Annuals such as Aira praecox and Arenaria serpyllifolia were frequently confined to ant-hills. However, many species, especially rosette or semi-rosette hemicryptophytes, were consistently uncommon on ant-hills, including Luzula campestris, Cirsium acaule, Poterium sanguisorba and Felipendula vulgaris. The pleurocarpous mosses Pleurozium schreberi and Pseudo-scleropodium purum were more abundant on north-facing than south-facing aspects of the mounds. Two main factors influenced plant patterns. Firstly, the heaping of soil over the plants by the ants often caused considerable seedling mortality. Secondly, species differed in their ability to disperse 'seed' onto the mounds. When I compared the viable 'seed' 'banks' present in equal weights and surface areas of ant-hill and pasture soils, by keeping the samples moist and counting the emergent seedlings, the 'seeds' of those species uncommon on ant-hills were much less frequent in ant-hill than pasture soil. Most seedlings on the mounds probably result from 'seed' dispersal onto the mound surfaces. Soil analyses show no large and consistent differences in nutrient concentrations between ant-hill and pasture soils and thus it seems unlikely that nutrient levels greatly influence the striking distributions of many species. The vegetation on and off two hundred and six wounds of a wide range of size and activity was recorded in detail on a 1.2ha plot at Aston Rowant N.N.R., Oxon. Mapping showed that species rarely occurred on ant-hills when they were absent from the sward nearby. From scatter diagrams of the abundance of each species in relation to ant-hill volume and ant activity on a mound, it was possible to deduce how succession of some species on ant-hills might occur, if ant-hill volume were proportional to ant-hill age. The following hypothesis is consistent with the evidence. 'Young' mounds are heaps of soil which smother existing vegetation. They are invaded in particular by Festuca rubra, Lotus corniculatus, Carex flacca, Leontodon hispidus and Helianthemum chamaecistus (if in the sward nearby). Thymus drucei invades the mounds vegetatively as they expand in basal area, and gradually becomes a more important component of the vegetation. Carex flacca and Hieracium pilosella rosettes become restricted to the edges of ant-hills because the continual heaping of soil onto the tops of the mounds causes their burial and death there. However, Leontodon hispidus and Plantago lanceolata, often present inside the margins of medium and large mounds, set abundant seed there and their seedlings sometimes become established on the bare soil on top. By this time, the summits of ant-hills tend to be occupied by Helianthemum chamaecistus and Thymus drucei, which can grow up through the heaped soil. Small mounds are often deserted by the ant colony. Their vegetation becomes closed and the relative abundances of species approach those of the surrounding sward more closely than on active ant-hills of the same volume. In the absence of burial, Carex flacca, Leontodon hispidus, Plantago lanceolata, Hieracium pilosella become established over the whole surface; H. pilosella rosettes are far more dense on abandoned mounds than active ones of the same size. When mounds are abandoned the chambers and channels disappear, organic natter accumulates, an A₀ horizon is formed and the previously structureless soil becomes aggregated into peds. The life cycles of the species with the most marked patterns la relation to ant-hills were studied more fully in field and laboratory experiments. The species characteristic of ant-hills usually have either vigorous vegetative growth when smothered, the ability to flower on the mounds in too year after germination or both. For instance, established plants of Helianthemum chamaecistus and Thymus drucei, woody chamaephytes which often dominate ant-hills, can grow up through 3-4cm of suddenly-heaped soil. Yet they rarely establish themselves from seed on the mounds. In Arenaria serpyllifolia, Veronica arvensis and other winter annuals, the flowers are predominantly self-pollinated, seed-set is high, the seeds are light (60 - 270 ug) and inhibited from germinating under only 5mm of soil, a high proportion (40) of those released onto the mounds germinate each year, and the seedlings rapidly decline in number. For instance, a cohort of 2078 seedlings of Arenaria serpyllifolia followed from August 1971 to July 1972 had a half-life of 0.14 years, the seedlings dying from rabbit-scraping, soil heaping, self-thinning, fungal infection, rainwash and so on. Nevertheless, almost every surviving seedling flowered in the late spring of the first year. The minute seed weights of these species are insufficient to allow them to become established in closed vegetation. Their seeds may be inhibited from germinating under swards by the higher ratio of far red to red light there; Arenaria serpyllifolia, Veronica arvensis and Myosotis ramosissima seeds did not germinate under two layers of Tilia leaves in a laboratory experiment. Cerastium holosteoides resembles Thymus and Helianthemum in being a perennial chamaephyte able to grow up through newly-heaped soil to some extent, and resembles the winter annuals in flowering the year after germination and producing abundant small (180 ug) seeds which hardly germinate under two layers of Tilia leaves. The species with roughly equal abundances on and off ant-hills are frequently stoloniferous and able to grow up through heaped soil to some degree (e.g. Agrostis stolonifera, Festuca rubra, Campanula rotundifolia, Galium verum) or short-lived and able to flower on the mounds (e.g. Euphrasia officinalis agg, Gentianella amarella, Linum catharicum and Medicago lupulina). All the species which are considerably more abundant in the sward than on the mounds are perennials, with very slow vegetative reproduction. Thus they depend on invasion by seed to become established on ant-hills. Only a small proportion of the rosettes in the sward flower each year, the flowers are cross-pollinated, the fruits are comparatively heavy (600 - 4000 ug), the seedlings are susceptible to burial, their development on the mounds is slow and they do not flower within the first two years. Their seedlings are robust enough to establish themselves in swards. Nevertheless, within this framework, the factors affecting their abundances on the mounds differ from species to species. For instance, in Cirsium acaule the number of fruits produced is reduced to 10-14 per capitulum per year by the weather, predation and parasitism. Their dispersal is hindered by the high frequency of pappi lost before dispersal, the heavy achenes and the weather. The production of vigorous seedlings on the mounds may be affected by fungal parasitism, and their survival is probably poor because they are susceptible to burial. Seedling development is slow on ant-hills and no flowering plants were seen there. In Leontodon hispidus, however, dispersal of achenes onto the mounds is efficient, and 5-9 of achenes germinate on ant-hills, so that seedlings are frequent on the mounds in autumn. Seedlings develop slowly, remain avail, and are susceptible to burial, but mature rosettes inside the perimeter of active ant-hills often flower profusely and reproduce vegetatively. Thus Leontodon maintains larger populations on ant-hills than Poterium sanguisorba, and Cirsium acaule. On the other hand, Carex flacca maintains high densities of rosettes around the edges of 'active' ant-hills because of vigorous vegetative reproduction; reproduction from seed is negligible. Poterium sanguisorba and Plantago lanceolata were also investigated. Thus the ant-hill environment acts as a sieve, letting through those species which can flower early on or grow up through heaped soil, and excluding those which cannot flower on the mounds, with poor dispersal and low seedling development. The flora of 'active' ant-hills is maintained at a lower stage of succession than the surrounding vegetation by the constant addition of soil to the surface. The vegetation resembles that of fixed calcareous dunes and of Breckland grass-heath before myxomatosis, two cowaunities in which sand movements also provide an element of environmental 'unpredictability'. Those species which invest a high proportion of their total net assimilation in reproduction ere pre-adapted to colonise and maintain themselves on ant-hills. Those species selected for efficiency in maintaining themselves in a closed community, with heavy fruits and an inability to flower early in life, cannot maintain themselves at a high density in a community of an earlier successional stage.