The patterns of species establishment, persistence and spread in a long‐term experiment investigating the re‐creation of chalk grassland using different seed mixtures on ex‐arable land were analysed with the intention of providing enhanced re‐creation/restoration prescriptions and a critique of experimental designs. Species had a wide range of behaviours; (1) species that did not establish from seed e.g. Blackstonia perfoliata, (2) establishment but poor persistence in closed sward e.g. Hieracium pilosella, (3) establishment and persistence but little spread e.g. Onobrychis viciifolia and (4) good establishment and spread over all treatments e.g. Centaurea nigra and Trisetum flavescens. Persistence appeared greater in species with lower germination rates in both light and dark. Spread was greater in species with higher seed mass and lower dark germination rates.

However, results from small‐plot experiments should be critically analysed. Though initial results show the potential for re‐establishment of vegetation, subsequent vegetation dynamics may be more closely related to the invasion of species from other treatments than the original species sown. Experimental designs should employ either large plots and/or wide guard rows to extend their useful life span. Long‐term data show which species are useful for cost‐effective restoration. Some species need to be sown over the whole area to provide cover (a ‘grass matrix’) and competition with arable weeds. Others need to be sown over wide areas as they spread poorly. Some species can be sown at low densities or in small patches as they have the ability to spread in closed vegetation.

Nomenclature: Clapham et al. (1987).