We studied the natural colonisation of new species in experimental grasslands varying in plant species richness (from 1 to 60) and plant functional group richness (from 1 to 4) in either regularly or never weeded subplots during the first 3 years after establishment. Sown species established successfully, with no differences in species richness or their relative abundances between the regularly and never weeded subplots during the study period. Aboveground biomass of sown species increased with increasing sown species richness in both treatments. While a positive relationship between sown species richness and total aboveground biomass (including colonising species) existed in the 2nd year after sowing in the regularly and never weeded subplots, this positive relationship decayed in the 3rd year in the never weeded subplots because of a higher biomass of colonising species in species-poor mixtures. Total aboveground biomass varied independently of total species richness 3 years after sowing in both treatments. Jaccard similarity of coloniser species composition between regularly and never weeded subplots decreased from the 2nd to the 3rd year, indicating a divergence in coloniser species composition. Coloniser immigration and turnover rates were higher in regularly weeded subplots, confirming that weeding counteracts species saturation and increases the chance that new colonisers would establish. Although our study shows that low diversity plant communities are unstable and converge to higher levels of biodiversity, the effects of initially sown species on community composition persisted 3 years after sowing even when allowing for succession, suggesting that colonising species mainly filled empty niche space.