Field experiments to investigate the efficacy of flame weeding were conducted in southern Sweden on a variety of natural weed flora at different developmental stages. The susceptibility of different species and stages was compared by modelling the dose–response with logistic models. Weed species with unprotected growing points and thin leaves such as Chenopodium album L., Stellaria media (L.) Vill. and Urtica urens L. were susceptible. When these plants had 0-4 true leaves, complete kill was achieved at propane doses of 20-50 kg ha-1. Species with protected growth points such as Capsella bursapastoris (L.) Medic and Chamomilla suaveolens (P.) Rydb. were tolerant due to regrowth after flaming, and they could be completely killed only in the early stages. Poa annua L. could not be completely killed with a single flame treatment, regardless of developmental stage or propane dose. Plant size had a major influence on the lethal dose requirement. Propane doses of 10-40 kg ha-1 were required to achieve 95% control of plant numbers for sensitive species with 0-4 true leaves, whilst plants with 4-12 leaves required 40-150 kg ha-1. When flaming naturally emerged weeds at early developmental stages, split applications of two half-dose treatments 1 week apart did not reduce plant numbers as effectively as a single late flame treatment with the same total dose.