To produce hybrid seeds, the cytoplasmic male sterile line must be free of contamination by maintainer seeds. The hybrid will then contain heterogeneous plants which will decrease the yield. To specifically identify the maintainer cytoplasm, we propose the use of a molecular method. A mitochondrial plasmid, specific to the maintainer cytoplasm is used as a molecular probe to reveal the possible presence of plasmid sequences in the CMS seed stocks. The method appears to be simple, at a reasonable cost, very fast and more accurate than field controls. For sugar beet, we first determined the level of the minicircle c in one seed by comparison with the probe loaded as a ladder between 5 pg to 1 000 pg, (fig 1). Because of a slight homology between minicircle c (130 pg per seed) and the minicircle a, the hybridization signal of dot blot does not allow 1 maintainer seed to be identified in less than 10 CMS seeds, (fig 2). In order to conciliate a deep hybridization signal and a low number of sub-samples, we divided the 1 000 seed samples into 67 sub-samples thus allowing us to accurately determine the number of contaminated maintainer seeds. We routinely checked 1 000 dry seeds for sunflower (table I). The sample in an assay was compared to controls artificially contaminated (fig 3). The method appeared to be expensive and relatively uncertain when CMS controls were also contaminated with maintainer seeds. By dividing the stock into 6 sub-samples, we were able to determine the maintainer contaminant seeds within each sub-sample (fig 4). We applied this method to 20 commercial samples from various companies (table II). The results of the molecular control fit the controls in the field.