In Cyclamen, a protocol for somatic embryogenesis, applicable to a large number of genotypes, was established in order to produce artificial seeds. The use of 2,4D to induce embryogenic competence is essential in this particular case. RAPD analysis was performed on embryogenic Cyclamen callus in order to study, in this propagation system, the variability related to the different 2,4-D concentrations and exposure times. The primary callus was induced on medium containing 2,4-D (2 mg L -1 ) and 2iP (0.8 mg L -1 ) from immature ovaries of one mother plant. From each ovary, one cell line was established. The embryogenic callus was then transferred in presence of increasing 2,4-D concentrations. After several subcultures, half of the calli were transferred onto hormone-free medium in order to detect the embryogenic potential related to the 2,4-D levels. The second half was used for DNA analysis. Twenty-five random primers produced 161 reproducible bands. Six primers produced band patterns that differed among the samples. The four cell lines used in the experiments were divided into two groups showing different phenotypic and molecular behaviour. In one cell line, a different RAPD pattern was observed in the callus grown at the higher 2,4-D concentration. RAPD analysis appeared to be a useful tool for the early screening of genetic variation during callus culture, and allowed precocious selection of the most stable embryogenic cell lines.