Somaclonal variation (SV) is a major concern in all micropropagation systems. It is described as the phenotypic variation displayed in in vitro-derived regenerants and it is believed to be originated from a large array of genetic and epigenetic mechanisms. Highly productive Coffea arabica hybrids are clonally disseminated in Meso-American region through somatic embryogenesis (SE). The objective of the present work in coffee is to evaluate the trueness-to-type of SE, to understand better the mechanisms involved in SV and further optimize SE conditions. We assessed the variations in the propagated plants at the phenotypic, cytogenetic (chromosome counting), genetic (mutations/AFLP, activation of transposable elements/S-SAP) and epigenetic (methylation/MSAP) level by using two complementary approaches. First, with 2 hybrids we studied industrial culture conditions expected to be weakly mutagenic thanks to the combined use of short term proliferation period (6 months) and low auxin supply (0-1.4 µM 2,4-D). Two proliferation systems i.e. secondary embryogenesis and embryogenic suspensions were compared, the latter being more productive and economic. AFLP and MSAP molecular analyses on 145 somatic seedlings showed that genetic and epigenetic polymorphisms between mother plants and emblings were extremely low, i.e. ranges of 0–0.003% and 0.07–0.18% respectively, with no significant difference between the proliferation systems. No plant was found to cumulate more than 3 methylation polymorphisms. For the two hybrids tested, massive phenotypic observations in nursery and field plots showed very low levels of SV (0.9% from 800,000 plants). Cytological analysis showed abnormal chromosome numbers (41-43, 45) in most of coffee somaclonal variants and normal numbers (44) in phenotypically normal plants. Stressful experimental conditions were also applied by using extended proliferation periods (4, 12 and 27 months) for three independent embryogenic lines established for the Catur