Abstract Seed accessions from 7 plant families and 28 species stored for above 20 years in the National gene bank of Bulgaria were evaluated. All seed accessions were maintained as base collection under long-term storage conditions with low moisture contents (5±2%) in hermetically closed containers at −18°C. On the basis of experimental data, the seed storage characters σ (standard deviation of seed death in storage), P50% (the time for viability to fall to 50%) and P10% (the time for viability reduction of 10%) were determined allowing the prediction of seed storage life and the regeneration needs. The results showed significant differences in loss of seed viability among species and within the species. After 20–24 years of storage, eleven crops showed minimal viability decline under 5% as compared to the initial viability (oats, barley, maize, bread wheat, durum wheat, smooth brome grass, faba bean, chickpea, sunflower, cucumber and pepper). For the same storage time, another group of crops (sorghum, triticale, orchard grass, tall fescue, common vetch, grass pea, lentil, common bean, rapeseed, tobacco, flax, cabbage and tomatoes) presented 5–10% reduction of seed viability. More significant changes in seed viability – above 10% – were detected for peanuts, lettuce, soybean and rye. The σ values varied from 20.41 years (Arachis hypogaea L.) to 500 years (for Avena sativa L. and Triticum aestivum L). There was wide variation across species, both in time taken for the viability to fall to 50% and in time taken for the seed viability reduction of 10%. The study illustrates the positive effect of both seed storability early monitoring and prediction of regeneration needs as a tool for limiting undesired losses.