 So last year biologists celebrated the 150 year anniversary of Mendel's discovery of the laws of inheritance. From the time of Mendel, the population of the world increased from 1.2 to 7.5 billion and we expect that it will reach about 10 billion by 2050. But how are we going to feed so many people thinking that we have to double food production in the next 30 or 40 years? While crop breeding has been very successful over the last 50 years, the gains in crop yield have plateaued and it's very difficult to increase them further. This is because most crops went to a very severe genetic bottleneck when they were domesticated. And the industrialization of agriculture has further decreased the variety of crops that are grown on farms. It's very difficult to introduce new genetic variation into a crop without losing some of the traits that it was chosen for. But what about epigenetic variation which is not dependent on the DNA sequence and which is therefore not genetically encoded? Epigenetic variation can be stably maintained over many, many cell divisions during the lifetime of an individual. But is it stable enough to be passed on from one generation to the next? Is it heritable just like genetic variation? In Mendel's, epigenetic inheritance is very controversial. That's very different in plants where we know many traits that were thought to be caused by a mutation. That is a genetic change and turned out to be epigenetic in nature, namely due to changes in DNA methylation. As in plants, epigenetic variation is actually heritable over several generations and could in principle be used in a breeding program. Unfortunately, epigenetic variation is very unstable and it changes back very often, much more often than genetic variation does. And so many scientists think that epigenetic variation is not stable enough to be used in a breeding program. To test whether we can select novel traits in the absence of any genetic variation, we did selection studies using the model plant Arabidopsis talliana, which is the best-studied plant at the molecular and genetic level. So using genetically identical strains, we selected for seed dispersal at the distance. And after five generations of selection, we grew all the plants in a uniform environment to get rid of confounding effects and then assessed about 20 traits in the next generation. Among other traits, we found that the selected plants flowered one or two days later than the original non-selected plants. And this was stable for two or three generations without selection. When we looked at their genome, however, we didn't find any sign of a genetic change. So we performed a genome-wide study of DNA methylation and we found that approximately 50,000 cytosines had changed their methylation pattern in comparison to the original non-selected line. And this was consistent in all three independent, completely independent selection experiments. When we analyzed gene activity in these plants, we found that indeed many of the genes that are known to affect flowering time were altered in their activity and their methylation pattern. So they were genetically identical but they were epigenetically different, resulting in a delay in flowering time. We have shown that we can select novel traits in a complete absence of genetic variation. Thus, epigenetic variation could be a source in a breeding program. And genetic variation is not the only source of variation. What fraction of epigenetic variation is actually stable enough to contribute to the diversity of crops to feed a growing world population?