Quantitative traits, such as size and weight in animals and seed yield in plants, are distributed normally, even within a population of genetically identical individuals. For example, in plants, various factors, such as local soil quality, microclimate, and sowing depth, affect growth differences among individual plants of isogenic populations. Besides these physical factors, also epigenetic components contribute to differences in growth and yield. The network that regulates crop yield is still not well understood. Although this network is expected to have epigenetic elements, it is completely unclear whether it would be possible to shape the epigenome to increase crop yield. Here we show that energy use efficiency is an important factor in determining seed yield in canola (Brassica napus) and that it can be selected artificially through an epigenetic feature. From an isogenic canola population of which the individual plants and their self-fertilized progenies were recursively selected for respiration intensity, populations with distinct physiological and agronomical characteristics could be generated. These populations were found to be genetically identical, but epigenetically different. Furthermore, both the DNA methylation patterns as well as the agronomical and physiological characteristics of the selected lines were heritable. Hybrids derived from parent lines selected for high energy use efficiencies had a 5% yield increase on top of heterosis. Our results demonstrate that artificial selection allows the increase of the yield potential by selecting populations with particular epigenomic states.