Individuals with a history of previous training of are more easily retrained even after prolonged de-training, and this phenomenon has been dubbed “muscle memory”. This has been an unfortunate term since it was attributed solely to motor learning in the CNS. Motor learning has, however not been a satisfactory explanation for why muscle force is more easily required if you have once been strong. Until recently muscle hypertrophy and atrophy was considered reversible processes as illustrated in Fig. 1. When muscle mass was built myonuclei was added to the growing fibres from activated satellite cells to support the larger cytoplasm. During atrophy the “excess” nuclei was believed to be lost through a selective nuclear apoptosis within the intact fibres. With observations based on in vivo imaging in rodents we have challenged this view. Thus, during a variety of inactivity conditions we observed no loss of myonuclei during atrophy, apoptosis was confined to other cell types. Muscle is a postmitotic tissue and myonuclei are very stable; a low estimate for humans would be a half-life of 15 years. Thus, the increased number of nuclei could represent a long lasting “memory” of previous size. We also investigated the ascending limb of the model in Fig. 1, and confirmed that during de novo overload hypertrophy the number of myonuclei are increased, but seemed to precede the growth in size. In recent experiments with satellite cell ablation we found that such cells are obligatory for de novo hypertrophy. In contrast, in muscles that were first atrophied by hindlimb suspension and then grew back to the previous size , new myonuclei were not recruited during regrowth. These observations led us to propose a new model (Fig. 2) in which there is a first training route where new myonuclei are added, and then a re-training route where the fibres can grow without adding new nuclei. We then asked if the retraining route was faster. Female mice were treated with testosterone that created hypertrophy. When the testosterone was withdrawn the fibre size, but not the elevated number of nuclei reverted to control levels within 3 weeks. After a 3 months washout period the muscles were subjected to overload, the testosterone group then grew 36% within the first 6 days, while the control group grew insignificantly (6%). We hypothesize that the elevated number of nuclei represents a novel form of epigenetic memory in the form of an elevated number of myonuclei aiding muscle strength building in muscles that has previously been strong. If applicable to humans this mechanisms has important implications for public health and for evaluation of doping exclusion times.