[영문]This paper sets out an empirical model of learning with forgetting and uses it to estimate how much hospital quality improves with experience. The size of the learning effect and the depreciation rate are estimated for two cardiac procedures in Maryland and Arizona. Models are estimated using patient survival as the outcome of interest. The results show that learning does not appear to be a factor in hospital quality for either procedure or for surgery generally. From a policy standpoint, based on these results, regulations in Maryland that seek to concentrate these two procedures among a small number of providers could not be justified on the grounds that higher volume would increase the quality of care