Methods of operating non-volatile memory devices are provided including receiving program data and a program address. Memory cells that correspond to the program address are selected from among memory cells in an erased state. The selected memory cells are programmed based on the program data such that each of the selected memory cells is programmed to one of a plurality of programmed states, where threshold voltage distributions of the programmed states are different from each other and are higher than a threshold voltage distribution associated with the erased state. By programming all or a portion of the memory cells corresponding to the erased state to have positive threshold voltages, degradation of the data retention capability of the memory cells may be reduced.