A nonvolatile semiconductor memory device and a programming method thereof are provided. The programming method includes first programming a cell among a plurality of adjacent memory cells to the highest threshold voltage distribution corresponding to a data state, and subsequently programming the other adjacent cells to the lower threshold voltage distributions corresponding to second and third data states. The second data state and the third data state may have the second highest threshold voltage distribution and the third highest threshold voltage distribution, respectively, or the third highest threshold voltage distribution and the second highest threshold voltage distribution, respectively.