In　this　paper,　we　propose　a　stochastic　win-stay-lose-learn　in　which　an　individual　is　more　likely　to　imitate　one　of　its　nearest　neighbor＇s　strategy　if　its　aspiration　is　not　achieved.　The　results　on　the　spatial　public　good　game　show　that　the　cooperation　can　be　greatly　enhanced　when　the　aspiration　is　moderate.　Besides,　we　have　studied　the　time　evolution　of　the　spatial　distribution　of　strategies　and　the　probability　that　cooperators　and　defectors　choose　to　learn　respectively.