A　learning　model　based　on　the　operant　conditioning　mechanism　(OCLM)　is　presented　in　this　paper　to　deal　with　the　autonomous　learning　problem　in　cognitive　robotics.　The　model　is　described　by　9　elements,　including　the　space　set,　the　action　set,　the　bionic　learning　function　and　the　system　entropy　etc.　To　describe　the　learning　mechanism　which　is　the　core　of　the　model,　a　new　notion　＂negative　ideal　degree＂(NID)　is　defined.　We　also　prove　the　convergence　of　OCLM　to　indicate　that　the　model　is　a　self-organization　system.　OCLM　has　been　applied　to　simulating　the　Skinner　rat　experiment.　The　results　show　that　this　model　can　well　simulate　the　animal＇s　operant　conditioning　behavior,　acquire　the　cognitive　skills　through　the　interaction　with　the　environment　and　achieve　self-learning　and　self-adaptability.