I wonder how you'd control for the fact that the Marshmallow Test is widely known and that parents may have explicitly trained their children on similar problems in the hope that they'll benefit later from the ability to delay gratification?
The correlation was found weak.