Why does adaboostM2 produce only stumps as weak learners

Hello, I noticed the same problem as sedar sedar. adaboost-stump When adaboosting a classification tree, the weak learners are all slumps. When fitting a classification tree with the same parameters directly on the training set, I get a much deeper tree. Can somebody tell me why I do not get deeper trees then stumps when using matlab ?

In the mean time, I will have to implement my own adaboosting algorithm...