Linguistic universals, markedness and learnability: comparing two different approaches