The result f(B) approximate to f(D), which had been obtained from sum
rules and lattice calculations, and which contradicts the naive predic
tion f(p) similar to 1/root M(p), is rederived using a simple argument
based on semilocal parton-hadron duality. It is argued that both here
and in sum rules the key point is the use of semilocal duality instea
d of the bound-state wave functions.