An explanation is proposed for the fact that Lepage-Mackenzie tadpole impro
vement does not work well for staggered fermions. The idea appears to work
for all renormalization constants which appear in the staggered fermion sel
f-energy. Wilson fermions are also discussed.