We investigate the validity of approximations that are sometimes made in ca
lculating the nuclear induced breakup of halo nuclei. We find that a trunca
ted coupled-channels calculations, in which the nuclear couplings between c
ontinuum states are ignored, gives almost the same result as a first-order
calculation. However, the couplings are much too strong to justify these ap
proximations. This is demonstrated in the frozen limit of a semiclassical d
escription, where one can compare to exact results. We find in this limit t
hat the one-neutron removal cross section of Be-11 obtained in the approxim
ate treatment is much larger than the exact result. This trend is also indi
cated at low energy by comparing a perturbative calculation of the breakup
of B-8 to the result of a more realistic treatment, which evolves the wave
function of the valence proton essentially to all orders in the target fiel
ds. [S0556-2813(99)01406-5].