Seismic imaging techniques can be subdivided into inversion and migrat
ion. The object functions for inversion and migration are, respectivel
y, the medium contrast parameters and reflectivity. In this paper, the
relationship between inversion and migration is approached by analyzi
ng the underlying representations (the forward models). It appears tha
t the ''two-way representation'' (which underlies inversion) as well a
s the ''one-way representation'' (which underlies migration) can both
be expressed in terms of a volume integral over the appropriate object
function. In their linearized form, these representations account for
primaries only. In this case, the one-way representation in terms of
reflectivity is the most accurate of the two, which implies that prope
r migration is more accurate than linearized inversion. Internal multi
ples can be taken into account by the nonlinear representations. As an
alternative, however, the ''generalized primary representation'' is i
ntroduced. In its explicit form, this one-way representation is linear
in the reflectivity (opposed to linearized). Nonlinear effects are im
plicitly accounted for by the generalized primary propagators. The gen
eralized primary representation is a suitable basis for true amplitude
migration, taking the angle-dependent dispersive effects of fine laye
ring into account.