Based on recent first-principles computations in perovskite compounds,
especially BaTiO3, we examine the significance of the Born effective
charge concept and contrast it with other atomic charge definitions, e
ither static (Mulliken, Bader, etc.) or dynamical (Callen, Szigeti, et
c.). It is shown that static and dynamical charges are not driven by t
he same underlying parameters. A unified treatment of dynamical charge
s in periodic solids and large clusters is proposed. The origin of the
difference between static and dynamical charges is discussed in terms
of local polarizability and delocalized transfers of charge: local mo
dels succeed in reproducing anomalous effective charges thanks to larg
e atomic polarizabilities but, in ABO(3) compounds, ab initio calculat
ions favor the physical picture based upon transfer of charges. Variou
s results concerning barium and strontium titanates are presented. The
origin of anomalous Born effective charges is discussed thanks to a b
and-by-band decomposition which allows us to identify the displacement
of the Wannier center of separated bands induced by an atomic displac
ement. The sensitivity of the Born effective charges to microscopic an
d macroscopic strains is examined. Finally, we estimate the spontaneou
s polarization in the four phases of barium titanate.