Electromigration has assumed considerable technical significance since its manifestation as a cause for integrated circuit failure. The driving force in electromigration is generally assumed to consist of a "direct force" due to the applied field and an "electron wind" term due to momentum exchange with the current carriers. In the case of the "direct force" it is uncertain how effectively the charge is screened. In the case of the "electron wind" it is unclear whether the moving atom senses the carrier's change in lattice momentum or real momentum or some more complex compromise. The prevailing viewpoint in the literature favors lattice momentum, and this is criticized. A more meaningful prescription is obtained, following Bosvieux and Friedel, by considering the forces on the unscreened lattice defect, due to the total localized charge polarization in the presence of current flow. The nature of this polarization pattern is still the subject of considerable disagreement. This author's views on the subject are outlined. © 1975 American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc.