The idea of a primitive system in Kron's theory is shown to involve two distinct concepts and to be not well defined. An attempt is made to give a sharper definition. The discussion is confined to electrical networks. The anomaly is stressed that though Kron's primitive network and the given network are physically different, the transformation laws imply that their instantaneous power inputs are equal. Ways proposed by various authors of avoiding this anomaly are briefly discussed. Some ignore the primitive network; others make the primitive network physically equivalent to the given network, either by changing the primitive network or by changing the given network. Since this does not accord with the method by which Kron forms his primitive network, the view is put forward that the primitive networkisphysically different from the given network. A new network, the “connected primitive network,” is introduced which, among other things, clarifies the physical significance of a singularC-transformation. The given network and primitive network are considered to stand for infinite families of networks; the singularC-transformation then selects from the family of primitive networks a subfamily that coincides with the family of given networks.