Two pseudorabies virus vaccine strains (Bartha and Norden) that have a similar deletion in the short unique (Us) region of the genome have been identified previously (B. Lomniczi, M. L. Blankenship, and T. Ben-Porat, J. Virol. 49:970-979, 1984). These strains do not code for the glycoprotein gI, a glycoprotein that has been mapped on the wild type virus genome by T. C. Mettenleiter, N. Lukacs, and H. J. Rziha (J. Virol. 53:52-57, 1985) to the sequences deleted from the vaccine strain. Restoration of these deleted sequences to the Bartha strain genome restores to the virus the ability to specify the gI glycoprotein. The Bartha vaccine strain grows as well as wild-type virus in pig kidney and in rabbit kidney (RK) cells, but is not released efficiently from and forms small plaques in RK cells. The rescued Bartha 43/25a strain (which has an intact Us) is released considerably more efficiently than the Bartha vaccine strain, but less efficiently than wild-type virus from RK cells; it also forms larger plaques on RK cells than does the parental Bartha vaccine strain. The Norden vaccine strain, which has a deletion in the Us, is released better from RK cells than is the Bartha strain, but not as well as is wild-type virus. We conclude that whereas the sequences in the Us that are deleted from the Bartha and Norden strain genomes specify functions that play a role in the release of virions from some cell types, at least one other function (which is defective in the Bartha strain but not in the Norden strain) also affects release of virus from these cells. Since restoration to the Bartha strain of an intact Us restores to the virus both the ability to grow in chicken brains (B. Lomniczi, S. Watanabe, T. Ben-Porat, and A. S. Kaplan, J. Virol. 52:198-205, 1984) and to be released from RK cells, the possibility that the lack of virulence of the Bartha vaccine strain may be related to its limited release from some target cells is discussed.