Genetically engineered viruses and viral genes inserted into retroviral vectors are increasingly being considered for experimental therapy of brain tumors. A primary target of these viruses and vectors is human gliomas, the most frequently occurring primary human brain tumor. To investigate the potential of genetically engineered herpes simplex viruses (HSVs) in the therapy of these tumors, we compared the attributes of two viruses, a recombinant from which the γ134.5 gene had been deleted (R3616) and a recombinant in which the γ134.5 gene had been interrupted by a stop codon (R4009). Previous studies have shown that these recombinants were completely devoid of the ability to multiply in the central nervous system of rodents. To pursue these studies, we developed a scid mouse glioma model. Tumor cell response (survival) for 103, 104, and 105 implanted MT539MG glioma cells was 38, 23, and 15 days, respectively. The results were as follows: (i) both R3616 and R4009 replicate and cause cytolysis in diverse glioma cell lines of murine and human origin in vitro, and (ii) in Winn-type assays 105 MT539MG cells coinoculated with R3616 or R4009 as compared to saline significantly prolonged survival in a dose-dependent fashion. Mice that received only tumor cells or the wild-type parent strain of the recombinants, HSV-1(F), died within 15 days. Survival was greatest with R4009. These experiments define both a model for screening oncolytic viruses and a genetically engineered virus of significant potential use as an oncolytic agent.