In a brief historical discussion the authors point out that the 54-year-old Rl strain of human tubercle bacillus possesses a certain small virulence for guinea pigs, producing spreading tuberculosis, followed normally by resolution of the lesions. In occasional abnormally susceptible animals, fatal spreading lesions occur. They then describe a method of producing totally avirulent variants of Rl, by a process of selecting subculture material from "resistant'' growths in old aged lysed cultures. They find such avirulent strains or dis-sociants incapable of producing spreading lesions even in animals previously or concurrently subjected to silica inhalations, or intracerebral inoculations. Intracerebral inoculation of the typical Rl variant plus silica killed animals within 2 months; non-silicotic controls survived 4-5 months. The original Rl strain inoculated intracerebrally is uniformly fatal, as a result of local disease. They explain the effect of silica as producing a lowered resistance similar to that occasionally occurring spontaneously, and suggest the use of silicotic animals in virulence tests of other dissociated variants of other strains. "Avirulent" variants of Rl reads awkwardly at first, as this organism is already considered avirulent in most common parlance.