[PSI+] is a prion of Sup35p, an essential translation termination and mRNA turnover factor. The existence of lethal [PSI+] variants, the absence of [PSI+] in wild strains, the mRNA turnover function of the Sup35p prion domain, and the stress reaction to prion infection suggest that [PSI+] is a disease. Nonetheless, others have proposed that [PSI+] and other yeast prions benefit their hosts. We find that wild Saccharomyces cerevisiae strains are polymorphic for the sequence of the prion domain and particularly in the adjacent M domain. Here we establish that these variations within the species produce barriers to prion transmission. The barriers are partially asymmetric in some cases, and evidence for variant specificity in barriers is presented. We propose that, as the PrP 129M/V polymorphism protects people from Creutzfeldt-Jakob disease, the Sup35p polymorphisms were selected to protect yeast cells from prion infection. In one prion incompatibility group, the barrier is due to N109S in the Sup35 prion domain and several changes in the middle (M) domain, with either the single N109S mutation or the group of M changes (without the N109S) producing a barrier. In another, the barrier is due to a large deletion in the repeat domain. All are outside the region previously believed to determine transmission compatibility. [SWI+], a prion of the chromatin remodeling factor Swi1p, was also proposed to benefit its host. We find that none of 70 wild strains carry this prion, suggesting that it is not beneficial.