Previous studies of Toxoplasma gondii, based on samples dominated by clinical isolates, have concluded that its population structure is clonal, despite the sexual reproduction that occurs in cats. To determine whether this applies to non-clinical isolates, we compared patterns of linkage disequilibrium (LD) among seven loci in samples of T. gondii from Brazil and the US. LD was detected in both locations, but it was substantially lower in Brazil. The lower LD in Brazil can be explained by a higher rate of sexual reproduction between different genotypes (outcrossing) because of a higher rate of transmission. The extent of LD between pairs of physically unlinked loci varied significantly in each location. Moreover, the magnitude of LD between corresponding locus pairs in Brazil and the US was correlated, despite minimal gene exchange between the continents (mean FST = 0.19). The heterogeneity among locus pairs and the correlation in LD between physically unlinked locus pairs from different continents suggests that locus-specific factors, such as epistatic selection are involved in maintaining LD in T. gondii. Possibly, the unique life cycle of T. gondii with its unpredictable transmission among diverse host species and distinct ecological habitats requires specific combinations of alleles from multiple loci. The usefulness of typing isolates based on physically unlinked loci is questioned not only by the geographic variation in the reproductive population structure, but mainly by the low overall predictability of the genotype of one locus based on the genotype in another (unlinked) locus. This predictability ranged between 23 and 45%, but was close to nil for a considerable fraction of locus pairs.