We used genotyping-by-sequencing (GBS) to investigate the evolutionary history of domesticated tetraploid wheats. With a panel of 189 wild and domesticated wheats, we identified 1,172,469 single nucleotide polymorphisms (SNPs) with a read depth ≥3. Principal component analyses (PCAs) separated the Triticum turgidum and Triticum timopheevii accessions, as well as wild T. turgidum from the domesticated emmers and the naked wheats, showing that SNP typing by GBS is capable of providing robust information on the genetic relationships between wheat species and subspecies. The PCAs and a neighbour-joining analysis suggested that domesticated tetraploid wheats have closest affinity with wild emmers from the northern Fertile Crescent, consistent with the results of previous genetic studies on the origins of domesticated wheat. However, a more detailed examination of admixture and allele sharing between domesticates and different wild populations, along with genome-wide association studies (GWAS), showed that the domesticated tetraploid wheats have also received a substantial genetic input from wild emmers from the southern Levant. Taking account of archaeological evidence that tetraploid wheats were first cultivated in the southern Levant, we suggest that a pre-domesticated crop spread from this region to southeast Turkey and became mixed with a wild emmer population from the northern Fertile Crescent. Fixation of the domestication traits in this mixed population would account for the allele sharing and GWAS results that we report. We also propose that feralization of the component of the pre-domesticated population that did not acquire domestication traits has resulted in the modern wild population from southeast Turkey displaying features of both the domesticates and wild emmer from the southern Levant, and hence appearing to be the sole progenitor of domesticated tetraploids when the phylogenetic relationships are studied by methods that assume a treelike pattern of evolution.