The dog, Canis familiaris, has classical domestication attributes, possibly derived from selection for tameness, except one: the capacity to share a long gaze (30+ seconds) with a human. All dog breeds have this capacity, even puppies prior to human socialization. No other domesticate can do so. Only Great Apes, among over 5000 mammals, can. Shared gazing replaces cortisol-based fight/flight hormone response with an oxytocin governed phobia/philia response. It involves mutations in neural hormone receptors, iris visibility, periorbital muscle control, and mate selection. Shared gazing contradicts domestication induced adolescent behavior fixation, as it requires apparent cognitive control of distractions. Domestication usually amplifies and suppresses existing traits. Humans still cannot induce complex beneficial mutations in animals. Therefore shared gazing likely evolved prior to domestication. Recent genomics evidence, comparing contemporary dogs and other canids, as well as sequencing archaeological canid remains, pushes back dog origins between 20,000 and 200,000 years BP. No evidence of human domestication efforts exists prior to 15,000 years BP. I review the genetic, archaeological, and experimental evidence of dog origins, which reveal tension in a dominant explanatory paradigm. Many researchers assume dogs evolved only through artificial selection. Evidence of ancient origins requires positing a far more ancient domestication activity than seems reasonable, or else reframing (or dismissing) evidence. Neither is necessary if 1) dogs evolved prior to domestication, 2) this pre-domesticated canid had a shared-gaze capacity, and 3) during the period of hunter-gatherer population increase, this capacity was exploited because it helped humans develop emotional attachment skills. Hunter-gatherer reports suggests the shared-gaze is key to canid domestication. Others have suggested this dog capacity might be a factor in their domestication. This paper's hypothesis makes it central to the process.