Teosinte, the putative ancestor of maize, is rarely found in early Mesoamerican archaeological deposits, and maize later took a very long time to develop big ears and to become an effective staple grain. These patterns are best explained by the hypothesis of early nongrain use. Consumption of teosinte or maize for stalk sugar or as a green vegetable created reverse bottlenecks at odds with the default assumption that early selection targeted grain. Such bottlenecks were especially potent in small teosinte/maize populations outside their upland habitats. The ultimate products of plant domestication should not be confused with the selection behaviors of ancient people. Such behaviors are sometimes counterintuitive, producing an archaeological record that fails to match the expectations of our reconstructed logic.