Rice (Oryza sativa L.) is consumed by more than half of the world’s population, but despite its global importance the mechanisms of domestication remain unclear. During domestication, wild rice (O. rufipogon Griff.) was transformed by acquiring non-seed-shattering behaviour, an important genetic change that allowed humans to increase grain yield. However, we show previously identified loci, sh4 and qSH3, are individually insufficient to explain loss of seed shattering nor increases in harvest yield in wild rice. We identify the complementary interaction of key mutations for abscission layer interruption and panicle architecture that were causal in the early domestication of Asian rice. An interruption of abscission layer formation requires both sh4 and qSH3, which presents an apparent barrier to selection of shattering loss. We identified the causal single nucleotide polymorphism at qSH3 within a seed-shattering gene OsSh1 conserved in indica and japonica subspecies, but absent in the circum-aus group of rice. We demonstrate through harvest experiments that seed-shattering alone does not significantly impact yield. Instead, we observed yield increases under a SRR3-controlled closed panicle formation, which is augmented by the integration of sh4 and qSH3 alleles causing a slight inhibition of abscission layer. Complementary manipulation of seed shattering and panicle shape result in a panicle structure that is mechanically stable. We propose a stepwise route in the earliest phase of rice domestication in which selection for visible SRR3-controlled closed panicle morphology was instrumental in the sequential recruitment of sh4 and qSH3 and leading to loss of shattering.Significance StatementRice is one of the most important crops worldwide. Loss of seed shattering in domesticated rice, previously attributed to single mutations such as in sh4, is considered the principal genetic change which resulted in yield increases. However, we show that sh4 is insufficient on its own to cause abscission layer disruption and other genes, such as qSH3 are required, making mechanisms for the initial selection of non-shattering unclear. We show that shattering loss in wild rice genetic backgrounds does not increase yields. We identify an interaction in which a second trait, closed panicle formation controlled by SPR3, both increases yield and facilitates recruitment of sh4 and qSH3 which synergistically augment yield, leading to a stepwise route for rice domestication.