A new model for the earliest stages in the evolution of subduction zones is developed from recent geologic studies of the Izu-Bonin-Mariana (IBM) arc system and then applied to Late Jurassic ophiolites of Cailfornia. The model accounts for several key observations about the earliest stages in the evolution of the IBM system: (1) subduction nucleated along an active transform boundary, which separated younger, less-dense lithosphere in the west from older, more-dense lithosphere to the east; (2) initial arc magmatic activity occupied a much broader zone than existed later; (3) initial magmatism extended up to the modern trench, over a region now characterized by subnormal heat flow; (4) early are magmatism was characterized by depleted (tholeiitic) and ultra-depleted (boninitic) magmas, indicating that melting was more extensive and involved more depleted mantle than is found anywhere else on earth; (5) early arc magmatism was strongly extensional, with crust forming in a manner similar to slow-spreading ridges; and (6) crust production rates were 120 to 180 km3/km-Ma, several times greater than for mature arc systems. These observations require that the earliest stages of subduction involve rapid retreat of the trench; we infer that this resulted from continuous subsidence of denser lithosphere along the transform fault. This resulted in strong extension and thinning of younger, more buoyant lithosphere to the west. This extension was accompanied by the flow of water from the sinking oceanic lithosphere to the base of the extending lithosphere and the underlying asthenosphere. Addition of water and asthenospheric upwelling led to catastrophic melting, which continued until lithosphere subsidence was replaced by lithospheric subduction. Application of the subduction-zone infancy model to the Late Jurassic ophiolites of California provides a framework in which to understand the rapid formation of oceanic crust with strong arc affinities between the younger Sierran magmatic arc and the Franciscan subduction complex, provides a mechanism for the formation and subsidence of the Great Valley forearc basin, and explains the limited duration of high-T, high-P metamorphism experienced by Franciscan mélanges.