We present U, Th, and Pa isotope data for young lavas from Costa Rica and Nicaragua in the Central American arc. Thorium isotopic ratios for Costa Rica and Nicaragua differ dramatically: Costa Rican lavas are characterized by low (230Th/232Th) (1 to 1.2) and, for four out of five lavas, (230Th/238U) greater than unity. Nicaraguan lavas have high (230Th/232Th) (2.2 to 2.7) and, for five of six samples, (230Th/238U) less than unity. All lavas have (231Pa/235U) greater than unity, with initial values ranging from 1.27 to 1.77, but those from Costa Rica have larger 231Pa excesses. There is a broad positive correlation between (231Pa/235U) and (230Th/238U) similar to the worldwide trend for arcs outlined by Pickett and Murrell (1997), although many of the Nicaraguan lavas skirt the high end of that trend. In greater detail, the Central American data appear to divide into separate high-(231Pa/235U) and low-(231Pa/235U) tiers. These tiers may be different because of either different residence times in the crust or different proportions of sedimentary components from the slab. Substantial (231Pa/235U) excesses (>1.5) in both Costa Rica and Nicaragua require a melting process that allows for enhanced daughter (231Pa) ingrowth. With increasing U addition, (231Pa/230Th) increases in a manner that cannot be explained adequately by aging of fluid components before partial melting and eruption. Thus, either some 231Pa is added from the slab, or melting-enhanced 231Pa ingrowth is greater in sources that have experienced a larger amount of slab-derived flux and a higher extent of melting. These observations can be explained if regions that have undergone greater extents of fluxing and melting have experienced these processes over a longer time interval than those that have had little flux added and little melt extracted. We propose a flux-ingrowth melting model in which corner flow in the mantle wedge supplies fresh hot mantle into a zone of slab fluid addition. Partial melting occurs in response to this fluxing. We assume critical melting at low porosity (1̃0-3), rapid fluid flux to the melting region, and rapid melt transport. Solid mantle traverses the melting region over 105 to 106 yr, thereby allowing 231Pa and 230Th ingrowth from U retained in the residues of melt extraction. Magmas are aggregated from all parts of the melting regime, mixing melts from incipiently fluxed regions with those from sources that have experienced more extensive fluid addition, partial melting, and daughter nuclide ingrowth. With suitable assumptions about component addition from the slab, this flux-ingrowth model matches a wide range of U-series and trace element data from Costa Rican and Nicaraguan lavas, with required average extents of melting of 1̃ to 3% and 7 to 15%, respectively. Upwelling and/or extensive melt-rock reaction are not required to explain large (231Pa/235U) excesses in Central America or other arcs. On Th isotope equiline plots, the model produces linear arrays that resemble isochrons but that have no age significance. Instead, these arrays are generated by mixing of melts from sources that have experienced fluid addition and partial melting over a range of time intervals, as seems likely in arc source regions. Finally, the flux-ingrowth model predicts considerable 226Ra excesses for integrated magmas. If we assume that 226Ra is added continuously with the slab-derived fluid, the model predicts large and increasing (226Ra/230Th) with increasing melting and slab-component addition, without requiring the addition of a distinct late fluid.