Mineral weathering rates are determined for a series of soils ranging in age from 0.2-3000 Ky developed on alluvial terraces near Merced in the Central Valley of California. Mineralogical and elemental abundances exhibit time-dependent trends documenting the chemical evolution of granitic sand to residual kaolinite and quartz. Mineral losses with time occur in the order: hornblende > plagioclase > K-feldspar. Maximum volume decreases of >50% occur in the older soils. BET surface areas of the bulk soils increase with age, as do specific surface areas of aluminosilicate mineral fractions such as plagioclase, which increases from 0.4-1.5 m2 g-1 over 600 Ky. Quartz surface areas are lower and change less with time (0.11-0.23 m2 g-1). BET surface areas correspond to increasing external surface roughness (?? = 10-600) and relatively constant internal surface area (??? 1.3 m2 g-1). SEM observations confirm both surface pitting and development of internal porosity. A numerical model describes aluminosilicate dissolution rates as a function of changes in residual mineral abundance, grain size distributions, and mineral surface areas with time. A simple geometric treatment, assuming spherical grains and no surface roughness, predicts average dissolution rates (plagioclase, 10-17.4; K-feldspar, 10-17.8; and hornblende, 10-17.5 mol cm-1 s-1) that are constant with time and comparable to previous estimates of soil weathering. Average rates, based on BET surface area measurements and variable surface roughnesses, are much slower (plagioclase, 10-19.9; K-feldspar, 10-20.5; and hornblende 10-20.1 mol cm-2 s-1). Rates for individual soil horizons decrease by a factor of 101.5 over 3000 Ky indicating that the surface reactivities of minerals decrease as the physical surface areas increase. Rate constants based on BET estimates for the Merced soils are factors of 103-104 slower than reported experimental dissolution rates determined from freshly prepared silicates with low surface roughness (?? <10). This study demonstrates that the utility of experimental rate constants to predict weathering in soils is limited without consideration of variable surface areas and processes that control the evolution of surface reactivity with time.