The development of seismic building codes in El Salvador is described including the incorporation of a probabilistic hazard study into the latest regulations. This hazard study is compared with three other assessments that cover El Salvador and it is shown that discrepancies arise from differences in each step of the hazard evaluation, leading to considerably divergent results. Although it may have been expected that the major uncertainty would result from the attenuation relationships employed, this review shows that the definition of the seismicity model is also poorly controlled. Attenuation relationships for spectral ordinates in Central America are also discussed.