Magnitude, distance, and amplitude corrections (MDAC) made to observed regional amplitudes are necessary so that what remains in the corrected amplitude is mostly information about the seismic source type. Corrected amplitudes can be used in ratios to discriminate between earthquakes and explosions. However, source effects remain that cannot easily be determined and applied as amplitude corrections, such as those due to depth, focal mechanism, local material property, and apparent stress variability. We develop a mathematical model to capture these near-source effects as random (unknown), giving an error partition of three sources: model inadequacy, station noise, and amplitude correlation. This mathematical model is the basis for a general multistation regional discriminant formulation. The standard error of the discriminant includes the variances of model inadequacy and station noise, along with amplitude correlation in its formulation. The developed methods are demonstrated for a collection of Nevada test site (NTS) events observed at regional stations (see Fig. 1). Importantly, the proposed formulation includes all corrected amplitude information through the construction of multistation discriminants. In contrast, previous studies have only computed discriminants from single stations having both P and S amplitudes. The proposed multistation approach has similarities to the well-established mb versus Ms discriminant and represents a new paradigm for the regional discrimination problem.