Bayesian SED-matched source detection with a single-SED prior, for illustration. The foreground model is that sources have fluxes that lie along a “Red” SED with color g − r ≃ 1. Left: probability contours for the foreground model are inclined in the direction of the prior. The flux prior prefers small fluxes but has some probability mass extended toward larger fluxes. The background model (null hypothesis) is that there is no source, only noise; these contours are circles centered on zero. The probability contours are spaced in powers of 10. Right: probability ratio contours for the foreground divided by background model. When performing source detection, these lines mark decision boundaries; points above and to the right of a decision boundary are considered to be detected, at increasing significance levels. The contours are spaced in powers of 10.