Effects of likelihood, prior, and response noise on predicted response patterns as a result of modeling the directional localization of broadband noise bursts. (a) Likelihood obtained by sensory evidence (i.e., no spatial prior and no response noise). (b) Bayesian inference (with the spatial prior but no response noise). (c) Full model (with prior and response noise). Gray: actual data of NH16 from [23]. Black: estimation obtained by the model considering spectral gradient profiles (GPs). Red cross: frontal position. Blue dashed lines separate regions of front-back confusions.