Explosive data from a carefully controlled transmission‐loss measurement in the Northeast Atlantic Ocean are used to evaluate, in detail, the capability of the AESD parabolic equation model. The analysis is performed on a path‐by‐path basis by comparison of computed (using a ray‐tracing program) and measured travel‐time differences. Sensitivtiy studies show that residual discrepancies could be attributed to uncertainties in source detonation depth and the assumption of piecewise‐linear bathymetry. The data show an arrival with no ray‐tracing counterpart; evidence, based upon its mean speed of propagation and the frequency dependence of its energy content, is presented to argue that it is an arrival ducted in the main sound channel. The behavior of this arrival with range is determined by the range dependence of the sound‐speed environment.Subject Classification: [43]30.20; [43]20.20.