We present data from a measurement of angular distributions of the reaction n + p {yields} {gamma} + d (1a) for neutrons with energies between 300 and 700 MeV. These data are compared with recent data for the inverse reaction {gamma} + d {yields} n + p (1b). Time-reversal invariance implies that apart from a normalization factor the angular distributions should be the same. We find an intriguing discrepancy in the vicinity of the well-known peak in the total cross section of reaction (1b) and apparent agreement elsewhere. The discrepancy occurs in the shape of the angular distributions and does not depend on the normalization of our data or those of the inverse reaction experiments.