The applicability of sum‐coincidence measurements of two‐step cascade &ggr;‐ray spectra to the determination of photon strength functions at intermediate &ggr;‐ray energies (3 or 4 MeV) is discussed. An experiment based on thermal neutron capture in Nd was undertaken at the Brookhaven National Laboratory High Flux Beam Reactor (BNL HFBR) to test this model. To understand the role of various uncertainties in similar experiments a series of model calculations was performed. We present an analysis of our experimental data which demonstrates the high sensitivity of the method to E1 and M1 photon strength functions. Our experimental data are in sharp contradiction to those expected from an E1 photon strength distributed according to the classical Lorentzian form with an energy invariant damping width. An alternative distribution of Kadmenskij et al., which violates Brink’s Hypothesis, is strongly preferred.