A model of spatial release from masking in reverberant environments has proved successful in predicting a wide range of empirical data, particularly for noise interferers or where there are multiple interfering voices. Here, we show how the model can be used to successfully predict optimal behaviour in such environments. Head orientation of around 30° away from the target speaker was predicted, and then shown empirically to improve speech intelligibility for normally hearing listeners, hearing-impaired listeners and cochlear implant users. This degree of head orientation was also shown to be compatible with lip-reading and robust to reverberation. However, location within a room, even one uniformly distributed with interfering sound sources, was predicted to have a strong influence on the effective signal-to-noise ratio and on the benefits of head orientation and directional microphones. Certain locations were predicted to bring all these benefits together and others to largely confound them. These differences were confirmed empirically for the benefits of signal-to-noise ratio and head orientation using virtual acoustics.