In psychoacoustics there is an increasing demand for more realistic testing environments that better capture the real-world abilities of listeners and their hearing devices. However, there are significant challenges involved in controlling the detectability of relevant target signals in realistic environments. We conducted an extensive detection study in a simulated real-world environment to understand some of the important dimensions influencing detection. A multi-talker cafeteria scene was generated using room simulation software and played back by means of a 3-D loudspeaker array. Detection thresholds for the target word "two" were measured adaptively for 8 different target directions in the horizontal plane. Performance was then measured for fixed signal-to-noise ratios around these thresholds to obtain a psychometric function for each direction. To examine the effect of target-location uncertainty, psychometric functions were also measured with randomized target directions. Detection thresholds depended on the target direction, consistent with changes in signal-to-noise ratio caused by the head shadow. Target-location uncertainty increased thresholds globally by a small amount. These findings provide a framework for controlling the detectability of target sounds in future experiments aimed at measuring localization, identification, awareness etc. in realistic listening environments.