We present a study that tests the ability to remember room acoustics – a cognitive skill that is one of the guiding mechanisms behind plausible virtual acoustics for extended realities. Room acoustic memory was tested by assessing a person’s ability to recognise sound samples, convolved with room impulse responses of everyday rooms presented in a preceding training session. To test a common assumption of detection theory, we conducted two listening tests using both a yes/no and a 2AFC paradigm. Results show that subjects can recognise different rooms above chance level, but even with relatively large differences between the rooms, the accuracy is low in general. Furthermore, the relation between the two test paradigms follows the prediction of detection theory when averaging over all participants, but less so for individual participants.