A classical study by Jones and Macken (1995) revealed that segregating a changing-state sequence of sounds into distinct steady-state sequences emanating from different spatial locations decreases the detrimental effects of irrelevant speech on short-term memory. Since this study is important in asserting that spatial auditory stream segregation determines whether a steady-state or a changing-state percept is formed, and since it has never been replicated by another laboratory, a preregistered replication was initiated. Participants were exposed to mono and stereo versions of task-irrelevant steady-state ('J-J-J…') and changing-state utterances of letters ('V-J-X…') while attempting to memorize a sequence of visually presented letters for serial recall. In the monophonic condition, all sounds were played diotically (i.e. through both headphone channels) resulting in a single auditory stream localized in the center of the head. In the stereophonic presentation mode, by contrast, the sequence was played through the left, right, and both headphones, thereby allocating each letter to a fixed spatial position, thus potentially forming three distinct auditory streams. The results show that while the stereophonic condition significantly attenuates disruption by irrelevant speech, as in the original study, this beneficial effect is not as large as reducing it to the level of a monophonic steady-state control condition, thus suggesting that spatial auditory stream segregation may not have been perfect.