All parts of an audio processing chain introduce latency. Previous studies have shown that high audio latency may negatively impact human performance in different scenarios, e.g., when performing live music or when interacting with real-time human-computer systems. However, is not yet known where the human perception threshold for audio latency lies, i.e., what the lowest amount of latency is that musicians might notice. Therefore, we conducted a user study (n=37) using the PEST method to estimate the just noticeable difference (JND) for audio latency under different base latency settings.
Our results suggest that base latency influences the perception threshold in a non-linear manner: Participants achieved a mean JND of 49 ms for a base latency of 0 ms, 27 ms for a base latency of 64 ms, and 77 ms for a base latency of 512 ms. Furthermore, the JND was lower for participants with high musical sophistication.