<h1>Replacing the Orchestra? – The Discernibility of Sample Library and Live Orchestra Sounds</h1> <h2>Reinhard Kopiez, Anna Wolf, Friedrich Platz, Jan Mons</h2> <h3>Abstract</h3> Recently, musical sounds from pre-recorded orchestra sample libraries (OSL) have become indispensable in music production for the stage or popular charts. Surprisingly, it is unknown whether human listeners can identify sounds as stemming from real orchestras or OSLs. Thus, an internet-based experiment was conducted to investigate whether a classic orchestral work, produced with sounds from a state-of-the-art OSL, could be reliably discerned from a live orchestra recording of the piece. It could be shown that the entire sample of listeners (*N* = 602) on average identified the correct sound source at 72.5%. This rate slightly exceeded Alan Turing's well-known upper threshold of 70% for a convincing, simulated performance. However, while sound experts tended to correctly identify the sound source, participants with lower listening expertise, who resembled the majority of music consumers, only achieved 68.6%. As non-expert listeners in the experiment were virtually unable to tell the real-life and OSL sounds apart, it is assumed that OSLs will become more common in music production for economic reasons. The complete research article has been published at PLoS ONE and can be found for free (open access) at: [http://dx.doi.org/10.1371/journal.pone.0158324][1] ![Figure 1: Results.][2] **Fig 1. Results from the auditory discrimination task between orchestra sample library and live orchestra recording.** (a) Histogram of the overall sensitivity (N = 602) in the discrimination of orchestra sample libraries (OSL) and live orchestra recordings (LOR) in a single-choice paradigm. The dashed line represents the mean discrimination performance, whereas the value of 0 indicates discrimination at chance level. (b) Distribution of response bias with a mean close to 0. Negative values indicate answering response in favor of the OSL; the positive values indicate answering response in favor of the LOR. (c) Discrimination performance for groups of low sound-discrimination expertise (non-musicians, amateur musicians, musicologists, and music teachers) and high sound-discrimination expertise (orchestra musicians, audio engineers, conductors, composers, and arrangers). (d) Correct response rates (hits and correct rejections) for the total sample (72.5%), the subgroups of low vs. high sound discrimination expertise (68.6% vs. 80.0%), and Turing's criterion of 70% for correctly identifying the sound sources to prove AI. Only the group with low sound-discrimination expertise was “cheated” more easily by the samples in that they could not identify correctly the sound source above a rate of 70%. ---------- <h3>Citation</h3> Kopiez, R., Wolf, A., Platz, F., & Mons, J. (2016). Replacing the Orchestra? – The Discernibility of Sample Library and Live Orchestra Sounds. *PLoS ONE 11*(7): e0158324. [doi:10.1371/journal.pone.0158324][3] [1]: http://dx.doi.org/10.1371/journal.pone.0158324 [2]: https://mfr.osf.io/export?url=https://osf.io/bfq5t/?action=download%26direct%26mode=render&initialWidth=709&childId=mfrIframe&format=1200x1200.jpeg [3]: http://dx.doi.org/10.1371/journal.pone.0158324