A scientific test of 440 Hz vs 432 Hz
Over the summer I wrote a blog post about people who like to detune music from standard concert pitch (440 Hz) to a lower tuning frequency (432 Hz) claiming it sounds better. Search for “432 hz” in YouTube and you’ll find plenty of examples where people have applied a pitch shifter to alter music ranging from a Mozart Requiem to Oasis’ Wonderwall.
But some claim that there is more to this than just preference. There are several blogs that claim that using a normal concert tuning of 440 Hz is bad for our health, but then what do you expect from something first promoted by the Nazis?
I decided to put this to a test with a web experiment. I couldn’t test the health effects, but it is straightforward to test for preference, which is what seems to the motivation for most of the YouTubers.
The experiment had ten pieces of music chosen from the Free Music Archive. Pieces were chosen which have a very different ‘feel’. Each of these was pitch shifted to simulate seven tuning frequencies of 416, 424, 432, 440, 448, 456 and 465Hz. So far people have listened to 1396 pairs of clips, where for each the listener simply says which one they preferred. That is a couple of hundred people doing the test. Thanks to all those who did it.
A higher mean score shows a frequency that was liked more. Changing the frequency has a significant effect. (This was statistically confirmed using a Kruskal-Wallis test, p<0.001). Pitch shifting the music to a higher pitch made the music worse (all p<0.03, Mann-Whitney U test). Pitch shifting the music to a lower pitch did not make any difference to preference.
People may think that music sounds better at 432 Hz and therefore applying a pitch shifter to their favourite tunes will improve quality, but for people who took part in my experiment this wasn’t true. 432 Hz and 440Hz were rated with equal preference. This doesn’t surprise me, because when we hear a melody it is mostly about relative pitch. I suspect that pitch shifting to higher frequency makes thing worse because of artefacts introduced by the pitch shifting plug-in I used. I wonder if the result might have been different for those with absolute pitch?
You can still try the experiment, it is the one labelled “Two Snippets of Music…” on www.sound101.org.
The data from the experiment can be downloaded from here.
 I took the songs from this paper: 1000 songs for emotional analysis of music doi>10.1145/2506364.2506365 I then grouped the songs into 10 clusters according to the emotional responses to the music that had been measured in the study reported in the paper. I used a k-means algorithm on the arousal and valence score. I then took the songs nearest to the centre of each cluster which had a license that allowed me to use it in the experiment.