I read a study that claims to have debunked the myth that only children can learn absolute pitch, and got 12 musicians who’ve not previously had absolute pitch to improve significantly at having absolute pitch.

On average, they spent 21.4 hours over 8 weeks, making 15,327 guesses. All learned to name at least 3 pitches with >90% accuracy, having to respond in under 2.028 seconds; some learned all 12. The average was 7.08 pitches learned.

Notably, the results on the new instruments were worse than on the instruments they were trained on, suggesting people can somewhat learn to rely on the cues from the specifics of the used instrument’s timbre:

The way it works is simply by having very short feedback loops. You hear a sound (played on a piano in the study) and have 1-2 seconds to make a guess for what pitch it is.

You learn new pitches gradually: first, you need to identify one (and press keys for whether it’s that pitch or some other pitch), and then, more pitches are gradually added.

In the study, before testing without feedback, to reset relative pitch memorization, a Shepard tone is played for 20 seconds. (It’s an auditory illusion that makes you feel like the pitch is perpetually getting lower or higher.)

I asked an LLM to make a web app version of it. I asked it to additionally use the Shepard tone more often for a shorter amount.

I also asked it to add colors to maybe produce some amount of synesthesia. I think there’s research that shows that synesthesia and absolute pitch correlate; I don’t know whether it can be induced to some extent, or would only be helpful for some people, but it seemed good to add in case it works. Later, someone on Twitter told me that they were taught the tones of Mandarin using colored cards, and it worked for them. People who experience synesthesia to at least some extent might have an easier time learning more pitches, though I’m not sure if it would be helpful to others.

I tried to minimize the time between recognition and feedback, so the web app reacts to the starts of the key presses, clicks, and touches, not to their ends; and immediately shows whether you were correct, and what was correct.

Finally, I added more instruments than just piano, hopefully, for better generalization.

With the first version, I posted it on Twitter:

It got a surprisingly high amount of engagement, which made the post a bit unfortunate in retrospect, because I made it before I actually fixed the bugs produced by the LLMs (now all fixed); on the other hand, the engagement meant that now I actually had to fix the bugs for people to be able to use the tool.

Two people shared that they have already learned to identify three pitches!

I now want to do experiments with a bunch of things (including the order of pitches presented: can it improve the learning curve and allow people to learn more than three more easily?), to collect the data on people’s progress, and maybe ask them questions (like whether they’ve played music or sang before).

Would appreciate recommendations for how to collect the data well without having to do anything complicated to manage it.

Would also appreciate more ideas for how to improve it for better pitch learning.

If you want to try to acquire perfect pitch, it might take you quite some time, but try it:

perfect-pitch-trainer.pages.dev