- Author
- Posted by izhak, with replies from many commenters
- Published
- 30 October 2025
- Kind
- Hacker News discussion thread about a blog article on hearing
- Original
- news.ycombinator.com/item?id=45762259
- Subject
- Audio models and analysis
- Themes
- What the senses do
What it is
The thread is on an article at dissonances.blog. Its web address says the ear does not do a Fourier transform. The thread had 484 points when collected. Replies come from engineers, musicians and a self-described neuroscientist.
What it says
- One commenter summarizes the article: the ear does a time-localized transform between wavelet and Gabor.
- Several commenters call the title clickbait or a strawman.
- Commenter anyfoo says a strict Fourier transform is continuous and infinite in time.
- Commenter fat_cantor cites a 2002 paper: the optimal code varies by sound class.
- Commenter antognini says the ear senses phase information that spectrograms often discard.
- HarHarVeryFunny says the cochlea favors frequency detail when low and timing detail when high.
- The thread repeats the article's idea that speech fills an unoccupied time-frequency niche.
What evidence it gives
The thread offers opinions, textbook definitions and a few citations. Commenters name a 2002 paper, a hearing-model book and other links. A self-described neuroscientist describes how hair cells fire. No commenter runs an experiment.
Interests and limits
Comments are unverified and often disagree. Many concern naming and definitions, not hearing itself. The linked article was not read for this page. Its claims come through commenters' summaries.
Connections
These links are this wiki's own reading, not claims the sources make.
- Resonate. The thread argues over whether the ear really performs a Fourier transform. Resonate updates a bank of resonators on each sample, with no window. The page makes no claim about ears. Still, the thread's account suggests testing it as a model of listening.