#auditoryscience #musicscience #neuroscience

These complementary computational strategies highlight the flexibility of hearing in its most important task: using traces of the past to better deal with the present – even when the “past” is just a particular patch of rain at the start of a song.

Here, we suggest an additional trick your auditory system could use: randomly sampling a few auditory features in time, to create a compact but distinctive “fingerprint” for the memorised sound.

Storing all details of complex, stochastic sounds would be biologically unreasonable. Earlier work showed that “summary statistics” – long-term averages of sound features – are an efficient way to identify broad classes of natural textures.

This memory trace was highly specific to the repeated recording: it distinguished that particular texture from many other recordings coming from the same physical source (same type of fire, or water...).

This raises a deeper question: what kind of internal representation is being stored in memory?

The key is repetition. As with artificial sounds previously studied (like noise bursts and clicks), the re-occurrence of a natural texture was enough to create a memory trace.

Hi all! Can I be added please?

Hi there, Please add me!