In this page, we illustrate examples of "neuronal melodies", where the simplest translation from spikes to music was used, namely each neuron produces its own note, at the moment it fires. It is surely possible to use more sophisticate ways of translating spikes to music, such as for example using more complex rules of harmonics and associate neuronal activity with more complex musical sequences; see the Spikiss Page.
Human medial-temporal cortex
80 to 100 neurons were simultaneously recorded from human temporal cortex from patients prior to neurosurgery. We succeeded in identifying excitatory and inhibitory neurons (see details in Peyrache et al., Proc. Natl. Acad. Sci. USA, 2012). The recorded spikes were converted to MIDI format, by associating each neuron to a given tone, and triggering the tone each time this neuron fired. The MIDI files were then converted to MP3 using freeware programs.
The "melody" produced by neuronal spikes gives an idea about the distributed firing activity of those neurons. MP3 files were generated for different cases: when the subject was awake ("Awake Melody"), during slow-wave sleep ("Sleeping Melody") or during REM sleep ("Dreaming Melody"). It was also done during an epileptic seizure (same patient), starting with awake type of activity then the switch to epileptic activity can be heard very well.
The different audio files available are:
The wakefulness can also be viewed as an animated video file, where the colors represent the LFP, the crosses represent the excitatory (FS) neurons, and circles represent the inhibitory (RS) neurons:
In these files, the time base is real time; all recordings are from the same subject, same electrodes and same set of recorded neurons (each recording lasts one minute). Two instruments are present, a woodblock for excitatory neurons and a xylophone for inhibitory cells. From these melodies, one can hear that the distributed firing activity is almost identical during dreaming compared to wakefulness, which emphasizes the high similarity between these two different brain states (see also Destexhe, Curr. Opinion. Neurobiol. 2011). Slow-wave sleep can be heard as an activity similar to wakefulness ("Up" states), with regular "pauses" of the firing activity ("Down states", which are associated to the slow-waves). During the seizure, the activity is very different and the melody produced is clearly impoverished...
Cat parietal cortex
The same procedure was followed for cat experiments. In this case, 8 multiunit recordings were obtained with a system of 8 pairs of tungsten microelectrodes. Spikes were extracted using the BrainWave software. They were converted to MIDI, by associating each neuron to a given tone, and triggering the tone whennever this neuron fired. The MIDI files were then converted to MP3 using freeware programs. The music scores were generated by importing the MIDI files into the "Guitar Pro 5" program.
MP3 files were generated for 4 cases: when the animal was awake (Wake-Neurons), during slow-wave sleep ("Sleeping-Neurons") or during REM sleep (REM-Neurons), where most dreams occur. The file "Poisson-Wake" is a randomly-generated stream of notes with the same statistics as for "Wake". Interestingly, the firing of one isolated neuron during wakefulness is undistinguishable from that of random (Poisson) activity (compare the audio file generated by one neuron during Wakefulness with that generated by a Poisson spike train with same statistics). However, the distributed activity (ie, the "melody") of several neurons is clearly different (listen to the difference between Wake Neurons with its Poisson equivalent, Poisson Wake - see below). This suggests that what makes our brains non-random is not in the firing pattern of individual cells, but it lies in the respective timing of the firing activity of different neurons...
The different audio files available are:
(in all audio files, the time base is four times slower than real time; all recordings are from the same experiment)
(This type of copyright is similar to the well-known GPL licences for software)
Unité de Neurosciences, Information & Complexité
back to main page