Sleep Simulation Results

I've blogged on-and-off for the last few years about why I think we sleep and dream. For example, here's the last video I did on this subject (with squinty eyes due to short-term LASIK sensitivity):

I've added a new twist to my theory since last time:

The human sleep cycle is 90 minutes long, and if you map it over a night, it looks a lot like what you'd see if you plotted a 90-minute long sine wave that gradually moves upward after every cycle. The peak of the sine wave would corresponds to REM sleep, which gradually increases in duration as the night wears on. For example:

The area of the brain called the PONS emits waves every 90 minutes during sleep, and there's even some evidence that these waves are generated during the daytime as well. These waves seem to radiate throughout the brain and stimulate various areas.

So I hypothesize that there's a regular 90 minute cycle going on all the time (there are 16 of these cycles in every day) that stimulates neurons. I'm not sure what the purpose of the cycle is, but perhaps it's important to prevent individual neurons from sleeping continuously for more than a few hours? Anyhow, my theory is that the stimulating effect of the 90 minute cycle only becomes apparent when most neurons are sleeping, since at its peak, the cycle causes many neurons to wake up for a while (and as I mentioned in my video, dreams are simply the thoughts of a partial mind).

In lab experiments, they've confirmed that peak PGO waves immediately precede REM sleep, which is exactly what my theory predicts.

To test my theory, I decided to write a little "brain simulator" using Ruby on Rails. The simulator models about 500,000 neurons that gradually build up metabolic waste while they're active and then dump waste while they're sleeping. The probability of a particular neuron being active is determined by the amount of metabolic waste it has accumulated, the activity of its neighboring neurons, the amount of external stimulation (e.g. daylight), and the level of the 90 minute cycle.

Here's the plot that I got (click to see it full-size):

 

The plot starts at 4 pm and the green line shows the metabolic waste level slowly rising. The blue 90 minute cycle has a small effect before night time, and you can see a few neurons drifting asleep before 10 pm. When the external stimulus goes away at 10 pm, the effect of the high metabolic waste combined with a downward 90 minute cycle causes most neurons to go straight into a deep sleep. The metabolic waste then starts to be dumped, and the neurons sleep state is modulated during the night by the 90 minute cycle. The peaks of the cycle correspond to gradually increasing amounts of REM sleep and the depth of each sleep cycle gradually lessons. This is exactly what is seen in experimental results. When external stimulus appears (usually in the form of daylight), enough metabolic waste has been dumped to allow the entire brain to immediately become active.

Overall, I'm very happy with the results, since it reproduces most of the measured effects of sleeping and dreaming from just a few simple rules. The next step for me is to send these results to some sleep researchers to get their opinion!

Comments

[Stephen Downes]

> I'm not sure what the purpose of the cycle is, but perhaps it's important to prevent individual neurons from sleeping continuously for more than a few hours?

I think it's a Boltzmann mechanism of annealing.

[Fantasy Writer Guy]

Can't hear your video yet 'cause I'm in a public place and mindful of courtesy. Your scientific testimony sounds convincing and legitimate to me. From a poetic point of view - I feel like I dream constantly - like it is some inevitable bi-product of having two minds - consciousness and instinct, but that I rarely (but sometimes!) experience my daytime dreaming because it is overwritten by the sights and sounds of the daylight world.

Like a two way mirror where the light may shine on different sides. Know what I mean...