Determining the “natural region” of neural activity
Now that is the million-dollar question. If we were to have the hypothetically perfect machine for creating whatever neural activity patterns we want, what should we create?
The challenge is that in our experiments we only ever observe a tiny fraction of the neurons in a brain for a tiny fraction of the animal’s lifetime. So to make mapping out the “natural” activity of neurons a feasible idea, we can do one of two things.
First, to counteract the number-of-neurons problem, we can use an animal with as few neurons as possible, but which still has interesting enough behaviours to be worth studying. Many labs are doing this. There is no agreement on what such an animal should be: a zebrafish? (Simple vertebrate, with the benefit of being transparent when a baby). A turtle? (rich set of motor behaviours that can be accessed easily — and walks). A bee? (Smart, can learn stuff, has 1 million neurons). A leech or a sea-slug? (Thousands of neurons; can swim and crawl, feed and defend) A nematode worm or A hydra? (Hundreds of neurons).
Second, to counteract the fraction-of-a-lifetime problem, the answer is to record from, well, the whole lifetime. Naturally, this goes better if your animal is simple and short-lived. The technical barriers to recording neural activity for a long time are formidable. If you use electrodes, then they move (changing the neurons they record) and the brain rejects them (it grows scar tissue over them, eventually stopping them from working). If you use imaging of dyes injected into the neurons, then the light used to record bleaches the dye over time, and it stops responding to activity (usually on a time-scale of hours). So it is likely that truly long-term recordings will take new types of recording device; or electrodes coated in material that the brain does not reject.
But, hey, we’ll get there eventually. The question then is: what will we do with these peta- or exabytes of brain-wide, lifetime-long neural activity data once we have it?
