|
| |
|
|
|
|
||||
| Home Help Feedback Subscriptions Archive Search Table of Contents | ||||
First published online October 19, 2007
Journal of Experimental Biology 210, iv (2007)
Copyright © 2007 The Company of Biologists Limited
doi: 10.1242/jeb.001164
Outside JEB |
LEARNING WITHOUT A BRAIN
University of Maryland tytell{at}umd.edu
|
You don't need your brain to walk. You don't even need it to catch yourself after you stumble. And now, it appears, you may not even need it to learn some new skills. A recent report suggests that the spinal cord itself, without the brain at all, is able to adapt to a new environment – and possibly even anticipate how its environment has changed.
Chad Heng and Ray de Leon of California State University in Los Angeles studied neonatal rats whose spinal cords were completely severed at the mid-thoracic level, cutting off connections between the rats' brains and hind limbs. Despite the spinal cord damage, many of the rats spontaneously recovered their ability to walk. This didn't surprise the researchers: it's well-known that the spinal cord contains the neural networks that drive regular walking, as well as reflex pathways that allow it to respond to simple perturbations like tripping or stepping in a hole.
But the researchers wanted to see whether the spinal cord can adapt to more complex effects – in essence, whether it can learn. So they built a system to alter the forces on one hind leg. They connected a small robotic arm to one ankle and programmed it to resist the leg's forward motion with a backward force proportional to its velocity – a `viscous' perturbation, similar to dragging the leg through a vat of honey.
When they turned on the viscous force, it threw the rats' stepping off. At first, they didn't quite manage to put the leg attached to the arm (the `perturbed' leg) on the ground before they tried to lift up the other one, so that the support phases of the two limbs didn't overlap at all. By the second step after the force came on, the support phases started to overlap slightly, and by the sixth step, the rats had recovered normal overlapping support between the two limbs. They still had some problems – steps on the perturbed side were slow and short – but, overall, they were able to produce an effective response to a rather unnatural, complicated stimulus.
How was the spinal cord adapting to the stimulus? The changes might have been some sort of complex reflex – a `feedback' strategy, which involves comparing the leg's position with an internal model of where it should be. Or they might have involved learning – a `feed-forward' strategy, which would mean that the spinal cord was altering that internal model of the leg's motion.
The fact that it took several steps for the rats to adapt suggests a feed-forward strategy. But if learning really occurred, the spinal cord should overcompensate once the viscous force was stopped, taking several steps to relearn the normal force regime. Unfortunately, here the researchers' data are mixed. Four of their 10 rats overcompensated when the researchers turned the force off, overlapping the support phases of the two limbs much longer than normal. The other six rats, though, didn't step at all when the force went off – they tripped – which made it impossible to determine whether they were overcompensating or not. So there weren't enough data for a statistically significant effect, but the trend suggests learning.
Regardless of the mechanism for the adaptation, whether it's a complex feedback pattern or feed-forward learning, Heng and de Leon have shown that the spinal cord is even more complicated than we thought. More than just transmitting information from the brain to the body, more than providing reflex patterns, the spinal cord may cooperate with the brain in learning new environments to produce effective behaviours.
References
Heng, C. and de Leon, R. D. (2007). The rodent
lumbar spinal cord learns to correct errors in hindlimb coordination caused by
viscous force perturbations during stepping. J.
Neurosci. 27,8558
-8562.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
