| The art of neural networking (cont) Raised in Northern California by hippie parents—she spent some years living in a teepee without electricity—Turrigiano had a nontraditional education. “My parents,” she explains, “were bohemian intellectuals. They read lots of books, and I grew up surrounded by books. So I was reading novels by the time I was seven, but when I first went to school in fourth grade, I didn’t know how to add a column of numbers. It was a spotty education.” Growing up in the woods, though, had an early and profound, if not quite realistic influence on her career. Turrigiano explains: “From the time I was about nine, I knew I wanted to be a biologist. But at that point, I thought what biologists did was go out in the wild, track wolves, study animal behavior, and things like that, not sit in labs and do these reductionist experiments. But in any case, the love of biology and the natural world was born quite young, and I used to spend a tremendous amount of time, as a kid, poking around tide pools and canoeing up rivers.” Reed College saw the potential in the staggering SAT scores and the unconventional transcripts, eventually graduating Turrigiano with a B.A. in biology. She went on to earn a doctorate from the University of California–San Diego, marry Nelson, give birth to their first child, and seek to resolve the seemingly impracticable matter of finding a place where two neuroscientists might find jobs within reasonable proximity to each other. Unfortunately, their initial prospects would have put them on opposite coasts, their son hovering somewhere in between. Brandeis, however, offered Nelson a position, and with its hallmark pioneering spirit of improvisation, its eye for talent, and its nose for a bargain, took Nelson and Turrigiano up on their unorthodox offer to share the position—50-50. Two neuroscientists for the price of one. Two years later, as though by some administrative form of cell division, each had acquired her and his own full-size status, and one of them . . . so far . . . has won a MacArthur.  Cliff Hauptman is
director of publications at Brandeis and editor of the Brandeis
Review.
|
 |
|
|
|
|||
Homeostatic
plasticity in neuronal networks explained
There are two incredible truths that strike anyone who has raised children. The first is that their personalities seem apparent almost from birth—shy or gregarious, willing to eat anything or refusing any but three or four foods of a particular color. No amount of parental anguish seems to alter these essential personality traits. The second truth is that they learn at an astounding rate. Clearly their brains are changing daily and in dramatic ways. To a neuroscientist, this raises a paradox; how can there be so much stability in the face of such rapid and profound changes in the basic wiring of the nervous system?
It would seem that there must be mechanisms that promote stability in the properties of neuronal circuits, just as there must be mechanisms that promote change. But for a surprisingly long time, neuroscientists studying development and learning have focused almost exclusively on mechanisms that progressively alter the properties of neuronal circuits and have largely ignored mechanisms that act to keep circuit properties stable.
Our brains are composed of billions of neurons that are electrically active and are connected to each other via synapses—electrochemical connections that allow one neuron to excite or inhibit the activity of other neurons. Any given neuron may receive up to 100,000 such synapses, and our abilities to think, feel, and remember are due to the particular patterns of connectivity between these individual neurons. Most neuroscientists believe that learning occurs through changes in the strength of the synaptic connections between particular neurons—if a synapse increases in strength, one neuron will be able to excite another neuron more strongly, and communication along that pathway in the brain will be enhanced.
Fifty years ago Donald Hebb suggested that such changes in synaptic strength might occur as a function of how correlated the activity of any two neurons are: if there are synaptic connections between two neurons and they are both active at the same time (that is, their activity is correlated), then that connection should be strengthened. For example, if a child touches a hot stove, the neurons that respond to the sight of the stove and the neurons that transmit the sensation of pain will both be active at the same time, and the synapses between them will be strengthened. In other words, an association between the qualities “stove” and “hot” will form in the child’s brain. This simple learning rule has tremendous power, because it allows a network of connected neurons to store information about many different correlations, or associations, in the environment.
But “Hebb rules” also have a dark side to them, because they are essentially unstable. If two neurons are active at the same time, this rule will strengthen the connection between them. Now one neuron will excite the other more strongly, and the correlation in their activity will go up. When the correlation goes up, the connection between them will be strengthened again, and the correlation will go up again, and so on, and so on, until the connection reaches some maximum value. You can see that this is problematic, because if correlations tend to drive all synapses to their maximum values, then the differences between synaptic strengths that encode information about the world will be obliterated; to the child, everything will become a hot stove.
