Wed February 27, 2013
Dr. Samuel Sober, Emory University – How Birds Learn to Sing
In today’s Academic Minute, Dr. Samuel Sober of Emory University reveals how birds listen to themselves to get their songs right every time.
Sam Sober is an assistant professor of biology at Emory University in Atlanta, Georgia. His current research on singing behavior in finches investigates the relationship between neural activity, muscular activation, and task performance by using a range of techniques to describe how neural circuits drive vocal output. He holds a Ph.D. from the University of California San Francisco.
Dr. Samuel Sober – How Birds Learn to Sing
Throughout our lives, our brains use input from the senses to correct our mistakes. During speech, for example, we listen closely to the sounds of our own voices to correct speaking errors. It turns out that songbirds are also experts at listening to themselves when they vocalize, and research in our laboratory uses songbirds to ask how the brain corrects its errors. We fit songbirds with tiny headphones that let us trick the birds into thinking they are making a mistake by changing the pitch of how they hear their own songs. In response, birds change the pitch of their songs to correct the apparent error.
In speech, birdsong, and many other behaviors, the brain has to rely on sensory information to correct errors. However, sensory information is inherently unreliable, because our senses can make errors in reporting information to the rest of the brain. Also, sensory signals are often noisy – imagine trying to hear your own voice in the middle of a noisy party. Therefore, the brain must compute how much to trust the senses when correcting its mistakes.
We recently used our headphone-wearing songbirds to test a mathematical theory of how the brain performs this computation. We hypothesized that the bird’s brain compares each piece of sensory information to the songs it has heard itself sing in the past. This comparison allows the bird to compute the probability that each sound could have come from its own voice and use this information to decide how much to change its song. By using the headphones to present pitch errors of different sizes, we found that high-probability errors – errors that birds are used to hearing themselves make – lead to rapid learning, whereas low-probability errors led to poor learning. These data suggest that birds use a simple mathematical trick to guide learning, and we hope that these results could be used to understanding human speech learning as well.