Lost in music - how birds learn songs

Songbirds stop listening to themselves when they start to sing...
04 March 2014

Interview with 

Richard Mooney, Duke University


Songbirds stop listening to themselves when they start to sing.


Within the brain of finches, the premotor cortex contains nerve cells that appear to be crucial for birds producing and learning the song they sing.  Recording from finch brainCritically, the auditory input to these cells is switched off when the bird sings itself; otherwise, owing to delays in processing, the auditory signals would arrive after each part of the song has been sung, as Duke University scientist Richard Mooney explains to Chris Smith...

Richard -   A longstanding problem we are interested in solving is how the brain uses auditory information to guide vocal learning.  In the region of the bird's brain we were recording from, neurons receive input from auditory parts of the brain.  In fact, it can show that when the bird is quietly resting, those nerve cells will respond to complex auditory stimuli such as the bird's own song.  But those same neurons are also active when the bird sings.  And so the question was, are these neurons actually integrating auditory information as the bird sings, or are they flipping back and forth from an auditory function to a motor function as the bird goes from listening to singing?  The latter turned out to be the case, but we had to go a long way to find out that answer because it's a really tricky problem to answer in a singing bird.

Chris -   Yes, indeed.  So, did you record and show that there was a sort of temporal relationship in terms of the activity in those sites to prove that?

Richard -   Yes.  So in fact, the recordings involved were recordings from the same nerve cell in both states.  Recording with a method that is really, really challenging to use, even in a cultured neuron that's sitting in a dish, and that is something called intracellular recording, where we use a really, really fine glass micropipette.  When I say 'fine', the tipped diameters are maybe a 50th of a micron. We can use these fine glass pipettes to poke inside of a small nerve cell and we can record the little chatter of connections that are made under that nerve cell, from auditory and motor regions.  So, we can actually probe whether the auditory inputs are active when the bird was singing.  That's the real breakthrough in this study.

Chris -   So, in other words, it is the bird listening to itself as it sings.

Richard -   Well, we certainly know from behavioural experiments that it does listen to itself.  But what we didn't know is whether or not these nerve cells, which seem so well-suited to do that kind of self-monitoring, are actually listening as the bird sings.  And they're not. Even though we can show they get auditory input when the bird is quietly resting and listening to its song played through a speaker, those nerve cells will get excited.  But when the bird sings, the cells are active again, but the auditory input is switched off.

Chris -   So, what do you think the point is, of having this feedback loop if most of time it's suppressed when sound is coming out?  Are the modifications based on audition to those cells made when there is not a motor programme being executed?  Is that how you think the system works?

Richard -   So, there are two parts to that answer.  I mean the first is that we think these nerve cells are doing something more than just controlling what the bird sings, but also, helping the bird to identify and recognise the songs of other birds that have songs very, very similar to its own song.  So, that's part of the answer.  The other kind of curious twist is that these nerve cells, their auditory properties are clearly shaped by what the bird hears.  Over development, they become tuned to what the bird sings.  And other birds' songs that are very similar to the bird's own song will excite those nerve cells.  Because the animal learns how to sing, that tuning property must reflect some kind of learning.  So, they're clearly responding to auditory experience, but when the bird is singing, the auditory input is turned off.

Chris -   I read a study previously where scientists were looking at basketballers and they showed members of the audience a ball leaving a thrower's hand.  Some of those audience members were professional basketballers.  Others were coaches, others were just members of the public.  The professional basketballers they found were effectively executing in their mind's eye, the same throwing manoeuvre through their motor cortex that the pro was doing on the basketball court to make predictions.  They were getting it right whether or not the ball would go in the basket about 90% of the time, compared to no better than chance for the coaches and members of the public who weren't doing that.  Do you think you've got a similar situation here where these birds are effectively executing a sensory stimulus through a motor circuit to ask, "Would my song sound like this?"? And in this way, they are marrying up their song with what the other birds like them, songs sounds like and doing a different comparison.

Richard -   I bet it is.  Just like that. This is a very, very powerful idea and there's a so-called motor theory of speech perception that listening to speech sounds engages the motor circuitry in our brain that produces similar speech sounds and that forms the basis of perception.

Chris -   And if we were to put the whole thing together, can you sort of build us a model of actually how you see all this fitting together?

Richard -   Right.  So, in the region that we're looking at, we think that this is a critical node for producing song and you can demonstrate this; by switching this brain region off chemically with a local anaesthetic, the bird will stop singing.  So, we know that it is essential for the bird to make a song.  What we guess, is that this song motor region produces a copy of the motor command and sends it into the auditory system and that copy contains information about what should happen.  If the two signals are highly congruent then nothing happens.  But if there's a mismatch, then the brain can say, "I made a mistake" and can learn from it.


Add a comment