NLC 2010 talk: Birdsong as model of language (via R. Kluender)

13 November 2010

This weekend was the 2nd annual meeting of the Neurobiology of Language Conference – NLC this year held at the Rancho Bernardo Inn, San Diego, November 11-12.  In addition to paper talks and posters, there were two (invited) lively debates and two invited talks. This blog piece is Prof. Robert Kluender’s (Linguistics and IDP Cogsci, UCSD) notes on one of them.

 “Evaluating the strengths and limitations of birdsong as a model for speech and language”  by Daniel Margoliash (Neurobiology of Language Conference, San Diego, November 12, 2010)

Potential pitfalls:

scala naturae  (vs. The Tree of Life)

Medieval version: 

God--> angels --> demons --> man --> woman --> animals --> plants --> minerals

“The ice cream theory of brain evolution” (David Linden: just add a scope for each cerebral addition/innovation)


The only way you can talk about humans as the “most advanced” is as the “most derived”

The Tree of Life encompasses all the life forms that exist as well as all the life forms that previously existed, and that’s a lot!

So what if I want to study language and there are no extant primates with vocal behavior that complex (i.e. that don’t show signs of vocal learning)?

We can look at other species and ask questions about common mechanisms.

Language is highly dependent on human motor and perceptual abilities (i.e. derived mammalian characteristics).

And in birds we can find parallel structures.

And this is what evolutionary biologists do all the time, it’s a common research strategy.

It’s easy to fall into the trap of thinking that humans are unique and can do all of these special things that non-humans can’t do (e.g. tool use in animals, Caledonian crow manufactured tool use in particular).

If we look at a broad range of animals we see evidence for a wide range of cognitive skills.

Corvids (crows and ravens) are highly derived species themselves,

So we need to be careful with a linear model of human evolution (both neurally and cognitively)

Early views of the organization of avian forebrain: birds don’t have a cortex, but only what came to be called the striatum

The pallium is arranged according to several interconnected nuclei

Harvey Karten (UCSD, Neuroscience) proposed that these regions within avian forebrain are related to layers of cortex in mammals, and that there is a common neural circuitry in birds and mammals (common forebrain plan, the “Karten hypothesis”)

Cortical layer markers correspond to where there are thalamic inputs in the avian forebrain

Karten: laminar organization in the chicken forebrain

Andy Bass: vertebrate forebrain organization has been preserved for millions of years since the split from fishes

So be careful of using the term “bird brain”!

If we’re going to talk about language, we have to talk about the unique specialization that we see in humans, and how can we do that from an evolutionary perspective if we just don’t know that much about it?

To explain human behavior, we need to address specialization and uniqueness

Since we don’t know the mechanism, we need to use first principles like Occam’s razor (and not cut ourselves in the process)

Specialization is a common feature of evolution

Weakly electric fish can “jam” each other, and so have developed a jamming avoidance response: one fish turns up the frequency, the other dials it down

It’s a simple mechanism, but how they do it is amazing and fantastic – it’s a very complicated operation (some neurons vote to go up, and others to go down, but there are more neurons voting to go in the right direction – what Walter Heiligenberg [former professor at SIO and UCSD tragically killed in a bizarre 1994 USAir vertical plane crash during approach while on his way to deliver an invited lecture in Pittsburgh] used to call “noisy democracy”)

This system evolved out of a communication system – the jamming avoidance response was built out of it – and it’s a beautiful solution

This evolutionary example tells us something

Another example: echolocation – complex sensitive auditory neurons in echolocating bats that are sensitive to different components of the auditory signal that must occur in conjunction in order for the neuron to fire

“How could that have possibly evolved?”

“Human language couldn’t have evolved” – that’s a remarkable statement to me; for one thing it feeds into creationism, and just exemplifies a lack of knowledge and creativity about how complex traits evolved

With this long-winded introduction:

[zebra finch songs audio]

Any one zebra finch male sings one song variant – how could they have learned that?

There isn’t just one species of songbird – if we study just one species of white rat, our conclusions will be limited

If we could look at hundreds of hominid species, we might come to different conclusions about the evolution of language

[starling songs audio – two of the apparently 3500, but not sure if that meant zebra finches or starlings? I think zebra finches…]

Starlings take advantage of independent control of the two sides of the syrinx; they have to keep track of the calls of their neighboring conspecifics as well, so there’s a large memory load

These animals can solve all sorts of auditory perception problems in the lab that I don’t have time to go into

You can ask about lots of species with complex behaviors: how general is the solution?

If you take the comparative approach…

Sensory and sensorimotor developmental song learning: the template theory

[three different zebra finch songs audio – one adult song, one male past the subsong stage (“plastic”), and one isolate (critical period casualty) song]

Mark Konishi: the template theory (with or without an external template)

It turns out that a whole host of things depend on this

Some features of the template theory:

  • innate predisposition
  • innate template
  • critical (sensitive) period for memory acquisition
  • acquired sensory template
  • sensorimotor learning
  • subsong – plastic song – crystallized song
  • normal development required for learning
  • (one more thing I missed)

Delayed auditory feedback (DAF) disrupts zebra finch song (how it manifests differs from individual to individual), reduces song to one syllable

If you subject the bird to DAF for a couple days and then turn it off, the bird will still “stutter” afterwards

About one third of zebra finches show clear effects of DAF, about half of that third shows subsequent syntactic disorders, and about half of those recover from it while the other half don’t

Let’s now look at the nervous system finally

There are a whole lot more “wires” than I’m showing in this diagram

Descending motor pathway

Striatum (basal ganglia)

My assertion is that everything I’m telling you is related to the mechanistic solution that humans adopt when they use human language

[video of single-unit recording in freely behaving (but caged) adult male zebra finch singing to a female in an adjacent cage: when the male sings, the neuron stops firing]

Role of AFP in maintaining adult zebra finch song

You can deafen an adult zebra finch, three days later the song has severely deteriorated, and a year later there’s still a deficit

However if you deafen AND lesion the bird in the basal ganglia output system, the song isn’t affected

Social context modifies motor activity

Genetic expressions are being turned on and off in your brain right now

We know that the basal ganglia are involved in speech and language, and maybe this is how

Sparse coding in the descending pathway – neurons fire at some point in the motif

At this level of the motor system, there is pure timing information (whether in one or multiple dimensions is unclear – if you heat up or cool down the brain, you can stretch out the bird’s song)

How you would teach a network to produce the sounds you want to produce?

If you look at the projection to the syringeal motor neurons, you see that the nervous system has taken complete advantage of the auditory signal (precision and bandwidth in motor system)

Mirror neurons exist in songbirds, and you can argue that these are stronger observations than in primates

Mirror neurons tell us about sensorimotor coordination: role of sleep in song learning

On the day after a bird was first exposed to a tutor model, the bird started to show a pattern of circadian rhythms that is not driven by regular circadian mechanisms

In the morning the quality of the song drops, and it increases in the afternoon

This is probably happening in humans – we just need to go and find out

There’s a compelling sleep component in speech perception in L2 learning

In adult birds, there are spontaneous neuronal replays of song during sleep that replicate the firing that occurs during singing while awake – not 100% but highly significant

The developing bird has a lot of information about the adult song while it sleeps

During sleep, patterns of discharge change – then spiking during singing changes accordingly after sleep

If this firing pattern lacked structure, the song would fall apart

There’s good reason that this is the mechanism that the bird uses to maintain the song via feedback mechanisms via rehearsal during sleep

You can record from neurons of a bird first exposed to a tutor song – there are changes in neuronal firing the first night before the bird has started to display corresponding neuronal firings in behavior (Shenk & Margoliash 2008, Nature)

Two plastic mechanisms, one during the night and one during the day

They’re in the same network but not very closely coordinated at first

Auditory pathways vs. song system pathways

We can find interesting patterns in the auditory pathway systems

Neurons that are in the auditory pathways are sensitive to….

[audio of zebra finch isolate “creole” experiment: isolate song is the only input to the next generation, that generation’s song (based on isolate) in turn becomes the only input to the next generation, which becomes the input to the next generation, and so forth]

The recursion controversy: central core of human language

Fitch and Hauser: cotton top tamarins recursion experiment (failed)

Tim Gentner (UCSD Psychology) experiment: ethologists don’t care about it and linguists reject it

Conclusion: What have we learned?

  • Humans are animals
  • Comparative biology is informative
  • A comparative approach to language
  • Recursion evolved (or at least the structures that support it)
  • Language is cognitive
  • Brain size matters


Seek you humans through all your days for all the ways you may be unique
and tho language may be one you thought you had under your belt
the writing is on the wall and no matter the language in which it's spelt,
the answer is remarkably few if not NO
So embrace the animal within; it's not necessarily a sin!
And since size matters and yours is big
remove the fig from your brain and open your mind
to find much continuity with all manner of animal-kind -- m.k.