The universal grammar of birdsong is genetically encoded

Human cultural traits such as language, dress, religion and values are generally said to be passed from one generation to the next by social learning. And in animal species which have language, the same is true; male song birds, for example, learn the songs with which they serenade potential mates from older male relatives.

A new study, published online in the journal Nature, shows that the songs of isolated zebra finches evolve over multiple generations to resemble those of birds in natural colonies. These findings show that song learning in birds is not purely the product of nurture, but has a strong genetic basis, and suggest that bird song has a universal grammar, or an intrinsic structure which is present at birth.

Birdsong shares similarities with, and is considered by some to be analagous to, human language. For example, both have grammar and syntax, and the songbird brain contains brain areas which are analogs of the speech centres in the human brain. Also, birdsong is passed down from one generation to the next, just like human speech. Male zebra finches learn their song by imitating an adult male relative – usually their father, or an uncle. The song is based on a template which consists of stereotyped syllables that are repeated in a fixed order. However, each individual bird introduces small variations into this template, and thus has its own unique song.

In the new study, Partha Mitra of Cold Spring Harbor Laboratory in New York and his colleagues investigated the songs of zebra finches which were raised in isolation, and thus not exposed to singing males during development. They placed juvenile finches in sound chambers, between 30 and 130 days after hatching, a period of development which is known to be critical for song learning. As expected, the isolated birds produced songs which were markedly different from those of the wild-types, or birds raised in natural colonies or with other birds in the laboratory. The songs had an irregular rhythm and were less structured, containing noisy broadband notes and high-pitched upsweeps. Some of the syllables were also prolonged, and often monotonous or stuttered.

Newly-hatched finches were then placed into the sound chambers with the isolated males. These “pupils” readily imitated the songs produced by their “tutors”, producing accurate copies. However, closer examination showed that they changed certain of the songs’ characteristics – copies of syllables which were longer than a given length (about one quarter of a second) were, on average, 30% shorter than the originals, and the relative frequency of highly abundant syllables was reduced.

When this first generation of pupils matured, a new batch of hatchlings was placed into sound chambers with them. This second generation of pupils also imitated their tutors’ songs, but again, they introduced minor variations into their otherwise faithful copies. The variations thus accumulated over successive generations, such that, over the course of three to four generations, the songs had evolved to sound more like the songs of wild-types than those of the isolated colony founders.

The authors note that their findings resemble the well-known case of a large group of deaf children in the Nicaraguan capital Managua, who in the 1970s and 80s spontaneously developed a unique form of sign language called Idioma de Señas Nicaragüense (ISN). They conclude that the results show that zebra finch song is not just a product of environmental influences, but is at least partly genetically encoded. Evidently, it is an extended developmental process, which emerges over multiple generations. In other words, zebra finch songs seem to exhibit what Noam Chomsky referred to as universal grammar:

…it is natural to expect a close relation between innate properties of the mind and features of linguistic structure; for language, after all, has no existence apart from its mental representation. Whatever properties it has must be those that are given to it by the innate mental processes of the organism that has invented it and that invents it anew with each succeeding generation, along with whatever properties are associated with the conditions of its use.

Essentially, Chomsky argues that the brain contains a limited set of rules for structuring language, which are not learnt, but are present at birth. These rules are flexible, but ultimately constrain the diversity of human language. Thus, all of the approximately 6,000 human languages share a basic grammatical structure, which facilitates their acquisition. Applied to the new study, this innate language structure sets limits upon the variations in the pupils’ songs, and perhaps drives those variations – towards the wild-type song structure.

The study focused on the acoustic properties of the songs produced by the isolated birds and the multiple generations of birds they tutored. The effects of song variation on mating behaviour were not explored, so it remains to be seen whether females have an increasing preference for the songs of successive generations of pupils over those of their tutors. And, as it is well established that the vocal centres in the song bird brain are regenerated anew every mating season, it would be interesting to investigate whether the progressive changes in the pupils’ songs are associated with changes in neurogenesis, cellular organization, or gene expression.

Loosely speaking, birdsong serves as a biological model of culture. It is learnt by social interactions, but it also shares another important feature with human language: it exhibits diversity, with geographically separated groups of finches having “dialects” of song which are distinct from those of other groups. As language is a defining feature of culture, the wider implication of these new findings is that  it and other aspects of culture – in birds, and perhaps even humans – may be partly encoded in the genome.

Fehér, O. et al (2009). De novo establishment of wild-type song culture in the zebra finch. Nature. DOI: 10.1038/nature07994.

14 thoughts on “The universal grammar of birdsong is genetically encoded

  1. I’m not sure if this experiment necessarily constitutes the existence of a genetically-encoded universal grammar, even in song birds. For instance, the successive change of the song over several generations, to the extent that it sounded like the wild variety, may just be the song adapting to fit the brain (and shaping it too). There is clearly more of a selective pressure on the transmission itself to adapt to be easily processed, with the most efficient song having a greater fitness value (in that it is mutually intelligible).
    This is not to say the brain is not have any genetic-biases that slightly fate neural circuitry in a specific developmental pattern; however, I’m very sceptical about the degree to which the brain, specifically in regards to language, is primarily the product of biological natural selection (as Pinker would argue, not necessarily Chomsky… so maybe this argument is slightly of the strawman variety). For more on this position, see Christiansen and Chater’s paper (2008) ‘Language as Shaped by the Brain’. Or for a good overview, visit Babel’s Dawn:

  2. One thing to remember is that this is for the zebra finch which is well known to have sing relatively unique songs, so this isn’t *too* much of a surprise. The other bird used as a model for birdsong is the starling which *does* learn new songs and motifs, and combines these together. Genetic learning in starling would be infinitely more surprising than in the zebra finch.

  3. You forgot to put your peer-reviewed blogging logo up there – or is this not considered peer-reviewed blogging? Good blog though.

  4. Ian: All my posts are about peer-reviewed papers. I’ve just stopped using the logo, but the posts still appear on the Research Blogging site.

  5. I think this zebra finch example is just as subject to criticism as Chomsky’s universal grammar arguments on the basis of its failure to eliminate a (very) plausible alternate hypothesis: The universal features of language may be a product of functional convergence, not underlying genetic universality. Basically, the objection is that language must have certain features in order to do the job that language does, which is the capacity for infinite representation and expression. Without features like tenses, clauses, objects, subjects, etc. – that is, without the features of universal grammar – language does not fulfill its function. The Nicaraguan sign language example reinforces this argument rather than undermines it, because the first-generation sign language that developed among the (older) deaf children was effectively a pidgin or proto-language which satisfied only some of the functions of language, and the next generation of younger children exposed to the pidgin were the ones who developed a fully functional new grammar and thus a new full-blown language. (See Derek Bickerton’s work on creoles and pidgins, esp. Language and Human Behavior, not his whackier recent arguments.)
    The fact that different languages satisfy these functional needs using different specific means – often radically different means – suggests that what is fixed is an endpoint, a function to be realized rather than a genetic/neurological substrate which determines the resulting structure. Of course, the neurophysiological potential for and drive towards fulfilling that function must still be present – but that’s a rather different claim from what the Chomskian universal grammar people make.
    And the same sort of alternative explanation seems quite plausible in these (and other) birds. In birdsong, there may be a neural potential for and drive towards (and thus neural selection for) a song which satisfies the functional criteria of being (1) a unique identifier for an individual and (2) discernible as such over appropriate distances in the appropriate ecosystem.

  6. It would be good to see a similar study on a species like starlings (as suggested) but focusing on the syntax. Yeah, significantly harder to quantify, but the emergence of a “complete” syntax very similar to spoken language was what made ISN such a fascinating case.
    Fortunately I know a few people who are actually working on something like this… so thanks for pointing out this article. I’ll make sure my colleagues are aware of it.

  7. I just want to recommend a great book on neurophilosophy to all the readers in here.
    It’s called “emotional amoral egoism”: A Neurophilosophical Theory of Human Nature and its Universal Security Implications.
    It offers a synthesis of philosophical and scientific approaches to human nature and a strong plea for a set of universal human values.
    You can check it out here.

  8. The grammar of bird songs is known to be recursive in the
    English sense. Technically, it is a context free grammar,
    and can express patterns like a a a a ‘n times’ followed
    by b b b ‘n times’; something a weaker form, called a
    regular expression grammar, cannot.
    Humans can understand context sensitive grammar, more
    powerful than a CFG. Human languages themselves are more
    powerful than CFGs. Various attempts have been made to fit
    them between a CFG and a CSG, but none works. As of now,
    one has to think of them as CSGs.
    So, no, bird song grammar is not a good model for
    understanding human language capacity. Also,
    neuroanatomically, looks like the neuronal circuitry is
    rewired every season, quite unlike our neural development.

  9. Birds which use birdsongs can recognize two different
    tones repeated an equal number of times. That t1, t1, t1
    … ‘n’ times followed by t2, t2, t2, ‘n’ times. This
    represents what Chomsky called a CFG (context free grammar),
    because it can be described with a set of grammar rules
    which apply independent of surrounding context.
    Interestingly, primates other than us cannot do this
    Humans can remember arbitrary rules. Most human languages
    do not use arbitrary rules, but are still more powerful
    than a CFG. They typically fall between a CFG and a CSG
    (Context Sensitive grammar).
    So, birdsongs are not the right model for human languages,
    computationally. They are at different levels of the
    Chomsky hierarchy of grammars. Also, neuroanatomically,
    the article says the song-related neural circuitry is
    regenerated every season. Not so for humans.

  10. For the most part I am uncomfortable with this sort of study being inferred or declared to be tangentially related to philosophy with enough intellectual proximity to actually usurp the ancient word “philosophy”.
    Maybe I am just -old fashioned?
    The study might be interesting to philosophy. But- everything is interesting to philosophy. Philosophy is the pinnacle and moral overseer of all knowledge, or so I thought -anyway.
    I would not call my interest in art -art philosophy. Nor would my interest in watching Tiger Woods -prompt me to call that interest -golf philosophy. Would it?
    This sort of open range word foraging leads to all sorts of complications, unless we just express ourselves like generations of birds. Perhaps?
    No. Words like “philosophy” mean something that makes a difference in our lives.
    This is perhaps a trivial point. No one should however, miss the more important picture, -that philosophy is meant to lead humanity by asking the right questions -but- not necessarily answering them, -and thus misleading humanity yet again -off on some wild tangent -we might for a while call a “science” -before we found it all too dangerous, misleading or simply wrong-headed.
    I do not mean to imply wrong-headedness here. Phrenology was wrong-headed, or at least in retrospect it seems wrong-headed.
    In a philosophic sense -there is such a thing- as too much knowledge, and even -knowledge-born social diseases, like Darwinism is in some forms -and social Darwinism is in every form. But this is nowhere near the extent of knowledge-born social disease. (Try singing that song for a while.)
    And I personally find the sort of psychological-neurological study -in this work- bordering on heading into that sort of anti-philosophic direction, as every science generally has the tendency to head toward.
    Care should be given not to attempt to make of philosophy, some mere science. Look at what happened to all of Aristotle’s science, -into the garbage can it went.
    No one should be in the least bit surprised to hear expressed the view that most -if not all- 20th Century psychology can be understood as the expressions of authors intent upon inflicting academia and humanity with the song that made the most sense to those psychological explorers at the time he/she made the effort to impress us all with their sick genius and -sometimes -a somewhat contagious mental disease. Such heinous aberrations are clearly not philosophic, unless they might be in someone’s diseased mind.
    I do not mean to imply a knowledge born mental disease in this article.
    As for searching for genetically encoded overlays -and these expressing themselves in stylistic display, we can easily find one by examining how each of us, -when we use our imagination- do not exactly conjure reality for our perception, and how this is the beginning of limiting what we can say (or sing?) about our memories (of reality?).
    Memory does not exactly correspond to reality for humanity, (not even close) which no doubt is bound up in our genes. No… We are not gods. We often times cannot even remember what happened just a minute ago, let alone tell someone else about it.
    Here it seems we are limited in our ability to mime, mimic and regurgitate (either for ourselves or others) what we have previously experienced in a much fuller -real sense, -or as real as we might be able to sense it in our first shot at it -in the present tense.
    Some might say here, -well of course not!
    But the process of differentiating what seems to make sense to some, -and what should make sense- becomes less clear when we consider, in my dreams, I am completely capable of experiencing the full, rich, landscape or reality, and I can get the all words out too!
    So it is not such an obvious limitation as we might presume, -rather one of perception that genetically limits the ability of human language to express reality, which forces us all to adopt some of what might otherwise be interpreted as arbitrarily halting grammatical-perceptual constructions.
    One such necessary approximation might be that I generally do not say in my dreams, “I think…” for in my dreams I generally KNOW, regardless the reality or unreality of the situation I find myself in, as real as these always seem to be in my dreams, -till I begin to wake up!
    I any event, I just thought I’d drop something in your hat as I passed by. Best! And, don’t trash the place. There are others coming in here after we leave, and I see no reason they should have to deal with the mess we might otherwise leave behind. Tat-ta!

  11. The first person to make a sound with a string did not know what Paccabel would do with it. The wild birds are Paccabel.
    I love listening to the magpies talk to each other where I live and yes after a while you can tell the young males that are practicing or mum is kicking the crap out of a lazy teenage bird that won’t look for it’s own food. I also know that when I hear them at 2am, it means there are lots of bugs around the street light – clever little dinosours…

  12. If the bird song is the result of an innately coded universal grammar, then why do they have to learn it by watching their father or uncle in the first place?

Comments are closed.