Starling syntax

4 minute read

Generative grammar -- the theory of language structure originated by Noam Chomsky -- is lost on most people. It has a unique place in biology, because it is a self-consistent system that entails specialized brain systems to learn and use it, but Chomsky himself has shown great skepticism about the power of natural selection to create it.

The theory has several elements, which it predicts are shared by all human languages. A question has been whether these elements may in fact be prerequisites of any equivalently complex form of serial communication. If so, one or more of them might spontaneously appear in species with complex vocalizations.

This week, Nature has a paper by Timothy Gentner and colleagues that shows that a humanlike syntactic structure can be learned by starlings. From the abstract:

Recent hypotheses make the central claim that the capacity for syntactic recursion forms the computational core of a uniquely human language faculty. Here we show that European starlings (Sturnus vulgaris) accurately recognize acoustic patterns defined by a recursive, self-embedding, context-free grammar. They are also able to classify new patterns defined by the grammar and reliably exclude agrammatical patterns. Thus, the capacity to classify sequences from recursive, centre-embedded grammars is not uniquely human. This finding opens a new range of complex syntactic processing mechanisms to physiological investigation.

The "self-embedding" part here is the crucial descriptor. A LiveScience article by Sara Goudarzi explains it fairly well:

A common characteristic of human grammar is inserting words and clauses within a sentence, without limit. For example, "Oedipus ruled Thebes" can become "Oedipus, who killed his father, ruled Thebes" or "Oedipus, who killed his father, whom he met on the road from Delphi, ruled Thebes," ad infinitum.
More simply stated, you can insert as many brackets as you want within a sentence as long as there are as many brackets on the right as there are on the left.
Chomskian linguists believe that this characteristic, known as "recursive center embedding," is a universal feature of human language, and the ability to process it forms the core of human language ability.

There is also an accompanying "News and Views" feature by Gary Marcus.

Gentner and colleagues rewarded European starlings for pressing a bar in response to AnBn strings of starling-generated sounds, such as rattle rattle warble warble, and withheld the reward for responses to the (AB)n grammar (and vice versa for another group of starlings). Although learning was not instantaneous, nine of eleven birds eventually (after 10,000-50,000 trials) learned to discriminate reliably between the two grammars, succeeding where the monkeys had failed. An extensive series of control comparisons strongly suggests that the ultimately acquired grammar is robust. Notwithstanding some minor worries, this is strong evidence that humans are not alone in their capacity to recognize recursion.

This is not a challenge to the uniqueness of human language as a communication system, but like several other observations, it does suggest that some of the interesting characteristics of human communication may be generally necessary of any complex form of serial communication. If you are going to have any kind of complex vocalization -- that is, one that includes more than around three signs -- then you will need some way to arrange those signs to denote meanings among them. If it is possible to have descriptors of one or more signs, there must be some syntactic (i.e., sign-order-related) conventions to relate them. Recursion may be a logical necessity for such communication, and from that aspect, there is every reason to think that species with relatively long vocalizations might be able to use it. I would expect cetaceans might have such a structure also, which is speculated by Marcus:

Can other varieties of birds that don't (in contrast to starlings) naturally acquire new songs also acquire self-embedded structures? Are humans alone among primates in their capacity to do so? Might the capacity for recursion be general across great apes, even if it were absent in monkeys? An intriguing possibility is that the capacity to recognize recursion might be found only in species that can acquire new patterns of vocalization, for example songbirds, humans and perhaps some cetaceans.

According to the LiveScience article, Chomsky isn't buying it:

"The article is based on an elementary mathematical error," said Chomsky, professor of linguistics at Massachusetts Institute of Technology. "They are overlooking the fact that there are many intermediate systems that are ignored in mathematical linguistics because their properties are empirically irrelevant."
Based on other work done 50 years ago by George Miller, Chomsky thinks further research would show that the birds are not grasping linguistics in the way the new study concludes.
"It has nothing remotely to do with language; probably just with short-term memory," Chomsky told LiveScience.

Yes, that would be where these kind of syntactic abilities should begin, I would think. In humans, there are undoubtedly specialized adaptations to build out syntactic structures quickly and over long utterances with many recursions. For most other species, one or two recursions depending on generalized short-term memory might be all there is. But the chance of continuity between the two is very interesting.


Marcus GF. 2006. Language: Startling starlings. Nature 440:1117-1118. DOI link

Gentner TQ, Fenn KM, Margoliash D, Nusbaum HC. 2006. Recusive syntactic pattern learning by songbirds. Nature 440:1204-1207. DOI link