News & Views | Published:


Startling starlings

Recursion, once thought to be the unique province of human language, now seems to be within the ken of a common songbird — perhaps providing insight into the origins of language.

Man the tool-maker. Man the cultural animal. Man the mimic. It's tempting to summarize the differences between humans and other species in a concise phrase, but most posited differences have turned out to be overstated. Chimpanzees and gorillas use sticks to fish for termites; orangutans use sticks for autoeroticism. And many of these capacities seem to be culturally mediated; they are transferred from one primate to the next by illustration and observation, rather than learned afresh by trial and error1.

The report by Timothy Gentner and colleagues on page 1204 of this issue2 challenges one more putatively uniquely human adaptation: the capacity to recognize complex ‘recursive’ structure. Gentner et al. showed that at least one non-human species, the European starling (Fig. 1), can be trained to acquire complex recursive grammars such as the AnBn language (in the case of the starling, rattle rattle warble warble; see below).

Figure 1: No bird brain.


The European starling, Sturnus vulgaris, which Gentner et al.2 show is capable of recognizing complex grammar.

Recursion, or self-embedding, is without question a hallmark of human language. For example, one can take a phrase such as love conquers all and embed it in a frame such as X knows Y, yielding, say, Chris knows love conquers all. The output of that process can then be fed back into the X knows Y frame, yielding, say, Terry knows Chris knows love conquers all. Embedding also makes relative clauses possible, as in the bracketed part of the paperback [on the coffee table] is hilarious. Marc Hauser, Noam Chomsky and Tecumseh Fitch3 have speculated that recursion might be unique to humans — and perhaps even the only contribution to language that is human-specific. Consistent with this, Fitch and Hauser4 found that cotton-top tamarin monkeys could not distinguish the AnBn language from an ostensibly similar language (AB)n that need not be constructed recursively.

The AnBn language (see Fig. 1 of the paper2 on page 1204) is generally assumed to be recursive because new sentences can be formed by successive insertion into the frame AXB, for example AB, AABB, AAABBB and so on. Gentner and colleagues2 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 worries5, this is strong evidence that humans are not alone in their capacity to recognize recursion.

What explains the discrepancy between the starlings' success and the tamarins' apparent failure? One possibility is methodological: Fitch and Hauser4 tested whether tamarins could acquire AnBn spontaneously from a relatively brief exposure, whereas Gentner et al.2 asked whether starlings could acquire similar structures from considerably longer exposures, enhanced with positive feedback, in a more active task. Only further experimentation can clarify whether tamarins' apparent inability to recognize the AnBn structure is context-specific or genuinely absolute.

If the split between tamarins and starlings does prove robust, further comparative work will clearly be necessary. 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.

Whether or not tamarins can be prodded into recognizing recursion, the analogous human capacity seems robust. Humans are quick to notice recursion and are able to do so without explicit reinforcement; perhaps most importantly, they can generalize recursive structures broadly. Starlings have thus far been shown to be able to extend AnBn only to new sequences of familiar sounds. Humans can clearly go further; once you recognize the pattern in AABB and AAABBB, it is a trivial matter to extend that pattern to new vocabulary (for example, CCCDDD or JJJKKK). Taken together with the tamarins, there actually seems to be a three-way split: some species may generalize recursion only to items that have already been instantiated in a given pattern; some species can generalize recursion freely to newly acquired vocabularies (arguably the essence of human language); and some species apparently cannot recognize recursion at all.

The “abstract computational capacity of language”3 may consist not so much of a single innovation as a novel evolutionary reconfiguration of many (perhaps subtly6 or even qualitatively7 modified) ancestral cognitive components, genetically rejigged into a new whole. Contemporary research suggests that the human brain contains few if any unique neuronal types, and few if any genes lack a significant ancestral precedent8. At the same time, humans show much continuity with their non-speaking cousins in dozens of ways that might contribute to language, including mechanisms for representing time and space, for analysing sequences, for auditory analysis, for inhibiting inappropriate action, and for memory.

None of this challenges Chomsky's long-held conjecture9 that children are innately endowed with a universal grammar — a set of mental machinery that would lead all human languages to have a similar abstract character. But that shared abstract character may have as much to do with our lineage as vertebrates as with our uniquely human innovations. In Charles Darwin's immortal words, “throughout nature almost every part of each living being has probably served, in a slightly modified condition” in some ancestor or another.


  1. 1

    Whiten, A. Nature 437, 52–55 (2005).

  2. 2

    Genter, T. Q., Fenn, K. M., Margoliash, D. & Nusbaum, H. C. Nature 440, 1204–1207 (2006).

  3. 3

    Hauser, M. D., Chomsky, N. & Fitch, W. T. Science 298, 1569–1579 (2002).

  4. 4

    Fitch, W. T. & Hauser, M. D. Science 303, 377–380 (2004).

  5. 5

    Kochanski, G. Science 303, 377–380 (2004).

  6. 6

    Pinker, S. & Jackendoff, R. Cognition 95, 201–236 (2005).

  7. 7

    Marcus, G. F. The Birth of the Mind: How a Tiny Number of Genes Creates the Complexities of Human Thought (Basic Books, New York, 2004).

  8. 8

    Fisher, S. E. & Marcus, G. F. Nature Rev. Genet. 7, 9–20 (2006).

  9. 9

    Chomsky, N. Language and Mind (Harcourt, Brace & World, New York, 1968).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.