Recursion: what is it, who has it, and how did it evolve?

Hello Hello and Happy New Year,

So a new article appeared on the internet late last year by Coolidge, Overmann and Wynn (2010) (hereafter referred to as COW because it makes me smile). It’s a really short sweet little paper and you should read it as recursion is perhaps one of the hottest topics around language evolution. This generally stems from Hauser, Chomsky and Fitch’s (HCF, 2002) claim that it is the only feature of language unique to humans. I thought it would be useful to outline some of the issues surrounding it as put forward by the COW paper due to its high-profile, controversial and important position within current issues in language evolution.


Recursion was first talked about within the field of linguistics by Bar-Hillel in 1953. This was before Chomsky included the concept in his Generative Grammar in 1956.

It wasn’t until 2002 that HCF made the claim that recursion was the only feature of language which was included in the faculty of language in the narrow sense (FLN) and was therefore unique to humans.


The article outlines two definitions of recursion (within linguistics):

(1) embeddedness of phrases within other phrases, which entails keeping track of long-distance dependencies among phrases

(2) the specification of the computed output string itself, including meta-recursion, where recursion is both the recipe for an utterance and the overarching process that creates and executes the recipe

I always worry when there is more than one definition for a thing because this often results in people talking past eachother or getting confused within their own arguments. These definition are also important to define before one starts making claims about whether recursion is present in species outside of humans or what people are talking about when referring to the evolution of recursion.

Evolutionary Scenarios

The paper also outlines two evolutionary scenarios for the adaptive value of recursion in human language.

(1) The gradualist position posits precursors, such as animal communication and protolanguages, and holds that the selective purpose of recursion was for communication.

(2) The saltationist position assumes no gradual development of recursion and posits that it evolved for reasons other than communication

The latter of these is the stand point taken by the HCF paper. Reasons for recursion evolving if one discounts communication could include the increase of working memory for other reasons or spacial navigation.

Pinker and Jackendoff (2005) argue that since recursion only exists in language to express recursive thoughts it must have pre-existed language.

COW (2010) points out that this is all very well but the question remains of what are recursive thoughts and why are they adaptive? These recursive acts may exist for the purposes of diplomatic speech, perlocutionary acts or for prospective memory and cognition (these are discussed at greater length in COW). These assume that the adaptive force was a social one which before Pinker and Jackendoff was not considered because recursion is often understood away from the social context of speech acts in the realm of mathematics.

Unique to Humans?

An often cited example debunking recursion’s importance to human language is the Piraha tribe who apparently do not have it (Everett 2005). The data from Everett is anecdotal, from one source and sketchy. Even if one was to accept the claims of lack of recursion they can be attributed to other factors such as cultural constraints or (although I think this is going a bit far, but then Bickerton always does go a bit too far) claiming the Piraha tribe have an underlying neurophysiological deficiency such as a limited working memory capacity or an extreme case of acquisitional delay.

COW then covers several animal studies which claim that recursion is present in animals including starlings and various monkeys. These are subject to the claim that the ability to acquire a phrase structure grammar means the presence of recursive ability (which is bollocks). These studies also fall short when one considers that starling’s songs are used to communicate emotional states, not recursive thoughts.


Bar-Hillel Y. (1953) On recursive definitions in empirical science. Proceedings of the 11th International Congress of Philosophy, Brussels. 19535:160165.

Coolidge, F., Overmann, K., & Wynn, T. (2010). Recursion: what is it, who has it, and how did it evolve? Wiley Interdisciplinary Reviews: Cognitive Science DOI: 10.1002/wcs.131

Hauser MD, Chomsky, N, Fitch (2002) The faculty of language: what is it, who has it and how did it evolve? Science, 298:1569-1579

14 thoughts on “Recursion: what is it, who has it, and how did it evolve?”

  1. I couldn’t find it anywhere but if you email hanachronism at gmail got com (or DM on twitter) I can send you a PDF. But shush, don’t tell anyone.

  2. Hi Hannah:

    I have a standing blog search/notification set up on this subject (pinker everett recursion language), and it led me here.

    Here’s my question, which I haven’t found answered — hence my standing search:

    I find Everett’s argument lacking: suppose recursion is a mental feature/module that’s independent of language. Does that mean there’s no language instinct? Couldn’t there still be mental modules that understand subjects, objects, verbs, etc. and their interactions, but that can (but doesn’t necessarily) draw on the recursion module? Likewise with counting and colors?

    In find Pinker’s response to be quite lame — way below the standards I usually expect to see from him. He seems to be defending recursion as a language module, even though it need not necessarily be one.

    As such his response seems reactionary and defensive of a perhaps-untenable position (cf Dawkins on group selection), rather than exploratory and open to correction via new evidence and ideas. Not what I expect from him. This happens as scientists get older, but I’d always expected that it wouldn’t happen with him. (Or Dawkins — “we clapped our hands red” — but I seem to have been wrong about him.)

  3. Hi Steve,

    So – yes I agree that the capactity to concieve things like nouns and verbs probably is seperate from out ability to apply recursion to these things but I think what people like Hauser, Chomsky and Fitch (2002) are saying is that it is our ability to interface these two things is what makes humans unique linguistic powers. Everett’s arguement is lacking because it’s probably not the case the Piraha doesn’t have embedding let alone recursion (I’ve written a presentation about the subject here: So HCF’s claims about this being a universal and uniquely human capacity still stands despite Everett’s claims to the contrary. Basically, there’s no evidence that piraha lacks embedding, recursion or colour terms (the counting thing is true though, which is slightly crazy) and so there is no evidence to suggest that we should discount recursion as a linguistic universal.

    In summary, I don’t think that recursion is unique to language or necessarily selected for language, but what is unique, and what is arguably the language faculty, is our ability to interface recursion with symbols, not as you suggest Pinker says, recursion alone.

  4. Hi Hannah,

    thanks for the link to your cool presentation.
    Just one quick question: isn’t it a bit too strong to say that Everett claims that Piraha is incapable of recursion? As far as I recall he only said that it is constrained but not absent.
    But some of the inconsistencies Nevins et al. point out really do look like glaring errors on the side of Everett. However, I think the Nevins paper in general is rather murky and Everett seems to adresses some of their points quite well in his rebuttal (although I have to admit I only skimmed it, TL;DR).

    But there is one “major” error in the Nevins et al. paper I as a native German speaker feel obliged to point out 😉
    “Hansens Auto” does not seem standard German to me but instead colloquial/regional/archaic. You would say “Hans’ Auto” or “das Auto von Hans.” So they should at least have mentioned that or picked another example without the “ens”-suffix.

  5. Thanks Michael!

    Yeah, it probably is a bit strong to say that but this was a presentation in response to a presentation which did make that claim, hence why it’s in there.

    I would have gone into the rebuttal but the pres was meant to be representative of the Nevins et al paper alone (and I also only had 25mins) but afterwards we did discuss some of Everett’s responses at length you’ll be please to hear!

    Thanks for pointing the ‘-ens’ thing out. If I end up extending this to an essay I’ll be sure to mention it.


  6. Good post. I’m currently making my way through the COW paper… So I might have more to say once I’ve finished reading it.

    One point though: there is no evidence to suggest that we should discount recursion as a linguistic universal. I’m not sure we should take a default position on this, other than what know: that most languages we see do have recursion, but there is no a priori reason to suggest that this is something other than a statistical tendency — reflecting stable engineering solutions. For instance, Evans & Levinson (2008) mention several examples where many languages are structured to minimise embedding, like Kayardild, which allows recursion, but caps it at one level of nesting: it forms sub-ordinate clauses through either nominalising the subordinate verb or by using a finite clause. Here’s Evans & Levinson’s conclusion on the subject:

    In discussions of the infinitude of language, it is normally assumed that once the possibility of embedding to one level has been demonstrated, iterated recursion can then go on to generate an infinite number of levels, subject only to memory limitations. And it was arguments from the need to generate an indefinite number of embeddings that were crucial in demonstrating the inadequacy of finite-state grammars. But, as Kayardild shows, the step from one-level recursion to unbounded recursion cannot be assumed, and once recursion is quarantined to one level of nesting it is always possible to use a more limited type of grammar, such as a finite state grammar, to generate it […] The clear conclusion that these languages point to is that recursion is not a necessary or defining feature of every language. It is a well-developed feature of some languages, like English or Japanese, rare but allowed in others (like Bininj Gun-wok), capped at a single level of nesting in others (Kayardild), and in others, like Piraha it is completely absent. Recursion, then, is a capacity languages may exhibit, not a universally present feature.

  7. Hi Wintz, I think Evans and Levinson’s point comes back to this definitional problem again. Nevins et al. (2009) argue that even operations such as merge have recursion (which is far more likely to exist as a universal than things like embedding).

    Merge is an operation which takes two linguistic units as input and combines them to form a set (a PHRASE), in which one element is designated as the phrase’s head. Two kinds of linguistic units may serve as input to Merge: (i) lexical items, and (ii) phrases formed by previous applications of Merge. Since Merge may take previous applications of Merge as input, the rule is RECURSIVE.

    Even though the existence of things like merge are probably dependant on to what extent you buy into things such as generative frameworks it certainly highlights the point that care needs to be taken when kicking words like ‘recursion’ about.

    Hansens Antwort: I’m not going to get into the German debate as I’m not a native speaker but thanks for chipping in.

  8. @ Hansen
    I wrote that Hansen does not seem standard German to me. And I’m pretty sure it isn’t standard German, but either archaic/regional/colloquial, and there might also be generational difference. I simply have never heard it used before.

    Luckily, the Duden agrees with me. As early as 1975 the “ens-Genitive” (as in “Hansens”)is described as an “altertümliche Endung -ens”(archaic suffix -ens). The 2006 Duden-Edition (§ 96) doesn’t even mention this possibility of forming a genitive anymore but explicitly states that it works like this ” Felix’
    Vorschlag, Heinz’ Geburtstag.”
    See also this interesting post

    But you’re right Hansens, it’s definitely German, it’s just that in a linguistics text I would have expected that they mentioned that its not the standard form, especially given that generative grammarians are kind of infamous for getting their non-English examples wrong. (For example, in Charles Yang’s 2006 “The Infinitive Gift” there is a German example that all German speakers in my course rejected as wrong and a Russian example that Speakers of Russian did not understand.)

    @Hannah: sorry for sidetracking your really interesting discussion on recursion 😉

  9. @Michael

    Following your links it becomes clear that you are confusing prescriptive vs. descriptive linguistics. Hansens used to be a form favored by language-arbiters, now it’s not. That’s all. Also looking at the footnote attached to this section of their paper (note 13) it looks like they were rather careful with their German data collection after all.

    And generative grammarians are no more infamous for “getting their non-English examples wrong” than other fallible (or confirmation-biased) human linguists. Look at some of the published commentaries on the Evans and Levinson paper, for example, if you want to see how a pair of aggressively non-generative linguists “got their non-English examples wrong”. Including ones that matter.

  10. @Hannah: sorry, still off-topic 😉
    @Hansens Antwort
    I didn’t see the work fo the Dudenredaktion and the People at the Institut für Deutsche Sprache as prescriptive but rather as descriptive of major trends in Standard German (which of course is a very fuzzy notion, especially if you don’t want to use the term in a prescriptive manner).
    I think this descriptive intention of the Duden is made evident, for example, by the fact that Gisela Zifonun and her colleagues work extensively with corpora. In addition, I especiially think the work done by the IDS on regional variations in Standard German is really impressive.
    I just would argue that most people in Germany would not say “Hansens”. For example, I’d say the “Deutsch heute” Corpus integrating sample data from 192 places where German is spoken should yield significantly less results for this form than for another. Framing “Standard German” in this way is IMO a descriptive not a prescriptive understanding of the term.

    I just would have preferred them to use the more common form or mention that this is a less commonly used one, although, you are of course right, its not substantial to their general argument, especially given what they write in the footnote you mentionen. And it really was just a minor quibble, and meant as a tongue-in-cheek remark.

    Picking out generative Grammarians was of course, a cheap shot, so sorry for that. It’s just that I’m more familiar with generative typological literature than with that of other frameworks, so I wrote about the errors I was familiar with. But of course getting examples wrong is not only limited to Generativist and I agree with you that Evans & Levinson don’t seem to do a very good job in their paper.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.