Why do children but not apes acquire language?

In his introduction to Thom Scott-Phillips’s Speaking Our Minds, Olivier Morin mentioned my review of the book in the TLS. For reasons of length I could not include more substantive objections to chapters 3 and 4 of Thom’s book in that review. However, since the gaps in his argument undermine his claim to have explained why humans but not apes acquired language, I don’t think the issues are trivial. While I develop some of these points elsewhere (see the footnote for details), since some of those papers are not yet out, this discussion seems like as good a place as any to point out the reservations that I have with some of the central claims of the book.

Central to Scott-Phillips’s explanation of why humans alone evolved language are, he claims, two cognitive abilities: the ability to act with and understand fourth order meta-representations, and the ability to distinguish between ‘informative’ and ‘communicative’ intentions. Scott-Phillips argues that humans but not apes acquired language because they possess both of these abilities in ways that apes do not. I don’t think this claim is defended adequately in his book.

As I mentioned in the review, there is currently no evidence that pre-verbal infants could understand fourth order meta-representations. Indeed, there is evidence – not discussed in the book – that some six-year-olds struggle to entertain even second order meta-representations (Perner & Wimmer, 1985). Scott-Phillips appeals to the existence of O’Grady et al. (2015) to show that human adults can entertain seventh-order meta-representations. However, since this ability may be a function of their language acquisition and not a pre-requisite of it, and since Scott-Phillips’s claim must be true of pre-verbal children if it is to support his conclusions, the O’Grady study just doesn’t tell us what we would need to know.

This takes us to the discussion of the significance of the informative-communicative distinction. According to Scott-Phillips’s view, ostensive inferential communicators (including pre-verbal children) must be capable of each of the following:

(1) the expression of informative intentions

(2) the recognition of informative intentions

(3) the recognition of communicative intentions

(4) the expression of communicative intentions

I won’t say anything about (1), (2) and (3) here. (There is more to be said about each one, but that may require a great deal of further unpacking. For example, there now exist data for (3) that show inconsistent roles for ostensive cues in both human and ape gesture interpretation; and while I agree with Thom’s empirical claims about (1), I would have liked to see more theoretical support for the benchmark that he sets.) However, I do want to raise some issues with this discussion of (4), understanding of which Thom takes to be manifested in engaging in hidden authorship. I found his handling of this point to be disingenuous.

On Scott-Phillips’s view, which he repeats in his 2015 Current Anthropology paper, it is a pre-requisite of grasping the informative-communicative distinction that one should be able to act with hidden authorship, since that requires recognising the significance of the ability to inhibit one’s communicative intention.

He supports the claim that pre-verbal children could distinguish between informative and communicative intentions by pointing to his own hidden authorship study (Grosse et al., 2013) on children. Although he refers to the subjects in this study only as ‘children’, the youngest children tested in this study were three years old, and so the study does not show that pre-verbal children are capable of hidden authorship in the way that his argument requires. It is not appropriate to draw conclusions about what pre-verbal children can do based on the abilities of three-year-olds. Indeed, when he and Gerlind Grosse ran their hidden authorship study in Leipzig at the end of 2009, they presumably also did not expect that preverbal children would show hidden authorship in their paradigm – since they did not attempt to test any children younger than three. It is a problem throughout Thom’s book that he is inattentive to the ages of subjects when making claims about human ontogeny.

The fact that Scott-Phillips over-interprets his own hidden-authorship data might be more forgivable if he did not also claim that the absence of evidence of apes performing well in the same paradigm should be interpreted as an ‘implicit collective acknowledgment’ by primate researchers that apes would fail the task. I agree with him that in this case, they probably would. However, since we don’t even know whether pre-verbal human children would succeed in this paradigm, it’s hard to know how to interpret this. I am sceptical that they would do so – but I also think this inconsequential, since hidden authorship is the wrong marker to use (Moore, 2015; under revision). Moreover, I am troubled that Scott-Phillips interprets an absence of evidence favourably in humans but unfavourably in apes, despite having reasons to doubt that the finding would really be present in pre-verbal children.

Since he thinks these abilities necessary for ostensive inferential communication, and since his attributions of the relevant abilities to pre-verbal children remains empirically unsupported and tendentious, we ought to be sceptical of his conclusion about why humans alone evolved language.* There may be many reasons why apes did not acquire language. My preferred explanation is that, unlike humans, they simply failed to grasp the coordinative potential of their existing communicative abilities – perhaps because they did not face the same ecological challenges as our ancestors, which forced them to overcome more challenging tests of coordination. I suspect that many of the abilities that he thinks are pre-requisites of ostensive-inferential communication are really acquired only after language. (I defend this view at length in the submitted paper mentioned in the footnote.) Even if my view is false, though, I don’t think Scott-Phillips’s discussion is sufficiently detailed to warrant any strong conclusions about the origins of language.

To give another example of his too quick treatments of important issues, there is no question that Mike Tomasello’s hypothesis about the cooperative foundation of human language is a hugely important one. However, Scott-Phillips dismisses it in a few paragraphs at the end of chapter 3, following only a very superficial discussion of the relevant literature on joint action. In fact, Tomasello’s claims about the cooperative foundations of communication follow from his acceptance of claims that Thom also accepts – namely, the need for pragmatic interpretation of utterances. On Tomasello’s view, pragmatic interpretation is possible only because speaker and hearer together take themselves to be engaged in a joint project of bringing the hearer to grasp the speaker’s communicative goal. Scott-Phillips’s cursory treatment of this point fails to get to grips with Tomasello’s underlying (and admittedly not entirely clear) motivation, and so does not make it clear why Mike’s view should not be accepted.

I hope that these comments won’t be interpreted as showing hostility to Scott-Phillips’s work. While I disagree with much of what he says, he is a wonderfully clear thinker, and my own ways of thinking about these issues have benefitted hugely from the perspicacious ways in which he carves up the conceptual terrain. At the same time, his preference for simple, clear answers leads him to ignore details. While it is too much to expect a book of 200-odd pages to answer all of the problems that it raises, it is not unreasonable to expect it to temper its conclusions in the absence of attention to detail. Scott-Phillips’s claim to have explained why humans alone evolved language is unsupported by the arguments that he offers.

On a general level, it may be that given Thom’s fondness for big pictures over details, he won’t worry about these complaints. As he says in his book:

“There is nothing wrong with lumping in science. On the contrary, it is how we develop our major theories and paradigms.”

The problem with this view is that ultimately it is details that will falsify our major theories. Ignoring them is therefore not an option.
I develop these points further in three related papers – my response to his Current Anthropology paper (Moore, 2015, ‘A common intentional framework for great ape and human communication‘), my response to his recent Animal Cognition paper (‘Meaning and ostension in great ape gestural communication‘ currently under revision at the same journal) and an original paper (‘Gricean communication and cognitive development’, under review). In both of the response papers I spell out an alternative test of the informative-communicative distinction. In the cognitive development paper I argue that neither fourth order meta-representations nor sophisticated folk psychological abilities are a pre-requisite of ostensive inferential communication. If anyone is interested, all of these can be downloaded from my page on ResearchGate: https://www.researchgate.net/profile/Richard_Moore14.


  • comment-avatar
    Archibald Haddock 23 June 2015 (09:51)

    In these critical comments, Richard takes issue with one of the central claims of SOM, namely that humans but not any other great apes engage in ostensive communication. A full discussion of the issues he raises would require a response that is longer and more detailed than I can give right now, and so here I focus only on defending SOM from some of his more serious charges.

    First, some context. I defend in SOM the claim, made by others before me, that engagement in ostensive-inferential communication proper depends upon the competent use of recursively-embedded mental states (see in particular §3.4). There is a tension between this theoretical position, on the one hand, and a common sense intuition, on the other, that processing recursively-embedded mental states is cognitively difficult, and certainly not something that can plausibly be involved in everyday communicative interactions. Something, then, has to give. Either accept the theory and reject the intuition, or vice versa. As I mentioned in my précis to this book club, “most researchers who have grappled with this problem have chosen to accept the intuition, or some version of it, and argue against the theory”. Richard is one such researcher, and he has made a number of substantial contributions to this discussion. In SOM I take the opposite position – I argue in favour of the theory and against the intuition – and it is this view that Richard takes issue with, both here and elsewhere. Although there are several important points of disagreement between us, I would like to stress that I am grateful for Richard’s detailed critique of my work, since he rightly brings attention to several points that do require further elaboration and defence.

    In particular, Richard is correct to emphasise that we do not (yet) have the evidence necessary to fully accept some of my claims. For instance, Richard questions whether young children, who are competent users of ostensive communication but who have not yet acquired language, are able to process the sort of recursively-embedded mental states that, I argue, are in fact involved in ostensive communication. We do not know at present whether this is true.

    However, I was not concerned in SOM with establishing this claim, and related ones, beyond doubt. My agenda was more modestly to argue for its plausibility. To this end, I reviewed in chapter 3 evidence that even pre-linguistic children engage in mindreading, and briefly mentioned one adult experiment that suggests that recursively-embedded mental states pose no particular cognitive challenges for adults. There is, as Richard points out, no way that such data could ever establish that my claims are fact. Such data can however establish that my claims have some degree of plausibility. I also, moreover, presented arguments that my claims are more plausible than alternative positions. This led me to conclude that “the idea that both adults’and children’s communication really does involve recursive mindreading is not only plausible, but is in fact the most parsimonious interpretation of the data at the time of writing”(p.74). Richard’s comments give the impression that I believe that these matters are fully resolved, but as I hope this quote illustrates, I do not believe this, and SOM makes no such claims.

    Subtle misreadings of my views recur throughout Richard’s comments. Regarding hidden authorship, Richard writes that “On Scott-Phillips’s view,…it is a pre-requisite of grasping the informative-communicative distinction that one should be able to act with hidden authorship”. Yet what I actually wrote is that command of hidden authorship “comprises good evidence of an understanding of what a communicative intention consists of, and its relationship with informative intentions”(p.89, italics added). I hope I am not being pedantic here. Good evidence is not the same thing as a prerequisite. Indeed, in the present context one is a fair presentation of my view, but the other is not.

    Still, this is incidental to the main point that Richard wishes to make with hidden authorship, namely that SOM does not pay sufficient attention to the exact ages at which children pass the various social-cognitive tasks under discussion. There is some partial truth to this. It is at least possible, as Richard suggests, that children in fact pass these tasks only after they acquire language. Moreover, if this is true, it would undermine my claims, and it is probably fair to say that SOM should have made this point more explicitly.

    What is not fair, however, is to fault a book for not being a different book. Academic output should be critiqued on its own terms. Richard believes that SOM should have paid more attention to several other claims –he offers Mike Tomasello’s work as an example – but it was never the agenda of SOM to provide a detailed review of the entire terrain. On the contrary, in fact. Here is what I say in the Preface to SOM: “I have tried, as much as possible, to make my arguments positive ones, in favour of a particular view of the origins of language. I am of course critical of other perspectives where it is necessary or useful to be so, but in some cases I have avoided direct confrontation with some views that differ from my own, in order to maintain a focus on the positive case for my own views.”(p.xiv). I make related remarks at several other places, including just before my discussion of Tomasello’s work (p.75). I see this focus on making a positive contribution as a strength rather than a weakness of SOM, but either way, these are the terms on which SOM should be judged.

    It is unfortunate that Richard’s comments end with a objection that comes close to ad hominem criticism: “it may be that given Thom’s fondness for big pictures over details, he won’t worry about these complaints”. The irony is that to support this assertion, he ignores detail himself: he quotes SOM directly, but out of context. The quote he uses –“There is nothing wrong with lumping in science. On the contrary, it is how we develop our major theories and paradigms”–comes from §3.3, where I defend the claim that Relevance Theory provides a proper scientific paradigm for pragmatics. As Richard knows, the development of new scientific paradigms is all about the search for generality in one’s explanations, and it is in this context alone that I argue in favour of lumping. I certainly not do not make any arguments in favour of ignoring details.

  • comment-avatar
    Richard Moore 23 June 2015 (22:55)

    Just a quick reply, because I’m in Greece all week and working with a terrible internet connection …

    I appreciate Thom’s comments and am happy concede some of them. It’s true that in the case of hidden authorship he talks about ‘evidence for’ rather than ‘necessary for’ – and I misrepresented him to the extent that I didn’t make that clearer. At the same time, it’s still not true that evidence of a certain behavior in three year olds is evidence of anything in pre-verbal children; and so it’s not clear that three-year-olds’ ability to engage in hidden authorship can tell us much about why humans but not apes acquire(d) language.

    I also appreciate Thom’s point about the scope of his book. However, as Olivier notes, it is still a book that claims to have solved every major problem in the study of language evolution. Since this is a somewhat immodest claim, I don’t think it’s inappropriate to hold someone who makes it to a robust burden of proof! I’m also genuinely puzzled by his presentation of his view as unorthodox (for example, in the sentence he quotes from his precis). In fact, versions of Thom’s view have been defended by Michael Corballis, Sperber and Wilson, and – most extensively – Tomasello. By contrast, the ‘standard view’ re-interpretations of Grice have come in a neglected book chapter by Juan Carlos Gomez, and mostly unpublished work by me. What Thom adds to the Tomasello et al. view – and does with great clarity and nuance – is a clear acknowledgement of the precise cognitive abilities that they implicate; and he replaces Tomasello’s evidence on the need for cooperation with an appeal to the explanatory mechanisms posited by Relevance Theory.

    As for the final comment, I didn’t intend for it to come across as a dig, and I’m sorry if that wasn’t obvious. The text I quoted was intended just to be a reflection on the differences in Thom’s approach and mine to the questions that interest us: we have had the splitters vs. lumpers argument many times over, and we fall on very different sides. Of course, I don’t think Thom has ever argued that differences should be ignored – even if I think he’s sometimes insufficiently attentive to them.

    Thanks for taking the time to respond, and greetings from Aigina.