Rethinking ostension (1. A terminological issue)

Relevance theory was developed in the 1970s and 80s. Over the years, there have been a various modifications—hopefully improvements. In this and in posts to follow, I want to engage in some further rethinking. Today, I start, as a warm-up, with a terminological issue.

In in our 1986 book, Relevance: Communication and cognition, Deirdre and I drew a sharp contrast between two forms of communication, which we called “coded communication” and “ostensive-inferential communication.”

In coded communication (exemplified by animal signalling and various cultural artefacts such as the Morse code), a communicator encodes a message into a signal that the addressee decodes. In ostensive-inferential communication (well exemplified by standard human pointing), a communicator ostensively requests the attention of the addressee and provides evidence from which the addressee can infer what information the communicator intends to convey.

There are combined forms:  a communicative behaviour can involve both the production of a decodable signal and an act of ostension. Verbal communication, we argued, is a case in point. It is ostensive but it uses the senses encoded by linguistic expressions not to encode the speaker’s meaning but to give evidence of this meaning.

On the other hand, there are no intermediate forms, behaviours that would qualify neither as signalling nor as ostension but that would be somewhere between the two. In particular, we deny that, in evolution, earlier primate signalling could have morphed into human ostensive communication through a series of intermediate forms (for a detailed discussion, see Thom Scott Phillips’ 2014 book, Speaking Our Minds). This does not mean, however, that there are no precursors to ostensive-inferential communication. I want to share some speculative ideas about these precursors. Before doing so, however, I need to do a bit of terminological house cleaning.

There is a problem with the label “ostensive-inferential communication” and more precisely with the word “inferential” in it. When we wrote our book, we roughly accepted Jerry Fodor’s view of the architecture of the human mind (presented in his 1983 Modularity of Mind): a variety of input modules combined with non-modular central processes. Inference, we assumed, was one of these central processes: it could be applied to any kind of task (in particular to the task of comprehension). We departed, however, from more standard views of inference in assuming that every concept (with the possible exception of concepts of individuals expressed by proper names) contributes its own deductive rules to the inference process. In other terms, we saw the inference process itself as quite domain-general, but inference rules as highly domain-specific.

Partly under the influence of then nascent evolutionary psychology, we soon came to see inference in a different light: not as a process performed by one central cognitive system using deductive rules, but as a function performed by a wide variety of cognitive mechanisms or modules, each drawing inference in its own specialised domain and using procedures adapted to this domain (for a recent detailed picture, see Hugo Mercier & Dan Sperber, The Enigma of Reason, 2017).

Comprehension, which attributes a communicative intention to a communicator, is a form of mindreading and might be viewed as one type of output of a mindreading module. We argued, however (in our 2002 “Pragmatics, modularity and mindreading”), that comprehension is performed by a more specialised inferential module – maybe a sub-module of a more general mindreading module. Unlike mindreading in general, comprehension can use a specialized procedure that takes advantage of the fact that the communicator is conveying a presumption of relevance and is actively helping her audience infer her meaning.

Talking, as we used to, of “ostensive-inferential communication” failed, then, to capture the very special character of the inferences that, we now realised, are involved in the comprehension process.

Thirty years ago, we had also been assuming that, in coded communication, inference plays at most a marginal role. Seeing inference, as we do now, not as one central system but as a function performed in different ways by most cognitive mechanisms, there is no clear reason anymore to contrast inference and decoding the way we did. Decoding itself can be viewed as a one type of specialised inferential procedure among many.

Even more importantly, much work on animal communication (in particular Dorothy Cheney & Robert Seyfarth’s, How Monkeys See the World: Inside the Mind of Another Species, 1990 – see also their 2008 Baboon Metaphysics) has shown how much animal communication involves context-sensitive inferences: the same coded signal can be disambiguated differently in different contexts.

So, for a while now, we have been talking of “ostensive communication,” dropping “inferential.” It seemed like a good and simple solution. But actually, it turned out to raise more problems than it solves. To talk of just “ostensive communication” forced us to enrich the meaning of “ostension” in ways that are not truly justified.

In comparative and developmental psychology, “ostension” is often used in a wide sense – arguably too wide –  to describe behaviour the function of which is to catch the attention of others. We had used it to refer, more narrowly, not just to the act of catching attention but of catching attention in an intentional and overt way. Now, to be able to use “ostensive communication” simpliciter, we had to further narrow down the meaning of “ostension.” “Ostension” would now denote intentional and overt behaviour of the communicator aimed at attracting attention not just to something she would be producing or pointing at, but also, ultimately, to her own communicative intention. We had good reasons to give up “inferential” in “ostensive-inferential communication” but as a result, we now had “ostensive” do not only its attention-related earlier job, but also the job that “inferential” had been intended to do. This, we soon realised, was terminologically awkward and theoretically question-begging.

Yesterday, I was, once more, discussing all this with Deirdre. The best solution we can think of at the moment to this terminological conundrum is to replace “inferential” with “interpretive.”

In Relevance, we characterised “interpretation” as a relationship between two representations, one representing the other in virtue of a resemblance of content. This is a fairly standard sense of “interpretation”: an interpreter from French into English, for instance, is given a text in French and produces a text in English that represents the French text by being as similar as possible to it in content. A summary or an exegesis are interpretations that either reduce or expand the content of the text they represent. To interpret the thoughts of others is to represent them by means of mental representations of one’s own of similar content, and so on. In the same spirit, we had written:

“We see verbal communication as involving a speaker producing an utterance as a public interpretation of one of her thoughts, and the hearer constructing a mental interpretation of this utterance, and hence of the original thought. Let us say that an utterance is an interpretive expression of a thought of the speaker’s, and that the hearer makes an interpretive assumption about the speaker’s informative intention.”  (Relevance, pp. 230-1).

So “ostensive-interpretive communication” highlights the two perspectives, that of the communicator who acts ostensively and of the audience who interprets. It does so more informatively and accurately than did “ostensive-inferential.”  Still, it is a mouthful. Any better idea?

In a week or so, I will move to more substantial issues regarding ostension itself and its place in evolution and in development.


  • louisdesaussure 7 February 2018 (22:14)

    I don’t like “ostensive-interpretive” very much because it’s opaque, at least more than “ostensive-inferential” was. Formerly I could say that I interpret ostensive utterances thanks to inference. With this new wording I should say that I interpret ostensive utterance through interpretation, which is not informative if not even trivial unless one gets the technical meaning of ‘interpretive’. Since interpretation involves some kind of inference* or rather perhaps “processing”, I wonder if there’s something we could do starting with “processing”. “Ostensive-processive” doesn’t seem nice; “ostensive-derivational” is not exactly good for decoding. Why not “ostensive-procedural” after all? Would perhaps fit with other aspects of the theory – but one has to admit that decoding is procedural by nature, which is still another matter of possible discussion.

  • Nima Mussavifard
    Nima Mussavifard 8 February 2018 (01:48)

    Ostension is in the eye of the beholder
    “…behaviour the function of which is to catch the attention of others.”

    This problem which has lead some primatologists to take primate attention-getters as cases of ostension (or to discard the notion altogether) is I believe to some extent due to lack of any behavioral criterion that would exclude cases of catching attention that are not in any useful sense ostensive. I can think of no behavior in humans with the “function” (innate, learned, or inferred) of attracting attention that is not indeed ostensive. Overtness, as I understand it, does not seem to be a very testable criterion. However, this doesn’t mean that the concept of ostension is to be dropped. Rather I think that ostension is to be seen as a property of the recipients’ interpretation of behaviors. On the part of the signaler, a behavior with the function of getting the attention is, I would say, ostensive. On the part of the receiver, however, it seems to be only humans who take such behaviors as signals of the intention to further communicate something. A behavioral criterion would then be interpreting an otherwise non-communicative signal as communicative when preceded by ostension. For instance, studies by Waxman and her colleagues show that beeping sounds could be understood as communicative by young infants only if embedded in an ostensive context; or Cornelia Schulze shows that 18-mo babies take an action as communicative if precedded by eye-contact.
    My point here is that comparative psychologists might be right in defining ostension as “behaviour the function of which is to catch the attention of others,” but humans happen to take actions with the “function” of catching attention as informing of an informative intention. Can we show such a phenomenon in (other) apes? Would primate attention-getters such as shaking branches or drumming elicit a significantly different interpretation of following actions?

  • Billy Clark
    Billy Clark 8 February 2018 (09:12)

    Ostensive-interpretive . . . (replying to Louis)
    I can see the awkwardness you identify, Louis, but I don’t think it’s so bad. I used to say to students that ostensive-inferential acts were acts of ostension by communicators designed to be understood by inference. With this formulation, we can say that they are designed to be interpreted and then go on to say what’s involved in interpreting them (so we don’t need to make the awkward circular statement). The new formulation also makes more salient that both communicators and addressees produce interpretations and think about how they differ, which I think might make it easier to focus on the pragmatics of communication as well as interpretation.

  • Diana Mazzarella
    Diana Mazzarella 8 February 2018 (15:53)

    Duplication vs coordination of meanings
    Thank you very much for sharing this, Dan. I am thinking about the issue of replacing ‘inferential’ in the label ‘ostensive-inferential communication’ with something that would capture the special character of the comprehension process while distinguishing it from decoding (which is also ‘inferential’ in the sense you define above). I like your proposal of ‘ostensive-interpretative communication’, but here is another idea.

    You write: “To interpret the thoughts of others is to represent them by means of mental representations of one’s own similar content”, and report the characterization of ‘interpretation’ given in Relevance as “a relationship between two representations, one representing the other in virtue of a resemblance of content”. My concern is that if identity is a limiting case of resemblance, decoding would appear as a limiting case of interpretation (where the relationship between the mental representation of the sender and that of the receiver are identical to each other).

    For this reason, I think it would be worth shifting the focus from the contrast resemblance-identity, which characterizes the output of the process of interpretation vs. decoding, to the process itself. With regard to this, I find the following passage from The mapping between the mental and the public lexicon (Sperber & Wilson, 1998) very interesting:

    “It is an illusion of the code theory that communication aims at duplication of meanings. Sometime it does, but quite ordinarily a looser kind of understanding is intended and achieved. The type of co-ordination aimed at in most verbal exchanges is best compared to the co-ordination between people taking a stroll together rather than to that between people marching in step.”

    Here we find the opposition between duplication of meanings and coordination of meanings. It is the coordination of meanings constrained by the pursue of relevance (and achieved by deploying the relevance-guided comprehension procedure) that is central to the notion of interpretation you put forward. So why not ‘ostensive-coordinated communication’?

  • Tim Wharton
    Tim Wharton 9 February 2018 (09:37)

    Baboons don’t do irony
    Sorry to be boring, but how about we go back to ‘ostensive-inferential’?

    1) Regardless of the fact that early RT work viewed inference in terms of Fodorian central processes, it’s now well understood that the term ‘inferential’ applies equally to domain-specific processes. Much of the literature on cognitive science describes the processes that underlie both vision and hearing (paradigmatic Fodorian input modules) as ‘inferential’. So what’s the problem with continuing to use the word and making it clear that we understand it in this sense (one that is slightly different to the ‘central processes’ interpretation in the original version of Relevance)? When I introduce the RT inferential model to students, I always mention the inferential ‘quick way’ described by Grice in Aspects of Reason, in which intermediate steps may be omitted in well-formed inferential processes (discussion pre-dating Gigernezer’s ‘fast and frugal heuristics’). As an aside, it’s far from clear to me that that even in ‘Logic and Conversation’ Grice was committed to a view of inference as a conscious, laborious process. Yes, he spelled it out explicitly in that talk, but at the same time was working with a more sophisticated idea of what inference and reasoning actually were.

    2) No matter how much we finesse the relationship between coding and inference – and they are surely, as Dan suggests, two sides of the same coin (this is what lies at the heart of Grice’s account, after all, in which non-natural meaning arises from natural meaning) – we still need to maintain his original insight – that the code model is not an appropriate one to describe human communication. The point applies equally to Grice’s ‘myth’ of the evolution of communication (and, ultimately, language) and his work on creature-construction too (which builds ethics and value so effortlessly into the picture also).

    As for the context-sensitivity of monkey and baboon calls, well, yes, of course. Bird-song is also sensitive to context, as are the alarm calls of domestic chickens – though they’ve never developed the one alarm call that would help them most… But there is a veritable chasm between *that* kind of context-sensitivity, which amounts to choosing from two or three possible interpretations, and the kind that is omnipresent in human communication, in which the possible number of interpretations are infinite: Metaphor? Implicature?

    Baboons don’t do irony.

  • Martin Stehberger 9 February 2018 (23:54)

    what if only the communicator was used in naming?
    If “interpretive” is the best word but still does not completely get rid of the problems of “inferential” (as Diana Mazzarella notes; also, in software, I think, decoding is called “interpreting” when followed by direct execution), then could you perhaps switch from addressee to communicator? For the contrasting model, the coded communication, you don’t have to say “coded-decoded”. In that case, what the addressee has to do is already implied in the term.

    What about “indicated” communication as the contrast to coded communication. The communicator indicates what he/she means. Thus ostensive-indicative communication? (Or indicated/-cating/-catory, etc?)
    Of course there may be other, better words for what the communicator does.

  • Dan Sperber
    Dan Sperber 11 February 2018 (15:51)

    I am grateful to Louis de Saussure, Nima Mussavifard, Billy Clark, Diana Mazzarella, Tim Wharton and Martin Stehberger for their insightful and challenging comments.

    Nima’s raises an important issue about the function of ostension, which I will discuss in the following posts in this series. I look forward to engaging the discussion with Nima then.

    The five other commentators are directly addressing the terminological issue I was raising in this introductory post:

    Tim suggest that the problems with our original “ostensive-inferential communication” are not so bad and that we should keep the label. I am not convinced. On the other hand I agree with his remarks about inference in animal communication, and the fact that Grice’s rational reconstruction of an inference to the speaker’s meaning was not intended as a processing model.

    The four other commentators share my qualms about “ostensive-inferential.” Billy is the only one to think that “ostensive-interpretive” might do, but I sense a lack of enthusiasm, which I share. Louis points to genuine difficulties with “ostensive-interpretive”: “interpretive” can be interpreted in too many ways. He ends up suggesting “ostensive-procedural” while recognising that it might have its problems too. Diana points out that identity is a limiting case of resemblance and fears that interpretive resemblance and decoding identity might not stand in a proper contrast relationship for our terminological purpose. She suggest as an alternative “ostensive-coordinated communication.” I see her reasons, but I fear this too is open to a variety of interpretations. Martin suggests “ostensive-indicative communication,” with the hyphenated terms both referring to what the communicator does. I agree that “indicate” is an underused term, but it too is open to a variety of interpretations. It would not be inappropriate, for instance, to say of a vervet monkey alarm call that it is indicative of the presence of a certain kind of predator.

    So I fear, we have, so far, found more problems than solutions. If, helped by these comments, we find a way to a better solution, we might return to this terminological issue in the last of this series of posts on ostension. For the time being, however, I hope you are like me and eager to talk less about terminology and more about substance, as I will try to do in the next post.