Rethinking ostension: (2) Attention manipulation

I believe we relevance theorists missed something important in considering ostension only in the context of what we called ostensive-inferential communication. Ostension, I want to suggest, is more diverse and widespread.

In relevance theory, we have been contrasting two mechanisms of communication, coded communication, also commonly known as “signalling,” and what I will, for the time being, call “O-I communication” (O-I for “ostensive-inferential” or “ostensive-interpretive” – I hope we soon find a better solution to the terminological problem I raised in the first post of this series). Unlike coded communication or signalling, which we took in agreement with the relevant literature to be the standard form of animal communication, we argued that O-I communication is a recent adaptation relying on typically human mindreading abilities. We underscored the discontinuities between these two forms of communication: it is a mistake to look for precursor forms of human O-I communication in animal and in particular in non-human primate signalling (as argued in detail by Thom Scott-Phillips in Speaking our Minds).

Ah but there is a third form of communication (in a very broad sense of “communication”), a very ancient and very widespread one: attention manipulation! (Arguably attention manipulation is itself a special case of an even wider category: sensory manipulation; sensory manipulation doesn’t necessarily need the attention of the target; however, here I focus on attention manipulation). Attention manipulation itself takes many forms. Ostension, I will argue (in the next post) is a recent form of attention manipulation.

Today’s post is about forms of attention manipulation older than ostension proper. But to begin with, what is attention, and how might it be advantageous to manipulate the attention of others? Attention has been intensively studied in psychology and more recently, in neuropsychology, with competing models and approaches. Judging from my limited acquaintance with this vast literature, the idea that attention aims at selecting the most relevant inputs is a commonplace. I have not, however, come across a worked-out definition of relevance in this literature, let alone a shared one. What one encounters, rather, are common sense examples: inputs that are evidence of potential reward or danger, inputs that may contribute to the agent’s goals are typically relevant and worth processing. True, no doubt. In the next posts, I will argue for a more precise notion of relevance and show how it matters to a better understanding of ostension, but here such a vague and intuitive notion of relevance will do.

So, back half a billion years ago. During the Cambrian explosion, animal cognitive mechanisms became complex to the point where selection of the most relevant inputs became a precondition for efficient cognition and attention mechanisms evolved. When attention first appeared, nothing had yet evolved to manipulate it. Nothing in what animals could attend to in their environment was designed to attract (or evade) their attention. However, organisms stand to benefit or suffer from the behaviour of other organisms, which may be partly determined by attention mechanisms. Hence, once attention-capable organisms evolved, ways to take advantage of their attention and its behavioural effects (or not to suffer from them) evolved in turn. In many species, for instance, the chances of mating are greater if you can attract the attention of potential mates. Cross-species mutualism (say between cleaner fish and their bigger fish clients) often requires attracting the attention of potential partners. And so on.

(Camouflage, used by many prey and predators, is not a means to attract or distract the attention of the other, but to escape it altogether. Sure, it is an adaptation to the risks presented by others’ attention, but still, it may be better not to categorise it as a manipulation of attention proper, nor as a form of communication.)

Evolved means to attract the attention of others need not involve any intentional behaviour, or, for that matter any behaviour at all. After all, plants have no intentions but still, many do produce flowers the function of which is to attract the attention of potential pollinators. In many animals, changes of colour, which is generally not something the animal can intend, have evolved to attract the attention of potential mates: the belly of male three-spined stickleback fish, for instance, turns red at the time of breeding, attracting females (and also, sometimes, the aggression of other males, as famously demonstrated by Niko Tinbergen).

Manipulation of the attention of one animal by another is typically an adaptive behaviour. It performs its function by influencing the targeted animal’s cognitive states and behaviour in a way that is advantageous to the manipulator. The highly salient red belly of male sticklebacks contributes, by attracting the attention of females and thereby influencing their behaviour, to the reproductive success of males.

Manipulation of attention (or sensory manipulation more generally) is a form of communication in a very broad sense of the term, where communication consists in one organism influencing others by modifying their cognitive states. Often in evolutionary and comparative studies of animals, communication is defined more narrowly as co-extensive with signalling. Signalling itself is defined as an interaction where not only does the communicator produce a signal the function of which is to influence a receiver, but where the receiver decodes the signal and produces a reaction adapted to it.  For signalling systems to evolve, two conditions must be fulfilled: producing a signal has to be advantageous to communicators; reacting appropriately to the signal has to be advantageous to receivers (see David Harper and John Maynard Smith’s 2003 book, Animals signals, and Thom Scott Phillips’s discussion of the issue).

The male sticklebacks’ red belly is both a way to attract the attention of females and a signal to these females. Female sticklebacks’ reproductive success benefits from paying attention to the red belly of males ready to breed (not just its presence, but also its intensity, see here), and from reacting appropriately.

Similarly, many flowering plants in need of pollination attract insects by advertising themselves as a source of valuable nectar. The bright colours and subtle scents of their flowers signal the presence of nectar. In visiting flowers and unwittingly carrying some pollen from male stamina to female pistils, pollinators, for their part, help the plants reproduce: the insects’ reaction to the signal is advantageous both to the insect and to the plant.

All naturally evolved flowers have the function of attracting pollinators, but not all of them do so by means of signalling. Some species of orchids, for instance, mimic the shape, colour and odour of a female wasp, tricking male wasps to try and copulate with them (see here). In the process, some pollen gets stuck on the wasps, who carry it to another orchid that, again, they mistake for a potential mate. The wasps, in other words, are being tricked into acting as pollinators. From an evolutionary and functional point of view, orchids and wasps are in quite different predicaments. The orchids have evolved to influence wasps by providing them with misleading cues. These cues—the shape, colour and odour of the flower—are adaptations that are quite successful at performing their function. The cognitive mechanisms male wasps use when picking up these cues have evolved to detect genuine female wasps, not orchid mimics. Hence, in being activated by orchid mimicry, the male wasps’ female detector malfunctions. When a cue is an evolved adaptation for the organism that produces it—for instance an orchid—aimed at influencing the behaviour of an animal—for instance a male wasp—by causing some of its cognitive mechanisms to malfunction, this is a clear case of attention manipulation, but not of signalling.

As illustrated by the example of the deceptive orchids (or by the famous case of females birds such as killdeers feigning to have a broken wing in order to attract a predator away from their nest), manipulation of attention is a form of communication that need not involve signalling proper.

Conversely, signalling can be effective without attracting the attention of the receivers. Quite complex pheromone signalling has been demonstrated among Caenorhabditis elegans roundworms, even though, with their few hundred neurons, they don’t do attention (see here).

Among humans, Irenäus Eibl-Eibesfeldt demonstrated in a 1968 article the existence of a universal signal: when recognizing a friend at a distance, humans typically raise their eyebrows for a fraction of a second: mutual awareness is thereby achieved but the signal itself – the “eyebrow flash” – is produced and decoded unconsciously. Or to take an example closer to home, Guillaume Dezecache et al. demonstrated in a 2013 article the existence of unconscious signals of emotion, which help explain emotional contagion in crowds.

All this to argue that attention manipulation 1) is an ancient and widespread form of communication; 2) is quite distinct from signalling: it can work without involving signals just as signalling can work without involving attention.

One last point: this post (unlike the previous one) is not about terminology. It is about substantive distinctions and claims. In particular, if you don’t like my broad use of the term “communication” to include attention manipulation, replace it with “communication*” or whatever term you prefer. Same thing if you don’t like “attention manipulation,” and so forth.

In the next post of this series, I will suggest that ostension is a form of attention manipulation that involves a modicum of mindreading (but much less than full-fledged O-I communication). (Juan-Carlos Gomez, by the way, has long been advocating such an idea).


  • comment-avatar
    Pierre Jacob 8 March 2018 (17:22)

    Is not attention manipulation a kind of behavior?
    Dan’s last post on attention manipulation insightfully draws our attention to beautiful cases of dissociation between attention manipulation and signalling. Some cases of attention manipulation are not instances of signalling. For example, some species of orchids have evolved means of manipulating the attention and behavior of male wasps that are thereby tricked into acting as pollinators. These orchids trick male wasps by mimicking the shapes, colours and odours of female wasps. While the cues produced by orchids are natural adaptation selected by evolution to promote pollination that benefits orchids, the perceptual mechanisms that enable male wasps to detect these cues have been selected by evolution to enable male wasps to copulate with female wasps (which increases the fitness of male wasps), not to act as orchid pollinators. Conversely other cases of signalling are not instances of attention manipulation if e.g., the recipient of a signal is taken not have enough neurons for attending to the signal or if the processing of the signal is known to be unconscious (as illustrated by emotional contagion in humans).

    However much I admire this post, I would like to express one worry about what I see as a potential broad tension. On the one hand, Dan makes it clear that he thinks of attention manipulation as an instance of communication in a broad sense of the term, “where communication consists in one organism influencing others by modifying their cognitive states.” I take this to mean that Dan recognizes that attention manipulation rests on agency or equivalently that one organism’s attention could not be manipulated unless the manipulator produced some *behavioral cues* that the manipulated organism was able to pick up.

    On the other hand, Dan also writes that “evolved means to attract or escape the attention of others need not involve any intentional behavior, or, for that matter any behavior at all.” I fully agree (of course) that the relevant behavior need not be intentional. I also agree that the relevant behavior need not involve bodily motions, muscle contractions or motor instructions. But I detect a tension between the view that attention manipulation is a kind of communication in the broad sense (consisting in one organism influencing others by modifying their cognitive states) and the view that attention manipulation need not involve any behavior at all on the part of the manipulator.

    One way to resolve the tension is to acknowledge that flower production can be part of plant (e.g. orchid) behavior or that the reddening of the belly of male sticklebacks can be part of male stickleback behavior, even if neither flower production nor coloring of the belly of fish involves bodily motions, muscle contractions or motor instructions. After all, there are also, it seems to me, uncontroversial cases of animal behavior where the behavior consists in not moving.

  • comment-avatar
    Greg Bryant 8 March 2018 (21:19)

    Proper and actual domains
    I’m surprised you don’t invoke the proper/actual domain concept and error management. Orchids exploit the sensory systems of wasps – the flexibility in wasps that causes them to attempt to copulate with orchids is nevertheless potentially adaptive for other reasons connected to wasp mating (being indiscriminate because of high intrasexual competition). So maybe malfunction is not quite the right word there – unless you consider inadvertent, but low cost byproducts to be malfunctions.

    Also, anything wrong with the concept of dishonest signalling?

  • comment-avatar
    Mo 9 March 2018 (12:59)

    Signaling…I don’t quite see why orchids aren’t signaling. It’s not that the wasp is misfunctioning, except in the outcome, is it?. I.e., they are responding accurately within the limits of their sensory experience. You wouldn’t say pareidolia is misfunctioning, would you? And what about the eye-tracking experiments where infants attribute intentional-agentness to contingently moving watering cans? Do you really want signaling to exclude exactly the specific case where the signal is a specifically adapted cue to cause this sense of a “malfunction” in another species? (Because I can’t think of any within-species examples that would satisfy exactly this set of requirements).

    Fantastic blog series, btw!

  • comment-avatar
    Dan Sperber 9 March 2018 (23:31)

    It is not easy to be clear enough
    I am grateful to Pierre, Greg, and Mo for comments that illustrate the difficulty one encounters – I have encountered – in trying to describe in a clear and unproblematic way the kind of phenomena this series of posts is about.

    Pierre want to hold together two ideas (1) the idea, common in biology, that a number of apparently non-behavioural features, like flowers, which a part of the anatomy of plants, or the red belly of male sticklebacks at breeding time, can be signals, and (2) the idea, which may seem common-sensical in psychology or the social sciences, that signals are behaviours or products of behaviour. To hold to these two ideas, he proposes to extend the notion of behaviour so that flowers are products of the behaviour of a plant growing a flower, and red bellies are the product of belly reddenings.

    This looks to me more like a terminological issue, and I don’t really see the benefit of Pierre’s proposal. A signal is sufficiently defined by its double function of providing to a communicator a means to influence a recipient and providing to the recipient a means to be influenced by the communicator, both effects being sufficiently beneficial for the capacity to signal and that to be affected by the signal to co-evolve. Nothing in this definition hinges on signalling being a behaviour (or some product of behaviour) that can be perceived by the recipient.

    Greg points out that, in discussing such cases, I could invoke the proper/actual domain distinction. Indeed, in earlier work, I used similar cases to introduce this very distinction. I could also have invoked error management theory. In the present context however, I don’t see how these notions would have been particularly relevant.

    On the other hand, I agree with Greg that my use of “malfunction” (as when I write “in being activated by orchid mimicry, the male wasps’ female detector malfunctions”) may be misleading. The wasp’s reaction result from its mechanisms functioning as they should, but in a situation where, by doing so, these mechanisms not only fail to fulfil their function, but may even marginally compromise it. Given the ambiguity between “function” in the sense of functioning, and “function” in the sense of the effect that has caused the mechanism to evolve, I should have been more careful in my formulation.

    Greg also asks whether there is something wrong with the concept of dishonest signalling? Certainly not. But when an organism mimics a signal it is not even equipped to use in a normal and honest way, then it is not clear that this mimicry of a specific signal is an instance of the same signal. Does for instance a parrot producing the sound[ aɪmˈhæpi ] really produce an instance of the English sentence “I’m happy”? Even if we made the terminological decision to answer yes and to call the mimic of a signal an instance of this signal, this mimic version of the signal is asymmetrically dependent on normal instance of the signal, and would not have the same function (in the evolutionary sense of function). This is why I find it more perspicuous not to call such a mimic of a signal an instance of it.

    Mo raises a question similar to Greg’s and I will answer it in the same spirit. Pareidolia (as when seeing a face in a cloud) is the output of a mechanism of perception that functions well processing an input that meets its input conditions, hence belong to the actual domain of the mechanism, but that doesn’t belong to its proper domain (hey, Greg was right, the proper/actual domain distinction can be relevant in this context after all!).

    Pareidolia and the cases of over-attribution of agency that Mo also mentions are not what makes the mechanism that produces them adaptive; they are not instances of the mechanism performing its function.

    When Mo then asks, “Do you really want signaling to exclude exactly the specific case where the signal is a specifically adapted cue to cause this sense of a “malfunction” in another species?”, he actually begs the question of whether the mimic of the signal is really an instance of the signal. And as I said, even if you decide to call it a signal, then this is a signal that has a function quite different from that of the normal signal it mimics. Within a species, on the other hand, an animal using in a deceptive fashion a signal it already has in its repertoire and that it would normally use honestly is definitely using this very signal. Again, what makes the dishonesty work is that is asymmetrically dependent on the normal, honest use of the signal.