The image is from here.
1. Essential indexicality
In recent decades a number of Anglophone philosophers of the analytic persuasion have shown a lot of interest in indexicality. It has been argued that in certain contexts indexical expressions such as the pronoun I are “essential” (Perry, 1979) because they can’t be replaced by non-indexical expressions that have the same reference.
Here is how the story goes. Suppose that in “I am about to be arrested” and “Masha is about to be arrested” both the indexical I and the non-indexical Masha can refer to the same person (even if it is not known that they refer to the same person) namely a political activist named Masha who is about to be detained by the Russian security service. If Masha thinks, “I am about to be arrested”, this realization might prompt her to go into hiding or flee Russia. By contrast, thinking “Masha is about to be arrested” will have no such effect unless supplemented with an indexical thought, such as “I am Masha” or “Masha is me”. Thus, the indexical element seems to be indispensable, i.e. essential.
However, not everyone is impressed by such observations. Cappelen and Dever (2013) took it upon themselves to publish a book-length critique of the idea of essential indexicality, which they see as superfluous. One of their key points is that the features gathered under the umbrella “essential indexicality” are nothing more than particular cases of other phenomena, such as context sensitivity and referential opacity.
Here are two examples to illustrate their claim. It is well known – to the point of being trivial – that the reference of the same linguistic sign in the same sentence can vary with the context of the utterance. For instance, “You are about to be arrested on the subway by the secret police” picks out different referents for you, the subway and the secret police, as well as for the present tense of the verb, depending on where, when and to whom it is said. It is also well known – though, admittedly, less trivial – that in opaque contexts, such as those of belief attribution, co-referential terms can’t always be switched without a change in truth value. Suppose that Masha the political activist goes by the nickname Vorobey, and I don’t know that. If so, the statement “Konstantin believes that Masha is about to be arrested” is true while the statement “Konstantin believes that Vorobey is about to be arrested” is false, even though Masha and Vorobey refer to the same person. On Cappelen and Dever’s view, no one has ever shown that essential indexicality, i.e. the non-substitutivity of I in Masha’s statements about herself, differs from these examples in a philosophically interesting way.
Needless to say, what is or is not philosophically interesting can be moot and subject to taste. However, when so many philosophers are exercised by an issue, it is worth wondering why they find it interesting. According to Cappelen and Dever, the main source of fascination with indexicality is its presumed connection with action. They argue that behind much of the discussion there is an assumption that can be reconstructed as follows: “Impersonal action rationalizations … are necessarily incomplete because of a missing indexical component” (2013, p. 37). Cappelen and Dever call this the Impersonal Incompleteness Claim (IIC).
As an illustration, consider again the case of Masha, who has now left Russia and applied for political asylum in Lithuania. Here is a very schematic impersonal rationalization of Masha’s actions (based on the template in Cappelen & Dever, 2013):
Belief: Masha is about to be arrested by the FSB.
Desire: Masha not be arrested.
Belief: If Masha leaves Russia, she will not be arrested.
Action: Masha leaves Russia.
According to the IIC, it doesn’t matter how much we flesh out and refine this impersonal action rationalization. It will remain inadequate until we make it personal, i.e. introduce the indexical I and revise the schema along the following lines:
Belief: Masha is about to be arrested by the FSB.
Belief: I am Masha, so I am about to be arrested.
Desire: I not be arrested.
Belief: If I leave Russia, I will not be arrested.
Action: I leave Russia.
In other words, the claim here is that we can’t account for Masha’s action unless we assume that at some point Masha formed a representation of herself as the target of an imminent arrest, of herself escaping the arrest, of herself leaving Russia, and so on.
Now, Cappelen and Dever’s goal is to show that there is nothing interesting about indexicality, so it is hardly surprising that they do their best to dismantle the IIC. In their view, an action rationalization can be fully adequate without any indexical element. Moreover, they argue that an adequate rationalization doesn’t have to involve any representation of the agent at all, whether indexical or non-indexical.
I believe that Cappelen and Dever are right to deny the IIC. I am also inclined to agree that non-substitutable indexicals are no more (or less) interesting per se than any other cases of referential opacity or context sensitivity displayed by linguistic signs. This is why I assume that the philosophical appeal of indexicality ultimately lies not in linguistic signs or their use. Rather, indexicals seem more “special” than other lexemes to the extent that they are lumped together with a decidedly non-linguistic phenomenon, namely the first-person view of the world – what Nagel (1986) calls the internal perspective and what others (e.g. Quante, 2012) have called the participant perspective.
I believe it is perspective, rather than indexicality, that has an intrinsic link to action. My reasons for this belief are largely empirical; more specifically, they come from studies of mirror self-recognition in non-human animals. This empirical evidence against linking action to the use of indexicals or any other linguistic behaviour is the subject of the next section.
2. Self-awareness in non-human animals and its implications
In this section I draw on observations of non-human animal behaviour in an attempt to show that the interesting phenomenon behind essential indexicality has nothing to do with language. To this end, I present a brief overview of empirical evidence for a particular kind of self-awareness in non-human animals. I then discuss Perry-style cases in light of this evidence.
To distinguish the kind of self-awareness I have in mind from other, perhaps more sophisticated types of self-awareness such as introspection and mental state attribution, I will call it Perrian self-awareness, or P-self-awareness. Recall the oft-quoted opening passage from Perry (1979):
I once followed a trail of sugar on a supermarket floor, pushing my cart down the aisle on one side of a tall counter and back the aisle on the other, seeking the shopper with the torn sack to tell him he was making a mess. With each trip around the counter, the trail became thicker. But I seemed unable to catch up. Finally it dawned on me. I was the shopper I was trying to catch. (p. 3)
The realization that he was the shopper with the torn sack prompted Perry to stop and look in his cart; in other words, it caused him to modify his behaviour in an observable way. If, on that momentous occasion, we had been following Perry around the counter, we could have deduced by watching his actions that he must have identified himself as the cause of the sugar trail. Here then is my working definition of P-self-awareness: def an animal, such as a human being, is P-self-aware iff it modifies its behaviour in a way that suggests that it has identified itself as the cause, or a cause, of a feature of reality.
Note that this definition makes no mention of language. To be clear, I am not claiming that Perry did not form in his mind a linguistic proposition along the lines of “I am the shopper I’m trying to catch” once he had realised that he was that shopper. Perhaps he did; after all, he is a highly educated human Western philosopher of the analytic tradition. What I do want to stress is that, even though we sometimes choose to give a linguistic form to the realisation that we cause a feature of reality, we do not have to. P-self-awareness does not need language. It can certainly be described using language, but it is highly unlikely that language facilitates it in any crucial way.
Consider the following scenario. A non-human animal, namely an Asian elephant called Happy by her human caretakers, is led into an enclosure. On the wall of the enclosure there is a 2.5-metre-tall mirror that Happy can approach and touch. On Happy’s head, well out of her sight, there are two marks made by a caretaker. One mark is clearly visible; the other is completely invisible; both are made using an odourless chemical compound.
As Happy approaches the mirror, she sees an image of an elephant in front of her. The elephant she sees has a large white mark on one side of its head. After contemplating the mirror for a little while, Happy realises that she is the elephant in the mirror. She is the elephant with the white mark on her head. As a result, Happy decides to investigate the white mark. She lifts her trunk and proceeds to touch the side of her head with the visible mark a total of 47 times before she becomes convinced that the mark is inconsequential. She then loses interest in the mark and goes on to do other things.
The scenario described above closely follows an actual run of the so-called mark test, as used by Plotnik, de Waal and Reiss (2006) to demonstrate mirror self-recognition in Asian elephants. One difference between my description and the actual experiment is that Happy the real-life elephant had been accustomed to the mirror during earlier sessions, and she seemed to realise right away that the elephant with the white mark was her. In other words, there was no demonstrable period of misattribution, when Happy would have held the false belief that the elephant in the mirror was another animal. I will soon return to this difference.
Over the last few decades, versions of the mark test have been used to demonstrate mirror self-recognition (MSR) in a number of non-human animals. MSR in chimpanzees was first reported as early as 1970 (Gallup, 1970). Since then, MSR has been shown in orangutans (Suarez & Gallup, 1981), bottlenose dolphins (Reiss & Marino, 2001), and the European Magpie (Prior et al., 2008), as well as Asian elephants. It is worth stressing that MSR is no mean cognitive feat. It may take a human child up to 24 months to start consistently recognising herself in a mirror (Reiss & Marino, 2001). A capacity for MSR strongly implies that the animal can identify a causal relationship between itself and the image in the mirror. In other words, passing the mark test requires what I have called P-self-awareness.
Let us now revisit Perry’s sugar trail case. As far as I understand, the impression that there is some profound indexicality going on there is largely due to the combination of an initial misattribution (“Another shopper is leaving the sugar trail”) followed up by a correct self-attribution (“I am the shopper leaving the sugar trail”). Our language allows us to narrate retrospectively the cognitive process involved in Perry’s example in a potentially infinite number of ways. The kind of narrative that makes the introduction of the indexical I seem profound goes like this. First, Perry correctly identifies the cause of the trail as a shopper with a torn sugar sack. However, his knowledge is incomplete because he hasn’t realised that he is the shopper with the torn sack. Once he has come to know that fact – once he has equated the mess-making shopper with the indexical I in his belief about what is going on – something profound happens, and he is moved to action.
Here is a corresponding schematic rationalisation of Perry’s actions:
Belief: A shopper is making a mess.
Desire: That the shopper stop making a mess.
Belief: If I find the shopper and tell him he is making a mess, he will stop.
Action: I look for the shopper.
(The action fails because the first belief is missing an indexical element.)
Indexical epiphany: I am the shopper making a mess.
Desire: That I stop making a mess.
Action: I stop and look in my cart.
Note, however, that we could just as easily use language to narrate the sugar trail incident as follows:
New belief: I [or whatever linguistic sign the speaker uses to refer to himself] am making a mess.
Desire: That I stop making a mess.
Action: I stop and look in my cart.
In other words: at first Perry didn’t realise he was causing a mess, and then he did. Indeed, we could simply say “Perry realised he was making a mess” and still convey the key cognitive element of what moved Perry to action. As far as Perry’s actions are concerned, the crucial change in Perry’s belief boils down to an instance of P-self-awareness: apparently, he identified himself as the cause of the mess and acted accordingly. Everything else is a more or less arbitrary narrative choice, made possible by the productivity of our language.
Now consider a schematic rationalisation of Happy’s actions in front of the mirror:
New belief: I am causing the image of an elephant with a funny mark on her head.
Desire: That I know what the funny mark is.
Action: I lift my trunk and touch the mark.
We could even leave out the mirror. After all, the mirror is arguably a mere tool that we use to test the elephant’s capacity for self-awareness. Here is another rationalization of Happy’s actions:
New belief: I have a funny mark on my head.
Desire: That I know what the funny mark is.
Action: I lift my trunk and touch the mark.
In other words: Happy realised she had a funny mark on her head. This mundane formulation still conveys the key cognitive element of what moves Happy to action. Everything else is a more or less arbitrary narrative choice.
Thus, I want to argue that what happens to Perry in the sugar trail case is no different cognitively from what happens to Happy the elephant in the mark test. When Perry realises he is causing a mess and stops his cart, and when Happy realises she is causing the mirror image with the white mark and lifts her trunk to touch her head, both display P-self-awareness in equal measure. The initial misattribution or lack thereof is irrelevant. For rhetorical purposes, I chose to narrate Happy’s interaction with the mirror in a way that parallels Perry’s narrative. I could have built the same narrative arc (misattribution – indexical epiphany – action) by using Gallup’s observations of chimpanzees. As demonstrated by their initial social responses, Gallup’s chimps would start out by misattributing the mirror images to some other chimps. Only later would they come to see the causal connection between themselves and their reflections and begin to use the mirror to explore various parts of their bodies (Gallup, 1970). However, at the end of the day such narrative choices may have as little bearing on P-self-awareness as does the stylistic choice between “Happy saw herself in the mirror” and “Happy visually perceived herself in the mirror”.
To sum up the discussion so far, I have argued that evidence of mirror self-recognition in non-human animals is also evidence of P-self-awareness. I have also argued that, as regards action, the key cognitive event in Perry-style cases boils down to an instance of P-self-awareness. I have related evidence of MSR, and therefore P-self-awareness, in non-human primates, cetaceans, elephants, and corvids. It follows that that at least some non-human mammals and at least one non-mammalian species appear capable of the self-directed, action-generating cognitive feat displayed by Perry in the sugar trail case.
As far as we know, none of the non-human species capable of MSR – with the highly speculative exception of cetaceans – use a sign system that displays more than a small fraction of Hockett’s design features of language. Therefore, it is very unlikely, to say the least, that P-self-awareness has anything to do with language. A non-human animal can’t tell a post-factum Perry-style story even while it can apparently engage in the same kind of cognition that moves humans to action in Perry-style cases. It follows that whatever is essential for action in Perry-style cases is probably not a feature of language. Instead, it is likely to be a phenomenon that obtains outside language and prior to language. The best candidate for this role that I can think of is the phenomenon of the internal, or participant, perspective on the universe. Our common evolutionary origins and our striking behavioural similarities suggest that we share the participant perspective with at least some other animals.
If this reasoning is correct, the real challenge lurking behind so-called essential indexicals is the fundamental problem of reconciling our non-linguistic first-person perspective on the universe with a linguistically mediated third-person perspective – or (to put it in terms of the so-called hard problem of consciousness) of incorporating our private experience into our theories, made up of public signs. This, in turn, has important methodological implications. If perspective and its connection to action are thoroughly non-linguistic phenomena, they lie outside the traditional domain of philosophy of language. This is the subject of the final section.
3. The philosophical study of perspective
Early on in their book, Cappelen and Dever (2013) discuss what they see as a potential problem with much of the literature on essential indexicality. In their telling, what often happens is that the authors discuss language and thought in one fell swoop, transitioning between linguistic structures and mental states as if the two always had to be isomorphic. Even while noting that, Cappelen and Dever choose to do the same in their own critique. One consequence of their choice is that they end up lumping together indexicality and perspective. To be fair, they don’t quite use the two terms interchangeably, but they do declare perspective and indexicality to be “real but shallow” in the same breath and, as far as I can tell, for the same reasons.
Admittedly, Cappelen and Dever have a valid reason for this undifferentiated treatment, given that they frame their choice of method as non-empirical (philosophy) versus empirical (psychology, neurology, cognitive science, etc.):
There’s no such thing as “doing psychology about very general truths about human beings” without empirical backing. The kind of non-empirical approach found in the Perry-Lewis tradition we are engaging with cannot establish contingent truths about how humans act… Armchair reflections about us moving our fingers won’t get us such conclusions; nor will philosophical reflections about generalizations or opacity. (2013, p. 55)
If the choice is between empirical science and philosophical analysis, it is understandable that both indexicality and perspective should end up on the same side – that of armchair reflection – and receive the same analytical treatment. However, I believe that the choice doesn’t have to be framed that way. Instead, the line of methodological demarcation can be drawn between two different areas of philosophical investigation: philosophy of language and philosophy of mind.
I take philosophy to be the practice of creating, combining, clarifying, evaluating and discarding concepts in an attempt to understand what is, or what ought to be. To paraphrase the Kantian adage, this practice is empty unless it involves some form of feedback loop between armchair reflection and empirical findings. I believe this is true of any area in philosophy; what varies is not the general approach but the relevant facts. As regards philosophy of language and philosophy of mind, the demarcation line is heavily blurred because so many of the relevant facts are the same. After all, language is a product of the human mind, and much of our mental activity is taken up with language. It is tempting to assume that the same concepts can help us understand both. Indeed, considering how many great thinkers have assumed that parsing language can help us understand anything, this temptation seems more than excusable.
However, there almost certainly are minds without language, such as those of higher animals and feral children, and there is purely computational language processing without minds. In empirical terms, one is clearly distinct from the other. Thus, it is not unreasonable to hold that at least in some cases the way we conceptualize linguistic phenomena tells us little or nothing about any interesting features of the mind. Indeed, in some cases conflating language and mind may be counterproductive – even if we use reflections on language as a heuristic rather than a theoretical commitment. As I argued in the previous section, we have strong empirical evidence that the phenomenon of participant perspective and its link to action may well be such a case. To use a somewhat crude analogy, observing the use of pronouns in the hope of understanding the internal perspective may be akin to counting the phonemes in the word cat with a view to understanding feline anatomy. Perspective is a fascinating phenomenon worthy of thorough philosophical investigation, and language can – has to – be used to build conceptual models that may help us to understand it. However, we have good evidence that language should not be treated as part of the phenomenon of perspective or a stand-in for it.
Cappelen, H., & Dever, J. (2013). The inessential indexical: On the philosophical insignificance of perspective and the first person. Oxford: Oxford University Press.
Gallup, G. (1970). Chimpanzees: Self-recognition. Science, 167, 86–87.
Nagel, T. (1986). The view from nowhere. Oxford: Oxford University Press.
Perry, J. (1979). The problem of the essential indexical. Noûs, 13(1), 3-21.
Plotnik, J., de Waal, F., & Reiss, D. (2006). Self-recognition in an Asian elephant. PNAS, 103(45), 17053-17057. https://doi.org/10.1073/pnas.0608062103
Prior, H., Schwarz, A., & Güntürkün, O. (2008). Mirror-induced behavior in the magpie (Pica pica): Evidence of self-recognition. PLoS Biol, 6(8): e202. https://doi.org/10.1371/journal.pbio.0060202
Quante, M. (2012). Person. 2. erweiterte Auflage. Berlin: De Gruyter.
Reiss, D. & Marino, L. (2001). Mirror self-recognition in the bottlenose dolphin: A case of cognitive convergence. PNAS, 98(10), 5937-5942. https://doi.org/10.1073/pnas.101086398