In a recent essay I made some, in retrospect, rather excitable claims about the possibility that a potentially insurmountable epistemic barrier might be emerging as a kind of fault-line within our culture. I wrote, of course, while impassioned, which is never wise, and cooler reflection has led me to doubt my own conclusion. This essay is therefore an investigation of the whole question of epistemic barriers. It does this in three sections. First it takes a second look at my original argument (made in my piece The Other in Culture) which used Quinean radical translation to prove that epistemic barriers are impossible and examine it very carefully for hidden assumptions, which, it turns out, are rife. The next section then analyses these assumptions in order to see when they break down, the point being to answer two questions:
|(1)||How dissimilar can two individuals be in culture and Weltbild and yet still not suffer from an epistemic barrier to converse?|
|(2)||How dissimilar do two individuals have to be before they do suffer an epistemic disconnect?|
The second of these questions is, obviously, the harder to answer, involving as it does some extremely imaginative Gedankenexperimente: in fact a fully adequate answer can only really be achieved by approaching the barrier from the other side, which is clearly impossible, and so we have already found that there are facts about this area of enquiry that are in principal unknowable. If an epistemic barrier exists we can have no idea of what resides on the other side, or of how many further barriers may exist beyond it. Therefore question (2) needs to be reformulated less ambitiously as:(2’a) Is it theoretically possible for epistemic barriers to exist?
|(2’b)||If they do, what, if anything, can we glean from communications issuing from the other side?|
|(3a)||Is there any way of proving the existence of an epistemic barrier?|
|(3b)||Is there any way of detecting communications emanating from the other side of an epistemic barrier?|
Answering these questions leads naturally onto the final part, where I take up again the question from The Male Gaze Gone Wrong of whether there is, in fact, an epistemic barrier within Western culture, dividing the self-focussed culture of ego-reinforcement from the more old-fashioned outward-looking culture. My conclusion is that there is not, but there is a very hard problem in radical translation.
Now, some of you will have already been put off by my rather specialised terminology. My article will rely on some key concepts that I will define here, simply so we all know what is going on, and you understand why I feel unable to use plain English. So, here we go.
- Epistemic Barrier: essentially a barrier to the free flow of knowledge between two parties. This is more than simply a language difference, because language differences can (in principle) be overcome by translation (see below). It is a systemic separation that cannot be overcome by any act of translation or representation, so on both sides knowledge exists, but the knowledge on each side is in principle unknowable by individuals on the other side. Obviously epistemic barriers are subtle and mysterious, if only because (as far as we know) we have never actually encountered one, and yet (as I shall argue below) they must exist. Their enormous importance arises from the fact that they bound the realm of the knowable, and so put us, as it were, in our epistemic place: we will not, can not know everything.
- Translation: means rather more in philosophy than everyday language-to-language conversion. Rather it deals with the communication of ideas from one source to another which may not share the original’s episteme (ontology, knowledge, beliefs, etc). Thus, in a sense, when I write these words and you read them, an act of translation is going on even if we both have English as our native tongue, because your episteme is naturally different from mine (we are not the same person) and so the meaning I place on these words as I write them may differ from the meaning you draw from them as you read them. Quine’s radical translation argument is an attempt to understand the limitations that are inherently placed on this translation process which derive from the fact that extension (truth value) is public, whereas intension (meaning) is private.
- Weltbild: has a wider meaning than any one English equivalent, hence my use of the German term. It is world-view, conceptual apparatus regarding the world, ability to derive information from the world, and essentially everything about how an individual experiences the world they live in. This differs from world-view in that it is not about attitude, it is something deeper and more inherent in our nature. Thus you and I may come from absolutely differing cultural backgrounds, and have no world-view in common, but we share a common Weltbild essentially because we are both (I hope) human, and so have a common way of being in the world, common senses, and common cognitive apparatus. This is crucial, so let me just ram it home: Weltbild is not about ideas, it is about the state of being in the world, one’s (to continue with the Germanisms) Sitz im Leben.
- Language: has a much wider meaning than natural language. It is taken here as being the entire conceptual apparatus we use to form, convey and analyse knowledge, so it includes natural language, symbolic languages (e.g. mathematics, logic) and other less formal modes of dealing with processing knowledge, such as imagery and symbolism. Thus my Weltbild is my epistemic state, while my language is the agent I use to act upon my Weltbild
3.1 The Proof that Epistemic Barriers do not Exist
I will now rehash the Quinean argument for (at least limited) mutual intelligibility, with the intention of then proceeding to examine the hidden assumptions (if any) on which it rests. Readers of The Other in Culture may recall that there I used Quine’s example where you and I are trying to communicate and you are endeavouring to determine what I mean by the word ‘gavagai’, which you suspect inhabits the same region of epistemic space as your word ‘rabbit’. I shall not use that example here, the argument being more abstract, but it is useful to keep it in the back of the mind, in case the abstraction becomes overwhelming. Say you and I are communicating. Then, in endeavouring to understand what you say, the information available to me is:
- The sentences you utter
- The truth value you assign to them
- The collection of sentences I am able to formulate
- The truth values I assign to them
I also make two key assumptions:
- The principle of substitutivity: if S in your language translates into S‘ in mine then they should agree in truth value in all possible linguistic and epistemic contexts.
- A theory of truth: the rule whereby I can relate the truth-values of sentences in differing contexts to their structure and syntax and the truth values of their constituent objects and predicates.
The principle of substitutivity is fairly obvious: essentially it says that if S and S‘ correspond then one should be true or false precisely when the other is, and vice versa. This is little more than a statement of the principle of extension from set theory (if two sets have the same members and the same non-members then they are the same set). The theory of truth seems equally unproblematical, but in fact the meta-theory of theories of truth is immensely complex. Let it merely be said that there are many theories of truth, that it has been proven that no theory of truth can be complete (in that there will always be sentences that are neither true nor false), and that different theories can give rise to very different results. Those wishing to learn more should consult Understanding Truth by Scott Soames. So how to I apply these assumptions to the data available to me? I play a complex matching game, endeavouring to find correspondences between your sentences and a subset of the sentences available to me. This will, in general, be extremely hard, because if you happen to say (for example) Klaatu burada nikto, then unless I am given some clue as to your intention, or I have a very large body of your other sayings, pinned to precise epistemic contexts, I am not going to get very far in translating it. This is why when discussing radical translation we generally simplify the game by using ostensive sentences, that is, the sentence that you utter is accompanied by some gesture or other meta-linguistic communication that indicates that your sentence refers to some part of the epistemic context in which we both find ourselves. This reduces the number of sentences that I have to evaluate in order to guess at your intention. Obviously we start with very simple sentences like ‘there is a rabbit’ rather than with recitations of national epics! The goal of all this is to reach a point where you utter a sentence which is held to be true, and I am reasonably sure of what that sentence is. I do this by looking for substitutivity: that is to say, I seek out circumstances where your sentence is or is not true and compare it to my sentences in those circumstances. The correct sentence(s) should be substitutable into any context and give the right answer upon substitution. To cut out the technicalities (of which I will note only that they depend vitally on my choice of a theory of truth, because I need to be able to get from truth or falsity of sentences to truth or falsity of predicates and vice versa), we reach the point where you and I can form sentences and I will know that your sentence S will be true precisely when my sentence S‘ is true. The question now is, does this mean that I know what S means? And the answer is: most emphatically, no. I know the extension of (potentially) every term and predicate in your language (by which I mean I know its truth conditions and how to use it in the process of forming valid sentences). But as for its intension, which is that which makes it be true or not true in the first place, i.e. its meaning, there is no way I can know that, because all I have been able to use in my act of radical translation is truth values and substitutivity. So I know (it seems) little or nothing about meaning. For example, unless one lives in arctic regions the sentences it is daytime and there is sunlight are substitutable, but their meaning is clearly entirely disparate; one can distinguish them eventually, but that involves some quite deep analysis of pairs of sentences, and establishment of complexes of analogical concepts. That was a problem caused because we had not probed far enough before forming our translation model. So, now consider the German phrase gut Bürgerliche Küche. It translates straightforwardly as good middle-class cooking, and yet the meanings of the two phrases are entirely different. The German has strong positive connotations, while the English has, if anything, faintly negative connotations. Our different cultures create resonances that are simply not there in the words themselves, and can only with difficulty be established using substitutivity (to do so would be very complex indeed, requiring a comparison of emotional reactions, which immediately calls into doubt the validity of the translation process). Now consider the ultimate distinction: that between individuals. The sentence This is sweet is simplicity itself. And yet what does it mean? We can say This is triggering certain taste receptors on my tongue, but though the extensions of those two sentences are identical (they are completely substitutable), the second does nothing at all to illuminate the former. I know what I think sweet means, you know what you think sweet means; but how much of that can we communicate? The sensation of sweetness is too direct, too immanent for us to describe: rather than being decomposable into parts susceptible of discussion and analysis. It is atomic and thus indescribable, other than in terms of its presence or absence, even though it is a very intense sensory experience. What this means is that I can discern by radical translation, precisely as much of the meaning that entities and predicates have for you as you are capable of expressing linguistically and no more. But then, if you cannot express those missing shades of meaning linguistically, it is not only I who loses them, but also those who share your language. Thus in principle, anything you can communicate in your own language, you can communicate to me. Thus, if this argument does, in fact hold, there can be no epistemic barriers. Now, is this a realistic model? Yes. When you and I communicate, even if we do it in the same language, we have to assume that we assign the same truth values to linguistic units, which we verify by comparing the contexts they appear in, and also that we share a common way of building linguistic units (with truth values) into true sentences. We know nothing of each other’s internal state: only of that which can be discerned externally (and which can all be lumped together under the heading of ‘linguistic units’ with little violence) is available. And though this approach may be criticised for reducing everything to the black-and-white world of truth and falsehood, it is not clear that a more nuanced approach would work any better, given that by using truth-value and substitutivity I can translate adequately all that is expressible.
Progressing through the argument, the hidden assumptions more or less scream for attention, but it is still worth discussing them. So, I will start by listing and defining them and then progress to a general discussion in which I aim to crystallise the key points that may permit the existence of epistemic barriers after all. Let me first of all, however, point out that it should be clear from the analysis above that the argument is extremely strong, and moreover it is supported by our experience of translation to date, that is to say the creation of mutual intelligibility between speakers of all known natural languages. It is therefore not surprising that the existence of epistemic barriers – systemic barriers to communication – is generally unsuspected or even denied. So, the main assumptions are as follows.
Substitutivity seems, at first sight, to be watertight, as is little more than tautology. But there is in fact a meta-assumption underlying it. That is to say, we assume that given languages L and L‘ then there should be a consistent translation relation between the two languages, so sentences S and S‘ are related precisely when they are substitutable. But what if it proves impossible to define such a relation? That is to say, what if it is impossible to find a consistent relation between S and any sentence(s) in ‘L‘, so S appears, to a speaker of L‘ to have no fixed interpretation, in spite of the fact that it is clearly meaningful to a speaker of L? In this case the principle of substitutivity is true but useless, because we have nothing to apply it to. If sentences are not substitutable then we cannot compare them across epistemic contexts, and so the whole enterprise of radical translation collapses.
- We must consider the impact on substitutivity of a failure of two languages to form a consistent translation relation.
The theory of truth is clearly an enormous assumption, with many ramifications. There seems to be a rather naive assumption that all communicating beings will use a two-valued Aristotelian logic, and yet we find that three-valued logics (where the values are ‘true’, ‘false’ and ‘undetermined’) are almost essential if we are to deal with truth in any rigorous way, not to mention some of the other standard logical paradoxes, while early Indian philosophers made use of the dialethic four-valued logics (‘true’, ‘false’, ‘neither true nor false’, ‘both true and false’). Though this may seem very abstract, ones choice of logic is a key determining factor in how one sees the world (or possibly the other way round) and so may limit ones ability to comprehend the function of sentences in an alien language. For example, if I use three-valued logic while you are rigidly two-valued, many of my statements must seem to you perverse or unaccountable, for deductions that to you are obvious are not to me, and thus substitutivity starts to fail. We can go further. The theory of truth, in that it is a way of deriving the relation between the truth value of a sentence, the truth values of its components and the sentence’s structure, makes some very fundamental assumptions about the structure not just of the known language (which is legitimate), but of the unknown language. It so happens that the languages we know of follow broadly the same syntactic structures (though there have been sporadic reports of such oddities as languages without verbs, they remain unverified) but to insist that all languages must be like human natural language and are formalisable within the languages of logic that we have so far discovered seems unreasonable. It can, and no doubt will, be argued, in rebuttal of this claim, that it is inconceivable that there should be a language that is not based on our categories, but there are two rebuttals to the rebuttal. First, and rather obviously, it is all we know, so obviously anything else seems inconceivable. Second, and more insidious, language shapes our episteme, what we can know and conceive of, is shaped by our Weltbild, that is to say the world of ideas that we and our imaginations inhabit, the way we ontologise the world around us, etc. If I do not have the epistemic apparatus required to conceptualise an idea, then I cannot conceive it. That sounds trivial, but it has enormous consequences. So if L has one syntactic structure and L‘ another, incompatible, syntactic structure, then substitutivity is bound to fail, and it will be impossible to establish a translation relation. This is simply because in translating from ‘L‘ to L we will be forcing L‘ into the structures imposed by L. As I indicated above, even when translating between two such closely related languages as German and English some information is lost. If we go to the extreme case of two languages with incompatible grammars, then information will inevitably be lost, as well as false information being created, because concepts expressible in L‘ will be inexpressible in L, and hence translate into nonsense, or fail to translate at all. Another way of saying this is that if I am to translate you effectively, then my language must be large and complex enough to be able to express any concept expressible in your language. This is not necessarily the case.
- Our language’s syntax shapes the way we approach and translate other languages, and limits the extent to which we can comprehend ideas expressed in alien languages.
- If I am to translate faithfully from your language to mine then my episteme must encompass yours (else there will be untranslatable concepts)
- Our choice of logic shapes the way we will model truth values.
There is another, even deeper, assumption relating to truth. In the discussion of theories of truth above, there is a hidden assumption that there is essentially one truth, but many different ways of finding it, so we did not question atomic truth: what it meant to apply a simple predicate to an object and to assert that the result is true. All of our discussion simply took atomic truth as read and dealt with the problems inherent in building from atomic truth to the truth (or otherwise) of complex sentences. But there is no guarantee that there is only one kind of atomic truth. For example, consider the atomic predication this stone is red. In our language, L, that means that the stone is coloured red at the precise instant that the sentence is asserted. If it had been blue five seconds before and was green five seconds later, that would not affect the truth of the assertion. But it is equally possible to imagine a language L‘, that takes into account the whole life of the object subject to predication, according to which, this stone is red required that the stone (to the best of the utterer’s knowledge) has always been red, is red, and will always be red. L and L‘ are both internally consistent, and so lead to sound theories for complex sentences which are in themselves consistent and as complete as any theory of truth can be, and yet they are utterly incommensurate. Consider for a moment just how alien L‘ is. We could see a red stone, and so assign the value true to this stone is red, and yet two individuals whose Weltbild is defined by L‘ could disagree as to whether it was red if (say) one of them had seen it before it was painted, while the other’s experience of it is limited to times after it was painted. Similarly, aL‘ speaker could assert that a traffic light was red and amber and green simultaneously without any contradiction. The apparent incoherence of truth in L‘ arises because the truth of sentences evaluated now depends on events at other times or epistemic places, and is a characteristic of intensional logics, such as modal logic or (as here) tense logic, but our logic in L is extensional, and so, from our point of view, L‘ is incoherent and nonsensical. In other words: the principle of substitutivity fails because L‘ is intensional. In fact, in this case we could save the situation because though truth in L is extensional, L contains intensional elements, meaning that we can express L‘ within it, and so a suitably modified form of the principle of substitutivity (with substitution into equivalent intensional contexts replacing simple extensional substitution) would be true. But, it is, of course inevitable that our language can contain encompass L‘; if it could not then I would have been unable to specify it so precisely. My point was to show that even a concept of truth very close to our own can result in an apparent collapse of the mechanics of radical translation. Generalising, languages which make use of such concepts as metaphorical truth, symbolic truth, moral truth or mythological / religious truth have the potential to be completely disruptive, because though I can broadly describe each of them as having an intensional truth function, now, unlike L‘, I cannot necessarily describe them in L, because their Weltbilde will contain schemata of cultural assumptions which are not identical to my own, and to which I have no access, as they are hidden behind a linguistic barrier. Thus such a language must be incomprehensible to me, as I have no common point of reference (which I did in the case of L‘) against which to evaluate sentences expressed in it (an elegant example of this is the Ancient Egyptian concept of Maat, which means – amongst other things – truth, but is more complex than extensional truth, including as it does concepts such as cosmic order, and is now rather mysterious because the cultural background that made Maat precise to an Egyptian was never written down, and hence is lost to us). So, even within the bounds of the describable, it is possible to imagine concepts of truth that are sufficiently alien that radical translation may break down. But there is no reason to assume that every concept of truth must be describable within our culture’s Weltbild. And with these indescribable concepts of truth, we have to assume that radical translation must fail, for if we cannot even describe the conditions under which a language marks a sentence as being true, how can we achieve any consistent translation from it to our language? As I hope I have demonstrated, the concept of truth and cultural baggage, or Weltbild are inextricably linked, so it is naive to assume that radical translation will work under all circumstances.
- Radical translation fails if the Weltbilde of the two languages do not have a common concept of truth, which means there must be sufficient commonalty of cultural background.
Another hidden assumption of radical translation is that the speakers of the two languages share, if not a common ontology then at least commensurate ones, so there is a discoverable rule that links objects in the two Weltbilde. That this is true should be fairly obvious: we cannot even begin to analyse sentences to discover structure and truth-value if we have no way of expressing or comprehending the objects referred to within those sentences. So, for example, a Weltbild whose ontology identified all things of a kind would prevent the formation of n adequate translation relation. And this would apply in both directions, because clearly the possessors of this Weltbild would have no concept of individuality or self, while our Weltbild and languages are predicated on the distinction between self and other. Thus, though we and they could possibly communicate very simple concepts (though even those may be fraught with complexity: so simple a sentence as it is not raining may fail, because to our interlocutor it is always raining somewhere, so it is always raining) any sentence involving quantification or individuals would fail to translate. This is now going beyond a difference due to one truth-function being intensional and the other extensional; we are reaching the point where the languages are simply too different even for us to be able to make such a statement: the concept of truth is simply other. Before it be said that this is merely a Gedankenexperiment, consider an even wilder Gendankenexperiment which turns out to be something of a parable. Say we are trying to communicate with a being who perceives not, as we do, macroscopic objects, but rather quantum wave functions. So, if I point and say rabbit this will be absolutely meaningless to the quantum creature, because as far as they are concerned there is always a rabbit everywhere, albeit with shifting probability distributions. So our ontology, which is based on precisely located discrete objects, and their ontology, which is (at a guess) based on shifting globalised objects with no clear individual existence, are so much at variance that it is hard to see how we could even begin to establish a translation relation. Indeed, as I have observed elsewhere, much of the vaunted ‘difficulty’ of quantum mechanics is likely to stem from a misguided attempt to force the language of the quantum creature into the straitjacket of our definitely non-quantum Weltbild. We should be impressed that we have achieved even our rather limited understanding of the quantum realm, which suggests that even if we can never truly understand a language on the other side of a barrier, we might be able to abscond with some crumbs of knowledge.
- Incommensurate ontologies can prevent the establishment of a translation relation.
- Though full translation across such a barrier is impossible, we may be able to achieve very limited and fragmentary partial communication.
Finally, there are a number of meta-linguistic issues that focus on two areas. FIrst, how do we know when an ostensive sentence is being uttered? And how do we know what are the objects being picked out? Or, conversely, how do we know that if we point at an object and utter its name, our interlocutor will understand out intent? The answers are simply we can’t, we can’t and we don’t. We have to make some behavioural assumptions as regards body-language or physical behaviour. And while this is all very well when dealing with other members of our species, or even closely related species, it would not work if we were trying to establish communication with (say) the quantum creature of the previous sub-section, to whom the whole concept of ostention would probably be meaningless anyway. The second issue is more insidious. For radical translation to work I have to know when my interlocutor thinks that what they are saying is true or false. I could assume that if they point at something red and utter hokak then they are asserting that it is red is true, but that is my assumption, and, for all I know, they could be saying it is not blue or even (plausibly) it is green, knowing that the assertion is false, for why should they not have an apophatic language which works by making false assertions? By assuming the truth value based on how my Weltbild dictates I use language, I am imposing the structure of my Weltbild on theirs, and so, if I do eventually arrive at a translation, I will seriously misrepresent what they are trying to say. And – here is the key point – there is no reason to assume that I would ever discover my mistake. As an example of this, consider Athanasius Kircher’s extraordinary ‘translation’ of Ancient Egyptian, which foundered on his assumption that the Egyptian conception of truth was allegorical rather than simply extensional.
- Radical translation relies on at least some common form of communication, if only at the level of body-language, so in fact before we start we already assume something out the other.
- We must not assume that our interlocutors use their language to convey meaning in the same way that we do; to do so only results in a translation that misrepresents them.
There is little more that needs to be said. It is clear that radical translation will only work if the two parties involved have Weltbilde that are sufficiently close that concepts existing within one are expressible in the other. We have seen that it is entirely plausible that there should be Weltbilde sufficiently disparate that this is not the case, and thus that in these cases radical translation will fail, meaning that we will have discovered an epistemic barrier. The only crumb of comfort is the suggestion that some very low-level communication may be possible such a barrier, though inevitably information would be lost when passing even the simplest concept in either direction. We shall return to this in the next section.
4 On Epistemic Barriers
4.1 Language Shapes Episteme
I observed above that the structure of our language (interpreted in the widest sense as being essentially equivalent to Weltbild) limits what we are capable of knowing, and therefore shapes our episteme. A consequence of this is, naturally enough, that it is entirely possible that there should be things that we can never know because our Weltbild cannot contain them. That is to say that the existence of epistemic barriers is possible. Now this is a rather startling conclusion from what was, when first mentioned, treated as if it were merely obvious. Let us therefore examine how Weltbild shapes episteme, taking as our starting-point three examples: one from physics and two from mathematics. The example from physics is our old friend, quantum theory, specifically the infamous double-slit experiment. In this we shine a beam of electrons at a barrier with two slits in it and measure the density of electrons emerging from the other side. Now we assume that electrons are particles, because in other experiments they have behaved like particles. But in this experiment it turns out that they are behaving like waves, and apparently have the ability to pass through both slits at the same time. In our normal Weltbild this makes no sense at all: something either is a particle or a wave. When one switches to the Weltbild of quantum mechanics these problems go away: what was a sometimes a particle sometimes a wave is revealed as an entity in a vast (infinite dimensional) space, and it turns out that though there is a way to relate these entities to our Weltbild, it is neither simple nor very well-defined, so it is entirely possible for one entity to at the same time look like both a particle and a wave, depending on how you view it. Now to mathematics and the concept of higher-dimensional space. For someone not trained in mathematics, the idea of a space with four dimensions is utterly mysterious; they have no way of conceptualising it and no machinery with which to mentally take hold of it and analyse or envisage it. However, with sufficient mathematical training one reaches the point at which higher-dimensional geometry is seen as merely a natural generalisation of our three dimensions, and eventually one gains the ability to conceptualise geometry directly in terms of algebraic objects: every geometer will be aware of the moment where they stopped trying to draw pictures in their heads and started to think conceptually instead. So while mathematicians cheerfully talk about four, seven or even infinitely many dimensions, the lay-person is left floundering. Now consider something even stranger. For most lay-persons the concept of infinity in itself is rather troubling; indeed, the formal definition of an infinite set seems simply counter-intuitive. Thus the idea that there might be different kinds of infinity, with some larger than others is simply incomprehensible. And then when we move on to observe that given any infinite number one can always construct a larger infinite number, that therefore there is an unbounded hierarchy of infinite numbers which is itself infinite in extent, and that there are some infinite numbers so vast that they cannot be obtained by operating upon any of the infinite numbers smaller than them and yet even these numbers have others bearing the same relation to them as they do to smaller numbers, something akin to panic is likely to set in. And yet, with sufficient mathematical experience one can reach the point of being able to work with these ideas and to have some form of comprehension of what they mean. What lessons can we draw? The example of quantum mechanics shows a case where radical translation is perhaps failing, in that the translation relation is extremely ill-defined. In fact, it is entirely plausible that the quantum Weltbild is so alien to ours that it is on the other side of an epistemic barrier, and so our current difficulties with formulating a consistent, coherent quantum theory are inevitable given that, for us at least, comprehending the quantum realm is impossible. As another example of the pitfalls of this kind of inadequate translation, consider the case of the grammaticalisation of the Finnish language. Non-fennophone linguists approached Finnish as if it were Indo-European (which it is not) and attempted to systematise its syntax based on a Weltbild derived from the study of Indo-European languages. Thus we end up with a language in which apparently there are fifteen cases, four types of verbal infinitives, some of which decline (?), and syntactical rules so complex as to be nearly meaningless. As with quantum mechanics, difficulty is a consequence of badly-mismatched Weltbilde, and in both cases the result is that the language being translated from (the quantum realm, Finnish) becomes almost incomprehensible when translated. We simply cannot fit the ideas into our episteme because our language, our Weltbild does not permit it. A second observation from the quantum-mechanical example is this. The initial reaction to the double-slit experiment is a conceptual failure because we expect that something is either a particle or a wave. It cannot be both. So when something acts as if it is both then we have no recourse but to panic. But why not both? Largely because the concept that entities as we experience them are ‘real’ in some sense is deeply embedded within our Weltbild, and we do not expect ‘real’ things to be of two kinds at once. In spite of the startling ease with which it is possible to fool them, we believe our senses precisely because if we do not then what do we believe? Our animal inheritance tells us to act on what we sense and treat perceived things as being concrete. Thus the idea that a perceived thing is in fact only a representation of a larger and more complex reality, forced to take its particular form by the limitations of our senses, is not one that comes easily to us, let along the obvious conclusion that the entire realm of what we perceive is essentially a construct derived from outside reality (if there is such a thing) by our limited ability to translate from its language into one that we speak (in the widest possible sense). It is perhaps worth stepping back and taking stock, as we have entered some very deep waters. The conclusion I seem to have drawn is that what I am and what I know, so the totality of my Weltbild, limits what I am able to comprehend. Is this particularly surprising? I don’t think so. Just as people had no concept of micro-organisms prior to the invention of the microscope, if we do not have the machinery to detect something then we won’t. And if we have perceptual apparatus that can partially detect something, then we will perceive fragments of it which we will then translate as best we can, given that we are lacking some (maybe the crucial part) of the information required to make an accurate translation (think of what a foreign language sounds like to one who cannot speak it). But turning from senses to ideas, the two mathematical examples show that without the right conceptual toolkit certain areas of epistemic space are inaccessible to use, whereas once we have acquired the toolkit, we can navigate them with (relative) ease. Indeed, the example of geometry, where something in the mathematician’s head ‘clicks’ and they pass from having to try to model geometry using two or three dimensional analogies to being able to see directly the geometrical significance of the algebra, is a clear case of someone learning a language and reaching the point where they no longer need to translate back to their native tongue but can work confidently in the new language. So there are ideas that we are unable to grasp unless our language is expanded in an appropriate way (as a very mundane example, anyone who can speak German will know what gemütlich means, and yet it cannot be translated adequately into English), but then we will have difficulty translating those ideas back into the base language. In other words, we are seeing, not an epistemic barrier, but more a kind of epistemic one-way road: one can expand ones Weltbild to allow the conceptualisation of new ideas, but there is no way of taking ones new knowledge back home. In conclusion then, language clearly does shape episteme, in that it dictates what knowledge we are able to comprehend, and may render knowledge that lies beyond these bounds as nonsense. We can, in an irreversible process, expand our language to encompass new concepts, and hence expand our range of epistemic possibilities. Now, our base language is clearly not complete, where by complete I mean is capable of encompassing all concepts, or, to put it another way, it covers all epistemic space. So, as our base language is incomplete, why should the extension be complete? Clearly this is highly implausible, particularly as we cannot prove that a language is complete, for how can we prove that there are no ideas it cannot encompass, particularly in view of the fact that ideas alien to a Weltbild may well be entirely invisible from within that Weltbild? On the flip-side, we cannot, from within a language prove that it is incomplete, because to do that we would have to be able to conceptualise ideas that the language cannot conceptualise. Therefore we can only know that our language is incomplete once we have extended it. So we are in a state of permanent uncertainty as to whether there is or is not fresh epistemic space to be explored. Therefore we can never guarantee that there is not knowledge ‘out of reach’, and therefore epistemic barriers are not theoretically impossible, answering question (2’a).
4.2 The Existence of Epistemic Barriers
In fact, I can strengthen this. Introduce a relation between languages, so L < L‘ precisely when L‘ is an extension of L that grants access to greater epistemic space. Also, identify languages that are susceptible to perfect translation, so we are talking here of classes of mutually intelligible Weltbilde. Say that L is maximal if it is inextensible, so there is no L‘ such that L < L‘. Clearly a complete language C (if one exists) is maximal, but can a complete language exist? Let us restate the question: can maximal languages exist? Can a language have internal tensions that prevent it from being extended? Unless one takes a very essentialist view of knowledge, and seriously believe that there is, as it were, an epistemic terrain waiting to be explored, then the answer must surely be no. For example, some natural languages have highly restricted number systems, so it is impossible to count above some specific small number. So it would be impossible to express any truly mathematical concepts in such languages. But that does not prevent speakers of those languages from extending their epistemic range to include mathematics for the very reason that makes it possible for us to know of the language’s limitations in the first place: that is to say that translation between these languages and languages with unrestricted number systems is achievable. Certainly, individuals have to learn an unrestricted language at an early age, before the restricted Weltbild is locked in, but the point is that being a native speaker of (say) Mangarrayi does not place an irrevocable epistemic barrier in ones way. So, generalising from this, it may be than an individual natural language is inextensible, but that does not prevent it from being translatable into a more epistemically flexible language that can be extended. This leads to an interesting point. If the speakers of Mangarrayi had never encountered European colonists then an external observer might have thought that their language was inextensible and incomplete. But, as we know, it was extensible (this is where the distinction between the epistemic languages we are discussing and natural languages is crucial: the natural language Mangarrayi is inextensible, but the epistemic language of its speakers is extensible). This comes back to the point made above that we cannot make any predictions about the path forward; we can only look back. For all we know our current language may be maximal, but all the evidence suggests that it is not, for it has been extended before, and, as we have just seen, apparently inextensible languages can be extended, and moreover, just as we cannot prove that it is not maximal, we cannot prove that it is. All of which comes down to a strong argument for the in-principle non-existence of maximal, and hence complete, languages. We need a new definition of completeness. If the speakers of Mangarrayi had never met Europeans (or any other group whose language had a complex number system) they would most likely never have developed mathematical thought. And yet, clearly their Weltbild gave the potential for mathematics within their episteme (in the same way, anyone who understands simple arithmetic has the potential within their Weltbild for an understanding of transfinite arithmetic). Generalising, all we can know is where we are now and where we have been; we cannot know where we will go, because that is contingent on unknowable future events. Thus in speaking of epistemic potential, which is precisely what we are speaking of when dealing with epistemic barriers, we must consider as the completion of our current state that portion of epistemic space from which we are not separated by an epistemic barrier, which is precisely the union of all possible languages extending our current language. Strictly speaking this union may not itself be a language (the mathematically inclined can fill in the details if they so wish), for clearly not all of its constituent parts need be mutually intelligible, but it is a member of a larger category of things that I will call language*, whose precise structure I will not bother to elucidate overmuch, beyond noting that every language is also a language*. So, a language* is *-complete if it is the union of all languages extending some language. Call the complete language* derived from a language in this way its *-completion. Clearly this is inextensible (which shows that it is not itself a language); we can also prove that any extensible language* is a *-completion. Let C be inextensible in language*; it contains some language L. Then C must contain the *-completion of L, because if there was some language L‘ that extended L but was not contained in C then we could extendC by adjoining L‘ to it (this extension is legal precisely because L‘ is the extension of something already in C). So all inextensible elements of language* are *-complete (note that going to the larger category of language* allows us to make statements that are unprovable, or simply untrue, of languages simpliciter). So we can now formally state what it is for there to be an epistemic barrier between two languages: it is simply that their *-completions are not equal. Or, in other words, the potential epistemic spaces, given the two languages as starting points, are distinct: there are ideas accessible in one that are inaccessible in the other. So we can divide languages into epistemic classes, which are languages which have the same *-completion, and hence which are potentially intercomprehensible. And now, here is the point (at last) of all this formalism: there is no a priori reason why there should be a unique inextensible language*. In fact we can know little about the category language*. For example, the only members of language* we can know are the languages simpliciter that are comprehensible to us; to know a non-trivial language* we would have to know an infinite series of languages, which would require us to predict unknowable future events, so it is very hard to determine the characteristics of a *-completion. Now, De facto the only elements of language* we know are those in the same epistemic class as our own. Anything else is either a language, which cannot be comprehensible to us, as it has a different *-completion, or the limit of a series of such languages. Putting this together, we conclude that there is no way we can determine what is an inextensible language* or how many of them there might be. This means that though the concept of language* is theoretically powerful, we cannot, say, use it to discover alien *-completions and inspect them to discover the form (if not the content) of languages on the other side of an epistemic barrier. Which is interesting, because, though it is clear that we cannot understand a language on the other side of an epistemic barrier, we might have hoped to at least be able to recognise one as being a language. This suggests that even that is not possible, a point to which I shall return. So, if a language lies on the other side of an epistemic barrier, then, by virtue of it having a different *-completion to our language, there can be no point of contact between it and any language we have or might some day have. We must remain forever mutually unintelligible and unrecognisable. Thus I am close to answering question (2’b) in the negative, and to giving decidedly discouraging answers to (3a) and (3b). Answering these questions solidly is the purpose of the next section.
4.3 Seeing Across the Barriers
This section takes stock of what we have seen already and then tries to answer one complex question (which has questions (1), (2’b), (3a) and (3b) inherent within it), that is: what, if anything can we know about languages and interlocutors on the other side of an epistemic barrier? I shall discuss this under three headings. First, what could we do if confronted with something that we knew was a communication, but which emanated from the other side of an epistemic barrier? Second, can we detect epistemic barriers? And third, assuming that the communication were not served up to us on a plate, is there any way we could detect communications from the other side of an epistemic barrier? That is, not translate them, but simply know that they are there.
Suppose that we have, somehow or other, obtained access to source of information emanating from the other side of an epistemic barrier. Let us not inquire as to how we could do this, and how we could know that that was what it was: those issues are the subject of the next sub-section. The purpose here is to examine question (2’b): having obtained this source of information, what can I do with it?
Let me start out with an analogy. Suppose that when we looked at the world, what we saw was pixellated. What I mean by this is that our visual system would divide up the world into cubes, and then within each cube we would see, not detail of the things within it, but a shade of grey whose brightness depended on the amount of ‘stuff’ in that cube, varying from black (empty) to white (completely full). To give some idea of what I am talking about, here are two images.
The left-hand image is what we usually see; the right-hand image is the result of pixellating over 16×16 squares, so every square of 256 pixels is averaged to a single shade of grey. As you can see, if you have seen the original picture, you can spot the parts of the processed picture that correspond to the hat and Madame Vigee Lebrun’s decolletage, hand and (with a little imagination) easel (for some reason her head is largely lost). But that is the point: you are using the original picture as a guide to discern features in the pixellated version. What would you be able to determine about the image if you had only the pixellated version to go on? Well, there’s one way to find out. Look at the next image.
Here we have nothing to go on. If one felt very brave one might hazard a guess that the square was a person, with the white blob at the top being their head, and the truly adventurous might hazard that the mass of white on the right-hand side of the square was something that the person was holding. And that’s about as far as we can go. Certainly there is nothing to indicate that this is, in fact, a pixellated version of the infamous Madonna with bared breast, said to be a portrait of Agnes Sorel. Now the point of this is not just to show how easy it is to see what you already know is there (though that does, undoubtedly play a part in the translation process when dealing with languages that are within your epistemic class, and so which are, to a greater or lesser extent, familiar in their structure), but rather to draw an analogy to the problem of communicating across an epistemic barrier. We on our side of the barrier are, with respect to the alien speaker of the far-side language, in the position of the viewer of pixellated images relative to the viewer of the whole image (though, counter-intuitively, the same will be true of the alien with respect to us, so there is a key distinction between the analogy and the epistemic situation in that in the epistemic case both parties lose information). That is to say, much of what is being communicated will appear to us to be total nonsense, as we cannot apply any form of the substitutivity principle, meaning that translation fails, or worse still, we may not even realise that it is there. But, if we work on a coarse-grained enough level, we may be able to abstract a very partial, fragmented and distorted notion of what is going on. So it may be possible to extract some knowledge from the communication, but there is no guarantee that it is anything that the original speaker would recognise should it be possible for them to examine it, which, of course, it is not.
This, last observation leads into what starts as an aside, but becomes crucially important to our argument. Say my language is L and yours L‘ and we have established a translation from L‘ to L. There are only two ways for you to verify the accuracy of my translation of one of your communications. First, you can translate it yourself, and compare the two translations, but then any distortions inherent in the process of translation will remain invisible, as they will apply equally to your translation and mine. So the test whereby you examine my translation into L and consider it as a communication in L is worthless precisely because you have then (unconsciously) taken on my Weltbild simply by electing to use L. So, the second approach is for you to translate my translation back from L into L‘ and then compare it with the original. But this is a notoriously risky business, even with closely related languages: any text distinguishing between pigeons and doves or slugs and snails will become quite garbled in the course of the round-trip from English to German and back again. Also, there is the completely undecidable question of whether the difference between the original and the doubly-translated version arose in the original translation from L‘ to L or the second translation from L to L’. Undecidable because in order to make that judgement we need to have a reference point in both languages, hence a guaranteed ‘correct’ translation of the original information into L. But it is precisely the ‘correctness’ of the translation that we are trying to determine. Therefore, there is no way of judging the correctness of a translation. The best we can do is to seek to reduce the ’round-trip’ distortion in both directions (so L‘ -> L -> L‘ and L ->L‘ -> L). Thus a best possible translation is one that minimises this error (quantified in some pre-determined way), which I will call the epistemic differentiation. Note that as it is the minimum over all possible translations, the epistemic differentiation is not a measure of the adequacy or otherwise of any one translation (indeed, there may not be a realisable best possible translation): it is systemic, a measure of the epistemic distance between two Weltbilde as expressed in their languages, with 0% corresponding to identity and 100% to complete isolation. So, for English and German, the epistemic differentiation is tiny: there are only a few concepts, such as those I have mentioned in passing in the course of this essay, that either do not translate at all or do not translate uniquely. For English and Japanese the epistemic differentiation is far greater, as English cannot express the relative social standings of speaker and spoken-to. But that is not to say that translation between English and Japanese is bad, rather it is the best possible given the epistemic distance between the two Weltbilde. So we must give up entirely on the idea of having perfect, or even good, translations: a translation can lose 98% of the information in the original communication and still be the best possible. If we now take L and L‘ to be on opposite sides of an epistemic barrier, then the epistemic differentiation will be close to or equal to 100%. If it is 100% then there is nothing more to be said: communication between the two Weltbilde is impossible. Say it is at 99.6%, and let us assume that we have a translation that achieves the minimum distortion set out by the epistemic differentiation (I do not propose to discuss how we might make such an optimal translation: for the purposes of this Gedankenexperiment let us simply assume that we have), so we can extract, in a way such that it would be recognisable to the speaker of L‘ if only we could work out how to translate it back, 1 part in 256 of the original information (exactly as is the case with the pixellated version of Agnes Sorel above). Then there are two conclusions to be drawn from this discussion. First, as we saw above, even at this extremely high epistemic differentiation, some information can be extracted. Second, this is, to all intents and purposes, a perfect translation of L‘ into L as it is the best possible translation we will ever achieve given the distance between our respective Weltbilde. So it is perfect given the constraint of our different Weltbilde; we need to relativise our concept of quality of translation.
Here we enter into very murky territory. There is a simple and concise answer: ‘no’. The definition of an epistemic barrier was a barrier to translation. Taking that as a starting point, let us consider this more carefully. If there is not a total barrier to translation between L and L‘, but there is a very high systemic epistemic differentiation, of the order of ninety-something percent, then does this count as being an epistemic barrier? Consider what I, as a speaker of L can ever glean from L‘ under these circumstances. I may gain some vague general notions, but I will never be able to tell how much of them is inherent in the communication I am trying to understand, and how much is simply my mind constructing artefacts from my own Weltbild in order to make sense of what is, essentially, senseless garble. Because if I can tell then that means I somehow have a means of telling whether my translation is correct or not, so in fact I have available to me a better translation than the one I am already using, so I just use that instead. Going back to the visual analogy, when you look at the pixellated Agnes Sorel, you can begin to read all kinds of things into it, but they have as much reality as the faces that we see in the clouds; they are artefacts resulting from our minds trying desperately to find the familiar among the unfamiliar. The point of this, is that under such circumstances, I cannot honestly say that I understand L‘; all I can really say is that when I encounter a communication in L‘, it puts certain ideas into my mind, but I have no real idea how many of them are strictly conveyed by the communication. Now, I can say, quite justly, that this is true of any translation, but when the epistemic differentiation is small, then at least the two languages are close enough for some form of dialogue to develop. There will come a point at which, with increasing epistemic differentiation, we cease talking to one another, and start talking past one another. Once we are in the talking past mode, though we may each of us understand some small part of what the other is saying, there is no way of verifying our understanding, as that would require a translation of greater fidelity than we are capable of. Which means that communication between such widely separated Weltbilde is strictly one-way, and its accuracy is entirely unverifiable: the ideas I gain from you could be inherent in your speech, or they could be faces in the clouds; I have no way of telling. This seems to me like a very fair description of an epistemic barrier. Thus, in response to question (2’b) we have to say that we may be able to glean something from a separated language, but we can never know how much of what we gleaned was really there, or to put it in a slightly more disturbing way, the best possible translation into our language need not contain any concept recognisable to the originator of the communication. So we conclude that given a Weltbild we can arrange other Weltbilde on a scale, with those permitting perfect translation, whose epistemic differentiation is close to 0%, at one end, and those with which no adequate intercomprehensibility is possible, with epistemic differentiation close to 100%, at the other. We will obviously agree that no epistemic barrier separates us from the Weltbilde whose epistemic differentiation from ours is small, and I have just argued that for a ‘sufficiently large’ epistemic differentiation, there is an epistemic barrier.
So, turning to question (1), where does the barrier set in? It turns out that we cannot even know the answer to this question. What I have just described is the classic setup for the sorites paradox. Briefly, a sorites paradox considers a problem where I have a class of things, and a function that assigns to each of those things a number in some specified range. I then have a predicate that applies to numbers, and which is false at one end of the range and true at the other. The other features that are required to make the sorites paradox complete are two assumptions: that there are no gaps in the values that the function can take and that the truth-value of the predicate is locally constant (it does not change if the value is changed by a small amount). Now it is fairly easy to see what happens: start with an object with value at one end of the range, where the predicate is true, and apply the local constancy to progress to the other end, where the predicate is false. Thus unless one of the conditions breaks down, true=false. Therefore one of the assumptions breaks down. Let me give two examples, one classical, one directly relevant to our discussion. In the classical case the things are collections of grains of sand and the function assigns to a collection the number N of grains in the collection. The predicate is then ‘is a collection of N grains of sand a heap?’ The two assumptions become the assertion that for every Nit should be possible to find a collection of precisely N grains of sand, which is trivially true, and that adding or removing one grain does not change a collection’s ‘heapness’, which is intuitively obvious. In our case, the things are languages, the function assigns to a language its epistemic differentiation from L, and the predicate is ‘is a language with epistemic differentiation n separated from L by an epistemic barrier?’. So, the first assumption is that for all n there should be a language with epistemic differentiation n from L, which is not at all obvious, while the second assumption is that a small change in a language should not tip it over the edge to the other side of an epistemic barrier, which seems reasonable. So, we have established that languages and epistemic barriers form a sorites paradox, and so the two assumptions break down. There are two possibilities. The first is that in fact the range of values assignable to objects does, in fact, have a gap, so there are forbidden values. As noted, there is no a priori reason, in the case of languages, why this should not be the case. The second possibility is that the local constancy of the predicate breaks down somewhere, which, in the absence of a gap, means that somewhere there is a sharp boundary between true and false, heaps and non-heaps, epistemically commensurable and epistemically separated. But clearly we don’t know where this is, else we would not have asserted the local constancy of the predicate. As as we don’t know where it is, there must be some range of values, within which it lies, where the behaviour of the predicate is unknown to us. But now say we somehow discovered precisely where the boundary was. Then we could find intelligible languages arbitrarily close to unintelligible languages, and so we could, with arbitrarily small error, approximate a language which is separated from us by an epistemic barrier. So it is not, in fact, separated from us by an epistemic barrier, as the error in translating it must be arbitrarily close to that involved in translating a mutually intelligible language. This is a contradiction. Therefore, there may be a barrier, but we can never know where it is: there is a region of epistemic space about which we must forever remain ignorant even as to whether it is or is not intelligible, so this is a form of second-order unknowability. Thus, in essence, we are back with the gap. There are some values of epistemic differentiation about which we can simply say nothing at all; values below this range correspond to mutually intelligible languages; values about the range are separated from us by an epistemic barrier. So the answer to question (1), about knowing how far apart we can be before we hit a barrier, is a very firm ‘we have no way of knowing’.
After the long hard slog of the last sub-section, this should be something of a relief. We now turn our attention to questions (3a) and (3b) which are about proving the existence of epistemic barriers and of languages on the other side of them. Obviously, if the answer to question (3b), on detecting epistemically separated languages, is ‘yes’ then then answer to question (3a), on the provability of the existence of barriers, is also ‘yes’. However, it should be clear from the discussion in the previous sub-section that there is no other way of proving the existence of an epistemic barrier. The sorites paradox argument showed that we can never find one by going looking for it; all we can find is things that we can comprehend, which are de facto on our side of any putative barrier, and things that we cannot comprehend, which are either nonsense or languages on the other side, and we have no way of telling which. And even if we identify something as being a communication in an epistemically separated language, all that that tells us is that there is a barrier, it cannot (again, using the sorites argument) tell us anything about the nature of that barrier or its location. Thus, the barrier, even if its existence is proved, must remain a rather hazy, ill-defined entity, rather like a quantum wave-function (which analogy raises the fascinating possibility that perhaps detailed knowledge of epistemic barriers is separated from us by some kind of second order barrier). Thus the only hope of salvaging any form of positive answer to either question is to find a way of detecting epistemically separated languages. At first sight, our only hope might appear to be if a space-ship landed in Whitehall and an alien got out and nailed ninety-five theses to the door of Number 10, Downing Street. Such a clear ostensive association with a bona-fide alien presence could, if we assume that we were not simply on the receiving end of alien humour, or maybe even garbage disposal, be taken as being pretty strong evidence that the theses were expressed in an alien language. So, if the received communication failed to show any of the signs we would expect of communications that have not been deliberately obscured (observe that encryption can be thought of as the deliberate establishment of a transient epistemic barrier between two communities that otherwise share a Weltbild) then our degree of confidence that they originated from behind an epistemic barrier, and hence of the existence of epistemic barriers, is precisely equal to our degree of confidence that we can recognise languages in our epistemic class when we see them, even when we have never previously seen the language in question. As this is something that we seem to be extremely good at (consider how easy it is to distinguish noise from language, even incomprehensible language, when turning a radio’s tuning dial), therefore we conclude that if an alien committed a demonstrably ostensive act relating to some collection of information and the information had none of the properties normally considered to distinguish language, then we could conclude that we had the elusive epistemically separated language. The question then is, if we were presented with exactly the same information, but without the context of an ostensive relationship to an alien presence, would we be able to say that it was epistemically separated language, as opposed to, say, random noise? The answer has to be no. We just said that it had none of the properties normally considered to distinguish language. So it will have no natural-language or mathematical properties, which means that it is, to all intents and purposes, random. We cannot it distinguish it from random noise, because in order to prove that it wasn’t a currently indecipherable language emanating from within our own epistemic class, we had to remove all properties that we are capable of distinguishing; any evidence of such a property conveys information expressible within our language. Thus, if we are rigorous in our effort to ensure that the putative alien language is truly alien, and not merely mundane but incomprehensible, the act of ostention is the only means by which we can distinguish it from random nonsense. One could question whether we need to be this rigorous, and that we must (in the spirit of the preceding sub-section) allow that alien languages may have some properties, just very tenuous ones. But here again the sorites paradox bites home. It may be that an alien language has some very elliptical property that conveys only a tiny amount of information to us, but we then cannot establish a simple test for alienness, because to do so would require knowing we to draw the line between our epistemic class and the alien, and we cannot, because we do not know where the epistemic barrier is. All we can do is to assert that totally alien languages will be indistinguishable from random noise, and that less remote (but still alien) languages will have some tenuous properties. But from our side of the barrier we cannot say what those properties are, or how strong they can be unless we first have to hand an example of an alien language manifesting those properties, for to predict these qualities of alien languages on purely theoretical grounds would be tantamount to predicting the location of the barrier. So we can only start to detect tenuously non-random alien languages if we have, by way of a boot-strap, an example of a tenuously non-random alien language. And the only way to break out of this loop is via an ostensive act: some extra-linguistic evidence that clearly marks out the information as being an alien communication. Thus, until the day dawns when an alien does land its space-ship in Whitehall, we have to accept that the answer to both parts of question (3) is ‘no’. Note, by the way, that this means that the SETI programme of listening out in the hope of hearing alien communications is fruitless. Even so (apparently) universal a property of communications as modulating radio waves is more likely than not, a feature of our Weltbild and hence not necessarily a feature of alien communications. Adapting the argument above, SETI will only succeed once a friendly alien has told them what to look for.
Looking at the argument so far, we seem to have reached a generally pessimistic conclusion. We can say that there is good reason to believe that epistemic barriers exist, but barring exceedingly unlikely events (the ostensive alien) we cannot prove that they exist, and under no circumstances can we determine where they are. It is possible that we may be able to glean some limited information from languages on the other side of a barrier, but we cannot prove that the information so derived was part of the intent of the originator and is not just an artefact of our Weltbild (or worse, our imagination). So, on the whole things look pretty gloomy. There are whole realms of knowledge that will be forever forbidden to us simply by virtue of our nature; anyone seriously believing that they can arrive at a theory of everything might as well pack up and go home, as might the theorists of alien contact. Having reached this generally pessimistic conclusion, let us look again at question (1) where it turns out there is a (small) amount of good news. We can’t actually know where a barrier is, but we can know something about the island of epistemic space around us: though we cannot precisely delineate its coast, we can at least get some idea of how large it is. And the answer to that is quite surprising. First, every human natural language yet discovered fits within one epistemic class; though some languages are more flexible or more complete than others, translation relations have been established in every case where speakers of a language exist and in some cases where the language is now dead. Now, on the universal scale, one species on one planet may not seem like much, but given the sheer diversity of human culture, and the extent to which some groups have been physically isolated, it is rather impressive that simply being human, with the common somatic and mental apparatus that implies, is enough to guarantee a commensurate Weltbild. In fact, we can go further. There is a long history of deeply flawed work on communication with non-human apes, but there are now the beginnings of evidence that it is possible to establish a limited translation relation with chimpanzees, though the work done to date is not sufficiently precise for the nature of the relation to be clear. More impressively, it has proved possible to establish a relatively complex translation relation, including positional grammar, predication and symbolic sentences, with bottle-nosed dolphins. While we might expect our very close relatives, who share a basic body-plan with us, to have some elements of a Weltbild in common with us, it is rather impressive that we can communicate at all with animals who inhabit a three-dimensional world, vice our two dimensions, have senses that we have not and have a body plan very different from ours, let alone being able to formulate a basic generative grammar. And finally, the remarkable case of Alex, the Grey African Parrot suggests that it is possible to establish some form of intercomprehension with a non-mammal (and again an animal with a very different body-plan from our own). Thus, wherever the epistemic barrier surrounding us is, it is clearly not as close as one might think.
In The Other in Culture, I used the radical translation argument to rebut the claims of some special interest groups that gender or sexuality or some ill-defined racial category is sufficient to guarantee separation. Though the original argument has now been shown to be incomplete, the current analysis shows that my conclusion still holds true. We have seen that an epistemic barrier results not in subtle shifts and biases, but in total incomprehension. As the best that the special interest groups can lay claim to is subtle shifts and biases, unthinking prejudicial language and the like, that means that there is no epistemic barrier, and therefore the radical translation argument holds true. In The Male Gaze Gome Wrong I, somewhat incautiously, suggested that the split in western culture between the older communitarian spirit and the more modern culture of self-affirmation was the beginning of an epistemic barrier, citing as evidence the fact that it appeared to be, if not impossible, then at least extremely difficult, to establish any form of translation relation when dealing with concepts of eroticism and sexual attraction. In view of the fact that our Weltbild not only transcends culture and age (we can take it back to the fourth millennium BCE), but apparently even species, it is, I would suggest, unlikely that this is the case. That is not to say that our Weltbild is not undergoing considerable change, or that we do not have a very hard translation problem on our hands in relating the two cultures (exacerbated, I suspect, by the fact that the two groups use words like ‘sexy’ to mean two very different things). Both of those facts are true. But, as I said in the original essay, because we are one species we have one Weltbild, and therefore one epistemic class. Humanity is not epistemically divided.
This may seem a rather strange reading list, but by far the best treatment I have found of this subject is to be found in a number of works of science-fiction emanating from Central Europe. Thus, with no more ado, my suggestions for further study are as follows.
- Davidson, Donald; Truth and Predication
- Lem, Stanislaw; His Master’s Voice
- Lem, Stanislaw; Golem XIV, published in Imaginary Magnitude
- Lem, Stanislaw; Proffertinc, article on Prognolinguistics, published in Imaginary Magnitude
- Lepore, Ernest & Ludwig, Kirk; Donald Davidson’s Truth-Theoretic Semantics
- Quine, Willard; Word and Object
- Quine, Willard; Ontological Relativity, published in Ontological Relativity and Other Essays
- Soames, Scott; Understanding Truth
- Strugatsky, Boris & Arkady; Roadside Picnic