Epistemic Barriers – this time in English!
I admit it. On Epistemic Barriers is (though I say it myself) a rather profound piece of work, containing some very important philosophical ideas. Unfortunately, it is also profoundly hard to read. I made the mistake of writing it in the impenetrable style of some of the philosophers I was critiquing. Which was a mistake, in that complex ideas were made impossible by an overly academic style.
So this essay is an attempt to present the same ideas, but with a more transparent exposition. That doesn’t mean that what I have to say will necessarily be easy to understand. As I have already said, the ideas are quite deep, and as they deal with the bounds of knowledge (what is and is not knowable) they are inevitably going to be mentally stretching. However, I will do my best to ease comprehension. That means none of the quasi-mathematical language or excursions into Germanisms.
2 What is an epistemic barrier?
In this section I’ll introduce the concept of an epistemic barrier. It’s quite a complex idea, so I shall approach it gently from, as it were, both directions. So first I look at the facts relating to the mysterious mutual comprehensibility of all human languages, and then I look at those facts that limit mutual comprehensibility.
2.1 The enigma of mutual comprehensibility
If we look at our world, there is a certain very surprising fact that constantly stares us in the face, but is so commonplace that we don’t realise just how surprising it is. What I refer to is the fact that, in spite of there being hundreds of languages spoken by people on this Earth, we have yet to find a language that is wholly incomprehensible (I am ignoring here certain dead languages, as the are special problems inherent in trying to decode a language with no native speakers and no obvious descendant languages).
Why do I say this is surprising? Well, if two languages have arisen from different sources, and are sufficiently isolated (say one in Europe, the other in Australia), then there is no obvious reason why it should be necessarily the case that speakers of the two languages should be able to establish communications with one another, let alone that it should be possible to translate between the two languages. We take this for granted, but, in actuality, the fact that all human languages seem to be mutually comprehensible is something of a mystery requiring explanation.
Now there are a number of competing explanations. I will list those most relevant to my eventual goal, which is not mutual comprehension, but mutual incomprehension. So the theories are:
2.1.1 All languages are essentially the same
According to this theory, there is only one way that a natural language can be formed, and thus any two natural languages must be mutually comprehensible, because they are simply realisations of the same underlying structure.
This implies that there is only one class of conceptual categories, and only one way of ontologising the world around us. For if (say) multiple ontologies can exist, there is no guarantee that objects as described in a language using one ontology will correspond in any well-behaved way with objects as described in a language using another, and so mutual comprehension will be limited to non-existent.
Thus the theory imposes a very regimented view of reality and what is and is not permissible in terms of individual’s conceptual models of the world. Indeed, it is almost Platonic, for the limitation to a single class of ideas and kinds of things is, in all but name, an acknowledgement of the existence of forms that all things (to a greater or lesser extent) realise.
Now, I do not consider this theory to be viable, for a number of reasons. First, it is incredibly anthropocentric. What reason do we have to believe that ours is the only viable way of anatomising the world and the world of ideas? Surely, all human languages are mutually comprehensible, but what reason do we have to believe that ours are the only languages that exist? Or that our planet is an exemplar for all those where language-using life may arise?
Second, consider the following thought experiment. Say we discover a species of aliens who have evolved on or very near to a neutron star. Due to the extreme nature of physics in such a neighbourhood, they sense quantum wave-functions directly. So while we see (say) a stone as an isolated object, they sense a distribution spread across the whole of space-time. In other words, our senses are local, while theirs are global. Now, our entire world-view is predicated on objects being localised in space and clearly distinct from one another. Indeed, the idea of predication, fundamental to all human languages, requires a clear concept of thing or things which clearly possess one or more definite properties. But to our quantum creature none of this makes sense, as the world consists of an endlessly shifting mass of probabilities, with no clear barriers or delineations, and so their world-view, and therefore their language, must be alien to ours. The only recourse, if we want to salvage the theory, is to say that language-possessing aliens cannot differ significantly from us in their physical environment and sensory apparatus, which is clearly absurd.
Third, the theory is logically circular. For it argues that every language must take the form of something we recognise as a language. Its evidence for that is that all the languages we know are recognisable as languages. But we would expect nothing else; if a language were truly alien then we could not necessarily recognise it as a language, so we would not know that it was there as a contradiction to our theory. Thus the theory is unfalsifiable and, as such, worthless.
Before you ask, the reason why I spent so long on attacking an obviously flawed theory is twofold. Well, it may be obviously flawed, but that does not stop it from being deeply embedded in our culture. Indeed, the whole effort to communicate with aliens, whether in the form of pictures on space probes, signals beamed into space, or attempts to ‘listen in’ on alien conversations, is based on the assumption that alien languages are similar to ours, and that certain linguistic facts that are obvious to us will likewise be obvious to aliens (and vice versa). One of my conclusions is, in fact, that this belief is false, and that therefore the enterprise of SETI is futile.
2.1.2 All human languages are essentially the same
This is a less ambitious version of the above theory, so it accepts that alien languages may differ wildly from ours, but still insists that human language is essentially wired-in, in the sense that it is inherent in being human. So, though individual languages may be contingent things, there is a universal grammar that is built into our neural hardware, so to speak, and is specified by the blueprint for a human being (i.e. our genome). This is the theory most closely associated with Noam Chomsky and his school.
I do not consider this theory to be viable, though it is harder to dismiss than the first theory, the argument against it depending on balance of probabilities rather than obvious flaws. Though Chomsky and his followers point to the fact that we acquire rules about what is and is not grammatically correct without hearing (many) examples of incorrect usage, this argument has a number of flaws. First it ignores the fact that, in fact, as children, though it is granted we do not usually hear solecisms, we are repeatedly corrected when we commit solecisms. Second the phenomenon is more economically explained as being the result of our modelling the language we speak on the language we hear: this, after all is how we acquire dialect and accent, so why should it not be true for grammar? Third, if, as Chomsky would like us to, we interpret this avoidance of grammatical solecisms as being evidence of hard-wired rules at work, how to account for the fact that said rules are strongly language dependent (so, for example, this sentence is perfectly correct in English, but, if translated literally into German, would be grossly incorrect)? To posit that I am genetically an English speaker is nonsensical, but if we try to universalise the rules to the point where they apply to all known human languages, they become nearly vacuous.
This is the next point. The universal grammar, as it is presented, exists at the very high level of specifying that verbs link noun phrases, and descriptions can be predicated of objects. But are these really sufficient ideas to justify the theory? And even if these facts are hard-wired into our brains, they are so minimal in their content that they do nothing at all to answer basic questions about language acquisition. In other words, a universal grammar sufficiently general to account for all human languages is so vague as to be almost entirely without benefit to its owner, and as such it is more or less impossible to see how it could have evolved.
And finally, once again the theory is circular. Say it is true that all the human languages we know follow a Chomskian universal grammar. That could just be due to the fact that, used as we are to Chomskian languages, we cannot recognise non-Chomskian languages when we encounter them. So the evidence that might falsify the theory is inaccessible to us, and thus the theory is unfalsifiable.
Note that this is not to say that there is no hard-wired aspect to human languages. For example, there does seem to be a fixed sequence of colours, such that if a language has a word for one colour on the list, it must also have words for all the preceding colours. However, this tells us nothing about language, and is more a feature of the human sensory system.
2.1.3 All humans are essentially the same
This theory can be seen as a subtler version of the previous theory. Rather than saying that any features of language are hard-wired into human brains, we observe that human language is intrinsically a property of humans. And as humans, we are (to a reasonable approximation) identical. Thus we have similar senses and similar minds, and so see the world in similar ways. And this means that an idea formed in one human mind cannot be alien to another human mind. That is not to say that we could swap thoughts and expect to understand one another; rather that my thoughts should not be such that your mind cannot accommodate them, and vice versa. And thus, human languages are susceptible to translation into other human languages precisely because they have the common feature of being languages of humans.
This theory can withstand all the arguments we have made against the other theories. It is anthropocentric, but in a different way to the first theory. That theory made man the measure of the universe. This theory makes it clear that human languages are a property of humanity, but makes no pretensions to speak of the properties of alien languages. Moreover, it is clear that this theory can be generalised, so that quantum being languages are mutually comprehensible, but, due to the extremely different conceptual and sensory apparatus of humans and quantum being,s it is a corollary of the theory that human and quantum being languages are not mutually comprehensible, just as we concluded above.
Moving onto grammatical issues, one of the great strengths of the theory is that it makes no claims about the structure of language; it is concerned with the structure of ideas. Thus, such hard-wired features of the human sensory system as exist (e.g. the colour sequence) can be found reflected in human language, and some of these may influence grammar (e.g. the notion of predication, which depends on a distinction between things and properties that is meaningful to us, but may not be to the quantum creature) but they do not dictate grammar.
Finally, the theory is clearly not circular, because it makes no pretension to give a universal theory as to what make a language, or even what makes a human language. Therefore we shall accept this theory of language and see where it takes us.
2.1.4 Epistemic barriers at last
In fact, it leads us on to our goal: the epistemic barrier. As I have said, in my thought experiment with the quantum creature, there will be little or no conceptual common ground between me and it. Because of this, there is no reason to believe that my mind could house its thoughts and vice versa. So therefore, the theory of language says that with a very high probability, it should be impossible to translate between my language and its. Or, to put it more simply, it will be impossible for me and the quantum creature to share ideas, not just because we speak different languages (as should be clear by now, language differences are a barrier easily overcome) but because of a fundamental difference in our conceptual worlds.
I will therefore say that in this case there is an epistemic barrier between me and the quantum creature. More generally, I will say that two languages, or two conceptual realms are separated by an epistemic barrier if it is systemically impossible to set up a correspondence (e.g. translation) between them. That is to say, the reason for a failure to translate is not (as with, say, undeciphered ancient human scripts) because we do not yet have the necessary key, but because it is fundamentally impossible that there ever should be such a key.
2.2 Barriers to mutual comprehension
Now let’s start from the other end, and look at the various barriers to perfect translation that exist. It turns out that translation cannot be perfect, and there exist barriers of various kinds, ranging from those between individuals who share a language up to those between human languages. But the common feature of all of these barriers is that we can still capture a predictable proportion of the original intent of a thought as it crosses the barrier. This is abruptly lost when we hit an epistemic barrier, where something mysterious happens. I shall discuss exactly what in the next section.
2.2.1 Barriers between me and you
In epistemology, the main reason for interest in translation theory is not because philosophers want to be able to translate their work into foreign languages. The interest lies in the fact that the mere fact of human communication is an act of translation. If you think about it, whenever you talk to someone, you assume that they will be able to interpret your words in the way you intended them, and similarly you expect that the intent you perceive in their words was the intended intent. Thus, even if you are both speaking the same language, you are translating between two idiolects.
Now this is a potential source for loss of information and ambiguity. For a start, languages do not all cut up the semantic space in the same way. So, many languages do not distinguish ‘green’ and ‘blue’ as concepts, having instead a ‘grue’ colour concept. So if you come from that background, your mind does not distinguish ‘green’ from ‘blue’, and even if you now speak English, you are likely (this seems to be what happens in practice) to use the words inconsistently, causing confusion for native English speakers. Similarly, an English sentence that hinges on the distinction between ‘green’ and ‘blue’ may well be pretty meaningless to you. Likewise, someone brought up speaking German who has learned English, may well find the way that native speakers treat snails and slugs as different kinds of animal somewhat strange.
But all this is quite cosmetic. There is a more fundamental issue. Consider the word ‘rabbit’. Now, you and can talk quite happily about rabbits, and I can be absolutely certain that whenever you say ‘rabbit’ you are signifying a lop-eared rodent. But do I know what ‘rabbit’ means to you? The answer is ‘no’. I can know precisely as much about your views on the word ‘rabbit’ as can be expressed using our mutual languages, but it is a truism that there are non-linguistic thoughts, and another truism that I can never learn your non-linguistic thoughts through the medium of language. So what you take ‘rabbit’ as meaning, its emotional and other non-linguistic associations, are forever hidden from me.
So here is a (reasonably) small barrier. That is to say, in conversation we use words, and we can be confident that whatever was the linguistic component of our intent when we framed them will be comprehensible to a competent colinguist. But the non-linguistic part of our thought must remain private, or else be expressed in symbolic form, in which case we are actually inviting those who observe our symbols to substitute their own non-linguistic ideas for ours. So ordinary, empirical statements translate absolutely, and as they form most of our discourse, that means that most of the time this person-to-person translation works perfectly. But complex ideas with personal emotional connotations inevitably suffer some loss in translation: the loss of meaning, a loss that can only be rectified in a somewhat elliptic fashion through art.
2.2.2 Barriers between human languages
Now consider translation as we normally think of it, between two human languages. Here there is again some inevitable loss of information, largely due to the simple fact that, as noted above, different languages slice the cake of meaning in different ways. As noted above, colour words are a perennial source of confusion: for example, the Russian words голубой and синий both translate into English as ‘blue’ and yet, to a Russian they signify entirely different colours. Prepositions and conjunctions are a particular minefield, as it seems that even closely related languages divide shades of meaning up between words in different ways. Idioms are another place where translation often fails entirely: there is no simple English equivalent for l’esprit d’escalier or schadenfreude and so these expressions are generally carried across unchanged, and have even entered common English usage.
What is, however, more interesting is cases where it is not so much meaning that is the problem as connotation. Thus, Gut bürgerliche Küche translates literally as ‘good middle-class cooking’, but whereas the latter phrase would have distinctly negative, even satirical connotation to one accustomed to British culture, in Germany it is an indicator that one can expect a good meal. So translation between languages also requires a translation between cultures, and, if done incautiously, loses meaning given by cultural context. This is, in fact, the direct extension of the loss of personal meaning in person-to-person conversation: now we are translating between cultures and the private information that cannot be expressed purely linguistically is precisely the baggage that any culture carries with itself.
Even more interesting, however, is the first hint of an epistemic barrier, that is to say the fact that there are cases where a language can express concepts that are untranslatable. One such is the German word gemütlich. One can try to translate this into English, but attempts to do so generally end up with half a dozen vague and inadequate phrases. But that does not mean that just because your first language is English you can’t know what it means. I am perfectly aware of what gemütlich means, and can use it correctly, and yet I have no idea how to render it in my mother tongue. Another such word is kitsch (German seems to be good at this game).
So what we have here is not quite an epistemic barrier, for I can form these untranslatable concepts in my mind. This means that they are not systemically untranslatable, but rather simply that over the centuries Britons have never felt the need to express the concept of kitsch, and by the time we did it was easier to borrow a loan word from German. So this does not contradict our model of language; instead it shows that building vocabularies is not just a matter of slicing the cake, but also (sometimes) of accidentally missing parts of it (and lest it be thought that English is impoverished, English has distinct words dove and pigeon where German has Taube, snail and slug for Schnecke and so on).
Note finally, that much of my discussion has involved English and German, two extremely closely related languages. The further apart two languages are, the more meaning is lost in translation (so English and German are both unable to express the Japanese system of honorifics without complex circumlocution).
2.2.3 Epistemic barriers again
So what have we learned about epistemic barriers from this discussion? First, we see that language-to-language translation generalises person-to-person, and that the mechanism for information loss is the same. Surely, more information is lost, increasingly so, as I have noted, as the languages and cultures move further apart, but essentially the information loss is controlled, and we can estimate (roughly) how much indeterminacy is introduced in the process of translation.
Second, we know what they are not, that is to say they are not just the consequence of gaps in vocabulary, or more or less complicated grammatical inflections. If English speakers the world over were to decide that plof was the translation of gemütlich then that would be an end to that apparently untranslatable concept; the fact that we could not say succinctly what plof meant is neither here nor there: we happily and correctly use words like labile or liminal in spite of the fact that if asked what they meant we may well end up embarrassing ourselves.
So, say gemütlich involved a genuine epistemic barrier. That would mean that it would be impossible for a non Germanophone to even comprehend what it meant, which is a much stronger statement than simply that there is no word for it. The concept of gemütlich, rather than being merely elusive should be actually unthinkable. Now this is a really strong statement, so let me put it more formally: two languages are separated by an epistemic barrier if some or all of one of the languages is inconceivable by speakers of the other. Or, to put it another way: there is an epistemic barrier between me and some idea not just if I can’t understand it, but precisely when I cannot even accommodate it within my mind. This will have very strong implications as regards our ability to discover alien languages.
3 What can we say about epistemic barriers?
3.1 Can we prove that they exist?
The simple answer to this question is no. Let me expand on that, as the reason why is rather interesting. So, how could I prove that an epistemic barrier existed? There are two possible approaches: one is to identify an actual, concrete epistemic barrier, i.e. to find one that I can point at, and the other is to use a general argument that shows, on theoretical grounds, that such things must exist.
3.1.1 Finding a barrier
I’ll discuss the details of finding a barrier, in particular locating how close the nearest one is, later. For now let’s stick to the rather crude question of whether I can find any way of detecting the presence of such a beast. Well, there’s really only one way of doing it, and that is to find a genuine bona fide language that we cannot comprehend (at least in part). So how would I do that?
Well, there’s one approach, which is that it might just happen that one day a flying saucer lands in Trafalgar Square, an alien emerges, marches down Whitehall and into Downing Street, where it nails ninety-five theses to the door of Number Ten. Assuming that the alien was not an interplanetary juvenile delinquent or performance artist, we could safely assume that the theses constituted language and, assuming (as seems overwhelmingly likely) that we could make no sense of them, we would again probably be safe in assuming that they were beyond our comprehension. So we would then have, for what it was worth, a genuine example of language from the other side of an epistemic barrier. We would know nothing about said language apart from this fact, but it would at least establish the existence of the barrier.
Now before you laugh and suggest that this scenario is a little crude, the problem is that barring such a clear act, pointing at some thing, as it were, with a large sign saying ‘alien language’ we are more or less stumped. The problem is this. To recognise a language from the other side of an epistemic barrier, we have to know something about what things are like on the other side. But we can’t, because by definition, such things are inconceivable by us. To argue that there ought to be features of any alien language that we can detect is simply to revive the universal language or universal grammar models for language that I dismissed in the previous section, which would have the amusing consequence that in order to find an epistemic barrier one has to assert that they don’t exist!
This fact has quite amazing consequences. If the only way to be certain that something is a communication in an alien language is to, as it were, catch aliens in the act of communicating it, then that means that alien languages, when seen blind, with no prior knowledge, cannot be distinguished as such. In fact, we can go further. Even if we do catch the aliens in the act, as their language is systemically incomprehensible to us, most likely we will be unable to detect any of its grammatical features (if, indeed, it has any) and so we will not be able to learn anything about the language, and hence not be able to spot even further instances of the same language. Even looking for repeats must be called into question, as there is no guarantee that the language will handle repeats in the same way that our languages do (where repeated ideas leads to repeated text).
What this means is this: no matter how many hard instances of beyond the barrier language we may accumulate, we will never be able to learn anything about beyond the barrier language. So whereas the bigger the corpus of a dead language, the higher the probability of decipherment, here a change in the size of the corpus has zero effect.
We conclude that no test can detect the presence of an alien language. It is, simply by virtue of being alien, indistinguishable from noise. And therefore a language from the other side of an epistemic barrier is in principle undetectable by any means available to us, short of catching an alien in the act of communicating it. And thus, as I said above, efforts, like that of SETI, to detect alien communications are futile. Even to assume that alien radio signals are modulated in a way we can recognise is an anthropocentric assumption too far.
So, barring a sudden intervention from on high, we are not going to be able to find a barrier.
3.1.2 Postulating the existence of barriers
Failing that, I can motivate a belief in the existence of barriers as follows. I can easily conceive of a being (e.g. the quantum creature from the last section) so dissimilar to us in sensory apparatus and world-view, that it inevitably will suffer from a complete failure to communicate with us. And this will not be a result of failure to try hard enough, but will simply be because the models of how the universe works belonging to us and the creature will be so different from one another that there is no way of mapping between the ways we divide up the cake of knowledge. Indeed, it is arguable that we have separate cakes. In other words, the failure to communicate will be systemic, inherent in our humanness and the creature’s alienness, so there will be an epistemic barrier between us.
But now, though I chose my quantum creature as a deliberately extreme example of an alien obviously separated from us by an epistemic barrier, if we consider any alien creature, then it will have evolved not just with different culture, but within a completely different biological, geological and so on and so forth context. It seems extremely unlikely that its way of thinking about the world will match with ours. And, as I noted above, different world-views lead to radically different languages, so I can hypothesise that any alien creature must be separated from us by an epistemic barrier (that is unless something highly fishy, like a bunch of super-aliens going round seeding worlds, is going on).
And that is enough. Though I believe it is overwhelmingly likely that alien creatures exist, I don’t need that for this argument. The mere fact that the existence of such a creature is not impossible (for if it were, then why are we different?) motivates my claim, that in all probability epistemic barriers exist. For even if there are no aliens, the possibility of aliens, and with it the possibility of knowledge that I cannot comprehend, exists.
3.1.3 Predicting the existence of barriers
So it is extremely likely that epistemic barriers exist. Is there any argument that I can use to prove that they exist? The answer to that is a bit tenuous. I can develop an argument (and will, see below), but it is somewhat tendentious and relies on a large number of fairly questionable assumptions.
The argument is essentially a rehash of that old favourite, the ontological argument. It goes like this. Clearly it is unlikely in the extreme that the extent of the realm of possible languages should be coterminous with the class of languages that my mind can comprehend, for to assert otherwise would be to assert, essentially that all information is open to the human mind, which would seem to be a rather questionable assertion (after all, my thought experiment of the quantum creature shows that it is clearly possible that information not open to my mind should exist). Thus the languages that I can comprehend must be strictly a subset of all possible languages, meaning that there exist languages that I cannot comprehend.
Let us examine this. It is actually reasonably sound. The only contentious patch is the assertion that if I could comprehend all possible languages then I could comprehend all possible information. But what if there are chunks of the universe that are indescribable in any language? That would allow me to comprehend all languages without having to lay claim to godlike powers. There are two ways of dealing with this issue, one Platonic, the other consisting of a mixture of information theory and hand-waving.
The Platonic approach asserts that if something is genuinely incomprehensible, in the sense that no mind exists, or ever has existed, or ever will exist, that can comprehend it, then it has no reality. This is not as stupid as it sounds. It is actually quite reasonable to say that given that (as Wittgenstein said) we must remain silent of that of which we cannot speak, and that (as observed above) not only must we remain silent, but we must also remain ignorant of its very existence, then for all our purposes it might as well not be there. For if it had any effect we could detect then it would enter within the realm of things comprehensible by us. This is a rather neat idea, but it has one flaw: once again we have proved that the barrier exists by asserting that nothing on the other side of a barrier can exist. Back to the drawing board.
The second possible approach is as follows. Let us assert that all that is can be expressed as information. This is one of the basic assertions of information theory, and it isn’t too much of a stretch to assume that, even though its form may change, the basic principles of information theory are true throughout the universe. But now, all information can be expressed as a code of some sort, and therefore we could, with a certain amount of hand-waving, construct a language, not necessarily comprehensible to us, that encoded this information. Therefore all that is is comprehensible in some language (which may, admittedly, be a language that no-one actually speaks). This approach is full of holes, but I believe it has some promise.
3.1.4 Constructing a trans-barrier language
That was the argument for the existence of barriers: suggestive, but, at present, inconclusive. There is another constructive argument, in that I will try to present a recipe for building a language that we cannot comprehend. Now, this is somewhat abstruse, in that it involves concepts such as meta-grammar, and it doesn’t add much to the overall argument, so those of a nervous disposition can happily jump ahead to the next section.
I aim, in a thought experiment, to construct an incomprehensible language. But didn’t I say that we can know nothing of the grammar of a language on the other side of an epistemic barrier? Indeed I did, but I said nothing about meta-grammar, that is to say the rules that the specification of the grammar must satisfy, or the grammar of the grammar. It is entirely possible that all languages could share a meta-grammar, for the relationship between meta-grammar and grammar is a one-way street. We know how to go from meta-grammar to grammar (though we can generate only grammars that we can comprehend), but there is no reason to assume that we can deduce the presence of the meta-grammar from a sample of incomprehensible grammatical language.
Say I define a meta-grammar (how I would set about doing this, I am not entirely sure, but this is, after all, a thought experiment). So now I take a computer and I write a program that, given a meta-grammar, will use random decisions to build a grammar based on that meta-grammar (if I want to be really clever, I can actually write a program to write the program, thus introducing randomness into the grammar generating program itself). And then I can use another program to generate ‘text’ in a random language conforming to that random grammar. And then the programs delete themselves and all files used in the generation of the language, leaving only a corpus of ‘text’ in the language.
What I get is an entirely artificial language, which has a non-denoting vocabulary (so words have meaning relative to one another, but no meaning outside of the language itself; this is not a problem: the same is true, after all, of the languages of formal logic or of computer languages). However, as it has been thoroughly randomised, I will assert that I will be unable to grasp the principles on which it works. For, to continue the thought experiment, there is no particular reason why it should even conform to our notion of words. Surely, it will have semantic units, but (say) a communication may be encoded so that the nth symbol is the sum of the nth semantic unit and all preceding symbols (such a language would have the property noted above of not repeating even if the content repeats).
This argument is incomplete, but it is, I believe, highly suggestive, and indicates a fruitful area for possible future research.
3.2 Can we find where one is?
I have produced a shaky but interesting line of argument that barriers do, indeed exist. The next question, then, is can we ever see a barrier? That is, can we ever tell the point of linguistic or conceptual development at which we will run into a barrier? The simple answer is (again) no, but the reason why not is rather interesting.
In fact, what we’ll see is that epistemic barriers are very strange beasts. They almost certainly exist, but it is entirely possible that their location depends on who it is that is talking. Moreover, the are undetectable, being always infinitely far from us (so they, by their very nature, exist at the very bounds of knowledge), but we may, if we are very lucky, be able to make an educated guess about their existence. In other words, in a suitable irony, the bounds of knowledge are themselves unknowable.
3.2.1 Epistemic barriers are not fixed objects
Think, for a moment, about the statement ‘some or all of one of the languages is inconceivable’ in my definition of an epistemic barrier between two languages. What would it mean for part of a language to be conceivable by me and part not? First of all, is it possible? Well, we can imagine a hierarchically organised language, where each tier in the hierarchy extends the tier below and adds additional structure. As a concrete example of this, consider the relationship between the specification of what a computer program should do, high-level programming code, low-level programming code and machine code. So it is entirely conceivable that if we encountered such a language, we could comprehend the bottom tier(s) but not the upper ones. So, yes, this is possible.
Having established that such a barrier can exist, what does it mean? From my point of view, there is a barrier down the middle of the language, dividing the accessible and inaccessible parts. But from the point of view of a speaker of that language there can be no such barrier: obviously they can comprehend all of their language, else it wouldn’t be their language. So where I see a barrier, they see none, but they may see a barrier (say) somewhere down the middle of my language, or it may even be that I can comprehend some of their language but they can comprehend none of mind (and before you say that that sounds silly, that is precisely the situation with Finnish and Estonian: Finns can understand Estonian but Estonians cannot understand Finnish). So the key point is that the location of epistemic barriers depends on where you are standing.
3.2.2 Epistemic barriers are invisible
Let me revisit the Platonic observation from the attempt at a proof of the existence of barriers. I claimed that if some concept exists on the other side of an epistemic barrier from us, then we cannot know even that it exists. This sounds rather mystical, but is quite simple. If I am mentally unequipped to handle the concept, then I cannot conceive of it, so no I cannot know about it.
But what if, say, I could not conceive of the whole concept, but there was some part, or fragment, that I could conceive of? In that case, I could be aware of the concept to the extent that my mind could house it, but I would have no way of telling that (as it were) higher forces were at play. As with the language with a barrier down the middle, I may see holes in the ideas or communications, as I understand them, but I have no way of telling whether those holes are genuine or are caused by my inability to process that which is beyond my comprehension.
As epistemic barriers bound what is, for us, the totality of knowledge, there is no way of telling whether they exist, and we are locked in an epistemic bubble of our own making, or they do not exist, and we just happen to be privileged with the key to all knowledge. That there is no way of distinguishing between these two possibilities is the reason why my attempt to prove that epistemic barriers did exist was so fraught with vagueness and complexity. In other words, epistemic barriers are invisible.
There is a very neat real-life example of this phenomenon. Quantum mechanics has wasted more ink in its short history than many disciplines. One of the principal problems with it seems to be that some aspects of the theory can be more or less understood, using appropriate mathematical models, but other simply make no sense. And the manifold problems with the theory as it stands (from Schroedinger’s cat all the way up to the grotesque nonsense of Feynman integrals) comes from an attempt to force something that is simply other into a human conceptual framework. I therefore propose (without proof) that we are seeing indirect evidence of an epistemic barrier, in that we are making touch with a level of knowledge that is only partially comprehensible, and so we fill the gaps in our understanding with more or less baroque imaginings.
This suggests a possible way of detecting a barrier indirectly, rather like a black hole, by spotting its side-effects. I will return to this below.
3.2.3 Epistemic barriers cannot be located
I can’t point and say ‘look, an epistemic barrier.’ But even if I can’t actually see the barrier, can I at least get some idea of when I am getting close to the bounds of the knowable? Putting it like that, the answer is, as you should be expecting by now, no.
Imagine, for the moment, that I can describe the realm of ideas as some sort of map, so I stand on the middle of a plane, and as you move away from me, the points on the plane correspond to more and more complex ideas. Now, given what I’ve already said, an epistemic barrier corresponds to a gap, where the plane is interrupted, and I have no idea whether things are resumed on the other side of the gap. Can I tell how far away the gap is?
Okay, let’s say I can understand the ideas corresponding to some point on the plane. Then it stands to reason that similar ideas should be comprehensible. So if I walk a little way in any direction I am in no risk of falling into the chasm. So what does this mean? There are two possibilities: the plane goes on forever, or it has an edge but I can never reach it because the farther I get from my starting point, the harder it is to move away. So in either case, there is no way in which I can get to the edge and peer over it. Rather than stopping sharply, the plane corresponding to my conceptual realm will, as it were, peter out.
What this means is that, no matter how far I get in my journey, no matter how complex my language or concepts get, I have no way of telling whether I am going to bump into a barrier. Because I can’t ever reach one, and no matter how close I might get, it always seems that I have plenty of space to explore. So epistemic barriers are always infinitely far away.
3.3 Can any information flow across one?
Thus far I have concluded that epistemic barriers almost certainly exist, but they are rather elusive, being invisible and always infinitely far away. For the purposes of this section, let’s say that one exists, and there is a whole treasure trove of information on the other side of it. What I want to know is, just how much can I tell about what there is on the other side?
Now, obviously, this question is only meaningful in cases where we have languages that partially cross a barrier, so we can see part, but not all, of a concept. In the case where no concept crosses the barrier, it must, almost by definition, remain totally opaque and, like a black hole, with which the analogy is very apposite, invisible. And so, again, like a black-hole, we must look at cases where the mysterious something behind the barrier interacts, at least indirectly, with things that are within our conceptual realm.
This is where the example of quantum mechanics introduced above comes into its own. Let us say that my proposal that there is an epistemic barrier preventing us from acquiring complete comprehension of the quantum realm is correct, but that some of the concepts are partially accessible. So what happens?
Well, if we look at quantum mechanics, and especially quantum field theory, we get quite a clear picture. Some parts of the theory are very well understood, so we can solve the Schroedinger equation for simple systems. Moreover, we can make rather accurate predictions about many things, from the behaviour of transistors, to what happens when you smash particles together in a supercollider.
But what we do not have is any sound conceptual understanding of how we make these predictions. The problem with quantum field theory is not just that the mathematics it is based on is sloppy. The problem is that the mathematics simply doesn’t work, to an extent that you can actually make the theory produce whatever result you want it to. The theory is based on a plethora of assumptions that seem to have no logical basis in physics, mathematics or anything, but which are required in order to get the ‘right’ answers.
Now this baroque complexity is, to my mind, rather reminiscent of the Ptolemaic description of the solar system, with its cycles and epicycles. In other words, it is an attempt to describe that which cannot be described within the existing conceptual framework. So, just like the epicycles, it can predict the right results, but that is not surprising given the rather large amount of fine-tuning that has gone into building the theory in the first place.
This leads me to another analogy, which is, I think the crucial one. We humans are programmed to recognise faces, as recognising other people is a key survival skill. And as such, whenever we look at any random-ish pattern, be it clouds, wallpaper or a carpet, we see faces. We find patterns that are not really there. Looking back at the example of quantum mechanics, it is hard not to conclude that the same is happening. We have a collection of concepts that we can understand around the edges, but where understanding tails off into random noise when we approach the barrier. And so the baroque complexities of quantum field theory are simply the equivalent of faces in the clouds: imaginary patterns that we trace in the randomness to fill in the gaps in our understanding. We think that they tell us about the universe, but in fact they tell us about ourselves.
And this, I believe, is what will happen when we look at an epistemic barrier. We will not see a barrier, but rather gibberish in which, being the pattern-seeking creatures we are, we will find patterns. But we can expect such patterns to be unsatisfactory, and insusceptible to any deep understanding, because at best they are attenuated shadows of the the unknowable, and at worst they are nothing more than random nonsense dressed up with the trappings of reason, and as such our belief in them will eventually (as seems, regrettably, to be the case with quantum field theory) become a matter more of theology than epistemology, and so to derive its strength from faith and not reason.
Now that idea opens up a whole new, and very tempting possibility for philosophical exploration, that is to say, an examination of the epistemological location of the deity, but I believe that now is the time to bring this essay to an end. We have seen that there are bounds on knowledge, but that those bounds are themselves unknowable, and moreover that though information may leak across the boundary, we will never be able to tell whether it is true air from another planet, or just images drawn on the clouds by our imaginations.