My emails with Ayndryl
This is a series of excerpts from an email exchange I had with Ayndryl, site admin and one of several contributors to Forgotten Languages.
It’s a pretty fluid discussion that curvedly touches on: linguistics, computer programming languages, cognitive psychology, Turing machines, quantum mechanics, and philosophy [having the final word].
If you’re familiar with their website this wouldn’t strike you as odd given that Forgotten Languages’ articles seem to, as a whole, cover seemingly every field of human study. This is one reason why the website is so hard to summarize/explain (another being that most articles are seemingly encrypted in various ways, but that’s another topic of discussion…).
Nevertheless, I think the following exchange can be enjoyed context-free.
Note: I received permission from Ayndryl himself to publish this exchange.
Boldings are my own.
[Saša:]
I’m interested in hearing your thoughts on something, if you don’t mind, given that you clearly speak multiple languages and study language at a very fundamental level. It seems like the “mainstream” academic opinion is that, although natural languages aren’t perfectly translatable between one another, all natural languages are equally expressive. I haven’t seen this explicitly stated anywhere, but it seems implicit in the linguistics and cognitive science literature. (My academic background is in computer science and math/stats, but I’ve had exposure to linguistics and cognitive science through research in natural language processing; that is, I’m far from an expert in the latter subjects.)
Here’s my question: are there any existing natural/human languages that are strictly more expressive than others? To make sure we’re on the same page, I loosely define expressiveness as the number of relatively distinct/independent concepts that can be sufficiently communicated in a given language. I am, of course, assuming we agree that languages are capable of “pointing to” real things rather than the so-called postmodernist view that languages (as spoken by human beings) construct concepts/reality, in which case any language could produce and thus denote an infinite number of concepts.
Also: are you familiar with the work of Timothy Morton? Object-oriented ontology?
[Ayndryl:]
All natural languages are effectively equally expressive, though languages are meant to convey ideas that are relevant and pertinent to the speakers of those languages. This means a natural language is meant to naturally express those ideas that are strictly relevant for those speaking that language, though this does not mean that language is strictly restricted to the belief system of the speakers. Hence, expressing ideas of particle physics in Lakota is not done naturally, you need to engineer your vocabulary, as a Lakota speaker, to do it. Can this be done? Obviously, yes. So the answer to your question is this: any natural language can be crafted and modified to express new ideas, even those not initially belonging to the speaker’s belief system.
Saša, your academic background in computer science, mathematics, and statistics can be used to rephrase my answer above. See, we can ask whether any computer language can be used to solve one and the same problem. Can we use Lisp, J, F#, or Factor to solve, say, a computation in fluid dynamics? The question is yes, we can, though obviously some programming languages are better than others to solve this problem efficiently. For problems in natural language semantics we would better use Lisp, while for tough problems in statistical mechanics we would use K, or J, or even Fortran. But this does not mean we cannot use K or Erlang to solve problems in natural language understanding, something for which Schema or Lisp are better fit.
Therefore, your question on whether there are any existing natural/human languages that are strictly more expressive than others must be answered negatively. But you make an intriguing distinction. You define expressiveness as the number of relatively distinct/independent concepts that can be sufficiently communicated in a given language. The question here is that before posing this question we must agree on this: if language expresses thoughts, then your question reduces to whether any human brain is limited in the number and richness of concepts it can conceive, and whether this limitation is imposed by language itself because, as far as we know, thinking means expressing internally concepts and ideas and, as it happens, humans think using language itself! This is a tough research area in cognitive psychology.
Your hypothesis that languages are capable of pointing to real things rather than construct concepts/reality, is challenged by quantum logic, in the sense that we do not know what reality is or, at least, we agree that things out there can be differently interpreted by different observers. Therefore, we lack a solid concept of what is real and what is not. Usually, you navigate this problem by using deictic ontology, that is, by pointing to a tree and assigning a label to that object, a label called tree. Any speaker of any language will use a different label, but pointing at the same object makes mutual intelligibility possible for all speakers. However, what happens when you face a phenomenon, an event, or even an object which is totally new and unknown to each and every human? Language collapses, understanding fades away, and logic blows up. This is a known problem in quantum logic, and a serious and deep problem having to do with mind/matter interaction and the very essence of reality and being.
Yes, I am familiar with Timothy Morton’s object-oriented ontology, and I feel sad this ontology is not more wide spread among society, for it poses interesting questions that could induce a positive and radical change in the belief system of people.
[Saša:]
You answered my questions, but created several more (I suppose that’s how good discourse goes!). I respond inline below:
All natural languages are effectively equally expressive, though languages are meant to convey ideas that are relevant and pertinent to the speakers of those languages. This means a natural language is meant to naturally express those ideas that are strictly relevant for those speaking that language, though this does not mean that language is strictly restricted to the belief system of the speakers. Hence, expressing ideas of particle physics in Lakota is not done naturally, you need to engineer your vocabulary, as a Lakota speaker, to do it. Can this be done? Obviously, yes. So the answer to your question is this: any natural language can be crafted and modified to express new ideas, even those not initially belonging to the speaker’s belief system.
This makes sense to me, and is along the lines of what I already believed. But I do find it particularly intriguing that this seems to imply that natural language emerges “whole” and “complete” — that is, all of the necessary potential to express human thought is contained within every natural/human language. A better way to put it: there has never been a case (among all languages in human history) where some feature of the grammar or morphology or phonetics (etc) of a particular language prevented it from evolving to represent new concepts, like, for example, the existence of some fundamental bug in a program? You can infinitely amend any given (naturally emergent) language (in some internally consistent manner) and its expressiveness is only bounded by the limits of human thought?
Saša, your academic background in computer science, mathematics, and statistics can be used to rephrase my answer above. See, we can ask whether any computer language can be used to solve one and the same problem. Can we use Lisp, J, F#, or Factor to solve, say, a computation in fluid dynamics? The question is yes, we can, though obviously some programming languages are better than others to solve this problem efficiently. For problems in natural language semantics we would better use Lisp, while for tough problems in statistical mechanics we would use K, or J, or even Fortran. But this does not mean we cannot use K or Erlang to solve problems in natural language understanding, something for which Schema or Lisp are better fit.
I do appreciate the analogy, however I don’t believe it’s entirely correct. If you’re classifying “computer language” as a Turing-complete language, then yes, all computer languages are equally expressive, by nature of all being recursively enumerable (expressible by a Turing machine). But if you include non-Turing-complete languages then this isn’t the case. Are you suggesting that all human languages can be reduced to some primordial Turing machine-like thing? (I’m now touching on Universal Grammar stuff, but I do believe my question is a bit more broad…) Depending on how you look at it, I suppose this is trivially true given that natural language is produced by the human brain and thus reducible to neural activity … I guess I’d be more intellectually satisfied with a coarser-grained reduction.
The question here is that before posing this question we must agree on this: if language expresses thoughts, then your question reduces to whether any human brain is limited in the number and richness of concepts it can conceive, and whether this limitation is imposed by language itself because, as far as we know, thinking means expressing internally concepts and ideas and, as it happens, humans think using language itself! This is a tough research area in cognitive psychology.
Ah, I see. I think this touches upon what I asked above re: whether the expression of human thought is somehow restricted by human language. I guess the answer is: we don’t know?
Your hypothesis that languages are capable of pointing to real things rather than construct concepts/reality, is challenged by quantum logic, in the sense that we do not know what reality is or, at least, we agree that things out there can be differently interpreted by different observers. Therefore, we lack a solid concept of what is real and what is not. Usually, you navigate this problem by using deictic ontology, that is, by pointing to a tree and assigning a label to that object, a label called tree. Any speaker of any language will use a different label, but pointing at the same object makes mutual intelligibility possible for all speakers. However, what happens when you face a phenomenon, an event, or even an object which is totally new and unknown to each and every human? Language collapses, understanding fades away, and logic blows up. This is a known problem in quantum logic, and a serious and deep problem having to do with mind/matter interaction and the very essence of reality and being.
Ha ha, I knew there’d be some contention to my use of the word “real”. I can’t speak much on quantum mechanics; it’s high up on my list of “things to learn”. I do find it fascinating that it seems to throw to the wind almost everything modern humans thought was true and obvious about reality, and yet a classical physics perspective (i.e. “billiard ball causation”) is implicit in almost everything about our modern society. That is, mainstream society hasn’t quite caught up/grasped the implications of those scientific discoveries from last century. This is actually one of the many reasons why I like Timothy Morton’s work — one, I think it offers a nice resolution to the so-called analytic-continental divide, and two, he addresses things like quantum mechanics and bravely grapples with what philosophy can say in light of it.
[Ayndryl:]
You do find it intriguing the possibility natural language emerges “whole” and “complete”. Actually, it does not. Language emerges in a slow process of self-organizing dynamics, and contrary to what one may infer from the syntax of a given language this process is highly non-linear. It is affected by the environment, which in turn affects the neural setup of the speaker, a speaker that, thanks to his capability to use language, modifies the said environment. It is a complex interplay which explains why a given language, say English, is totally alien for a Middle Age speaker compared to a speaker of modern English.
However, I do agree with you in that language expressiveness is bounded by the limits of human thought, though this is not the only parameter limiting and constraining expressiveness. See, language requires a minimum of two agents: two speakers. It is a means to exchange ideas and therefore there are at least two limiting factors. One is your neural setup, your capability to have ideas; second is your partner’s capability to understand those ideas. This has to do with you biological and neurological setup. Language must be understood by both speakers, or otherwise it wouldn’t serve its purpose. You both need to share a common view of the environment, otherwise communication fails. Please, consider the case of two speakers, one speaking Samoan, and a second one speaking English. Though their languages are totally different this does not prevent the emergence of a shared code, and given the necessary time and effort they will come out communicating and exchanging ideas. Notice that they both share the same environment, and that they do share a common biology and neurology. They sense the environment using the same senses, thus they perceive the environment in an identical way.
Yet, what would happen if one of the speakers meets, say, a dolphin? Or a jelly fish? Would they be able to communicate, to exchange ideas? Actually, would they ever have the same ideas about the surrounding environment? Here, you see, there is a biological barrier. What about trying to communicate with a hypothetical alien intelligent species? Is intelligence enough to ensure fruitful exchange of ideas? Or would you both need to invent not just a new shared language, but rather a new way of communicating beyond language itself?
You ask me if I am suggesting that all human languages can be reduced to some primordial Turing machine-like thing. I am, indeed. Actually, I do also believe all life forms can be reduced to Turing machines. Alan Turing, too. This is a subtle statement: it means Turing machines can reflect and think about their being machines. However, by definition, being aware of being turns you into a quite different and special machine, a class of machines you call life forms and who have special properties standard non-aware machines don’t.
You write that natural language is produced by the human brain and thus reducible to neural activity. I would agree with this statement, provided you refer to human natural languages. But don’t forget language is an emerging property of complex systems, and complex systems are not just human (or biological) in nature. Exotic logical systems exist which have a language and are not biological in nature.
Finally, Saša, I ask for your indulgence for my contention to you using the word real. Certainly, quantum mechanics has a saying in what we must understand about Reality, but its bears to Philosophy to say the last word. Do not forget quantum mechanics will always be under suspicion for just one reason: it is the product of brains which are limited by how they sense the surrounding environment. In quantum mechanics the role of the observer is essential and fundamental. However, current formulations of quantum field theories and chromodynamics are totally anthropocentric. They wish to impose on us a description of reality as seen by humans. This is unacceptable to me. I cannot accept a description of reality derived by Turing machines.
Quantum mechanics do indeed require an observer, but it says nothing about the dreams of the observer. In fact, it cannot even explain why there is an observer in the first place.