The Future of Language

More from Henrich’s The Secret Of Our Success:

Linguists and linguistic anthropologists .. have often assumed that all languages are more or less equal, along all the dimensions that we might care about – equally learnable, efficient, and expressive. .. Recently .. cracks in these intellectual barricades have begun to multiply. .. Like [other kinds of cultural] toolkits, the size and interconnectedness of populations favors culturally evolving and sustaining larger vocabularies, more phonemes, shorter words, and certain kinds of more complex grammatical tools, like subordinating conjunctions. (p. 233, 259)

The most ancient languages we know of are visibly impoverished compared to modern languages today. It just takes longer to say similar complex things in those languages. Assuming that the size and interconnectedness of populations speaking the main languages continues to increase into the future (as they do in my em scenario), we can make some obvious predictions about future languages.

Future languages should make more distinctions such as between colors, and have larger vocabularies, more phonemes, and shorter words. They should also have more grammatical tools such as adjectives, tenses, prepositions, pronouns, and subordinating conjunctions. Technology to assist us in more clearly hearing the words that others speak should also push to increase the number of phonemes, and thus shorten future words.

For obvious reasons, science fiction almost always fails to show these features of future language.

If you search for “future of language” you’ll find many articles noting that the world is losing many unpopular languages, and speculating on which of today’s languages will be the most popular later. And this creative attempt to guess specific changes. But oddly I can’t find any articles that discuss the basic trends I mention above.

GD Star Rating
Tagged as: ,
Trackback URL:
  • zarzuelazen

    Depends what you mean by ‘language’. The three core ontological domains are (1) physical world , (2) social/teleological world, (3) logical/math world. Most people only use the word ‘language’ to refer to communication in the social domain (2) e.g. English, French etc., but course there are ‘languages’ for domains (1) and (3) as well.

    I think the most interesting future developments will be in domain (3) – math/logic languages.

    I should mention that there’s just been a big critical breakthrough in machine learning. No hype here, this truly is a huge leap forward, an ‘order-of-magnitude’ paradigm shift.

    It turns out that philosophers have been looking at ‘concepts’ all wrong. Until just recently concepts were thought of as features in a ‘feature space’. The big new breakthrough now shows that concepts are ‘models’ in the sense of a ‘computational model’ – concepts are programs! (the philosophy of object-oriented programming that programs are best thought of as models of the world appears to have been vindicated).

    Whereas previous deep learning techniques needed hundreds of training examples to learn new concepts, the new technique is matching or bettering human learning ; the programs are even able to generalize from a single new example! Search on ‘Bayesian Program Learning’ (BPL) framework: Concepts are represented as simple probabilistic programs.

  • isegoria

    Heinlein’s Speedtalk, from his novella Gulf (found in Assignment in Eternity), has a complex syntax and many more phonemes than current languages. It’s also logic-based.

  • lump1

    In the late 60’s, the French band Magma wanted to write music that sounds like it’s from the future, and did all the singing in an invented “future” language called Kobaian. Interestingly, Kobaian sounds vaguely Germanic but somewhat more terse and written phonetically if the liner notes are anything to go by. Consonants are even more prominent. It’s not something for everyone, but Magma are still at it and now others started to using the language and the style, called Zuehl (the Kobaian word for “music”).


    English has already stripped away much unnecessary complexity, I guess you notice this more easily when your first language is not English. It’s easy to learn if you already know a germanic or romance language. When people say English will be the global language they’re not saying it will be exactly the same as it is now.

    • The long term growth in complexity should consist of adding many complex elements and then taking most of them away. The process of innovation requires many attempts, few of which end up being long term successes.

  • Zhang Tingyu

    Why would future language need more colors? Russian has more color words, hasn’t done them any good.

    Ancient languages aren’t impoverished by any rational criteria. Classical Chinese takes substantially less to express complex things than in modern Mandarin. In fact the educated classes always despised the vernacular as being verbose and tacky. People in Hong Kong think the same way about mainland prose today.

    Languages aren’t nothing special.They are culture, nothing more. If more color words are needed it isn’t hard to come up with them; people who need them professionally already use them. The same with phonemes or anything else. Ancient Greek has much more detailed tense conjugation than most modern languages. So what? Context can disambiguate anything.

    Your high-complexity model seems to assume that future humans will share no life experience, no disambiguating context at all, hence the need to spell everything out.

    • Why would future language need more colors?

      If you can express the same idea in fewer words, it takes less working memory, which can be employed in understanding the utterance more thoroughly. A richer language should conduce to greater intellectual productivity.

      [I apply this reasoning to punctuation in “Writers should exploit all punctuation marks – . I discuss the comparison with lexicon.]

      • Zhang Tingyu

        While I don’t know what model you have in mind when you say “understand the utterance more thoroughly”, language is very very malleable as it is. People who want more precise terminology can produce it very little time. Intellectual productivity precedes innovative language use, not the other way around.
        That we don’t have more words shows we don’t need them. For that in change in future language you need to show why our language needs would change.

      • That we don’t have more words shows we don’t need them.

        1. If I need a neologism, I might invent one. But then I’d have to memorize it. We learn new words with much less mental effort than it takes to memorize our inventions.

        2. Using new words in communication is a social coordination problem. I can’t just start using a new term. I’ve invented new terms and have never enjoyed the slightest success in getting others to use them. Others have done the same. We are all limited to fewer words than we think optimal.

        By “understand the utterance more thoroughly” I meant merely spend more brain cycles thinking about its implications, as opposed to figuring out what meaning is intended.

  • kriszyp

    Isn’t the lesson of computers that less, more distinguishable phonemes (in the case of computers, 2) that can be more rapidly sequenced and understood is superior to more phonemes?

  • Ryan

    Perhaps this is my traditional linguistic training coming through, but I’d dispute the framing of this argument – it seems to portray a monolithic linguistic tradition that denies extralinguistic factors in language change and typological drift, but in fact linguists have been proposing such factors for hundreds of years. There’s certainly an oscillation in acceptance of the idea, and we’re on a swing back towards embracing it, but this is nothing new. The problem with the idea is that it’s really hard to show these effects with any level of confidence. In fact, while I certainly believe these extralinguistic factors play a role in shaping language, and predictions for future languages should take them into account, I’d argue that none of the strong claims you made are currently supported by equally strong data.

    Part of the problem is the number of confounding variables – in particular, language descent and areal effects have traditionally been problematic. Classic studies in the field like Hay & Bauer (2007) and Atkinson (2011) have been disputed on these grounds. There has been a bit of a rush in linguistic typology in recent years to ground sampling methods and statistical tests on firmer ground, but a majority of the field remains a confusing mess of half-thought-out-ideas, bad math, and bad linguistics. There are no standards, and so comparing different studies is difficult. This in itself is enough to make me cautious about accepting any single paper’s claim that, say, the size of an individual’s vocabulary is correlated with population size.

    There is also growing evidence that the social-linguistic correlates have more to do with amount of language contact than with population size. These are two obviously correlated variables; however, in a scenario in which the population speaking a certain language grows because of reproduction rather than because of diffusion, you might not get the same effects that we’ve seen in the past. This is part of a larger idea – that language trends correlate with factors more complex than population size (such as language contact, community network density, and population stability), which may not change in a single direction as we move into the future. There is little doubt the population will continue to grow, but – will the number of multilinguals increase? Will we interact with more people on a day-to-day basis? How long will those interactions be? Saying that “size and interconnectedness” will continue to increase may be underspecifying the trends.

    I’ve left this to the end because it’s not central to the kind of argument you’re making, but your claim that “the most ancient languages we know of are visibly impoverished” seems the most incorrect to me. The data we have for old languages are incredibly restricted, both in size and in number of genres, and so we’re not aware of the richness that may have been present in them. You’re not going to find examples of subordinating conjunctions in Mycenaean Greek, because we only have their old inventories and receipts. Furthermore, we don’t have a large enough number of ancient language corpora to make a strong generalization about them. I’d appreciate finding out which sources led you to make this claim.

    • I am relying on this book as a source; I’d be interested to hear book reviews or other rebuttals of this view. You are mainly saying that the issue is complex and data is uncertain and therefore no conclusion can be held with much confidence. That sounds reasonable, but I’d like to hear if someone is willing to make a stronger argument.

    • arch1

      Do you know of a list containing estimates of aggregate corpus size for various ancient languages? I have seen claims that our corpus of classical Latin, for example, is surprisingly small but haven’t found authoritative specifics.

  • Christian Kleineidam

    English is a language that lost it’s ability to distinguish singular `you` from plural `you`. These days `they` can also mean both singular and plural.

    German seems to slowly use it’s 4th case and get’s gramatically less complex. Contemporary German is a lot easier than the German Mark Twain complained about.

    Both English and German can’t distinguish maternal from paternal aunts and uncles while Latin could make that distinction.

    When it comes to color words Xkcd has a nice chart:
    Woman who think about the color arrangement of their dresses effectively speak a version of English with more words from colors than guys do.

    • IMASBA

      English also lost the formal version of “you” (the French still have “tu” and “vous”).

  • hk

    implication doubleplusungood

  • Pingback: The Secret of Our Success | Bayesian Investor Blog()