Language Erodes

I used to be a linguistic prescriptivist. I used to think there was one and only one way to correctly use language, and held on to rules like prohibitions on split infinitives and prepositions at the end of sentences. The word “whom” made me all warm and fuzzy inside, and I felt whipping out a dictionary was a valid way to resolve a disagreement.

In my defense, it’s hard to avoid adopting this view when you grow up as a natural know-it-all and therefore come to think of “knowing the correct answer” as the supreme virtue for any human being to possess.

Later, when learning more about etymology and the mechanisms of language and how it changes, I switched sides. My new ethos was “convention is just convention, not worth the paper it isn’t even written on; I’ll change anything if I feel it’s an improvement”.

I’m still mostly of that persuasion, having defended things like “literally” used nonliterally, and the phrase “could care less” as making perfect sense, but I have mellowed, and rejected the naive, Panglossian descriptivism that thinks any language change is for the better and nothing can ever be ~wrong.

That view confuses, or perhaps conflates, the futility of trying to halt language change with it never being a bad thing. You fool, don’t you understand that the grapes are sour?

“Language evolves” is its battle cry. It’s a glib, thoughtless dismissal of any and all complaints about the misuse of words. And it works — because it’s true. Language change is a lot like evolution: a blind, purposeless process that keeps what works. “Works” in the very narrow sense of successful self-preservation and propagation.

It works as an argument because the word “evolution” has certain connotations: it suggests progression and improvement, going from simple to complex, primitive to advanced, crude to sophisticated — from worse to better1. The hard-to-challenge implication becomes “language naturally improves, so objecting to any language change puts you on the wrong side of history”.

To be fair it’s easy to think of evolution as improvement when we see the ingenious feats of engineering it’s produced over many millennia. But it’s not true, not in the sense of there being some inherent progressive or improving force, principle or rule in evolution. There’s not even any momentum. Nothing makes it happen or continue to happen once it’s moving in a certain direction. “Evolution” is just a description of something basically tautological — genes that take part in building organisms more likely to survive and leave more offspring will survive and become more common over time.

Evolution is erosion

To get us to abandon this seductive but false idea of an evolutionary progression or “ladder”, evolution is sometimes compared to erosion. I think both top evolution pedagogs Daniel Dennett and Richard Dawkins have done this. Erosion is also a blind, unstoppable and mindless natural process but, compared to evolution, its destructive properties are more apparent. It produces smoothness and decay more than complexity.

While the “language evolves” metaphor suggests improvement, “language erodes” does not, and it’s just as apt an analogy for language change.

The typical case of language erosion takes a precise concept, with sharp, distinct edges, and uses it somewhat sloppily over and over again, making it fuzzier, communicating more of a holistic vibe than a specific thing. In a sense it is drained of meaning, becoming less and less useful for precise communication. It goes from sharp and narrow to smooth and broad — just like erosion.

I have examples. As a tech person in a consulting company it irritated me when I saw it happen in real time with one of the big jargon hits of the early 2010s. Reaching peak popularity around 2013, the phrase “big data” referred early on to datasets that were so large that traditional database management systems couldn’t store or process them effectively, and other techniques and tools were necessary. In the hands of MBA:s, consultants and marketers, however, with no clue about the sort of issues you deal with in database management, it got so worn out by being used in all kinds of senses that it soon meant little more than “data”, but with the added connotation of being “impressively much” in a general sense, plus invoking a feel that “data is big , as in, it’s really the hot new thing right now as there is so much of it, and we’re on top of that whole deal”.

No Bob2, 2000 records is not “big data” just because it looks like a lot when you open it in Excel.

Natural word erosion

This is all natural, of course. It’s built into how we learn new words. We aren’t typically given explicit, precise definitions. We hear words being used and infer what they mean using context and our own background knowledge. E.g. everyone understood what “cromulent” meant. This often works well enough, but not always. It doesn’t, for example, work well for technical terms coming from professional fields or social contexts we’re unfamiliar with.

In other words we’re good at understanding the function words fill but not necessarily what it is that they mean that makes them fill that function. Hearing the phrase “increasing exponentially” has made a lot of not very mathematically literate (and some mathematically literate but thoughtless or careless) people think “exponentially” means something like “a lot”. Textbook word erosion. In reality it means increasing (or decreasing, strictly speaking) at a fixed proportional (not absolute) rate over time. Sometimes this means “a lot, quickly” – since it involves the absolute rate of increase itself increasing – and crucially that is often the time when we will hear it being used (because it’s dramatic) so that’s the implied meaning we absorb.

“Statistically significant” is going the same way. It has a very precise meaning in statistics, that I will now attempt to state accurately in one sentence (it’s scientist’s equivalent of the offside rule): for an effect to be statistically significant there has to be a less than 5% chance for the observed result (or something even further in the same direction) to occur, given that the effect does not exist. I.e. given that what we’re looking for doesn’t exist, this result is really unlikely, so it might be more likely that it does indeed exist. You might, if you promise to be very careful, be allowed to shorten “is statistically significant” to “likely exists”.

I’ve seen it used a number of times in the exact sense you would predict if people didn’t know this and instead relied on their preexisting understanding of “statistics” and “significant” to make sense of the phrase. It looks like it means “something statistical that is significant” when used in expressions like “the amount of ventriloquists is Belgium is hardly statistically significant”. Given the technical definition this makes no sense at all, since even one single Belgian ventriloquist would provide enough statistical evidence that some do exist. It’s intended to mean, of course, that “they’re so few that they’re an insignificant group, and this information is of a statistical nature”. That’s word erosion3.

Now, this is evolution in a sense. Words evolve toward meanings that are more “fit”. Fit for being widely used. They shed meanings that aren’t likely be transmitted accurately because they’re difficult, unintuitive, or rely on specialized knowledge. In doing so they get more fit for propagating themselves, i.e. they get easier to use for more people, for more things, in more situations.

It’s evolution, but also erosion in that they get less specific and less able to communicate precise concepts. As words evolve and erode, they get better at some things but worse at others, and this is a real cost. It makes perfect sense to complain.

The response best for your mental health in equanimity, but one of these I cannot let go of is “troll”. It used to mean somebody who amused themselves by saying provocative things they didn’t mean to get a rise out of people thinking they were serious. By now it’s been diluted into little more than a generalized insult. I’ve seen it explicitly defined as “a person that’s being abusive online”, which it isn’t supposed to mean at all! Nor does it refer to people saying provocative things they actually do mean! Now what do we use to express the original meaning?

It’s easy to see how and why this erosion occurred. A troll is just there to provoke and should be ignored (“don’t feed the trolls”). What they say isn’t legitimate and they can be justifiably dismissed. But with that we have a valid reason to dismiss anyone we think isn’t acting fully in good faith! Even if they aren’t pretending, like a genuine troll. And then we’re only a small step from anyone who’s being mean or saying something we don’t like or acting illegitimately according to us in any of a number of ways. And meaning is sanded down.

Incentive-accelerated word erosion.

“Troll” is a good example of how natural word erosion (through imperfect learning) can be accelerated by social incentives. When there are rewards to using a word slightly outside its original meaning it can erode mighty fast. “Incel” went from a simple and clear-cut concept (anybody involuntarily celibate), via a specific subculture around it, to a vague insult potentially aimed at any unattractive man giving off kinda-creepy vibes, all in a matter of months (I’ve heard it used to insult a married man). Other words, originally referring to specific and worthy-to-point-out phenomena, like “gaslighting”, “cultural appropriation” and even “mansplaining” has also been eroded to near-meaninglessness due to powerful incentives to overuse. And of course, “racism” has been expanded to include anything that looks even a bit, from any angle, using any lens, like racism as defined in the previous step, for quite a few iterations now.

Those examples are from the political left, but that not because I think it doesn’t exist on the right. It certainly does but I’m less exposed to shitty right-wing discourse than shitty left-wing discourse so it’s harder to find examples (although “groomer” is getting there). I can think of “socialism”, which economically speaking has a pretty sharp definition but on the right can mean almost anything in the direction of valuing others or society as a whole over individual rights and self-interest. A review of the book Bringing up Bébé (written by an American woman living in France, describing how the French raise their children), complained about French children being taught “socialism” at an early age. By what I could gather this meant that they’re taught self-restraint. I get how she meant it, I really do, it’s perfectly natural word erosion, but also very far from, and less specific and clear cut than “the workers own the means of production they’re working with”.

The ~right are also responsible for eroding “Marxism” into meaning “any collectivist conception of society focused on inequalities” and “postmodernism” into meaning “any unconcern for truth and objectivity”. The ~left are similarly responsible for eroding “capitalism” into meaning “each and every way society is not an infinitely wealthy hippie commune”4.

Intelligent language design

Language evolves and erodes, with or without incentives. Words will evolve to be used more, demanding less specific conditions and less specific knowledge. When they do so they also erode, becoming less sharp tools of communication for those who do have the specific knowledge and are in the right specific conditions. As they’re used widely they acquire a bunch of hard-to-enumerate connotations, associations, nuances and vibes. Meanings turn more illegible, if you will. This accumulated cruft, bringing vagueness, ambiguity and instability with it, is a liability when you want to be precise.

What’s the solution? Intelligent language design: the deliberate creation of new words that mean only and exactly what you intend for them to mean. It’s one of the major reasons for the popularity of jargon — at least it is for me; I make up new words whenever I don’t find any that say precisely what I want, without risk of readers coming to it with varied and unknown preexisting personal associations and residual meaning fragments that affect their interpretation in unpredictable ways5. No thanks.

So in a way, this post is a response both to “language evolves” as a dismissal against any complaint, and to “just use words that already exist instead of making up new ones”.


  1. This idea of what evolution is has found itself into popular culture where it does some serious damage. X-men, Pokémon, and the dumb 2001 movie Evolution, I’m looking at you. ↩︎
  2. Not the real name. ↩︎
  3. I know this less well than statistics, but I think the same has happened to the phrase “radical feminism”. As a “technical term” it refers to a particular school of second-wave feminism, but has come to mean “feminism that is in a general sense radical”. “Critical race theory” is left as an exercise. ↩︎
  4. If people use words in these eroded ways it’s not a valid reason to just dismiss what they say. “But they don’t even understand the words they’re using!” Perhaps, but they’re referring to something, and that doesn’t cease to exist just because you can argue that the sign they use to point at it is more accurately pointed at something else. ↩︎
  5. See for example, use of “reality” by social constructionists. ↩︎

Did you enjoy this article? Consider supporting Everything Studies on Patreon.

9 thoughts on “Language Erodes

  1. I get the point and I agree with you; and the much of the pattern seems the same as in the Political Capital Flow/Yutting post (though much is also just laziness, stupidity, and ignorance).

    But consider the issue from a different point of view, not from lowest common denominator language but rather language as used by the thoughtful. Is this still true?
    I don’t think it is; I think that those who want to think are being given better thinking tools , not just over slow generations (English invents a future tense based on the word “will”) but within our lifetimes as we see words like yutting, concepts like skyhook, or phrases like “I can tolerate anything but the outgroup”. This is not exactly Sapir-Whorf, it’s not “literally” English that’s becoming richer so much as the Anglo communities set of tools for thinking.

    OK, so with this in mind, consider two questions.
    First, is it monotonic? Yes we enrich ourselves with new thinking tools, but do we lose older thinking tools? I’m not sure. Perhaps it’s not monotonic but is corrected within a generation or so? I think there was a backwards glitch when the 19th C (and earlier) sophisticated understanding of “the devil” as “those parts of our self that think or behave differently from as we’d wish” got buried under first psycho-analysis, then various social “science” theories that depended on a unitary self to make sense; this is being corrected in our time via more sophisticated concepts like “a non-unitary self”, though it’s still very early days for this particular concept.

    Second, and getting back to your essential point, does this ability to think better via a better language (and set of associated tools) percolate downward? Again I’m hardly sure, but I think that overall it does. It’s not perfect, but it happens. For example the language of 60s psychiatry started to become mainstream (for the mainstream of people like you and me) in the 70s, language like co-dependent, passive-aggressive, “what I hear you saying is…”, and is now so mainstream that it can be used on a network TV show and everyone’s expected to know what it means. Mostly I think this is progress, that the ways of thinking and behaviors engendered by this language are usually better than what went before.
    Of course consciousness is, as always, the PR agent, concerned primarily with self-justification and excuses, which limits the extent to which better language can lead to better behavior; and this is most prevalent in the examples you give, where people either want to perform tribal loyalty regardless of truth, or want to show off. But not all of life is tribalism and showing off… And even when it is those cases, having at hand concepts like yutting, or virtue signaling allow one to recognize behavior and tune it out.

    On the other hand, examples like evolution (or “theory”) certainly are depressing, and don’t seem to be on the way to getting better any time soon 😦

    Liked by 2 people

    1. Not quite following all the way but I think you’re agreeing with me? Language change follows certain patterns that can be described as both “evolution” and “erosion”, and this brings both benefits and drawbacks. Evaluating any particular change must be based on the specifics of the case.

      Like

  2. “Given the [above-given] technical definition this makes no sense at all, since even one single Belgian ventriloquist would provide enough statistical evidence that some do exist.”

    I would have chosen “the incidence of ventriloquism in the Belgian population matches the incidence of ventriloquism in other EU member states” as null hypothesis instead. (For some operationalization of “matches”.)

    An interesting concern for jargon design: assuming an essay with a very successful addition to ontology, what will happen to its future “readers coming to it with varied and unknown preexisting personal associations”? As in, to what extent are future associations predictable — either purely as a function of the concept independently of the symbol (i.e. the conceptual fitness landscape is recognizably tilted in some direction) or as a function of the symbol chosen?

    Also, for amusement, write a webpage with a gensym. As in, in the limit case of infinite effective list size, every reader would see the key concept introduced in the essay referred to by a different word.

    Liked by 1 person

    1. You mean the last thing as an experiment? Yeah would be fun. The ventriloquist example was, in the different example I was thinking of, not comparing any two things, but basically saying “there are so few ventriloquists that they’re insignificant” which a reasonable meaning but not correct use use of the phrase — and clearly resulting from coming across it and absorbing the wrong meaning.

      I agree that you can’t control the future meaning of terms, that’s just a fact of life and something you should remember when reading any old text (or reading any text honestly). You can predict it a little bit, I guess, and I’ve been thinking about how you create concepts that “erode gracefully” but I can’t say I have a solution.

      Like

  3. What’s the solution? Intelligent language design

    That we have access to expanded techniques in this non-physical domain doesn’t mean the old ones aren’t any good. If languages evolve, then they evolve in response to selective pressures, and the constant nagging of prescriptivists is just as valid a selective pressure as any.

    Like

Leave a reply to John Nerst Cancel reply