Change by mistake

« previous post | next post »

A couple of days ago, I tried to answer a journalist's questions about nonplussed ("Nonplussed about nonplussed", 8/6/2008). I wasn't entirely satisfied with one of my answers, and so I've tried again today.

The question was

Q: Is it possible for a word to become so commonly misused that the new (wrong) definition becomes acceptable? Has any word like this ever had its new meaning included in a dictionary?

The answer, of course, is that every word in every language has been through many cycles of "misuse" becoming "acceptable" — this process is otherwise known as language change. (Some English words have not changed a great deal in the few centuries since reliable dictionaries for English have existed — but that's another matter.) As examples, I cited the journey of silly from "happy, blissful" to "foolish, simple", by way of "pious", "innocent", "helpless", and "insignificant". I also mentioned dirt's shift from excrement to mere soil.

But in the comments, Ann objected:

I'm not sure that "silly" is really a good example of the process that might be happening with "nonplussed." In the former, each subsequent definition represents a slight shading of meaning, an expansion from meaning A to meaning B to meaning C, etc. But with "nonplussed," it seems as though people simply don't know what it means and are getting it wrong, probably from the negation you noted.

I worried about this myself in writing that post. Of course, it's usual to subdivide modes of semantic change more finely than just "an expansion from meaning A to meaning B to meaning C".  Thus Leonard Bloomfield's 1933 taxonomy used the categories of narrowing, widening, metaphor, metonymy, synecdoche, hypobole, litotes, degeneration, and elevation. And some of these are not at all a "slight shading" of change, but rather a quantum leap.

But it's true, the use of nonplussed to mean "unfazed" doesn't really seem to fall into any Bloomfield's categories as I understand them. And a quick glance at Andreas Blank's 1998 taxonomy — and the other taxonomies listed in the Wikipedia article — also fails to find any category that exactly fits. (Though maybe "cohyponymic transfer" is relevant, and perhaps for that matter "metonymy" if you interpret it so loosely as to mean reference to any associated concept.)

In contrast, the standard examples of words whose usage has shifted so far that they turn almost into their own opposites, and perhaps back again — peruse, fulsome, moot, nice — fit one or another of the standard categories of semantic change pretty well. So what should we do with cases like nonplussed"? Here it seems that people are using a word W that means X to mean Y instead, where the only real connection between Y and X is that both are in the same general area of abstract concepts? (At least in these cases, there are also some things about the form of W that promote the mistake, and this seems likely to be true in general, if only for functional reasons.)

There are many examples of historical change where people in some sense "got it wrong" as they did with nonplussed, as opposed to just promoting a connotation or a figurative usage to become part of the core meaning, or regularizing irony, or bleaching out some aspect of the core meaning to allow wider usage. But the obvious examples in the textbooks are folk etymologies (i.e. triumphant eggcorns) like hangnail, or morphological analogies as in the historical developments among abidden, aboden, abode and abided.

So am I missing something in the standard taxonomies of semantic change, or do we need a new category for examples like nonplussed ?

(I don't mean to suggest that the phenomenon is new, just that the standard taxonomies don't seem to have a place for it, unless I've missed something. Which I might well have done, since this is not an area that I know very well.)


  1. Casey said,

    August 9, 2008 @ 9:37 am

    "Egregious" used to mean the exact opposite, I'm pretty sure, although my OED isn't in this city. I remember reading a theory that a mix of people not knowing what it meant very well and sarcastic usage made the meaning switch. I think that would work as an example of this phenomenon, but I don't think they gave a name for it.

  2. jon kvanvig said,

    August 9, 2008 @ 9:37 am

    Besides "infer", another example may be the journalistic practice of using "refute" in place of "rebut".

  3. rootlesscosmo said,

    August 9, 2008 @ 10:14 am

    Because the prefix "in-" can have almost diametrically opposite meanings (cf. "pediculos incontusos" the other day) you get changes like "inflammable," "able to be set aflame," being replaced "flammable," with the same meaning, to prevent misinterpretation of "inflammable" as "not able to be set aflame."

  4. Mark Liberman said,

    August 9, 2008 @ 10:15 am

    @Casey: According to the OED, egregious in English originally just meant "prominent" (as you'd expect from the Latin etymological meaning "out of the flock"); the use "remarkable in a good sense" antedates the use "remarkable in a bad sense", but only by a few decades (1534 vs. 1573(. The OED suggests that the negative meaning "prob. … arose from an ironical use of 2, though our earliest quotations afford no evidence of this". The fading of "remarkable in a good sense" is recent.

    None of this is much like nonplussed or infer — it's just narrowing (the loss of the positive sense) with perhaps a bit of ironic inversion thrown in (for the development of the negative sense in the first place. Though it seems just as likely to me that the original meaning was neutral for some people, i.e. "salient in a way that might be good or bad way or indifferent", even though Latin egregius is glossed by Lewis & Short as "distinguished, surpassing, excellent, eminent".

  5. Theophylact said,

    August 9, 2008 @ 11:27 am

    How about "I could care less" for "I couldn't care less"?

  6. Tom said,

    August 9, 2008 @ 11:31 am

    Infer seems to me quite different from non-plussed, since we could lay this out:

    original meaning: to come to a conclusion not directly stated
    new meaning: to make your listener come to a conclusion not directly stated.

    Isn't there a whole class of verbs that act this way?

    For example:
    strike out (1): to get out by missing three strikes.
    strike out (2): to make the batter get out by throwing three strikes.

    Ah yes, with these verbs, the transitive sense means [to make someone X] where X is the intransitive sense. The only thing separating infer from this class is the transitivity doesn't follow through, at least I don't think so. I've never actually heard or used it to mean "imply", but I'm pretty sure you can't say "I inferred him to bring me a glass of milk". (Googling "inferred him" comes up with the unrelated "inferred him to X" where "him" is the subject of the embedded clause, and a medicalish usage I've unfamiliar with).

    Nonetheless, I still thing inferred can be understood as a (mis)generalization of the above pattern, which is pretty different from non-plussed.

  7. Mark Liberman said,

    August 9, 2008 @ 12:27 pm

    @Tom: If your analysis of new infer as a causative were correct, then you'd be absolutely right, it wouldn't be at all the same sort of thing as new nonplussed.

    Your analysis is a nice idea, and maybe something can be made out of it, but it has a problem to start with: English causatives are generally transitive. Thus "the water boiled" but "I boiled the water"; or for your example, "he struck out" but "I struck him out".

    A typical "misuse" of infer is

    The Airbus ads inferred that four-engined aircraft were safer for long trips over water.

    where the traditional usage would require "implied" rather than "inferred". The idea certainly is that the ads caused listeners to infer a certain conclusion. But the author didn't write

    *The Airbus ads inferred viewers that four-engined aircraft were safer for long trips over water.

    and I don't think that any users of new infer would ever use it that way. So whatever new infer is, it's not an ordinary causative along the the lines of "strike out".

  8. Craig Russell said,

    August 9, 2008 @ 1:39 pm

    @ Mark

    All right, how about this for an analysis of the 'misuse' of 'infer':

    English contains several pairs of verbs that have the following relationship: verb A means 'cause to B'

    set=cause to sit
    lay=cause to lie
    teach=cause to learn
    raise=cause to rise
    imply=cause to infer

    For each of these pairs, there are common "mistakes" or "misuses" where the boundaries between A and B weaken: A is used in the place of B, and/or B is used in the place of A. Presumably, this confusion arises because there are other verbs that can be used both ways (move=cause to move, burn=cause to burn, etc). The level to which these "mistakes" have become accepted varies.

    For some pairs, the bleedover can go both ways:
    e.g. "He sits (instead of "sets") the glass on the table" (B used for A)
    "The glass is setting (instead of "setting") on the table" (A used for B)

    Sometimes, A can be used for B, but not B for A:
    "If his temperature raises (instead of "rises"), I'll take him to the doctor."
    But never: "Global warming is rising (instead of "raising") the earth's temperature."

    And sometimes, B can be used for A, but not A for B:
    "The teacher learned (instead of "taught") us a lesson."
    But never: "The students taught (instead of "learned") the material really well."

    "Infer" would fall into this last category; B is being used for A.

    With this model, I don't see how transitivity is an issue; with each "mistake", the verb takes on the properties of transitivity of the verb it's being used instead of:

    "I'm going to lay (instead of "lie") in bed all day." –Transitive A ("lay") becomes intransitive when used for intransitive B ("lie").

    "He sat (instead of "set") the book on the desk."–Intransitive B ("sit") becomes transitive when used for transitive A ("set").

    Buying it?

  9. Russell said,

    August 9, 2008 @ 1:54 pm

    Benjamin W. Fortson IV's chapter in the Handbook of Historical Linguistics discusses phenomena that seem to be what we're after (you can find the original passages by searching the book on Amazon; try "hark" and "enervated")

    He brings up hark as meaning 'exclaim' rather than 'attend', at least for one speaker that he overheard. He says it does not fit into any of the traditional categories. He notes that possible causes of this change are: the ambiguity of interpretation in perhaps the most familiar use of the word (Hark! the herald angels sing), the association with verbal activity (hark back, originally said of hunting hounds), and perhaps the fact that it rhymes with bark. In any case he classifies this as simply "reanalysis."

    In footnotes, he talks about "a more complicated sort of reanalysis … when a word's meaning is assigned on the basis of a similar-sounding word." He gives enervated 'energized' and diffident 'indifferent' as examples. One might also wonder if livid 'flushed red' is another such case (could be due to similarity with 'vivid', or it might just be metonymic' (one's face may flush when angry).

    He never enters the area of morphological analogy, sticking simply to "similar-sounding."

    As to whether we need a new category (if "reanalysis" is too broad), I'm not sure. The classification set out in most intro textbooks (and related in the Wikipedia article) is interesting, but they've never seemed to be the atomic categories of change. Rather, something like invited inference is probably behind many of the categories, with particular linguistic and cultural factors leading to different sorts of inferences on a per-word basis.

  10. John Cowan said,

    August 9, 2008 @ 2:01 pm

    The OED's evidence suggests that the 'deduce' and 'imply' senses of infer are equally old in English, both landing in the 1520s. Indeed, none of the other (now obsolete) senses are any older,, so apparently the polysemy has been there from the beginning, sometimes repressed by prescriptivists, sometimes recrudescing.

    A better example might be the sense 'fortunate' of fortuitous, which the OED's revision process has not yet reached. Enormity is also interesting: it originally meant simply 'irregularity' (that is, something outside the norm); later 'moral irregularity, wickedness', and still later 'enormousness', the sense now condemned by prescriptivists, but still influencing the approved sense such that enormity is now rarely, if ever, applied to trivial sinfulness.

    [(myl) The point about infer is absolutely correct, and so I've removed the passing references to this word from the body of the post, so as not to confuse matters. It's possible that what you call "recrudescing" is an example of what we're talking about, but it's also possible that this is just another long and losing prescriptivist battle to impose order on a situation that was never orderly to start with.]

  11. Tim Silverman said,

    August 9, 2008 @ 2:27 pm

    @Tom, @Mark: the change in the semantics of "infer" isn't a causativisation, but it is (partly) a reassignment of semantic roles to syntactic roles, and so has at least a family resemblance to causativisation.

    The situation in both the old and new meanings of "infer", and also of "imply", involves three participants: a speaker, a listener, and a conclusion that can be drawn from what the speaker said to the listener, or communicated in some other way. (What the speaker actually said or did can be ignored here, although in a more detailed analysis it should be counted as a fourth role, since it can be made explicit in some cases, e.g. "Peter inferred from Jacey's shrug that Brian was now out of the picture." or "Silence does not imply consent.")

    One of the roles, the "conclusion drawn", appears in a that- clause as, I guess, a direct object, in all the cases we're looking at. But one of the things that does change in the move from old to new "infer" (or from old "infer" to old "imply") is the semantic role that slots into the syntactic role of grammatical subject: before the change, it is the listener; afterwards, the speaker.

    But we shouldn't stop there, because something else changes as well, something (potentially) in the situation itself: which person makes the connection between what is said and the indirect conclusion drawn. In both cases, old and new, this is done by the person in the semantic role filled by the subject.

    Before: "Sam inferred from Mark's words that … "
    After: "Mark inferred (to Sam) by his words that … "

    This second change is not surprising, since the subject is often associated with agency and the action of pulling or pushing the implied conclusion out of the expressed communication can easily be seen as a sort of agency. The other person—the one who is not the subject—may or may not draw the conclusion. That is (using the old meanings), the listener may infer something that the speaker didn't intend to imply (provoking the protest "I didn't mean to imply that … ") or the speaker may imply something that the listener fails to infer.

    There are certainly historical examples of reassignments of semantic roles among grammatical roles—for instance the change in the meaning of the verb "like" where the one doing the liking was originally the object and the thing liked originally the subject (e.g. "This apple likes me well," cf the modern verb "please"), but now it is the other way around ("I like this apple"). I seem to recall that something similar, but more complicated, happened to the meaning of the word "think". That is, "think" originally meaning something like "seem", with the proposition that was being thought as subject and the person who though it as indirect object (in the dative), whereas now thinker is subject and the thought is, I guess, a direct object.

    In the latter two cases, there is some inherent ambiguity about agency. Is the person who likes an apple an agent?—they are animate, rational, and aware, and can potentially do something to the apple as a result of liking it, such as eat it. Or is the apple exerting agency, by affecting the emotions of the person who likes it (and who is not in control of their likes and dislikes) despite itself being inanimate, non-rational, unaware, and incapable of action? Similarly, am I in control of my beliefs, or do they force themselves on me?

    On the other hand, with inference, the ambiguity is not necessarily inherent in the situation, although perhaps it is so in the case where the implication is intended by the speaker and the inference is drawn by the listener, since it is perhaps not clear whether the conclusion is forced (or strongly pressed) on the listener by the speaker's choice of words (or facial expression, or what-have-you), as opposed to being actively drawn out by the listener from its hidden place lurking among all the possible implications of what the speaker said.

    Hmm, I've rambled a lot without coming to a conclusion here …. Oh well, enough words.

  12. Q. Pheevr said,

    August 9, 2008 @ 2:30 pm

    Ullmann (1962) describes four sources of semantic change, characterized by two cross-classifying binary features, and I think one of his categories is a good match for at least part of what's going on with nonplussed and infer, although it's not perfect.Ullmann talks about changes based on {similarity, contiguity} of {sense, name} (where "name" refers to form). The four resulting categories are:Similarity of sense (=metaphor, as in the foot of the mountain)Contiguity of sense (=metonymy, as in bluestocking for 'intellectual')Contiguity of name (=clipping or ellipsis, as in porter for 'porter's ale')Similarity of nameThis last category includes various semantic changes based on reanalysis or folk etymology, and I think it would encompass nonplussed, at least to the extent that the new meaning is based on inferences about what non- and plus and -ed might mean when stuck together. But it's the slipperiest of Ullmann's categories, and most of the nicest examples seem to involve a change of form in addition to the semantic change (as in samblind 'semi-blind' > sand-blind 'having a defect in the eyes, by which small particles appear to fly before them' [in Dr. Johnson's definition], which is more like the agnail > hangnail case).However, Ullmann does also mention the tendency to interchange careen and career as an example of change based on similarity of name, and that sounds very much like what's going on with infer and imply. Still, it's clearly not just similarity of name that's involved here, because the words in question are (perhaps crucially) already in the same semantic field. (People might talk about a vehicle careering from side to side as it careens down a hill, but I doubt that anyone looking for a job would go to a careen counselor.)

  13. Q. Pheevr said,

    August 9, 2008 @ 2:35 pm

    Sorry about the absence of line breaks in my previous comment; I was using <p> and <li> tags that seemed to be working in the preview, but which have vanished in the posted version.

  14. Tim Silverman said,

    August 9, 2008 @ 2:40 pm

    Hmm, and I stuck in a <div align=centre> for my row of asterisks, which also worked in the preview but disappeared in the posted version. Though perhaps that just means I shouldn't write posts so long that they need to be typographically divided into sections.

  15. Tracy Hall said,

    August 9, 2008 @ 4:56 pm

    Maybe the reason that it is so difficult to find a clear precedent for direct reversal in word meaning is that such a process does not occur–by which I mean to imply, and you may reasonably infer, that writing "nonplussed" when intending to convey a meaning of "unperturbed" is just plain wrong.

  16. Adrian said,

    August 9, 2008 @ 5:07 pm

    What Fortson rather generously terms "reanalysis" (Russell, supra), is most often called "misanalysis" by eggcornologists. The "hark" example reminds me of Browning's misuse of "tw*t". Whether such abusages as "the butt-crack of dawn" (LL2232) fall into the same category, I'm not sure.

  17. Dave Bath said,

    August 9, 2008 @ 8:14 pm

    If there is one word where a change to the exact opposite annoys the hell out of me, especially as both usages are still "current", it is "Epicurean".

    Meaning 1 (and the original)
    * (Follower of | following) the philosophy of Epicurus, who argued we should trust our senses when deciding on truth or falsehood of propositions, and be satisfied with only the basics (eg plain cheese, chateux plonk). Very much against hedonism.

    * Someone who is heavily into sensual gratification, the fanciest food and wine, a hedonist.

    Is there a linguistic term for this, where the two completely opposite usages are linked to exactly the same origin, and no irony/sarcasm is involved?

  18. Jean-Sébastien Girard said,

    August 10, 2008 @ 2:02 am

    There is a term for the class of word "nonplussed" belongs to, although it is interested in the "usage" aspect (that is, the way these words polarize nitpickers). The word is "skunked" (i.e. skunked terms) and in it "nonplussed" is joined by words like "decimate", the aforementioned "infer", "fulsome", "enormity", "disinterested", and several others.

  19. Sili said,

    August 10, 2008 @ 10:43 am

    Tracy Hall,

    But it does occur.

    I think the informal name for the phenomenon in Danish is "pedulord" ('pendulum words'), but that's been objected to since there's usually only one 'swing' of the word.

    Here's a short Q&A with some examples:

    Forfordele: from Old High German "vervorteilen", that is to cheat or put at a disadvantage through unfair distribition. The new meaning is to be put at an advantage through unfair division.

    Bjørnetjeneste: a favour that does more harm than good (after the La Fontaine fable of the bear that tried to get a fly of its sleeping masters face, but killed him in the process. New meaning a very big favour, because the bear is such a big animal.

    Laps: dandy, fop, overdressed man. New meaning: shabbily dressed man, someone wearing mended clothes (influenced by "lap": patch)

    Godt og vel (sense 10.3): a bit more than. New sense: a little less than.

  20. Topher Cooper said,

    August 10, 2008 @ 12:39 pm

    As another example, what about "nimrod" (the Mighty Hunter of the Bible) to mean "idiot" probably from Bugs Bunny's sarcastic description of Elmer Fudd as "What a Nimrod!" in a Warner Brothers cartoon.

  21. John Cowan said,

    August 11, 2008 @ 10:09 am

    Tim: Think is not an example of semantic shift, but of two separate verbs colliding due to sound-change and analogy. In Old English the verb thyncan, past tense thúhte, meant 'seem'; the distinct though related verb thencan, past tense thóhte, meant 'think'. Due to sound-changes in Middle English, both thyncan and thencan became thinke, and although only thóhte would regularly become thoughte, thúhte ended up being dragged along with it.

    Consequently, the two verbs became identical in form, and the close semantic parallelism between him thought 'it seemed to him' and he thought 'he thought' eventually caused the latter, more unusual form to be dropped in favor of the former, with the sole exception of the archaism methinks 'it seems to me'.

  22. Anonymous Cowherd said,

    August 11, 2008 @ 3:55 pm

    Since nobody's mentioned it yet AFAIK: The actual meaning of "nonplussed" isn't really that far away from "unfazed", at least not in my mental dictionary. "Nonplussed" to me means "confused", but in a blank-stare sort of way: bemusement. A "huh?" reaction.

    The politician was nonplussed by his critics' attacks.
    The politician was bemused by his critics' attacks.
    The politician took note of his critics' attacks, but did not know what to make of them.
    The politician took note of his critics' attacks, but did not respond to them.
    The politician was unfazed by his critics' attacks.

    So, using "nonplussed" when you really mean "unfazed" is an error, and stupid; but there are plenty of circumstances in which a writer could use either word, depending on what subtleties he meant to convey.

  23. rkillings said,

    August 13, 2008 @ 12:48 am

    « not at all a "slight shading" of change, but rather a quantum leap »

    Since you bring it up, perhaps you can expound on how it comes to be that "quantum leap", which in physics is the *smallest possible* change of state (how small? Planck's constant is approximately 7 x 10^-34) can be contrasted with "slight shading"? Is there some linguistic reason that we confound discrete vs. continuous with the notion of magnitude?

  24. Faldone said,

    August 13, 2008 @ 8:44 pm

    Is there some linguistic reason that we confound discrete vs. continuous with the notion of magnitude?

    I think you've answered your own question. Whatever it means on the subatomic level, discrete changes go from point A to point B without visiting any of the places in between. Calling this a quantum leap in the macro world seems to me to be a perfectly valid metaphor.

  25. John Cowan said,

    August 14, 2008 @ 12:03 am

    A quantum leap is small quantitatively, but in its metaphorical use the emphasis is not on the size but on the discontinuity: by contrast, a "shading" is a continuous change.

  26. "You keep using that word." « To strive, to seek, to find, … said,

    January 22, 2013 @ 12:15 am

    […] of course, is simply the evolution of language through what linguist Mark Liberman calls "change by mistake", though embarrassingly, I have to admit to misusing both these […]

RSS feed for comments on this post