Resistance to Changes in Grammar Is Futile!

Nicola Davis writes in the Guardian about “Detecting evolutionary forces in language change,” by Mitchell G. Newberry, Christopher A. Ahern, Robin Clark & Joshua B. Plotkin (Nature 551: 223–226):

The authors of the study say that the work adds to our understanding of how language changes over centuries.

“Whether it is by random chance or selection, one of the things that is true about English – and indeed other languages – is that the language changes,” said Joshua Plotkin, co-author of the research from the University of Pennsylvania. “The grammarians might [win the battle] for a decade, but certainly over a century they are going to be on the losing side.”

Writing in the journal Nature, Plotkin and colleagues describe how they tracked different types of grammatical changes across the ages.

Among them, the team looked at changes in American English across more than one hundred thousand texts from 1810 onwards, focusing on the use of “ed” in the past tense of verbs compared with irregular forms – for example, “spilled” versus “spilt”.

The hunt threw up 36 verbs which had at least two different forms of past tense, including quit/quitted and leaped/leapt. However for the majority, including spilled v spilt, the team said that which form was waxing or waning was not clearly down to selection – meaning it is probably down to chance over which word individuals heard and copied.

“Chance can play an important role even in language evolution – as we know it does in biological evolution,” said Plotkin, adding that the impact of random chance on language had not been fully appreciated before. […]

The study also explores the use of negation in sentences, such as “I say not”, across English texts dating from the 12th to the 16th centuries, revealing that the placement of the negative word has changed more than once due to selection, possibly because of a desire for emphasis.

“There a was period of time where double negation … was the way to negate things, just as it is in French today,” said Plotkin.

Dave Wilton at complains:

They have artificially selected verbs that have both a regular and irregular past tense forms in modern English and examined those. In those, they found no overarching selective force over the last eight centuries. They’ve artificially picked out anomalous verbs and found there was no explanation for the anomaly. To really determine what is going on, you have to look at all the verbs (or a representative sample). If they had done that, I think the results would have said that, yes, there is a powerful selective force toward regularization. What they’ve done is cherry pick the odd ones and confirm that they’re odd.

I’ll be interested in the thoughts of my commenters, but of course I agree with the general point summarized by the title I swiped from Davis. Thanks, Trevor and Eric!


  1. What Dave Wilton is talking about is actually a rigorous mathematical result about Markov chains that have a possibility of permanently dying to zero. The fact that zero is an absorbing state exerts a obvious pressure toward small values. However, If you look at instances where you know that, at large times, the value of the variable is actually nonzero, you do not see evidence of the zero-ward pressure.

  2. Words tend to remain irregular when used often for simple reasons of classical conditioning: am, are, is are drummed into our heads from long before we can talk (at least in dialects that use them), so there are two countervailing pressures: analogy pushes (in the main) towards regularity, memory reinforcement pushes towards the status quo. When a regular verb becomes irregular, an explanation is needed.

    The trouble with the given explanation (regular verbs become irregular when similar verbs become more common) is not that it is implausible, it is that the tests employed were, at least as far as the article shows, one-sided. In order to see if the explanation truly explains, we need to know whether there are other regular verbs that are similar to irregular verbs that have increased in frequency, but have resisted any pressure towards irregularity themselves. The article tells us nothing about those.

  3. David Marjanović says

    Wasn’t there a study recently that found rare verbs tend to become more regular while common ones even tend to become more irregular?

    Analogy causing irregularity is by no means unheard of. Winken, wünschen, zünden (“wave, wish, ignite”; the latter with lots of prefixed forms like “set on fire/light up” or “inflame”) are regular in Standard German, but have (following the unrounding of the rounded front vowels) joined the i-a-u pattern of singen and finden in dialects like mine and in Viennese mesolect – the i-u pattern, rather, because the simple past is long forgotten.

    just as it is in French today

    In written French. From the spoken language ne has almost disappeared.

  4. How depressing.

  5. What, that ne has almost disappeared from spoken French?

  6. marie-lucie says

    Actually, double negation is more characteristic of Spanish than of French: No sé nada is literally ‘I don’t know nothing’, while Je ne sais rien was literally I don’t know something until the words used with ne acquired a negative meaning by ‘osmosis’ or ‘contamination’.

  7. But nada is from Latin [res] nata, so the same principle applies.

  8. Aha! So when I say “I don’t know a thing” it’s negative concord.

  9. I know shit about that.

  10. LH: nada is from Latin [res] nata

    I did not know that. But there are other n- initial words which are negative, like ningun ‘nobody, nunca ‘never’.

    What is nata ? I can only think of the feminine form of natus ‘born’.

  11. What is nata ? I can only think of the feminine form of natus ‘born’.

    I presume it is indeed the feminine form of natus ‘born’.

  12. marie-lucie says

    So res nata lit. ‘born thing’ must have been a set phrase, something like ‘anything’.

  13. Analogy causing irregularity is by no means unheard of

    Who’d have thunk it?

  14. David Marjanović says


    Also switches from a more irregular to a less irregular but still irregular class, like brang, brung in various & sundry Englishes.

  15. David M.:

    Winken ‘wave, beckon’ is actually an oddity, swinging between the strong and weak declensions. In OHG it was weak, in MHG strong, and today it is divided by geography, but Duden considers both gewinkt and gewunken to be standard (the “latter is said to be “besonders umgangssprachlich”, though), giving only winkte as the preterite. I wonder if that isn’t hypercorrection, though, for the Low German form is winken, wunk, wunken.

    Something similar has happened to English sleep, weep. In OE they were slǣpan, slēp; wēpan, wēop, both ordinary class 7 verbs (class 7 is pretty irregular already in OE). In ME they went weak (with some OE precedent), but what with vowel shortening before two consonants and the simplificof final clusters, the preterites and participles have come to be pronounced /slɛp, wɛp/ in running speech, as if they had been strong all along. The OED speculates that the weak form slǣpte may be older than the evidence shows, reflecting a cognate of dialectal German schläfen that was lost before OE was written. For wēpan it merely says ‘probably originally weak’ but gives no evidence.


    The joke version is that when the Roman Empire broke up, the phrase non rem natam ‘nothing’, lit. ‘no thing born’ broke up too: the Italians kept niente, the French got rien, and the Spanish received nada!

    In fact the etymology of niente < Old Italian neente is not known: ne gentem ‘no people’, ne(c) entem ‘no thing’, ne inde ‘not there’ have all been proposed. In any case it is cognate with French néant ‘nothingness’, but formerly ‘nothing’ as still in fainéant ‘do-nothing’ (apparently a parallel, not a calque) < faire néant.

  16. David Marjanović says

    giving only winkte as the preterite

    If only because the dialects that would have *wank happen to have lost the preterite wholesale.

    (That said, wank sounds vaguely ridiculous even if you don’t know that much English. Therefore, I’ve long suspected it’s actually the other way around: a paradigm winken – *wank – gewunken could not have arisen, but winken – gewunken is no problem… likewise gewunschen and entzunden.)

  17. marie-lucie says

    JC, Thank you for “nothing”!

    néant ‘nothing’ as still in fainéant ‘do-nothing’ < faire néant

    “Fainéant” ‘lazy’ can be a noun or adjective, as in les rois fainéants, a series of (Merovingian ?) kings known for their weakness and incompetence. The word is likely to derive from fait-néant ‘does-nothing’ rather than including the infinitive form, in keeping with the structure of a large number of compounds such as tournevis ‘turns-screw’ (screwdriver).

  18. Matthew J Roth says

    Yes, “ne” is disappearing from spoken French, and to my mind, it’s depressing. It’s clearly negative, as indeed, my own word choice shows, and it’s what’s inherited from Latin and shared in one way or another across many European languages. We have “not” in English, and while I can’t tell you the exact tree from whence it comes and the difference between it and “ne,” but it’s obviously doing the same thing.

    William Ashby, professor emeritus at UCSB, did two sets of recordings in Tour, one in the 1970s and the other in the 1990s. He studied liaison both times, and as I understand it, there’s also a study on negation from the first corpus.

    It’s the reverse of the historical trend, where in a certain elevated style one could drop “point,” which in the XVIIth century was far more common that “pas.” I believe this was a written phenomenon, though there’s not a hard-and-fast way to know. “Point” is still heard in Switzerland, perhaps because of the more frequent reading of the Decalogue historically. Sure, younger people exposed to French spoken French are likely more probable to say “pas” (and possibly drop “ne”) but you need to prepare yourself for “point” and for /ɑ/, among other things.

    Now, I am used to a more formal register of spoken French, not only as an L2 learner in America but even among native speakers. For example, “je ne sais pas” is still /ʒɛ nɛ se pa/. In middling contexts, it becomes /ʒn se pa/, and when I’m more casual or answering quickly, it is the ubiquitous /ʃe pa/.

    I’m also just happy I found a website that allows me to type IPA characters and copy and paste where I need them. Hooray!

  19. English had ne too, and lost it for good and all around 1600, though it had weakened to “n'” before that. First it was (neglecting spelling and pronunciation) “I ne said”, then “I ne said ne aught”, then “I ne said naught”, then “I ne said not”, then “I said not”, then “I did not say”, then “I didn’t say”. This is a process seen in many languages, known as Jespersen’s cycle: a pre-verbal negation word weakens and is supplemented by a post-verbal word; the pre-verbal word disappears; an auxiliary verb comes to be used, leaving the negation word between auxiliary and verb, efffectively in pre-verbal position again. English is going round the cycle for the second time now.

  20. The problem with much of the discussion around this is that it seems to confuse the whole concept of regularity (conformance to a certain constructional pattern) which is a universal characteristic found in all languages with the specific distinction ‘regular/irregular verbs’ found in the terminology used in pedagogic grammars such as English.

    In fact, there is nothing all that the irregularity of the English irregular verbs is a result of a computational view (non-native) of language. We can just as well look at them as majority and minority constructions in the morphology of the English verb. The -ed construction is the most common which makes it seem regular. But many of the other constructions are also perfectly regular, they just only have very few member verbs (e.g. keep, sweep, sleep, creep are very regular). What we conceive of as irregularity is just the fact, that we cannot determine the membership of a construction from the form of the verb (e.g. steep, bleep, seep). But that’s the case for most gender constructions across languages and we don’t necessarily call them irregular.

    All that paper seems to say is that there is some expected tendency for smaller constructions to disappear but they are just as likely to grow in size. Remember they also examined things that make English more complex over time such as the periphrastic do.

    I think there is interesting new thinking on the morphosyntactic complexity of languages that shows that some languages really are more complex than others (on the most intuitive notion of complexity) and that this complexity is somewhat related to size (smaller languages tend to have more complexity) and history involving large numbers of non-native adult speakers joining the linguistic community. Because we now mostly study large languages that are the result of huge contact trends it appears that there is a tendency towards simplicity but as we look at all the languages out there, this seems very implausible, because complexity is the norm and it is hard to imagine how it would have emerged if there was just one direction of travel somehow baked into the nature of language.

    So what this paper indicated is that there is more going on – which is what we already knew (e.g. from all the work on grammaticalization and creolization). They are simply trying to apply some evolutionary concepts here – mostly drift and random mutation. Whether that is fruitful or not, remains to be seen. But I’m not convinced by this line of critique.

  21. David Marjanović says

    you need to prepare yourself for “point”

    In 2004 I encountered someone my age from somewhere in France (neither Paris nor deep south is all I know) who routinely used point for “not at all”, like je sais point “I have no idea”.

    English is going round the cycle for the second time now.

    As seen in the American phonetic merger of can and can’t in many environments, and in I could care less.

  22. Lars Mathiesen says

    Scandinavian has done the same, of course. Negation is (specifically in Danish for simplicity) marked by an adverb in the post-finite-verb slot for “small particles” — ikke = ‘not’ or aldrig = ‘never’, older ej = ‘not’; the older ne which went in the front has disappeared.

    (The etymology of ikke is not as clear cut as it used to be, it seems. Wiktionary has two incompatible claims — the joy of crowd sourcing — that it is either ON eitt = ‘one’ or ON ei < *aiwaz = 'ever', extended with -gi < gin < *kʷos-ne which was a freestanding emphatic particle in older times. But other sources tell me that modern ej actually goes back to the second construction, while ikke is from the first with progressive devoicing).

    Also (a propos á niente) *ne-wait-ek-hwarjaʀ > Da nogen = ‘some/anybody’.

  23. And Swedish has progressed past icke to inte.

  24. Lars Mathiesen says

    Inte is not directly from ickeeitt/einn gi gives neuter ikke and common ingen allying with nogen (where the -n is an excrescent accusative ending). With the two disconnected, ingen (and nogen) grew new neuter forms inget (noget) — reformed in Danish to intet, and in Swedish via ingte to inte. (As a pronoun, sequentially replaced by newer neuters formed to ingen: intet and inget).

    (Danish ikke could be used pronominally once, Jeg har ikke at give = ‘I have nothing to give’. To a modern ear this is faintly familiar but sounds like there is a suppressed noget in there, Jeg har ikke noget at give. It’s actually equivalent to the more emphatic Jeg har intet at give in current grammar).

  25. Lars Mathiesen says

    Edit time ran out while I was checking this — but Old Swedish and Old Danish actually had ængte, not ængit as the neuter of ængin.

  26. PlasticPaddy says

    In Schwäbisch there is ed/edda meaning “not”. I think I heard a variant “it”. I had thought that the basic word was something like aught in English or the beginning of etwas in German, combined with a separable ne, which was then lost. But I do not know what the etymology is.

  27. Stu Clayton says

    In Schwäbisch there is ed/edda meaning “not”. I think I heard a variant “it”.

    # Schwäbischexperten können die Herkunft einer Person von der Schwäbischen Alb dank ihres Dialekts überraschend zuverlässig genau einem Ort (schwäbisch „Fleggâ“) zuordnen. Dabei spielt u. a. die Aussprache von „nicht“ als „nedd“, „nedda“, „edd“, „edda“, „nitt“, „idd“, „idda“, „itt“ oder „itta“ eine Rolle. #

    Found in Jewiki (!).

  28. Russian has:

    I nowhen nowho no-dick* nowhat ne say. “I never say anything to anyone.

    *Nodick for the sake of completeness. Dick-negatives like in dick there (“did you pass the exam?” “dick there:(“) or just dick “no” do not result from contration, they arise from semantical ambiguity of the pronoun itslef, which can stand for nothing and everything at once, depending on how you look at it.

  29. Jewiki ist auch bei Facebook und freut sich über viele Likes.

  30. David Marjanović says

    That’s simply an older version of the Wikipedia article.

  31. inte


    Alternative forms
    (shortened) tē
    (Witzapan) tēsu
    IPA(key): /inˈteː/

    Ina ka inte kimati
    She/he said she/he doesn’t know
    Inte niweli nitekiti yek kwak nimayana
    I can’t work well when I’m hungry


    no, used to deny or disagree.
    Tikneki chiupi chukulat? — Te, nikwaja wey
    Do you want a piece of chocolate? — No, I already ate a lot


  32. David Eddyshaw says

    Hah! A mere lookalike!

    Swedish is (of course) a Scandi-Congo language and unrelated to Uto-Aztecan. The true cognate is Lingala te “no, not.”

  33. Lars Mathiesen says

    I thought the most interesting bit was Proto-Norse *-gin (*-ɣʷin ?) from *kʷos-ne. Verner strikes again, if that holds water.

Speak Your Mind