David Robson (linked before at LH, most recently here) writes for Aeon about the linguistic phenomenon known as ideophones; he starts with a quiz about Japanese (e.g., does nurunuru mean ‘dry’ or ‘slimy’?) and continues:

One of the founding axioms of linguistic theory, articulated by Ferdinand de Saussure in the early 19th century, is that any particular linguistic sign – a sound, a mark on the page, a gesture – is arbitrary, and dictated solely by social convention. Save those rare exceptions such as onomatopoeias, where a word mimics a noise – eg, ‘cuckoo’, ‘achoo’ or ‘cock-a-doodle-doo’ – there should be no inherent link between the way a word sounds and the concept it represents; unless we have been socialised to think so, nurunuru shouldn’t feel more ‘slimy’ any more than it feels ‘dry’.

Yet many world languages contain a separate set of words that defies this principle. Known as ideophones, they are considered to be especially vivid and evocative of sensual experiences. Crucially, you do not need to know the language to grasp a hint of their meaning. Studies show that participants lacking any prior knowledge of Japanese, for example, often guess the meanings of the above words better than chance alone would allow. For many people, nurunuru really does feel ‘slimy’; wakuwaku evokes excitement, and kurukuru conjures visions of circular rather than vertical motion. That should simply not be possible, if the sound-meaning relationship was indeed arbitrary. (The experiment is best performed using real audio clips of native speakers.)

How and why do ideophones do this? Despite their prevalence in many languages, ideophones were once considered linguistic oddities of marginal interest. As a consequence, linguists, psychologists and neuroscientists have only recently started to unlock their secrets.

Being a journalist, Robson can’t resist the occasional overstatement (“Their results pose a profound challenge to the foundations of Saussurean linguistics”), but that’s a venial sin, and he provides lots of good examples. He also quotes Mark Dingemanse at Radboud University in the Netherlands, who turned up at LH back in 2008 when he was still a PhD student — I’m glad to see his blog, The Ideophone (“Sounding out ideas on language, vivid sensory words, and iconicity”), is still around. There’s some interesting discussion of Japanese ideophones in this LH thread from a few years ago; e.g., minus273 said “Words like pikapika, cognate to non-ideophonic hikar– seem to imply that ideophones escaped the p > f > h sound changes. (Thanks, Kobi!)


  1. I read that article a week or more ago. I’m afraid that the challenge to Saussurean linguistics bit put me off it from the start. I seem to remember seeing similar statements about Saussure’s doctrine of the arbitrariness of the sign at other places on the Internet. It looks like it’s another bit of disinformation doing the rounds.

    Incidentally, has anyone used Zotero (mentioned at Mark Dingemanse’s website)?

  2. Amazing. My comment got dropped because I linked to Zotero.

    Well, I wasn’t too impressed by that David Robson article precisely because he started with that bullshit about Saussure. It’s not the first time that I’ve seen the claim so it’s possibly another bit of disinformation spawned (or spread) by the Internet. And it will be believed by lots of people so that others have to continue argue against it. What a waste of time and effort. To coin a Trumpianism, it’s like ‘fake scholarship’. Someone spruiks a few lies and it takes all the king’s horses and all the king’s men to put it back together again. Possibly for decades.

    Back to Zotero: it’s mentioned at Mark Dingemanse’s blog. Has anyone used it?

  3. David Eddyshaw says

    In Africa, at any rate (where a large number of ideophones in a language is the norm and not the exception) the notable thing about ideophones is not so much their iconicity. Most ideophones are frankly pretty arbitrary. I mean, why (in Kusaal, natch) is “very (red)” wimm while “very (white)” fass? (They just are, that’s why.)

    And, given that iconicity is hardly unheard of in languages that don’t do ideophones, it isn’t surprising that it would turn up particularly often in ideophones. It turns up cross-linguistically even in words for “big” and “little”, where the “little” words much more often than not have higher vowels than the “big” words.

    The really striking thing about ideophones in very many African languages is their often weird phonology. The final double ss of fass, for example, is impossible in ordinary vocabulary. African ideophones quite often even have individual consonants not found elsewhere in the language.

  4. Well, I wasn’t too impressed by that David Robson article precisely because he started with that bullshit about Saussure.

    I’m confused. Are you saying that Saussure didn’t write about l’arbitraire du signe? Because he certainly did (Wikipedia). And in any case it seems odd to reject an entire discussion of ideophones because the reporter doesn’t get his history of linguistics exactly right.

  5. David Eddyshaw says

    Thinking about this in the bath (such dedication!) it seems to me that Robson has basically got this exactly backward. The odd thing about ideophones from the standpoint of a speaker of a language that doesn’t have them is precisely that they aren’t arbitrary. If you’re a foreign learner you need to commit them to memory just like other vocabulary. Crucially, they aren’t the same thing as onomatopoeia – if they were, they wouldn’t be ideophones. The business about iconicity is a red herring.

    (Come to that, onomatopoeia itself is also fairly arbitrary. Japanese dogs don’t go “woof.”)

  6. The same bullshit you pointed out, how ideophones overturn a basic tenet of linguistics.

    Of course it’s not good to reject an entire discussion because the reporter got some basic facts wrong in his introduction but there you are. That sort of thing doesn’t dispose you to read the rest of the article in a charitable frame of mind.

    This is what I wrote elsewhere for a different audience:

    “One of the founding axioms of linguistic theory, articulated by Ferdinand de Saussure in the early 19th century, is that any particular linguistic sign – a sound, a mark on the page, a gesture – is arbitrary, and dictated solely by social convention.”
    This now popular meme misrepresents what linguistics is about. It is, in fact, a straw-man argument. Linguists have never denied that “onomatopoeic words” exist, nor have they denied that there is often a link between the sound of words and their meanings, e.g., why ‘teeny’ (with the vowel ‘ee’) is used to emphasise smallness, voiced consonants are sometimes related to larger size, or “sl” is used in words denoting things that are smooth and wet (slimy, slush, slick, slobber, slide, or even sly).
    The Saussurian notion is on a more abstract level: it is designed to combat the popular idea that a word like, say, “black” exists because it just sounds so, well, “black”, or that a wall is a “wall” because it sounds so “wall-like”. Even the examples given tell us that there is nothing pre-ordained about the assignment of meanings to words in language. Sure, “nurunuru” means “slimy” in Japanese, but in German “nur” means “only”, in Mongolian “nuur” means lake, in French “nourriture” means “food”. Of course it is often possible to find a link between word and sound, but you would be daft to found a science of language based on the idea that the meaning of words is linked with the way they sound. The first step in linguistics is to describe how language is structured in its own terms, which means disabusing yourself of the notion that the way a word sounds is automatically associated with a meaning. Only once this principle has been accepted can you then go on to study possible connections between sound and meaning.

    Agree with David Eddyshaw that “onomatopoeia itself is also fairly arbitrary”.

  7. David Eddyshaw says

    (Where I said “precisely that they aren’t arbitrary” there, I meant “precisely that they are arbitrary.” That’s what comes of thinking in the bath. I won’t be doing that again in a hurry.)

  8. I was so confused!

  9. John Woldemar Cowan says

    The Saussurian notion is on a more abstract level: it is designed to combat the popular idea that a word like, say, “black” exists because it just sounds so, well, “black”, or that a wall is a “wall” because it sounds so “wall-like”.

    Or, on an even less sensible level, the notion that Adam called the tiger ‘tiger’ (or טיגרִיס‎ tígris) because it looked like a tiger.

    That’s what comes of thinking in the bath.

    Funny. I worked out a rather complex program (if I do say so) in the course of many showers on non-consecutive days; the shower is a basically undistracting environment with little to do but think, since not even I will bring a computer into the shower.

    In Africa, at any rate

    I assume that in Africa the ideophones are areal, whereas in the Dravidian languages they seem to be inherited, with cognates in different languages (though I can give no examples offhand). What is also notable is that Tamil is both fully diglossic (H about 7 centuries older than L) and heavily ideophonic: crucially, the ideophones cross the H/L barrier freely, as they do not in many other languages where they are confined to spoken forms only.

  10. It was interesting to me to learn that kuru-kuru means to twirl, and a little googling turned up that kuru-kuru-pa means “crazy.”

    When I was in Tanzania some years ago, there was a hit pop song called Mapenzi Kizunguzungu. In Swahili, I learned, kizungu means to turn, to spin, and Kizunguzungu means dizziness. So Mapenzi Kizunguzungu means “Dizzy Love.”

    (Yes, I know that Mzungu means “white person”: or “foreigner,” from the 19th c habit of white “explorers” to wander about, seemingly aimlessly – that’s unrelated to the meaning I’m talking about).

    Kuru-kuru and kizunguzungu, both having meanings relating to spinning or twirling and both being used in phrases describing disturbed mental states, seem to be some sort of evidence of the independent working of similar ideophones.

    Mapenzi a great song, by the way.

  11. David Eddyshaw says

    I assume that in Africa the ideophones are areal

    Unfortunately I don’t know much about the ideophones of languages closely related to Kusaal; dictionaries tend to neglect them as being not quite the thing linguistically. Poor relations to “proper” parts of speech. However, their frequent phonological bizarreness suggests that they are not inherited in the way most vocabulary is.

    One case which looks suggestive is the Kusaal intensifying ideophone for “white”, fass or farr, which looks very like it’s borrowed from Hausa. It has nothing in common with the Kusaal adjective for “white”, pielig, but resembles the corresponding Hausa ideophone fat, which does have some resemblance to the Hausa fari “white.”

    So yes, probably, on the basis of the very limited information I have.


    is an interesting paper, basically suggesting that the adoption of characteristically Khoisan phonemes like clicks into Xhosa etc was largely driven by the borrowing of ideophones.

  12. David Eddyshaw says

    In general, I think your implication (if I understand you right) is correct: ideophones aren’t all the same thing across different languages. Really rather different phenomena have got lumped together because from the standpoint of Standard Average European they all look a bit similar.

  13. The pika pika example that Hat referred to is a good example. AFAIK, apart from borrowings, initial /p/ in Japanese occurs only in onomatopoeic words. Pika pika might reflect the history of the word (failure of /p/ to turn into /h/), but there are also other onomatopoeic words that begin with /p/ that don’t seem to be related to “lexical words”.


    Chinese (Mandarin) 轱辘 gūlu ‘wheel’, ‘roll’ (Cantonese gu1 luk1, gu1 luk6) is enticingly similar, but Bauer proposes a relationship with Tibetan འཁོར་ལོ (khor lo ‘wheel’) and Proto-Indo-European *kʷékʷlos ‘wheel’. Maybe they are all sound symbolic?

  14. David Marjanović says

    marie-lucie has told us several times that kwel-/qwel- is all over the Pacific Northwest in similar meanings. In short, yes.

  15. Finnish has a pretty well-documented and researched onomatopoetic–ideophonic lexical hinterland. It does not come with much weird phonology, but does involve weird morphology: insofar as there are any lexical “morphemes”, they’re merely one phoneme long and will need to be compiled until meeting the usual two-mora minimum word requirement before additional morphology can kick in. Thus e.g. ulista, urista, ölistä, öristä, mulista, murista, mölistä, möristä (all referring to some sort of deep and unclear noise: ‘howl’, ‘growl’, ‘grumble’, ‘mutter’…) could probably be thought to be made up of morphemes m-, -u-, -ö-, -l-, -r- and the ideophonic-verb marker -is-. (Alternatively one could pick an entry at random and derive others from it thru “vowel mutations” and “consonant mutations” that do not really exist elsewhere in the language.) Regardless the full formations can have additional semantic load: öristä is more likely to be a drunkard than a bear, murista the opposite, and yet this will be of no help for figuring out the nuances of their “l-grade” equivalents.

    For a more obviously lexical example, consider kökkö, köllö, pökkö, pöllö, all referring to objects laying or standing about, and with various accompanying derivatives such as kököttää ‘to sit hunched-over motionless’; pöllö has additionally become specialized as meaning ‘owl’ in particular (and by now is even the standard Finnish term for all strigiforms).

    Research in how and when these could be connected to ideophones in other Uralic languages is taking its first firmer steps as well. One of my colleagues should be wrapping up soon his PhD on the topic.

  16. Sure, “nurunuru” means “slimy” in Japanese, in Mongolian

    “nurshuu” means “sticky, gooey”

    but maybe that’s just common Altaic, not sound symbolic.

  17. Thanks for the shoutout LanguageHat!

    Interesting to see the speculative discussions here. Interested readers might like to know that since 2008 there has been a lot of empirical work charting all these issues and more (check the link in my sig for papers). I will mention just three results that are most relevant to the discussion. In the 2016 Language paper we showed (using 200 ideophones from 5 languages) that people can guess ideophone meanings at levels above chance, though with plenty of room for error; we conclude that ideophones are at most weakly iconic and certainly have arbitrary aspects. In the 2016 Journal of Experimental Psychology paper, we showed that Japanese ideophones are easier to learn than adjectives (though both obviously are learnable): iconicity just gives a little nudge towards higher memorability. Finally, in the 2018 Glossa paper I review the history of ideophone research and address issues of language ideology and crosslinguistic diversity that come up in several comments above. All should be easy to find online and will provide pointers to other empirical work on the typology and iconicity of ideophones.

    Dingemanse, Mark, Will Schuerman, Eva Reinisch, Sylvia Tufvesson & Holger Mitterer. 2016. What sound symbolism can and cannot do: testing the iconicity of ideophones from five languages. Language 92(2). e117–e133. doi:10.1353/lan.2016.0034.

    Lockwood, Gwilym, Mark Dingemanse & Peter Hagoort. 2016. Sound-symbolism boosts novel word learning. Journal of Experimental Psychology: Learning, Memory, and Cognition 42(8). 1274–1281. doi:10.1037/xlm0000235.

    Dingemanse, Mark. 2018. Redrawing the margins of language: Lessons from research on ideophones. Glossa: a journal of general linguistics 3(1). 1–30. doi:10.5334/gjgl.444.

    For the record, Robson’s Aeon piece is obviously aimed at a popular audience and makes some choices that not everybody would make. I don’t think ideophones overthrow De Saussure, though they do call for more nuance; I don’t think it is helpful to see ideophones as “fossils of protolanguage”; and I would have stressed the communicative utility of ideophones in everyday interaction in languages as varied as Japanese, Siwu, Basque and Quechua. However, it is a good overview that introduces a broad audience to this fascinating topic so overall I’m happy with the good press for the topic.

  18. For what it’s worth, I guessed “dry”. Maybe it’s because I’m only reading rather than hearing the words, but I can’t say any of them evoke any particular sensations in my mind.

  19. Thanks for dropping by and sharing the references, Mark!

  20. I think WP’s examples of English ideophones are a mish-mosh of ordinary words with a sound-symbolic flavor, onomatopoiea, and actual ideophones. Here’s the examples with my classification of them: boing (I), boom (I), bang (I), bling-bling (O), criss-cross (SS), dilly-dally (SS), murmur (O), pitter-patter (I), sing-song (SS), swish (I), splish-splash (I), ta-daa! (I), thud (I), tick-tock (I), tinkle (SS), twinkle (SS), zigzag (SS). I made my judgements on ideophones on whether they fit into the frame “<NP> went ‘<I>’.” I introduce the inner quotes because He went zigzag is obviously cromulent, but I judge that zigzag is an ordinary adverb here. I also tried imagining them as comic-book sound effects.

    If anyone disagrees in detail, please post!

  21. Is “dilly-dally” sound-symbolic? Similarly with “criss-cross” and “sing-song”. They seem to me only partially sound-symbolic, namely the reduplication part, since “dally”, “cross”, and “song” are words with no sound-symbolic content. “Splish-splash” and “pitter-patter” I’ll allow.

  22. I agree that your examples are only partially sound symbolic, and likewise with shilly-shally (where the last part is < Shall I?) and many more.

  23. And your general point is irrefutable: Wikipedia has no idea what ideophones are. Wikipedia is picky about getting things right when it comes to pop-culture detritus and techie stuff, but anything beyond that is a mess.

  24. Okay, I’ve pruned the list.

  25. While we’re on the subject of (mild) synaesthesia, does anyone know if high notes are called high because they’re written higher on the music staff, or were they put high on the staff because the metaphor of high notes was already current?

  26. Hard to say. The earliest neume notation dates to the 9C, but the first recorded use of high in the musical sense in English only to the 14C (Gower’s Confessio Amantis). But that’s not saying what the case may be in other languages. L & S lists altisonus as meaning only ‘sounding from on high’, no reference to pitch.

  27. The way pitch production works in the human voice (the musical ur-instrument), there are several things a singer may do to make it easier to hit notes near the upper and lower ends of their range. Low notes can be produced more easily with the head tilted downward; it also helps to sing them quietly. In contrast, to hit the highest notes, it helps to throw back your head and sing loudly. Given these facts, the vertical metaphor associated with pitch seems very natural, and it could be almost as old as music itself.

  28. > pikapika, cognate to non-ideophonic hikar–

    *lightbulb moment* I feel stupid for never having noticed this before.

    > relationship with […] Proto-Indo-European *kʷékʷlos ‘wheel’.

    I’d heard of the possible relationship between /kuruma/ (車, car) and ‘wheel’, but again, it had never occurred to me till now that /kuruma/, /kurukuru/ and /kuruu/ (狂う, go insane/go wrong) are related. Thank you so much, kanji-equals-etymology-fallacy!!! (Which I don’t believe in, but which still clouds my view.)

  29. kurukuru

    Which brings up its voiced counterparts /guruguru/ and /gururito~gurutto/, and just a short way off /gorogoro/, /gororito~gorotto/, and /korogaru/.

  30. David Eddyshaw says

    Just noticed that the Numbami language that Joel Bradshaw has just been talking about on the thread about emerging sign languages does interesting things with ideophones:

  31. After reading this, I tried to explain it to some people over dinner, and thought that perhaps the Monty Python sketch with the ‘woody’ and ‘tinny’ words might help.

    However, I can’t find a single working example of the sketch on the internet.

    Can someone point me to a clip, or perhaps someone has it downloaded?

  32. Huh, I had linked to one here, but I see it’s been removed. Sad! However, I confess I don’t see how it would help with ideophones.

  33. John Cowan says

    Contra Saussure, this fine poem “El Golem” by Borges:

    Si (como afirma el Griego en el Cratilo)
    el nombre es arquetipo de la cosa
    en las letras de ‘rosa’ está la rosa
    y todo el Nilo en la palabra ‘Nilo’.

    This, however, seems to me a very Borgesian reading of the Cratylus, which as I understand it is not about explaining words by things, but by explaining words by other words; indeed, it consists mostly of etymologies of proper names (usually correct), more proper names (usually bogus), and ordinary words (entirely bogus).

    I do particularly like the rhyme Golem/Scholem, however.

    Update: Another fine verse, describing the experience of the golem after awakening as he is enmeshed in language for the first time:

    Gradualmente se vio (como nosotros)
    aprisionado en esta red sonora
    de Antes, Después, Ayer, Mientras, Ahora,
    Derecha, Izquierda, Yo, Tú, Aquellos, Otros.

  34. Seems to me I’ve read that the Cratylus was written to take essentially Saussure’s position, but the tradition tended to take Cratylus’ counter-position.

Speak Your Mind