Annals of Spectacularly Misleading Media

« previous post | next post »

If you were scanning science-related stories in the mass media over the past 10 days or so, you saw some extraordinary news. A few examples:

"Scientists discover a ‘universal human language’".
"The hidden sound patterns that could overturn years of linguistic theory" ("In a surprising new study, researchers have uncovered powerful associations between sounds and meanings across thousands of unrelated languages").
"Global human language? Scientists find links between sound and meaning" ("A new linguistic study suggests that biology could play a role in the invention of human languages").
"In world's languages, scientists discover shared links between sound and meaning" ("Sifting through two-thirds of the world’s languages, scientists have discovered a strange pattern: Words with the same meanings in different languages often seem to share the same sounds").
"Words with same meanings in different languages often seem to share same sounds" ("After analyzing two-thirds of the languages worldwide, scientists have noticed an odd pattern. They have found that the words with same meaning in different languages often apparently have the same sounds").
"Unrelated Languages Often Use Same Sounds for Common Objects and Ideas, Research Finds".
"Researchers Find the Sounds We Build Words From Have Built-In Meanings".

The trouble is, many of these reports are complete nonsense: no one "discovered a universal human language" or "overturned years of linguistic theory" or showed that "world languages have a common ancestor" or demonstrated that "the sounds we build words from have built-in meanings". And other stories simply trumpet as news something that has been known, argued, or assumed for millennia: "biology could play a role in the invention of human language", "words with the same meaning in different languages often have the same sounds", etc.) There may be a story out there that soberly presents the actual content and significance of the research — but if so, I haven't found it.

How did this happen? It seems to be the same old sad tale. Science writers, in search of sensational headlines and lacking adequate background to read and evaluate actual scientific papers, re-wrote wildly irresponsible press releases.  And as usual, it's not clear how complicit the scientists were, but there's little evidence that they tried very hard to tone down the hoopla.

Here's the start of the press release from Cornell University — Susan Kelley, "A nose by any other name would sound the same, study finds":

In a study that shatters a cornerstone concept in linguistics, an analysis of nearly two-thirds of the world’s languages shows that humans tend to use the same sounds for common objects and ideas, no matter what language they’re speaking.  

Published today in the Proceedings of the National Academy of Sciences, the research demonstrates a robust statistical relationship between certain basic concepts – from body parts to familial relationships and aspects of the natural world – and the sounds humans around the world use to describe them.  

“These sound symbolic patterns show up again and again across the world, independent of the geographical dispersal of humans and independent of language lineage,” said Morten H. Christiansen, professor of psychology and director of Cornell’s Cognitive Neuroscience Lab. “There does seem to be something about the human condition that leads to these patterns. We don’t know what it is, but we know it’s there.”

The referenced paper is Damián Blasi et al., "Sound-meaning association biases evidenced across thousands of languages", PNAS 2016. Its abstract:

It is widely assumed that one of the fundamental properties of spoken language is the arbitrary relation between sound and meaning. Some exceptions in the form of nonarbitrary associations have been documented in linguistics, cognitive science, and anthropology, but these studies only involved small subsets of the 6,000+ languages spoken in the world today. By analyzing word lists covering nearly two-thirds of the world’s languages, we demonstrate that a considerable proportion of 100 basic vocabulary items carry strong associations with specific kinds of human speech sounds, occurring persistently across continents and linguistic lineages (linguistic families or isolates). Prominently among these relations, we find property words (“small” and i, “full” and p or b) and body part terms (“tongue” and l, “nose” and n). The areal and historical distribution of these associations suggests that they often emerge independently rather than being inherited or borrowed. Our results therefore have important implications for the language sciences, given that nonarbitrary associations have been proposed to play a critical role in the emergence of cross-modal mappings, the acquisition of language, and the evolution of our species’ unique communication system.

The paper is a solid and careful empirical contribution to the exploration of some interesting issues — which have been discussed, debated, and explored experimentally for decades, centuries and even millennia.  For example, in Steven Pinker's 1994 work The Language Instinct, we read:

[M]ice are teeny and squeak, but elephants are humongous and roar. Audio speakers have small tweeters for the high sounds and large woofers for the low ones. English speakers correctly guess that in Chinese ch'ing means light and ch'ung means heavy. (In controlled studies with large numbers of foreign words, the hit rate is statistically above chance, though just barely.)

Or in The Cambridge Encyclopedia of Language by David Crystal (1997), there's 2-page-long entry on Sound Symbolism, which concludes

The examples of sound-symbolism are fascinating, but in the absence of frequency information about the phonological and lexical patterns in the various languages, it is not possible to arrive at a definitive interpretation. […] in the absence of historical data, drawing conclusions from sound symbolism about the origins of language would be premature (§49). Far more descriptive data are needed, accompanied by experimental investigation of the speakers' intuitions about the relationship between sounds and meanings.

I would characterize the Blasi et al. paper as a worthwhile contribution to the first of Crystal's requirements, namely "more descriptive data".

And discussion of these issues didn't all start in the 1990s. Friedrich Max Müller's 1861 Lectures on the Science of Language, Delivered at the Royal Institution of Great Britaini, famously discusses a related set of issues (starting on p. 371)

What we want to find out is this, What inward mental phase is it that corresponds to these roots, as the germs of human speech?

Two theories have been started to solve this problem, which, for shortness' sake, I shall call the Bow-wow theory and the Pooh-pooh theory. According to the first, roots are imitations of sounds; according to the second, they are involuntary interjections. The first theory was very popular among the philosophers of the eighteenth century, and, as it is still held by many distinguished scholars and philosophers, we must examine it more carefully. It is supposed, then, that man, being as yet mute, heard the voices of birds and dogs and cows, the thunder of the clouds, the roaring of the sea, the rustling of the forest, the murmurs of the brook, and the whisper of the breeze. He tried to imitate these sounds, and finding his mimicking cries useful as signs of the objects from which they proceeded, he followed up the idea and elaborated language. This view was most ably defended by Herder. […]

For instance, ' Man sees a lamb. He does not see it like the ravenous wolf. He is not disturbed by any uncontrollable instinct. He wants to know it, but he is neither drawn towards it nor repelled from it by his senses. The lamb stands before him, as represented by his senses, white, soft, woolly. The conscious and reflecting soul of man looks for a distinguishing mark; — the lamb bleats! — the mark is found. The bleating, which made the strongest impression, which stood apart from all other impressions of sight or touch, remains in the soul. The lamb returns — white, soft, woolly. The soul sees, touches, reflects, looks for a mark. The lamb bleats, and now the soul has recognised it. " Ah, thou art the bleating animal," the soul says — within herself ; and the sound of bleating, perceived as the distinguishing mark of the lamb, becomes the name of the lamb. It was the comprehended mark, the word. And what is the whole of our language but a collection of such words? '

There's more — Müller names and discusses the "Ding-dong" theory, the "Yo-he-ho" theory, and the "Ta-ta" theory, in each case referencing scholars of previous decades and centuries. All of these are ideas about "Sound-meaning association biases" that would naturally lead to the kinds of patterns across language that Blasi et al. explore.  (And Jespersen later added the "La-la" theory, as a name for Darwin's idea about the origins of language in song.)

Nor did these speculations begin with the 18th-century philosophers that Müller cites. Noah Jacobs, Naming Day in Eden: The Creation and Recreation of Language, 1958, points to relevant discussions going back to Democritus and Heraclitus:

Adam did not permit himself to be diverted by deceptive sounds or arbitrary combinations at the expense of good sense as did poets of a later age. He did not exploit "apt alliteration's artful aid" to suggest the clatter of hoofbeats (Virgil), hissing serpents (Racine) or the progress of rats (Browning's "The Pied Piper"). Sound alone, which beguiles the ear without engaging the mind, is an unreliable vehicle for thought. The commonplace must not be made alluring by tawdry adornment; the mule's head need not be hung with tinkling bells.  […] [Adam's] aim was not to copy reality in every detail in the manner of the "material imitation" of the pre-Platonic philosophers who regarded language as a passive stamp of reality, every sound having an innate quality which makes it suitable to represent certain ideas; he followed Plato's theory of "ideal imitation" where a word expresses the inner essence of a thing, the meaning behind the ever changing object. This view is rejected by those who hold to the "convention theory" according to which words acquire meaning not from their sounds but by common agreement. A word merely provides an appropriate mode of conveying our thoughts, and its meaning is defined by custom and mutual agreement. The sound of a word tells us as little about its meaning as a key about the contents of a room or a seashell about the marine life that haunts the deep. Adam could have given different names to the same creature (Eng. ewe, Arab. najat) or the same name (Eng. ewe, Hung. juh); or two different animals could have received the same name (Eng. dog, Heb. dag, fish). This is the ever recurrent argument advanced against the theory of "ideal imitation" since the days of the Sophists when Democritus refuted Heraclitus: there is no natural connection between a word and the thing it designates. But Adam's names could not have been submitted to anyone for agreement by the very nature of the case. They rang true of their own accord because his speech was a reflection of his reason, his oratio flowed from his ratio. That is, his basic agreement did not come from the arbitrary corroboration of mortal men but from God who inspired him with the breath of life and who vouched for his speech.

When I first took a linguistics course in 1965, the topic of "sound symbolism" was on the syllabus — as was a discussion of its limitations. Some of these limitations will become intuitively clear to you if you try to converse with monolingual speakers of a language you don't know, relying only on your shared inventory of "nonarbitrary associations" between speech sounds and meanings. The problem is that while such associations exist, they don't get you very far in creating a realistically shared vocabulary. In quasi-scientific jargon, the "percent of variance accounted for" is small; alternatively, we could say the average reduction in referential uncertainty produced by knowledge of the sounds of a word,, though non-null, is extremely small.

You should read the whole paper, but here's the result of a tiny analogous experiment. It deals with letters, not (remapped) phonemes, and it's entirely English-internal, but it may give you a point of reference for the effect sizes that Blasi et al. find. I took the inventory of synonyms for each of 9 probe words, and compared the frequency of all 26 letters in each set with the overall frequency of each letter in a large American English word list. There were many statistically-significant patterns, some of which made sound-symbolism sense — for example the frequency of 'l':

# Syn  # l # letters  % l %l in dict. Syn%/Dict%
sharp 43 9 317 2.84% 6.08% 0.47
short 43 12 386 3.11% 6.08% 0.51
shock 38 9 283 3.18% 6.08% 0.52
long 38 15 292 5.14% 6.08% 0.85
smart 40 14 252 5.56% 6.08% 0.91
flat 29 18 222 8.11% 6.08% 1.33
dull 37 23 280 8.21% 6.08% 1.35
stupid 50 35 382 9.16% 6.08% 1.51
round 31 21 229 9.17% 6.08% 1.51

Thus in 43 (thesaurus-grade) synonyms for "sharp", there are 9 l's in 317 total letters, for a frequency of 2.84% — less than half the overall frequency in the dictionary. But in 50 synonyms for "stupid", there are 35 l's in 382 total lettes, for a frequency of 9.16 — more than 1.5 times the overall l frequency in the dictionary. These ratios are quite similar to the ratios found in the Blasi et al. paper — for instance, the association they found between nose and the alveolar nasal was a ratio of 1.47 between the frequency of /n/ in the words corresponding to that concept and its frequency in other words. (And FWIW, I chose the letter 'l' and the cited set of nine words before looking at the numbers…)

Another, more technical, sort of limitation of sound-symbolism phenomena is that regular sound changes in the history of languages are rarely affected by such considerations. Thus the Great Vowel Shift moved the whole English vowel space around in a major way, severing many sound-symbolic connections while perhaps creating others; Grimm's Law had a similar effect on Proto-Germanic. There's little evidence that sound symbolism played a significant role in advancing or retarding these changes, creating exceptions, etc. (See Stan Carey's "Language change and the arbitrariness of the sign", 10/28/2013 , for a bit more discussion.)

Recognition of all these limitations has led linguists to focus on the relative arbitrariness of each language's mapping between words and sounds — but every linguist has always also recognized that patterns of sound symbolism exist, within and across languages.

So whoever made up that Cornell press release's lede — "a study that shatters a cornerstone concept in linguistics" — was seriously ignorant, deluded, or dishonest.  I certainly hope that the idea didn't come from Morten H. Cristensen, "professor of psychology and director of Cornell’s Cognitive Neuroscience Lab" — even a neuroscientist should know better.

I've reported many times on equally egregious misrepresentations of scientific results — see "Debasing the coinage of rational inquiry: a case study" (4/22/2009) for one striking example among many. But I have a more personal interest in this case — there was a section in my 1975 dissertation on phonetic symbolism, ideophones, and iconic meaning, engaging the question of how much of prosodic meaning is conventionalized, and to what extent.  (See"Ask Language Log: Sounds and meanings", 3/9/2008, for a long quoted passage.) As Dwight Bolinger  put it in 1978, “Intonation is a half-tamed savage. To understand the tamed or linguistically harnessed half of him one has to make friends with the wild half.”

But in contrast, I'm skeptical of the view that similar considerations apply in a meaningful way to the lexical and referential aspects of phonology. Certainly Blasi et al. don't make any such argument.

I can't resist closing with a pointer to the best popular-culture treatment of phonetic symbolism:

And finally — though a bit of an anticlimax — some LLOG posts on related topics:

"Ask Language Log: Sounds and meanings", 3/9/2008
"Waza waza", 4/20/2008
"Japanese (and Chinese) Onomatopoeia", 7/21/2008
"Skeevy", 6/22/2009
"Unce", 5/22/2010
"Phonetic marketing", 10/23/2010
"What English sounds like if you have Wernicke's aphasia", 10/22/2011
"Phonosymbolism and Phonosemantics in Chinese", 1/13/2012
"Huh?", 11/9/2013
"Labiality and femininity", 12/26/2014
"What does 'Schmetterling' sound like to a German?", 1/30/2016

Note: Victor Mair flagged the existence of this paper in "Common language", 9/14/2016, but he didn't note the orgy of spectacular over-interpretation that followed in the popular press.


  1. J.W. Brewer said,

    September 23, 2016 @ 3:16 pm

    To be mildly charitable to the over-interpreters, it does seem as if someone with just enough superficial knowledge of the history of linguistics to be dangerous might have picked up along the way a simplified narrative like "in olden times, people often believed words had inherent meanings based on their sound, because they hadn't yet separated mysticism from science (digression on Kabbala goes here …), but then Saussure came along with the revolutionary insight that the relationship between signifier and signified was arbitrary." The problem is in taking the Saussurean claim to be that the arbitrariness is universal and exceptionless rather than simply ubiquitous, so that any evidence of non-arbitrariness is taken as a dramatic/surprising breakthrough.

  2. D.O. said,

    September 23, 2016 @ 3:57 pm

    Prof. Liberman, your toy example doesn't prove much. Under the "null-hypotheses" of random distribution of letters within and between words (this would be a pretty extreme way to remove correlations) number of l-letters in a typical set of synonyms that you collected is about 20, which should create about 20-25% standard deviation just because of random fluctuations. Your set (last column) shows about 47% variation measured by standard deviation. I hope Blasi et al. has more firm statistical foundation.

    Let's make a quick and dirty calculation. Suppose a typical basic word has about 4 morphemes out of inventory of 50. It makes the probability of a typical word to include any given morpheme about 8%. Let's settle on 10%. Out of 4000 analyzed languages it should get expected 400 examples of a given word with a given morpheme. But only Indo-European languages number more than 400. If we think that basic vocabulary is pretty stable within language families, this fact alone will produce enormous correlations. Statistical difficulties to sift through all of it should be pretty steep.

    [(myl) (I think your use of "morpheme" may be a misunderstanding of something? and FWIW, by an exact binomial test the the 95% confidence intervals for the 'l' counts at both ends of my toy word list are inconsistent with the hypothesis that the underlying probabilities were the same as in the dictionary at large…) But the point of my toy example is not to prove anything. I just wanted to illustrate the scale of sound-meaning correspondences that Blasi et al. find — their featured examples (like nose) generally involve finding that a certain sound-class is 1.5 to 2.5 times more (or less) common in a certain semantic word-class than it is overall. On a quick read, it looks to me like their calculations of statistical significance are sound — though I'd like to hear what you think after a more rigorous inspection, since obviously there's a danger of "data dredging" in a case of this kind. But "statistical significance" is not necessarily "practical significance", which I guess here would be "communicative significance".

    As for the toy example, you could make it look better by combining "short sharp shock" with "dull stupid round", which was the pairing I first thought of, and which would multiple the observed N and the observed hit counts approximately by three. (I added "long" and "smart" and "flat" for associative reasons). But as I said, the only point here is to show that the "meaning association" of /n/ with nose in their data is about the same as the association of 'l' with "round"…]

  3. Rubrick said,

    September 23, 2016 @ 4:20 pm

    It's sadly very easy to impress people with findings that "overturn the accepted paradigm" in a field like linguistics, simply because vanishingly few people have the foggiest idea what the currently accepted paradigm is.

    Incidentally, thanks for the Python link! I've watched that sketch untold times, but couldn't resist watching it again when it was so convenient, and this is the first time I caught the "PVC sort of word" throwaway.

  4. Guy said,

    September 23, 2016 @ 4:21 pm

    The "universal human language" claim seems to be based not only on a misunderstanding of the research but also a sort of reasoning similar to what Pullum has called the "big bag of words" understanding of language.

  5. JS said,

    September 23, 2016 @ 4:49 pm

    As I noted in the other post, the "Significance" blurb that opens the paper says that "a careful statistical examination of words from nearly two-thirds of the world’s languages reveals that unrelated languages very often use (or avoid) the same sounds for specific referents. For instance, words for tongue tend to have l or u, “round” often appears with r, and “small” with i. These striking similarities call for a reexamination of the fundamental assumption of the arbitrariness of the sign." So blame for the nature of the popular media reaction lies pretty much entirely with the authors in this case. The last sentence in particular… shameful.

  6. Jichang Lulu said,

    September 23, 2016 @ 5:00 pm

    "English speakers correctly guess that in Chinese ch'ing means light and ch'ung means heavy."

    To nitpick: in the romanisation Pinker seems to be using, 'heavy' is chung (plain voiceless affricate, no apostrophe).

  7. D.O. said,

    September 23, 2016 @ 5:07 pm

    I am not qualified to evaluate this work myself and anyway it would take too much time to look in all the details. But 2 things made me uneasy

    If signals are inherited from an ancestral language spoken in remote prehistory, we might expect them to be distributed similarly to inherited, cognate words; that is, their distribution should to a large extent be congruent with the nodes defining their linguistic phylogeny(see Fig. 3 for illustration). A direct evaluation of this hypothesis is infeasible due to the absence of etymological dictionaries for all but a few families. However, it can be tested indirectly given that cognate words are expected to be more similar to one another than noncognates (45). We investigated whether the presence of the signal-bearing symbol was a better indicator of overall form similarity between words than other shared symbols, using a β mixed-regression model that distinguishes the effects of symbols, concept, and lineage (SI Materials and Methods). The model is heavily dominated by the effect of lineage, and signal presence (although significant) has a negligible effect in the opposite direction than predicted: the genealogically balanced average effect is less than a 0.5% decrease in similarity for those words sharing a signal-related symbol compared with those sharing some other symbol.

    If you have etymological data for a few families, why not to test it on these families? And second, if you find that your signal for basic vocabulary has no relationship to the known relationship for languages themselves, it means that something is awfully wrong. I was afraid that cognates would swallow any effect, but that they have no effect at all is a complete mystery.

  8. Phillip Minden said,

    September 23, 2016 @ 5:29 pm

    Sʲinʲij – gəɫuboj.

  9. Pflaumbaum said,

    September 23, 2016 @ 6:11 pm

    A beautifully clear explanation and a Python sketch I'd never seen before. Thank you.

  10. Bathrobe said,

    September 23, 2016 @ 6:14 pm

    This study has had one positive benefit for me. I now know how to distinguish the Mongolian words for 'autumn' and 'nose', which are khamar and namar. The best way to remember is that the one with the n isn't nose!

  11. FM said,

    September 23, 2016 @ 8:08 pm

    Unfortunately, from my interactions with neuro types, they frequently do believe in weird dogmatist strawman linguists. Someone should do an ethnography, really.

  12. Chris Button said,

    September 23, 2016 @ 10:35 pm

    Another, more technical, sort of limitation of sound-symbolism phenomena is that regular sound changes in the history of languages are rarely affected by such considerations. Thus the Great Vowel Shift moved the whole English vowel space around in a major way, severing many sound-symbolic connections while perhaps creating others; Grimm's Law had a similar effect on Proto-Germanic. There's little evidence that sound symbolism played a significant role in advancing or retarding these changes, creating exceptions, etc.

    Just taking possible shifts of a hypothetical proto-phoneme /r/, we could suggest the following two developments in opposite directions:

    /r/ > /l/ > /j/

    /r/ > /ʁ/ > /g/ > /ŋ/

    Linking /j/ and /ŋ/ to a common ancestor in /r/ (albeit in separate languages) is but one example of the massive variation that any work of a similar scope to the one being discussed here will encounter.

    When one encounters any deviation from established sound laws in comparative historical work, the first approach is to look for external influence. Onomatopoeia is usually only used as an explanation in the most obvious cases, but I wonder if such an approach is perhaps a little too cautious? Could many of the "exceptional" words that appear to have no discernible external influence, or conversely have so many apparent cognates that they appear to be "areal" in origin, actually be the result of an onomatopoeic resistance to regular change?

  13. maidhc said,

    September 24, 2016 @ 2:56 am

    Rubrick: I heard it as "BBC sort of word".

    Since the sketch was ostensibly set in 1942, I don't know that PVC had been developed yet. Although the only reason to have it in 1942 is to justify bringing the character from the Biggles sketch to reprise his "Cabbage crates over the briny" line.

    That's a funny sketch too. "I'm sorry sir, I don't follow your banter."

  14. Graeme said,

    September 24, 2016 @ 4:27 am

    Old hat. Bob Dylan proved the onomatopoeic and rhyming basis of language c 1979

    (I know it's only a thought experiment, but the idea of prelingual man gazing reflectively on a 'white soft wooly lamb' engaged in some Edenic gambol seems a tad anachronistic. And a claim for photo–vegetarianism).

  15. Stephen said,

    September 24, 2016 @ 5:19 am

    According to

    PVC was first produced in 1872, attempts to use it commercially in the early 20th century were unsuccessful because of processing issues and brittleness. In 1926 a method of adding plasticisers to was devised that resolved those issues.

    However in 1942 I would expect that rather than PVC someone would have said plastic. In the case of the people in the sketch I would think a long 'a' (a bit like plarstic) would be used.

  16. Damian Blasi said,

    September 24, 2016 @ 6:07 am

    Thanks Mark for the very balanced evaluation of our paper's coverage.

    First of all, I'd like to address the significance statement that we wrote that ends with: "These striking similarities call for a reexamination of the fundamental assumption of the arbitrariness of the sign."

    I do think this is true and that this is a consequence of the results of our paper. "Reexamination" doesn't imply saying it is false or useless: to me it implies that we can't just disregard these non-arbitrary associations as completely marginal if they appear in the proportion they do under the conservative assumptions we assumed. Is that really a shameful sentence, as JS suggests?

    As for the press release: I had to evaluate the press release in English of one of the institutions I'm affiliated with and my job was basically to prune down overblown statements like "the textbooks of linguistics need to be rewritten". My personal experience in the last two years is that the interests of the scientists, the institutions and the media do not necessarily align always. Probably the way to go with this would be to make scientists to produce informal accounts of their findings and their significance in "neutral" zones – like blogs or personal websites. That will rise the question whether scientists themselves are well-equipped to communicate their research to other fellow researchers and the taxpayers, though.

  17. Damian Blasi said,

    September 24, 2016 @ 6:10 am

    DO's comments:

    If you have etymological data for a few families, why not to test it on these families?

    Showing that the effect holds or doesn't within 1/5/10 families given the ~300 language families + isolates out there doesn't seem to be extremely relevant to our point, unless one wants to make strong assumptions about expected effect size.

    And second, if you find that your signal for basic vocabulary has no relationship to the known relationship for languages themselves, it means that something is awfully wrong. I was afraid that cognates would swallow any effect, but that they have no effect at all is a complete mystery.

    We don't imply that within families these things do not correspond to cognates – we say that, sharing a signal-related symbol is not a better diagnose of cognacy than two words of the same family and for the same concept sharing at least one symbol (but not the signal-related one).

  18. D.O. said,

    September 24, 2016 @ 9:26 am

    Dr. Blasi, now I understand more of what you are saying, but then it does not affect the basic concern that whatever correlations you see among ~4000 languages are, in fact, correlations within language families + noise. The word for nose has n in a lot of Indo-European languages because they are cognate. The thing I would have wished you've checked is that increased correlations are not driven by cognancy, not that the sound-meaning correspondence does not help to stabilize word sounds under language changes.

  19. JS said,

    September 24, 2016 @ 9:34 am

    "Shameful" is a strong word; something like "come on, guys" would have been a more face-to-face type choice. The problem is not at all that the authors are wrong in saying that "we can't just disregard these non-arbitrary associations as completely marginal" — that is trivially true, and supported by a century of research. The problem is that the implication of the significance statement is that these particular results have led directly to that insight. To paraphrase the blurb: "our study reveals [X] // for instance, [Y] // [X],[Y] overturn one of the most fundamental assumptions of modern linguistics."

  20. Damian Blasi said,

    September 24, 2016 @ 10:00 am

    DO: we explicitly control for language family in our testing scheme, so cognacy is not a concern.

    JS: The problem is not at all that the authors are wrong in saying that "we can't just disregard these non-arbitrary associations as completely marginal" — that is trivially true, and supported by a century of research.

    Can you please cite any published paper that, after controlling for all the confounds we have identified is able to show a sound-meaning association in a large number of unrelated languages? There is a large literature on sound-symbolism, iconicity, synaesthesia and other sources of non-arbitrary sound-meaning associations – a lit that, as you mention, has at least a hundred years old. But going from what we are able to detect in experimental conditions (or in a handful of languages) to a persistent association found in a large number of languages is not straightforward – and much less trivial, IMO. It would be possible to think in a scenario where humans are sensible to specific sound-meaning associations without that having any impact of the structure of the vocabulary, perhaps due to the multiple benefits of arbitrariness many authors have shown. In this respect, our work does provide that otherwise missing piece of evidence.

  21. JS said,

    September 24, 2016 @ 11:58 am

    @Damien Blasi, Sincere thanks for your responses. I would describe the language of your post above as positively disposed towards your project, as is only proper, but nonetheless fair-minded in its assessment of your specific contribution, which incidentally I do think is original and significant. But I find the path from the opening blurb to the popular media uptake very short and straight in this case. I am curious whether you find so, or no. For instance: I think it would be better to try to be clear regarding in what sense the sound symbolic association between /i/ and "small" that you note is a new discovery and in what sense it is going on 90 years old (the work of Sapir 1929, etc.) — Jonathan Smith

  22. JS said,

    September 24, 2016 @ 12:11 pm

    sorry, should be "Damian"

  23. Damian Blasi said,

    September 25, 2016 @ 4:46 am

    Thanks for the comments, JS.

    But I find the path from the opening blurb to the popular media uptake very short and straight in this case. I am curious whether you find so, or no. For instance: I think it would be better to try to be clear regarding in what sense the sound symbolic association between /i/ and "small" that you note is a new discovery and in what sense it is going on 90 years old (the work of Sapir 1929, etc.)

    If with association between /i/ and "small" you mean "a sensibility by humans to associate / relate /i/ with the concept of "small" in experimental conditions or in a anecdotal handful of languages" then we didn't discover that, and we never say so in the paper (btw: in all of the cases we could find someone making this sort of claim about the signals we found, regardless of the evidence they used, we cited them). What we did was to show that such association actually emerges as well when a large number of languages is considered, under rather strict testing conditions. This is new, and as I mentioned in the last post, the original experimental work doesn't entail the cross-linguistic pattern.

  24. David Marjanović said,

    September 25, 2016 @ 5:02 am

    "White, soft, woolly" sounds better in the original German (to coin a phrase), where it alliterates: weiß, weich, wollig.

  25. Chinook Man PhD said,

    September 25, 2016 @ 11:51 am

    A truly adequate linguistic theory, whensoever we linguists develop one, is going to account for and anticipate folk (mis-)understandings about the nature of language. O:-)

    Because folk-linguistics is a(n anthropological) linguistic phenomenon, it's not to be grimaced away.

RSS feed for comments on this post