Is “Huh?” a Universal Word?—And Why Is It Important?
According to the authors of the PLOS article, the starting point for their research on the universality of “huh?” was an observation made when something else was studied: “that almost all spoken languages provide two basic ways of signaling communicative trouble: an interjection like ‘huh?’ and a question word like ‘what?’.” However, while the words for ‘what’ vary widely across languages (e.g. que in Spanish, čto in Russian, etc.), the interjections “looked suspiciously similar across languages”. To investigate this seeming consistency more fully, the authors examined 196 tokens of ‘huh?’ (in relevant contexts) in ten languages: Cha’palaa, Dutch, Icelandic, Italian, Lao, Mandarin Chinese, Murriny Patha, Russian, Siwu, and Spanish. Moreover, the research also found examples of this interjection in 21 additional languages, including English. All 31 languages are shown on the map reproduced on the left.
Dingemanse and his colleagues make three key claims about “huh?”: (1) that it is a “universal word”, (2) that it is not innate, (3) that it is likely shaped by convergent evolution. Let’s consider each claim in turn. First, what makes “huh?” unique is that it seems to crop up in language after language, with the same meaning (or function), which the researchers call “other-initiated repair” (OIR): when speaker A said something that speaker B could not hear or did not understand, speaker B (“other”) initiates a repair by signaling that there is a problem in communication but without specifying what that problem is. The actual repair then follows in the form of simple repetition, sometimes with slight modification. A model of such a conversation is given below:
A: It’s not too bad.
A: ‘S not too bad.
The claim about the universality of “huh?” comes from the fact that similar sounding interjections are used for “other-initiated repair” in a variety of languages. Just how representative the list of languages with “huh?” is remains to be seen. Although the authors’ claims apply only to the specific set of languages studied, media reports greatly overexaggerated the findings. The host of the AirTalk show said that this word is found “in virtually every human language”, “in any number of languages”, and “in almost any language”. The headline of The New York Times article by Jennifer Schuessler states that it is “the syllable everyone recognizers”; the article itself states that this word is “universally understood, across all countries and cultures”. The author of the NPR article, Alva Noë, says that this word is “native to all languages”; the headline goes so far as to ask whether this one word could “unite the world”. The headline in The Atlantic states that “Huh Means the Same Thing in Every Language”, and the article itself, written by Olga Khazan, calls this expression “practically universal”. Yet, relatively solid data are available only for ten languages, which constitute about 0.1% of the world’s currently spoken tongues. The sample, moreover, is hardly representative, despite the authors’ claims, as it includes two closely related Germanic languages (Icelandic and Dutch), two closely related Romance languages (Italian and Spanish), and half of the languages considered belong to just one language family: Indo-European (and more specifically, to its “European” branches). The Americas, Africa, Australia, East Asya, and Southeast Asia are represented by one language each. No language from South Asia, Central Asia, the Middle East, or North America is included in the small sample.
The solidity the data from the ten languages studied can also be questioned. The authors claim to have at least ten (and on average twenty) videotaped recordings of conversations that illustrate the use of the interjection in each of the ten languages. The video posted on their website, however, contains only the recordings of the interjection itself, as it is pronounced in each of the ten languages, not the full repair-initiation dialogs (allegedly for privacy reasons). The claim that Russian uses “a?”, a supposed variant of “huh?”, for OIR, however, does not ring true to me, as a native speaker and a trained linguist. Not only have I worked on Russian syntax for over 15 years, but several of my research projects have examined conversational, colloquial constructions in Russian, including the so-called subextraction (e.g. the Russian analog of ‘Expensive he bought car’, which is ungrammatical in English but extremely common in colloquial Russian), on-the-spot lexical borrowing (e.g. perepostanut’ ‘to repost [on Facebook]’) and nominative topic constructions (e.g. kniga… ja pročital ‘Book I read’). In order to examine these phenomena, I spent months crawling through transcripts and video recordings of naturally occurring conversations, looking at structures other than those described in prescriptive grammars of the language. Yet, I found no instances of OIR “a?” in any of those records—and it seems very foreign to me. To test whether I have simply been “deaf” to this expression, I conducted a little—and admittedly not very scientific—experiment: over the past week, I have been mumbling to my unsuspecting family and friends, eliciting the OIR response. All of them came back with Čto? ‘what?’, not A?. But the latter is more commonly used in Russian in other discourse functions, as mentioned by Barbara Partee in one of her comments in the LanguageLog. For example, A? can be used in Russian to elicit a response from the second speaker (in parallel to the French n’est-ce pas? or the dialectal English init? and eh? tags) or to elicit an elaboration from the first speaker:
Eliciting a response from speaker B:
A: Ty ne znaeš kuda ja ix položila, a? (‘Do you know where I put them, eh?’)
B: V škaf, navernoe. (‘Into the closet, probably.’)
Eliciting an elaboration from speaker A:
A: Ja položila ix v škaf. (‘I put them in the closet’)
B: A? (‘And so?’)
A: No tam ix net. (‘But they are not there.’)
While this common use of the interjection does not prove conclusively that A? is not used in the OIR function in Russian, it certainly casts doubt on the commonality of its use and therefore on one of the researchers’ claims: that this sound template is so advantageous in fulfilling an essential conversational function that it has been arrived at by various languages independently, as the best possible solution to a common problem of miscommunication.
The Russian example also leads one to question how similar is the pronunciation of the “huh?” interjection in various languages. Dingemanse and his colleagues crucially stress that the interjection is not pronounced exactly the same in different languages. Quite the opposite: its phonological shape is in line with the phonetic inventory and prosodic (i.e. intonational) patterns of each given language. What the various versions of “huh?” in different languages have in common is a “template”: one syllable, containing at the most an unrounded lax vowel from the low-front quadrant of the vowel space (that is, some version of [a], [E], or [Ã]), and maximally a glottal stop [/] or a glottal fricative [h] as the onset. Moreover, the vowel is typically somewhat nasalized, and the intonation contour corresponds to the interrogative contour used for questions in the language. In most languages, it is a rising intonation, but in Icelandic and Cha’palaa the falling intonation is used for both questions and OIR interjection.* The authors explain this common template by recourse to the concept of “minimal effort”: this pattern is the easiest to pronounce as it involves all articulators—the tongue, the lips, the velum—being in their resting position. What is crucial is what the articulators do not do: the tongue does not raise to the high position and does not retract to the back position, the lips do not round, and the velum does not raise to produce a purely oral (i.e. non-nasal) sound. The principle of minimal effort, however, does not seem to apply to the vocal cords: their resting position is open, which is what we do when we simply breathe out without saying anything. This position produces [h] but not the glottal stop, for which the vocal cords must be closed shut—a arrangement opposite to the “minimal effort” position. In addition, the “minimal effort” theory does not explain why there should be an onset in some languages, as all languages allow onset-less syllables. In other words, why is “huh?” rather than “uh?” used for the OIR in English?
The similar yet not precisely the same pronunciation of “huh” across languages is important for two reasons. First, it serves to show that “huh” is a word rather than a biological sound, such as those made by sneezing, laughing, crying, or grunting, all of which sound exactly the same across languages and cultures. According to Dingemanse et al.’s definition, being “integrated into each linguistic system” is what makes “huh” a word. This does depart from conventional definitions of “word”, particularly from the classical definition of Ferdinand de Saussure that takes a word to be an arbitrary relationship between meaning and sound. In the case of “huh?”, the connection between its meaning and its sound is not arbitrary. Instead, the authors claim that diverse languages zeroed in on this particular sound template because it best fulfills the conversational function as “a simple, minimal, quick-to-produce” way to indicate a possible miscommunication about to happen.
Let’s now consider the second key claim of the PLOS article: that “huh” is not innate. Dingemanse et al.’s argument for non-innateness of “huh” comes from two sources. First, as mentioned above, the exact pronunciation of the interjection differs from language to language and therefore must be learned, and second, the authors claim on their website that indeed “huh” is learned: “babies don’t use it, infants don’t use it perfectly, but children from about 5 have mastered it perfectly, along with the main structures of their grammar”. It is not clear, however, if the researchers have actually examined the acquisition of “huh” or what evidence they may have to support the above claim. Moreover, the age of 5 seems to be quite early for the acquisition of discourse markers, as most previous research points out that such markers are acquired rather late. For instance, Champaud and Bassano (1994) claimed that French children acquire a mastery of discourse markers like mais (‘but’) and pourtant (‘nevertheless’) between the ages of 8 and 10. In the same vein, Scott (1984) claims that conjunct use typically emerges between the ages of 6 and 12. If children acquire the proper use of “huh?” later than the age of five, it does not contradict the authors’ main contention that it needs to be learned. Still, one wants to see some evidence for empirical claims such as that “huh” is indeed learned.
The third key claim, that “huh” is “likely shaped by convergent evolution”, in parallel to biological cases such as sharks and dolphins developing the same body shape despite coming from very different lineages, is not the only plausible explanation for the similarities found. The authors do not seem to consider an alternative theory: that the similarity (and in some cases, identity) of “huh” across languages is due to language contact. For example, is it really a coincidence that Spanish and Tzotzil (one of the 21 additional languages considered in the study) share an identical form of “huh”, /e/? After all, Tzotzil is spoken in Mexico and had been under Spanish influence for centuries. Such similarities could also stem from language-external factors, which have been called upon to explain the otherwise surprising similarity of sound-meaning pairs across languages in two additional cases: the “mama” words and onomatopoeic words. As it turns out, many of the world’s languages have words for ‘mother’ (or ‘father’, ‘grandma’, ‘grandpa’, and the like) that sound like “mamma” or “babba”. The similarity is not accidental and can be explained by the fact that such sound sequences—which can hardly be called words—are the first to be pronounced by babies when they begin to vocalize, or babble. Pronouncing sequences of open syllables with a voiced labial consonant ([m] or [b]) followed by a low mid vowel [a] is a way babies master control over their vocal apparatus; they do not mean anything. Yet parents the world over think that the child is naming the key caretaker, typically the mother. Similarly, the source of similarity among onomatopoeic words, such as the ones representing animal sounds across languages, is obvious, as they are trying to imitate non-linguistic sounds that are the same across languages or cultures. There is only so much variation in how members of different linguistic groups “hear” such animal sounds, and certain “templates” can be distinguished for various onomatopoeic words (for example, the words for cat mewing start with [m] and contain a lowering diphthong [ja], etc.).
This brings me to my biggest reservation about the PLOS article: so what? While the claims, to the extent that they are substantiated, are “mildly interesting”, as Tim Stowell put it, one may question whether all the globe-trotting and data analysis were justified by the scope of the findings: yet another sound template that is used for a certain meaning/function across languages because of language-external factors. The authors of the paper do not draw any major conclusions from their findings, although they do call for “closer attention to the infrastructure for social interaction that underlies language in use, and its possible influence on language structure”. Some journalists reporting on the article, most notably NPR’s Alva Noë, likewise refrained from passing judgment or making far-fetched implications. Yet others, particularly, Jennifer Schuessler from The New York Times, could not abstain from pushing a larger anti-Chomskian agenda. Citing an interview with Nick Enfield, one of the article’s authors, Schuessler claims that “the study… is part of a broader effort to challenge the dominant view that language is primarily a matter of inborn grammatical structure, as Noam Chomsky has argued”. However, it is not clear how the similarity of sound of the OIR interjection across ten languages challenges the view that language is primarily a cognitive computational capacity or that elements of grammatical structure—crucially not words!—are innate. Much more than one piece of evidence that social interaction can shape vocabulary needs to be proven in order to challenge the Chomskian paradigm.
*The description of the intonational patterns used in Icelandic interrogatives is actually more complex than that (for a detailed discussion see Árnason 2011: 317).
Árnason, Kristján (2011) The Phonology of Icelandic and Faroese. Oxford University Press.
Champaud, Christian, and Dominique Bassano (1994) French concessive connectives and argumentation: in experimental study in eight- to ten-year-old children. Journal of Child Language 21: 415-438.
Scott, Cheryl M. (1982) Adverbial connectivity in conversations of children 6 to 12. Journal of Child Language 11: 423-452.
« Questions for Readers Regarding Biblical...
Martin Lewis’ Lecture on Indo-European... »