Frequently Asked Questions

In this FAQ we cover some of the more outrageous interpretations of our results, point out the problem with doubts based on introspection, and reiterate that there are formats for initiating repair beyond ‘huh?’. Further, we stress that we talk about convergent cultural evolution and deal with various objections referring to sample size, innateness, inheritance and contact. All of these points are also made in the paper in one form or another — as indicated by the links to relevant sections [RS] — but if you knew that, you might not be here!

Dingemanse, Mark, Francisco Torreira, and N.J. Enfield. 2013. “Is ‘Huh?’ a universal word? Conversational infrastructure and the convergent evolution of linguistic items.” PLOS ONE. doi:10.1371/journal.pone.0078273.

Huh? #

We see what you did there!

Is ‘Huh?‘ the only universal word? #

No. Our findings suggest that ‘Huh?’ is likely to be attested in all spoken languages, but we don’t claim exclusivity for this word. In fact, we point out that it is unlikely to be the only such word, because the proposed mechanism —convergent cultural evolution— is a general one that may well apply to other aspects of linguistic structure. [RS]

Seen this way, the relevance of our study goes far beyond the humble ‘huh?’ — as we argue, we need a closer examination of language in actual use to understand how it  may be shaped and constrained by the fact that it is used first and foremost in social interaction. [RS]

Is ‘Huh?‘ exactly the same everywhere? #

No. In fact a crucial finding of our study is that it is subtly different from language to language: it is calibrated to the local language system. We argue that this is one of the things that show it is a word as opposed to some kind of innate prelinguistic grunt. For instance, in Icelandic, ha? has falling intonation, something that makes it sound like a statement (like ‘huh.’) to speakers of English — but this fits the system of Icelandic, where many questions also end in falling intonation. In Spanish, it is e?, with a vowel that is distinctly different from the English a? . This shows it’s calibrated to the vowel system of Spanish, where /e/ is the most common vowel[RS]

While we find the interjection is not exactly the same, in the 10 languages we study in detail, we do find that the bandwidth of the variation is exceedingly narrow, an observation that is confirmed in 21 additional languages. If ‘huh?’ were just a word like any other (e.g. ‘dog’ or ‘what?’), one would expect that in unrelated languages it would show the kind of variation we see for such words: different consonants, different vowels, different word shapes, et cetera. In our paper we show that it doesn’t: in the languages we study, it uses only glottal consonants out of a whole array of manners and places of articulation, only non-high non-back non-rounded vowels out of the whole vowel space, and only a monosyllabic shape out of all the options that many languages offer. That is the sameness we attempt to explain in the paper. [RS]

Was ‘Huh?‘ the first word of human language? #

We have no reason to believe this and we make no such claim in our study. In fact, we argue that to have such a word, specialized for clarifying matters of understanding, only makes sense when a flexible, fully functioning cooperative system of communication like human language is already in place. [RS]

We don’t say this in my language! #

Based on our research, we can say that if your native tongue is a spoken language, it’s likely to have a short and sweet huh?-like questioning syllable with the function of ‘initiating repair’, i.e. asking for repetition or clarification. It will probably have a vowel somewhere between (phonetic) /e/ and /a/, it will likely have a articulatorily minimal consonantal onset if any, and it will have an intonation that is understood as questioning. It might also sometimes be pronounced with the mouth (half) closed, e.g. m? or n? — see below.

The way to discover these things, though, is usually not by introspection: people are notoriously bad at reflecting on their own language use. If you really want to know whether something like this does or doesn’t occur in your language, there’s a straightforward way to find out: collect a corpus of everyday, informal, face to face interaction (5 hours should do) and systematically and exhaustively sample that for any and all occurrences of other-initiated repair, as we and our colleagues did for all 10 languages in our sample. If you don’t find it in there, let’s talk. [RS]

We don’t say huh?, we say { what?shto? / wie bitte? / pardon? } or maybe m? #

In our study (and in other published work coming out of this project) we make clear that every language has multiple ways  of signalling trouble in conversation. These include question words like what? in English, shto? in Russian or va in Swedish; or polite forms like sorryin English or wie bitte? in German. In all of the languages for which we have representative corpus data, though, we did find at least this huh-like word. Furthermore, corpus data shows that in most languages (though not all) the interjection is more frequent than the question-word format, and that polite expressions such as ‘sorry?’ and ‘wie bitte?’ are very rare indeed in informal interaction. For us in this particular study, that is reason enough to pay close attention to the form and function of this widespread and commonly used interjection, the intrinsic interest of the other options notwithstanding. [RS]

We also note in the paper that in every language we found a closed-mouth form like m? or n? that fits our explanation just as well. This wasn’t the most common form in any of the languages we studied, and so we analyse it as an underarticulated version — basically a version that takes even less effort than [a?] already does. One thing we noticed (and also report in the paper) is that forms like m? and n? tend to be used when speakers are in close proximity to each other. This makes sense, because it’s not as acoustically salient as a? and its kin. Meanwhile, if you really feel that something like a? or m? is marginal in your language, there’s a straightforward way to find out, as we note above. [RS]

For the bigger picture, see this publication: Dingemanse, Mark, Joe Blythe, and Tyko Dirksmeyer. 2014. “Formats for Other-Initiation of Repair across Languages: An Exercise in Pragmatic Typology.” Studies in Language 38 (1): 5–43. doi:10.1075/sl.38.1.01din (pdf).

It doesn’t sound like h-u-h in my language! #

Nor does it in English. Don’t be fooled by the vagaries of spelling. Our claim is not about the three letters ‘huh?’, nor is it about the precise phonetic shape [huh], nor is it about the English word ‘huh?’. What we found is universal is the function of this word along with a set of constraints determining its form — constraints that are due to the fact that it is found in a conversational environment common to all spoken languages.

So: every language has a short word approximately like this that performs the same function of ‘initiating repair’ — i.e. indicating that there is some problem with the prior talk. For instance, in Spanish it is e?, in Lao it is a?, in Siwu it is ã? (with a clearly nasalised vowel), in Icelandic it is ha? (with a clear /h/ sound at the start). What’s common to all languages is that it is a short questioning syllable with a vowel in the low-front-central region and a minimal consonant if any. [RS]

So you checked all 5000+ languages spoken today? #

We didn’t, of course, and that’s one of the reasons the title of our study asks a question (‘Is “Huh?” a universal word?’). We noted this item was strongly similar in everyday language use in the 10 languages we studied. These 10 are from 6 unrelated language families from all over the globe, so already that was an unexpected finding — all the more because the question word alternative “what?” did show strong differences across these language families. [RS]

Still, could it be that we happened to be working on just the 10 languages that coincidentally have the same kind of form for ‘huh?’? To check that it wasn’t just a sampling fluke, we added as many additional languages to our sample as we could find data for. We extended our sample to 31 languages from 16 unrelated language families from all over the world and still found the same thing. Words tend to be different across unrelated languages, and so when you find a similar form in so many unrelated languages, and you don’t not find it anywhere you look (get that double negative?), you know you’ve come across something in need of explanation. Measuring the extent of the similarity, and explaining it, is what our study focuses on.

Your sample seems small — is 31 languages really enough? #

More is always better. However, rigorous standards prevent us from just adding languages to the sample based on introspection or indirect sources like dictionaries or scripted conversations. We want to compare like with like. Before we consider an interjection comparable, we need to see evidence of it being used to initiate repair in actual records of conversation, available for repeated inspection. That is how we painstakingly got to 31 languages from 16 distinct language families spoken around the world. If you have this kind of data for other languages, we would very much welcome it. [RS]

Importantly, however, in those 31 languages, we never found evidence for a different form or for the absence of this kind of interjection. That is, in the 31 languages we had reliable conversational data for, we found this kind of form and not something else. Finding basically the same template —a monosyllable with questioning prosody and all articulators in near-neutral position used for initiating repair— in 31 languages from 16 distinct phyla, and not not finding it anywhere one looks, is a striking fact about language that is strongly in need of an explanation — which we provide in the paper.

Why didn’t you include language X? It has it too.#

You’re probably right. We limited ourselves to 31 languages for which there is reliable, published data from actual recordings of informal conversation. Given the proposed mechanism of convergent cultural evolution, we expect that an interjection like this will occur in virtually all spoken languages. But actually showing this is harder than you might think (see under sample size). Introspection is unreliable in matters of language use, so you’d need to show it occurring in a representative corpus of everyday conversation before we will able to accept it.

You listened to 196 recordings, what does that mean? #

By ‘recordings’ in the context of this study we mean single instances of the word ‘huh?’, taken from much larger video corpora of conversation. For each of the 10 languages in our study, we and our colleagues collected at least 5 hours of audio and video recordings of everyday social interaction. Each of us systematically sampled these corpora to find sequences of other-initiated repair, a subset of which featured the interjection. We made sound clips of 196 of these OIR sequences (about 20 per language), and then extracted the OIR interjection from them for the careful phonetic comparison we report on in our study. [RS]

To be perfectly clear: we did not travel around the world to record the word ‘huh?’ — the data for this study come from a much larger enterprise in which we carry out long-term fieldwork to gain a deeper understanding of the languages and peoples we study. The results of our fieldwork feed into fundamental linguistic research. The larger goal of our project is to contribute to a better understanding of what language is like, how much languages really have in common, and why only our species has it. Learn more about our work.

Of course it’s similar across languages, it’s just a { grunt / instinctive noise } ! #

Um… try doing a real grunt next time when you don’t understand something. You’ll probably get a different kind of response. What we show in the paper is that you have to pronounce this word according to the rules of your own language. We also find that people choose to say it — it’s not an involuntary reflex but a word chosen at a specific juncture in conversation. We also note that children are not born saying it, but have to learn it (they have perfect mastery over it only when they’re five, according to one study of American English). [RS]

Based on our findings, ‘Huh?’ seems very different from involuntary noises like a sneeze or cry of pain. Indeed, to have such a word, specialized for clarifying matters of understanding, only makes sense when a fully functioning cooperative system of communication (i.e., human language) is already in place. Does that mean we’ve demonstrated it’s not innate? No, merely that the evidence militates against it. As we note in the paper, “If there is a mechanism that can explain cross-linguistic similarity on a more proximate timescale, without resorting to genetic encoding, this is preferred on scientific principles of parsimony.” In the paper we put forward such a mechanism: convergent cultural evolution. [RS]

You use the term evolution but you also say this word is unlikely to be innate. What gives? #

Scientists distinguish between biological evolution (involving DNA and genes) and cultural evolution (involving cultural items like words or customs). Both are subject to evolutionary processes in Darwin’s sense of descent with modification. Language is interesting because it builds on our biology (our bodies and brains are essential for speaking), but is at the same time a culturally evolving system (our words and constructions are culturally transmitted and evolve over time). When we talk about the evolution of a word like huh? we mean cultural evolution. The idea is that huh? is part of the culturally evolving system we call language. [RS]

The term we use to explain the similarity of ‘huh?’ across unrelated languages is convergent cultural evolution. We observe that this word is used in a particular place in conversation where it needs to do a number of things at once: 1) signal a problem with what another speaker just said, 2) show that the problem has to do with a lack of knowledge, and 3) ask for a response. In that place, or ‘conversational environment’, there is little time for planning an elaborate response, precisely because you may not have caught anything in the first place. What you need there — indeed, what all speakers in all languages need there — is a short and simple questioning syllable that is easy to plan and produce. As Martha Gill summarises in the New Statesman, “You might say that environmental forces squeezed this word into shape.” [RS]

Couldn’t the similarity be due to inheritance? #

Languages evolve over time and they form evolutionary lineages such as “Germanic” (including Dutch and Icelandic), “Tai-Kadai” (Lao), or “Kwa” (Siwu). Within a lineage, languages inherit many properties of their common ancestor, which leads to them being alike in certain respects. Could the similarity that we describe just be due to inheritance?

Unlikely. Our sample is diverse: even the smaller subset of 10 includes languages from at least six distinct language families. While inheritance may explain why, say, the OIR interjections of Dutch and Icelandic are similar, it doesn’t explain why these two would be so similar to Cha’palaa (a Barbacoan language from Ecuador), to Murriny-Patha (a Southern Daly language from Australia), to Lao (a Vietic language from Laos), or to Mandarin Chinese (a Sinitic language from mainland China). To further control for possible inheritance effects we added data for another 21 languages, and found not a single language in which the interjection was strikingly different. Inheritance simply doesn’t cut it to explain the striking global similarity of this item. [RS]

In relation to this, some readers might ask: if language relatedness may explain some of the similarities, why bother to include multiple languages from one lineage at all? The reason is that we want to answer not just the question of possible universality, but also in the question of whether it is a word. From the diversity within a family —for instance the intonational difference between Icelandic and Dutch, or the consonant and vowel differences between those two and Spanish— we see that the interjection is a word (i.e. shows integration and conventionalisation), answering the second of our two research questions. [RS]

Couldn’t the similarity be due to language contact? #

Languages may take words and expressions from other languages with which they are in contact. Sometimes this results in cross-linguistic similarities even in unrelated (or distantly related) languages spoken in the same area. Could language contact be a feasible explanation for what we find?

This is highly unlikely. Contact is good for explaining areal similarities, not global ones. If contact were all there is to it, one might expect languages in one area to have one kind of form (say, ‘a?’), languages in another area to share another one (‘vuvuli?’), and language in a third area to have a third form (‘bibo?‘). What one wouldn’t expect would be to find the same form in unrelated languages that are unlikely to have been in contact. And yet that is exactly what we find. Our sample of 10 languages contains languages from 6 distinct phyla spoken in very disparate locations (Ecuador, Ghana, Western Europe, South-East Asia, Australia). The larger sample of 31 languages is even more diverse in both geographic and phylogenetic terms (16 phyla). To reiterate, even though contact may explain certain areal similarities, we still need a mechanism for explaining the global pattern. And that is what we focus on in the paper. [RS]

I read that okay and Coca Cola are pretty widespread too. Is this a similar case? #

It’s actually very different. There are places in the world where the influence of Anglo culture is small, and where Coke is hard to find. People in such places are unlikely to have picked up ‘Coke’ and ‘okay’. They will however have a long history of talking with each other, and therefore they probably have their own version of ‘huh?’. Simply put, people around the world don’t need Coke the same way they need techniques to clear up misunderstandings in communication.

If you think huh? spread like okay or Coke, think again. Are we to picture people waiting for a word like this to fill a gap like Coke fills the sweet commodity gap around the world? Unlikely — they already had a need for a word like this before before the English (or French, Spanish, etc.), came along, so there would be no reason for them to take it from another language. Not to mention that our evidence suggests that in most cases they didn’t, because every language seems to have its own version of huh? — see above.

(We didn’t see this one coming so we don’t talk about Coca Cola in the paper. A missed chance!)

Doesn’t huh? have other uses too? #

Most certainly. As we note in the paper: “In some languages the interjection, or an item similar to it, was also found in other sequential environments, for instance to mark surprise or to pursue a response. Such alternative (and probably derived) uses provide insight in possible paths of semantic change, but we exclude them here to make sure we are comparing like with like.” [RS]

In other words, of course words like huh? may serve other purposes as well; and some of these other purposes may even be interestingly similar across languages (such as the use of huh? as a kind of tag-question, likely derived from the interactive function of managing attention and pursuing a response). However, we don’t investigate those other functions of huh? here, because examining one of them in detail is more than enough for the kind of careful comparative study we pursue.

Need more info?

Try our summary of key findings. Or read what others have written about our work — good summaries appeared in the LA Times, the New Statesman, The Atlantic, or Alva Noë’s NPR blog.