Six months ago a linguistic factoid made global headlines: ‘huh?’ is a universal word. The New York Times described it as “the syllable that everyone recognises” and for the Süddeutsche Zeitung it was “the most important word in the world” because of its role in solving communicative mishaps. Some reports claimed it was the first human word or the only global word. The news was based on a paper by me and my colleagues Francisco Torreira and Nick Enfield, entitled, Is “Huh?” a universal word? Conversational infrastructure and the convergent evolution of linguistic items (see the paper or download the PDF). For us, the newsworthy part was in the second part of the title. For the rest of the world, it was in the first part.

The global huh-laballoo (as one commentator called it) was an interesting experience. Of course some important details were lost in the process. We did not, as some reporters thought, travel the globe just to record people saying ‘huh?’ (instead, our work was as much corpus-based as it was fieldwork based). Nor did we find that people everywhere produce the phonetic sequence h-u-h (instead, we found that the word is calibrated to local linguistic systems). But most importantly, we did not only study ‘huh?’ — the finding was just one of the first in a much larger project in which we are comparing conversational structures across languages.

Pragmatic typology

Typology is the comparative study of linguistic systems. Just like one can develop typologies of sound systems, syntax, and semantics, so one can develop typologies of pragmatics and conversational structures. At the Max Planck Institute for Psycholinguistics in Nijmegen we run several large-scale comparative projects aimed at contributing to an emerging field of pragmatic typology — the typology of systems of language use and the principles that shape them. Our work is part of a recent upsurge of comparative research on conversational structure (e.g., Selting and Couper-Kuhlen 2001; Enfield and Stivers 2007; Sidnell 2009; Stivers et al. 2009; Dingemanse and Floyd 2014, among others).

With the publication and media storm behind us, it is interesting to look a bit more closely at the reception of the ‘huh?’ paper as one instance of work in this emerging field. We were happy with the quality of most of the coverage — not a given in linguistic matters, as Language Log never fails to note. Not only were there hundreds of fairly accurate news stories around the world, but also thoughtful pieces in venerable venues like The Atlantic, the New Statesman, NPR, Süddeutsche Zeitung, Die Welt, and Smithsonian Magazine.

This is good news for linguistics: new findings can elicit international media attention, testifying to the avid interest that people take in our subject of study. The simple website with explanatory materials that accompanied the release also helped to steer media coverage in the direction of scientific accuracy, away from some obvious pitfalls. There were, of course, also critical notes. Some readers worried about about contact and inheritance as alternative explanations, matters that we were able to clarify in our FAQ. In this post I want to discuss three issues that seem particularly relevant to readers of Diversity Linguistics Comment: the matter of sample size, the problem of introspection, and the challenge of achieving comparability.

Sample size

In our paper we use two language samples: an extended sample of 31 languages based on our own data and published literature, and a focused subset of 10 languages for which we and our colleagues had access to rich multimedia corpora of informal conversations. Some critics appeared not to have noticed this two-pronged approach and take us to task for looking at only 10 languages.

What did we actually do? Our starting observation was the remarkable similarity of this interjection in the diverse set of languages covered in our project. However, we realised that 10 languages (of 6 phyla) is not the largest typological sample ever, and so we scoured published transcripts of conversation in a wide range of languages to locate further evidence. There were 21 additional languages for which we could find corpora and transcripts that included repair sequences, and in all of those we were able to locate a form of the interjection. That puts the sample at 31 languages from 16 distinct phyla.

Compared to enormous comparative projects like WALS or UPSID, this is still a modest sample. But is it too small? The chance that there are only 16 phyla where this form occurs and we randomly picked those approximates zero; and if we stick to a conventional significance level of 0.05, we can rule out the idea that the property is not universal but only occurs in 354 out of 426 Glottolog phyla (thanks to Harald Hammarström for this observation). What this means is that already from the 16 phyla covered, we can infer that the trait is very likely extremely widespread. Even if it may not be literally found in all 7000+ of the world’s languages (we may never know), the sheer fact of its suspicious similarity in so many unrelated phyla calls for an explanation. The central focus of the discussion in our paper is to provide such an explanation — in the form of a mechanism of convergent cultural evolution. For our argument, it is sufficient that the trait we study is widespread across independent lineages; it is not crucial that it is literally universal.

Of course, like everybody, we like large and diverse samples. And here we hit on the first of the growing pains of pragmatic typology: the data are not there yet. The questions we ask require a different kind of data than that of more traditional approaches to typology. There is no WALS for conversational structures, nor is this kind of data found in grammar libraries or collection of dictionaries. As we note in our FAQ, “rigorous standards prevent us from just adding languages to the sample based on introspection or indirect sources like dictionaries or scripted conversations. 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”. Such data are hard to come by. No doubt this will change in the future, as more linguists turn to the study of conversation (which after all is the ‘primordial site of language use’, Schegloff (2006)) — but for now and some time to come, the hard-won results of pragmatic typology will be based on relatively limited samples.

Introspection and the nature of data

A discussion of our paper at Language Log yielded some interesting comments. One “Martin” made a number of insightful points about our proposals concerning convergent evolution, which led us to clarify a number of points on our FAQ page, for instance concerning what exactly we mean by convergent cultural evolution.

Many of the responses on Language Log were of a different caliber. In the paper we pointed out (citing sources) that in matters of language use, what people think they do is often very different from what they really do. Yet several commentators did little more than express their disbelief about the findings. These included prominent linguists who were happy to rely on their own intuitions even when they ran directly counter to our corpus evidence. In one case, the speculations were so different from our findings that I shared some of our data to resolve the matter. This led to a productive exchange in which our critic conceded that our data indeed showed what we found it did, and not what she thought it should. (This admission was made in the comments to a critical blog post at GeoCurrents which has since been taken down.)

Apart from demonstrating the benefit of working with actual recordings of language use for reasons of accountability, what this exchange showed is that especially in matter of pragmatics, introspection is fraught with problems. All sorts of things get in the way: from lack of awareness (‘I’m sure we never say this!’) to language ideologies (‘that’s rude, we are taught to say excuse me’), to judgments derived from atypical situations (‘I would never say this to my boss’). While intuitions are interesting for many reasons, one cannot argue with corpus evidence.

Here then is the second of our growing pains: introspection often fails to give reliable evidence in matters of language use, and we need evidence from language in use to get a clear view of the kinds of structures we are interested in — be it individual items like ‘huh?’ or the larger paradigms of formats for other-initiation of repair they are part of (see Dingemanse, Blythe, and Dirksmeyer 2014 for a first report). That native speaker testimony should be treated with care in matters of pragmatics was already noted by Silverstein in his classic The Limits of Awareness (1981). As Silverstein noted, in the case of matters beyond these limits, “we have to depend upon cross-cultural analysis and the accumulated technical insight based upon this, for native speaker metapragmatic testimony is not going to be necessarily accurate for the general analysis of language” (Silverstein 1981:20). One goal of pragmatic typology is to furnish methods that provide precisely this type of cross-cultural analysis and technical insight: methods that enable rigorous and systematic comparison of language in use across settings and societies.

Natural control method

An important methodological aspect of our study deals with the question of how to achieve comparability. The underlying idea is simple. To do typology, we need systematic comparison. To do systematic comparison, we need to compare like with like. There are established methods for this in phonological, syntactic and semantic typology (Hyman 2009; Lehmann 1978; Pederson et al. 1998), but how to do this in conversation may seem less straightforward. Indeed, many linguists have been trained to believe that conversation, the basic mode in which we experience language, is irredeemably degenerate, ridden with irrelevant performance errors and devoid of clear structure. Decades of research in conversation analysis have refuted this idea (Sacks 1992; Schegloff 2007; Enfield and Sidnell 2014), but this literature is still not widely known in linguistics and it is based mostly on English and a few other major languages. This is the third of our growing pains: the search for rigorous and systematic ways to compare conversational structures.

The key is to find contexts that are stable across different languages. In our study of ‘huh?’, this context is a sequence of other-initiated repair (Schegloff, Jefferson, and Sacks 1977): A says something, B initiates repair, then A repeats the prior turn. This sequential context provides us with a way to keep one part of the system stable for comparison: the same structure can be located across corpora and across languages, and since the interjection is wedged in between the other two turns, we can be confident that we are comparing what is functionally the same item. Just like experimentalists use laboratory conditions to achieve experimental control, we can use the sequential structure of conversation to achieve natural control. In a recent review we call this the natural control method (Dingemanse and Floyd 2014).

The natural control method has been used, implicitly or explicitly, in a range of comparative studies. Fox, Hayashi and Jasperson (1996) study same-turn self-repair in English and Japanese, investigating how different syntactic structures enable different kinds of repair operations. Rossano, Brown and Levinson (2008) study gaze behaviour in question-answer sequences in Italian, Tzeltal and Yeli Dnye. Stivers et al. (2009) study the turn-taking system in ten languages around the world by comparing hundreds of question-answer sequences, keeping sequential environment and social action constant to capture subtle by systematic differences in timing. Sidnell and Enfield (2012) study a type of social action called ‘K+ second assessment’ in Caribbean English Creole, Finnish, and Lao, and trace the ‘collateral effects’ of different linguistic resources for carrying out this action. Finally, in our work on other-initiated repair, we isolate sequences of other-initiated repair in ten languages to study ‘huh?’-like interjections (Dingemanse, Torreira, and Enfield 2013) as well the larger linguistic system for repair initiation, in order to understand the range of linguistic resources that languages draw on for this function (Dingemanse, Blythe, and Dirksmeyer 2014, and further forthcoming publications by us and other project contributors).

Most work so far has focused on conversational structures that have been well-studied in English and a handful of other major languages — just like the focus of early studies in phonological typology on vowels and consonants, or of semantic typology on colour terms. However, the natural control method can easily be extended to other domains. Take evidentials. Most traditional analyses look at these as markers that encode speaker knowledge, but recent work on evidentials in conversation has brought to light markers that encode the distribution of knowledge between speaker and addressee (Clift 2006; Gipper 2011; Nuckolls & Michael 2014). This work has outlined a number of sequential environments in which evidentials can be found. A next step is to locate these same environments across in different languages, using the natural control method to study the use of evidential strategies in a comparative fashion. Sequential analysis offers powerful tools for comparative research.

In closing

Pragmatic typology is a relatively young field of inquiry. Its youthfulness comes with certain benefits —chief among them the excitement of new discoveries— but also with growing pains. Here I have outlined three of them: the matter of sampling, the nature of data, and the challenge of achieving comparability. With the rise of conversational corpora, changes in evidentiary standards, and the natural control method, these challenges can be overcome, and pragmatic typology will be able to grow into a fruitful field of linguistic inquiry.

*      *      *

P.S. The first version of our paper ‘Is “Huh?” a universal word?’ did not have correctly typeset IPA characters. The journal has since issued a correction. If you downloaded the paper before April 2014, get the new version here.

References

• Clift, Rebecca. 2006. “Indexing stance: Reported speech as an interactional evidential.” Journal of Sociolinguistics 10 (5): 569–595.
• 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).
• Dingemanse, Mark, and Simeon Floyd. 2014. Conversation across cultures. In Cambridge Handbook of Linguistic Anthropology, ed by. N. J. Enfield, Paul Kockelman, and Jack Sidnell. Cambridge: Cambridge University Press. (PDF)
• 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 (PDF).
• Enfield, N. J., and Jack Sidnell. 2014. Language presupposes an enchronic infrastructure for social interaction. In The Social Origins of Language, ed by. Daniel Dor, Chris Knight, and Jerome Lewis, 92–104. Oxford: Oxford University Press.
• Enfield, N. J., and Tanya Stivers, eds. 2007. Person Reference in Interaction: Linguistic, Cultural, and Social Perspectives. Cambridge, UK: Cambridge University Press.
• Fox, Barbara A., Makoto Hayashi, and Robert Jasperson. 1996. Resources and repair: a cross-linguistic study of syntax and repair. In Interaction and Grammar, ed by. Elinor Ochs, Emanuel A. Schegloff, and Sandra A. Thompson, 185–237. Cambridge: Cambridge University Press.
• Gipper, Sonja. 2011. Evidentiality and intersubjectivity in Yurakaré: An interactional account. PhD dissertation, Nijmegen: Radboud University & MPI for Psycholinguistics.
• Hyman, Larry M. 2009. “How (not) to do phonological typology: the case of pitch-accent.” Language Sciences 31 (2-3): 213–238. doi:10.1016/j.langsci.2008.12.007
• Lehmann, Winfred P., ed. 1978. Syntactic typology: studies in the phenomenology of language. Austin: University of Texas Press.
• Nuckolls, Janis B., and Lev Michael, eds. 2014. Evidentiality in interaction. Amsterdam: John Benjamins.
• Pederson, Eric, Eve Danziger, David Wilkins, Stephen Levinson, Sotaro Kita, and Gunter Senft. 1998. “Semantic Typology and Spatial Conceptualization.”Language 74 (3): 557–589.
• Rossano, Federico, Penelope Brown, and Stephen C. Levinson. 2008. Gaze, Questioning, and Culture. In Conversation Analysis: Comparative Perspectives, ed by. Jack Sidnell, 187–249. Cambridge: Cambridge University Press.
• Sacks, Harvey. 1992. Lectures on conversation. 2 vols. London: Blackwell.
• Schegloff, Emanuel A. 2006. Interaction: The Infrastructure for Social Institutions, the Natural Ecological Niche for Language, and the Arena in which Culture is Enacted. In Roots of human sociality: Culture, cognition, and human interaction, ed by. Nick J. Enfield and Stephen C. Levinson, 70–96. Oxford: Berg.
• Schegloff, Emanuel A. 2007. Sequence Organization in Interaction: A Primer in Conversation Analysis. Cambridge: Cambridge University Press.
• Schegloff, Emanuel A., Gail Jefferson, and Harvey Sacks. 1977. “The Preference for Self-Correction in the Organization of Repair in Conversation.” Language 53 (2): 361–382. doi:10.2307/412243
• Selting, Margret, and Elizabeth Couper-Kuhlen, eds. 2001. Studies in Interactional Linguistics. Amsterdam: John Benjamins.
• Sidnell, Jack, ed. 2009. Conversation Analysis: Comparative Perspectives. Cambridge: Cambridge University Press.
• Sidnell, Jack, and N. J. Enfield. 2012. “Language Diversity and Social Action.” Current Anthropology 53 (3): 302–333. doi:10.1086/665697 (PDF).
• Stivers, Tanya, N.J. Enfield, Penelope Brown, C. Englert, Makoto Hayashi, Trine Heinemann, Gertie Hoymann, et al. 2009. “Universals and cultural variation in turn-taking in conversation.” Proceedings of the National Academy of Sciences 106 (26): 10587–10592. doi:10.1073/pnas.0903616106 (PDF).

---

---