Frequently Asked Questions

If you’ve read about our work elsewhere, please note that nothing beats reading the actual paper — it’s open access! Here we answer some questions we’ve encountered. We may be updating this page, so if your question isn’t answered, check back later or shoot us an email.

Huh? #

We see what you did there! See here for our prior work on that word. Here, we present the bigger picture.

What? #

One of the best examples of a repair sequence in popular culture is the following scene from Pulp Fiction (1994):

So we say ‘Huh?’ every 90 seconds? #

Err, no. We ask for clarification on average every 90 seconds, so (1) that’s not just using ‘Huh?’ but also more specific questions and suggestions and (2) that’s an average across all the 12 languages that we have data for, so there will of course be spates of talk in which we proceed relatively smoothly (and other moments at which we really are struggling). Still, once every 90 seconds on average is pretty frequent — and this shows how crucial this system is. That is one reason we wanted to understand how it works across different languages and cultures.

Didn’t we know this already? #

Surprisingly, no. So far, comparative work in linguistics has focused on speech sounds, words, and grammatical elements. Only recently has it become practical to study language as it is used in conversation. With regard to repair, some people assumed there would be a great deal of cultural variation, while others hypothesised that the system might be universal. We set out to settle this issue empirically, using data from a wide range of languages.

Could it have been different? #

You’ve found that people do the same three things everywhere. But perhaps that’s just a logical consequence of having a communication system. Could it have been any different?

Yes, absolutely. For starters, language is not the only animal communication system. While one might think that methods for the recovery from communication problems are indispensable, in fact a great many other communication systems go without them — so the existence of an interactive repair system sheds new light on what’s special about human language.

Second, there is a great deal of diversity in languages at many levels. Take number words: some languages have complex systems while others get by with relatively simply “one, two, many” systems. It could have been the same with repair: it might be that some languages mostly or exclusively use only the most general repair tool (the “error signal” Huh?, which we’ve previously shown to be universal) while others use a wider range. Instead, what we found is that the three basic types are used in the same situations across all the languages in our sample.

What are possible applications of this research? #

The tools and principles found in our study may help us to make computers communicate in more ‘human’ ways, for instance when they don’t understand a voice command or search query. Right now, when computers don’t understand us, they do little more than produce an unhelpful error message. The three universal methods, and the order in which they are used, amount to a basic toolkit for designing ‘clarification dialogs’ that can be used in human-computer interfaces regardless of the language they are implemented in.

The findings also have applications in language teaching: the three basic types of repair —in their language-specific garb— are much more useful to know than stilted phrases like “I beg your pardon?” often found in learning guides. Finally, they may be useful in cross-cultural communication: knowing when and how to ask for ‘repair’, and knowing how to interpret others’ use of these methods, can help people to secure mutual understanding more efficiently.

Your sample seems small — is 12 languages really enough? #

More is always better. However, rigorous standards prevent us from just adding languages to the sample based on introspection or secondary sources like phrasebooks or scripted conversations. We want to compare like with like. For the kind of rigorous comparison we have carried out, what is needed is (1) a sizable video corpus of naturally occurring, informal conversations and (2) a language expert with a solid understanding of the methods of conversation analysis and interactional linguistics.

Those requirements are tough to meet, and in our project we’ve met them for 12 languages of 8 independent, unrelated language families around the world. It includes languages from urban and rural societies, languages produced in the spoken and signed modalities, languages spoken by millions as well as by a few thousand people, and languages which haven’t been in contact with each other. Finding the same basic system in all of these highly different languages makes it very likely that we are tapping into universal properties and principles of social interaction.

The paper: Dingemanse, Mark, Seán G. Roberts, Julija Baranova, Joe Blythe, Paul Drew, Simeon Floyd, Rosa S. Gisladottir, Kobin H. Kendrick, Stephen C. Levinson, Elizabeth Manrique, Giovanni Rossi, and N. J. Enfield. 2015. “Universal Principles in the Repair of Communication Problems.” PLOS ONE 10 (9): e0136100. doi:10.1371/journal.pone.0136100.