ELFA project

What’s so funny? More laughter in academic talk


Even real scientists like to laugh.
Photo by Ruth Orkin
Source: artnet.tumblr.com

Is it possible to fully experience humor when using a foreign language? This varies from person to person (you probably know someone with no sense of humor in any language), and maybe also from culture to culture. There’s a lot of culture-specific humor, so that even native speakers of the same language from different cultural backgrounds (e.g. Brits and Americans) are susceptible to misunderstandings when a joke is missed or a metaphor lacks a cultural reference.

Much intercultural research, even on academic talk, takes this monolithic approach – Culture A does it this way, Culture B does that that way, when Culture A goes to Culture B to study, there’s going to be problems. But lingua franca interaction adds additional variables, especially when English is spoken by second-language users outside of an English-speaking country. What then?

In an earlier post I presented some data from the Corpus of English as a Lingua Franca in Academic Settings (ELFA corpus), which I compared to similar spoken data from the U.S. When looking at the broad, corpus-wide frequency of laughter in the two corpora, there was no striking difference between the native and non-native speaker data. A laugh occurs 2-3 times per 1,000 words in each corpus, and laughter is concentrated in similarly interactive events like seminar discussions.

Climbing the XML tree

Like I mentioned in the earlier post, counting XML tags misses the point of XML. Structured data like this can be queried in imaginative ways with some programming skills, and additional metadata (data about the data) concerning the speaker and speech event can be associated with each occurrence of laughter. I performed such a query on the XML version of the ELFA corpus, outputting the results in a spreadsheet format and using Microsoft Excel’s Pivot Table feature to explore this mass of data – 3384 rows of data for each occurrence of laughter, with 13 speaker and event variables on each line.

Taking all instances of laughter in the corpus, 65% of these were “full” laughter (n=2191), with 35% (n=1193) being spoken laughter. Dividing the data in a different way, 70% of all laughter events were uttered by an identified speaker (i.e. who had accompanying speaker metadata). Of the other 30%, only 85 laughs were by unidentified speakers, and remaining 937 acts of laughter were tagged as “several speakers”, a choral laugh.

The data on laughter in the ELFA corpus. The analysis follows the laughter by several speakers, looking for figures on co-occurring laughter by main speakers. The chart below shows how many times the identified speakers laughed based on their first languages.

What about this laughter by more than one speaker? Where does it occur? By pivoting the tables once again, I find that 53% (n=495) of choral laughter occurs as a “nested utterance”, meaning the main speaker has somehow elicited this laughter, but continues speaking as part of a continuous turn. So then I wondered, does the main speaker laugh too? Is this laughter in the background, so to say, or do the speaker and listeners co-construct the laughter event?

It turns out they’re almost even. From the 495 incidents of nested, choral laughter, 47% have co-occurring laughter in the longer, subsuming utterance, suggesting laughter that is jointly produced by speaker and listeners (though this should be checked with qualitative analysis). These cases might provide evidence of collaborative, rapport-building functions of laughter. Yet, 53% of these cases of group laughter overlapping another utterance don’t feature laughter in the main speaking turn (i.e. others are laughing but the main speaker is not). Again, this is where qualitative analysis starts, but I’ll wager a hypothesis that these aren’t all cases of listeners laughing “at” the speaker; instead, I would bet that ELF users do indeed know how to make each other laugh, like ordinary users of natural language always manage to do.

Laughter and culture

Going back to the laughter and culture theme, I found a related article by Prof. Gertrud Reershemius (2012). She examines seven academic presentations delivered in English in a UK university, and eight presentations by German native speakers in a German university. Among the eight presenters recorded in the English corpus, four of these are native speakers of German who are presenting in English.

Reershemius’ quantitative data can be summed up like this: two of the English native-speaker presenters generate exactly 100 laughs from their audiences in total, far more than anyone else in the sample. The German presenters don’t generate much laughter in any language. I don’t say this to be funny, this is the take-home message: these numerical findings are followed up by qualitative, pragmatic analysis on how these British presenters deploy laughter strategically as a function of their academic culture, a culture not shared by the Germans. It’s not that the Germans don’t have a sense of humor, they simply come from an academic culture in which research is presented differently than in the UK.

I might criticise this conclusion just based on the data – having found two British presenters with a lively sense of humor doesn’t really say anything generalisable about “British culture” beyond those two researchers. But I’d rather draw attention to this culture-as-monolith idea: German culture and Finnish culture were having tea and chatting quietly. Then British culture burst into the room, started telling jokes and getting noisy. Finnish culture stared at the floor for 3.21 minutes. Then German culture…. you get the idea.

The mother of all melting pots

Something that gets missed in the monolithic, unidirectional conception of intercultural encounters is that much of this takes place on nobody’s “home turf”. In the ELFA corpus, 51 different first languages are represented in academic English speech events in Finland. In a mixture this diverse, in a frozen forest of northern Europe, what can be said about British and German culture? If we extend Reershemius’ findings a step further, can we expect to find striking differences between the occurrences of laughter among different cultures?

This is a more general question than audience laughter in lectures. Rather than considering the act of “making someone laugh”, I wanted to see who is laughing. I sorted all occurrences of laughter in the ELFA corpus by the first language of the person laughing. Getting raw counts, however, is not enough; discovering that Finns laugh 840 times is meaningless without taking account of their heavy representation in the corpus – just over a quarter of total words. Fortunately, original ELFA compiler Elina Ranta of the University of Tampere has calculated the total words for each first language in the corpus, so I used these figures to generate standardised frequencies.

The chart below shows the top 15 first languages in terms of frequency of laughter (laughs per 10,000 words). It’s interesting to note that native speakers of these 15 languages account for 80% of the words in the corpus, and 83% of the total laughter by identified speakers.

Instances of laughter by speakers of 15 first languages in the ELFA corpus. Rows are sorted by standardised frequency (laughs per 10K words).

As we clearly see, speakers of Kihaya are overwhelmingly the happiest people on earth. Actually these first two entries – Kihaya and Japanese – are good examples of why Prof. Mauranen generally discourages these types of first-language comparisons in ELFA. There isn’t enough data from each first language to make representative claims about speakers of those languages. As a result, these first two entries can be safely dismissed as a handful of people having an especially good time.

Keeping the limitations of the data in mind, the most striking finding is derived from the overall standardised frequency of laughter in this sample: 24 laughs per 10k words. Looking back up the chart, this same frequency is occupied by spots 8, 9 & 10 – native speakers of Swedish, English and German. Significantly, these are among the best represented first languages in the corpus, each having over 50,000 words of speech. Considering that the English category includes several varieties of native English, the German speakers’ frequency of laughter was higher than speakers from the UK.

The main finding seems to be that laughter is distributed evenly among the speakers of these three quite different linguacultures. Among native speakers of English, Swedish, and German, there seems to be no difference in the frequency of their laughter in academic ELF speech events. And who beats them all with 28 laughs per 10,000 words? The stereotypically silent, sullen Finn, as the first intercultural communication book that you find will tell you. Actually, we Finns are quite a humorous bunch; it’s just none of your business.

Our data, your research

These data and findings could be the first step toward a nice MA thesis or article. This is where the qualitative research begins – looking at the laughter in context. I don’t intend to pursue the topic further, so anyone else is welcome to use the raw data (in spreadsheet format) and the corpus. Just ask!

Reershemius, G. (2012). Research cultures and the pragmatic functions of humor in academic research presentations: A corpus-assisted analysis. Journal of Pragmatics, 44 (6-7), 863-875. DOI: 10.1016/j.pragma.2012.03.012.