The speaker shapes the language and the language shapes the speaker

Language changes. And an individual’s use of language changes, as they develop from child to teenager to adult. How might these two kinds of development be related?

Researchers at Stanford University have been looking at the question. They took as their data two discussion websites (both for the discussion of beer, though they assure us the topic wasn’t critical.) A nice thing about these websites was that they could download everything: all the discussions, with every contribution labeled with who made it. On both websites most contributions were in English, but, like any community, the community defined by each website developed a distinctive vocabulary and way of expressing itself. Each is a society in miniature. Both had been going for over ten years, and the researchers could look at how this shared language changed over time.   They could also look at each individual’s language, from their first post, their ‘birth’ in the community, to their last, their ‘death’. Both communities had several thousand contributors, some long-lived and some short-lived, some talkative, contributing several hundred posts, others meek, contributing just once or only a few times.

The core method for the research was to identify the recurring expressions in each community, observing their rise and fall over the lifetime of the community, and also to observe the expressions used by each individual. They could then plot how the two sets of patterns interacted. An example is the term used in one of the communities to refer to a beer’s smell. Two options were to use the word aroma, or the letter S (short for smell). What was the norm at any point in time, and how it changed, and which users used which option, can all be followed in the data.

When we join a new community, we do not know its distinctive vocabulary: if we want to become a fully integrated member of the community, we will need to learn and adopt it. Thus, new members’ early contributions were not ‘on the nail’ of the community’s norms, but over time they learnt the community’s vocabulary and fitted in better. This was their community ‘childhood’.

It often happened that, once they had fully learnt the language of the community, they started playing with it, contributing new expressions and being quick to adopt others’ innovations. So at this phase they were contributing to language change. This was their adolescence.

And then, typically a third of the way through their lifetime in the community, they tended to stick. They carried on using the terms that they and ‘their generation’ had coined, but were less inclined to adopt new terms, or to innovate themselves.   This was their adulthood.

Strange to say, the division was into these three stages, with the peak of creativity one third of the way through a lifetime, whether the lifetime was one year long or ten.

For the full paper, see

No country for old members: User lifecycle and linguistic change in online communities. Cristian Danescu-Niculescu-Mizil, Robert West, Dan Jurafsky, Jure Leskovec, Christopher Potts. Proceedings of WWW Conference, 2013.  (Winner of best paper award.)