So for some reason, I’ve come across three studies in quick succession based in mapping language. Now, if you know me, you know that nattering on about linguistic methodology is pretty much the Persian cat to my Blofeld, but I really do think that looking at the way that linguists do linguistics is incredibly important. (Warning: the next paragraph will be kinda preachy, feel free to skip it.)
It’s something the field, to paint with an incredibly broad brush, tends to skimp on. After all, we’re asking all these really interesting questions that have the potential to change people’s lives. How is hearing speech different from hearing other things? What causes language pathologies and how can we help correct them? Can we use the voice signal to reliably detect Parkinson’s over the phone? That’s what linguistics is. Who has time to look at whether asking people to list the date on a survey form affects their responses? If linguists don’t use good, controlled methods to attempt to look at these questions, though, we’ll either find the wrong answers or miss it completely because of some confounding variable we didn’t think about. Believe me, I know firsthand how heart wrenching it is to design an experiment, run subjects, do your stats and end up with a big pile of useless goo because your methodology wasn’t well thought out. It sucks. And it happens way more than it needs to, mainly because a lot of linguistics programs don’t stress rigorous scientific training.
OK, sermon over. Maps! I think using maps to look at language data is a great methodology! Why?
- You get an end product that’s tangible and easy to read and use. People know what maps are and how to use them. Presenting linguistic data as a map rather than, say, a terabyte of detailed surveys or a thousand hours of recordings is a great way to make that same data accessible. Accessible data gets used. And isn’t that kind of the whole point?
- Maps are so. accurate. right now. This means that maps of data aren’t just rough approximations, they’re the best, most accurate way to display this information. Seriously, the stuff you can do with GIS is just mind blowing. (Check out this dialect map of the US. If you click on the region you’re most interested, you get additional data like field recordings, along with the precise place they were made. Super useful.)
- Maps are fun. Oh, come on, who doesn’t like looking at maps? Particularly if you’re looking at a region you’re familiar with. See, here’s my high school, and the hay field we rented three years ago. Oh, and there’s my friend’s house! I didn’t realize they were so close to the highway. Add a second layer of information and BOOM, instant learning.
The studies
Two of the studies I came across were actually based on Twitter data. Twitter’s an amazing resource for studying linguistics because you have this enormous data set you can just use without having to get consent forms from every single person. So nice. Plus, because all tweets are archived, in the Library of Congress if nowhere else, other researchers can go back and verify things really easily.
This study looks at how novel slang expressions spread across the US. It hasn’t actually been published yet, so I don’t have the map itself, but they do talk about some interesting tidbits. For example: the places most likely to spawn new successful slang are urban centers with a high African American population.
The second Twitter study is based in London and looked at the different languages Londoners tweet in and did have a map:
Interesting, huh? You can really get a good idea of the linguistic landscape of London. Although there were some potential methodological problems with this study, I still think it’s a great way to present this data.
The third study I came across is one that’s actually here at the University of Washington. This one is interesting because it kind of goes the other way. Basically, the researchers has respondents indicate areas on a map of Washington where they thought language communities existed and then had them describe them. So what you end up with is sort of a representation of the social ideas of what language is like in various parts of Washington state. Like so:
There are lots more interesting maps on the study site, each of which shows some different perception of language use in Washington State. (My favorite is the one that suggests that people think other people who live right next to the Canadian border sound Canadian.)
So these are just a couple of the ways in which people are using maps to look at language data. I hope it’s a trend that continues.
makingnoiseandhearingthings.com 
Reblogged this on On Politics and Language and commented:
I think that maps can also be quite fascinating when related to language data, although I’d argue that they can sometimes offer representations of data that are misleading. Especially in dialectology, the use of dialect maps tends to give a very clear cut representation of a multifaceted and complex phenomenon. At least in dialectology, maps often present an issue as, on this side of the line or in this area they speak one way and on the other side of the line they speak another, which is troubling. Like I said though, I think they’re interesting and useful if they’re used properly.
That’s definitely true. Any type of language model is going to simplify the data to some degree. Ideally, we would like to remain faithful to all the intricacies of the data… but that would yield a model which was identical to the language itself and therefore not really usable. As long as you realize that that’s just part of creating a conceptual model, though, and that no model will ever fully account for all variation, I don’t see it as harmful.
I do worry that most people don’t really get a good foundation in what science can and can’t do, which I think is what leads to that sort of misconception. I know that certainly in my schooling, which was science-heavy, we never really talked about the theory or philosophy of science until I got to college. Memorizing formulas, yes. Analyzing what, exactly, formulas were and their relationship to the real world, not so much.