Can you talk about the sorts of data that you’ve worked with throughout your career and how the landscape of data collection has changed over time?
I’m a big fan of different data collection methodologies. I don’t think one is superior to the other, I think they can complement each other quite well. It also depends on the time you’re in. If you’re in a pandemic, I think it pays off if you’re open minded to different methods. I started off collecting traditional data through sociolinguistic interviews in the field, with a microphone setup or in sound attenuated booths. I did booth recordings as well as studio recordings. Eventually, I thought it would be interesting to explore what smartphone apps can do for us, because nowadays everybody carries a phone, which is basically a super good recording device, and a tablet that you can draw on, or do lots of other tasks. So, you can cut out the travel and the tedious organization of devices, and crowdsource the data that you need. However, all this comes with a trade-off. Usually, you miss the rapport-building with the participant that’s very important for sociolinguistic interviews. If you want to collect data from lots and lots of participants, you usually can’t go too deep with metadata on the participants. For example, if you want to collect lots of data, you only collect metadata for age and gender. But if you want to go for broader kinds of analyses, you’d also need data on mobility, how much participants move, social networks and all that stuff. And that’s hard to do with an app, because of attention spans. Participants opt out after five or six minutes. So, you have to be very careful not to overload them, and that can be a slippery slope, I find. And, of course, all this goes for the western world. You need to have these devices and good connectivity. But as soon as you do typological work through fieldwork in places where people don’t have these phones, then these ways of data collection are no longer an option.
In your paper, you discuss the issue of data quality in the context of working with phonetic data. What do you think would be some potential issues when it comes to other types of features, such as morphological, syntactic, semantic or discourse-pragmatic? Do you think the new ways of collecting data through apps or web-based software would affect the data on those kinds of features?
Good question. The most risky bit is definitely phonetic material. Especially because there are so many different types of devices. If you work with an iPhone, you’re almost guaranteed to have a pretty good sound quality. They are very expensive devices, and so the quality is phenomenal. If you work with Android devices, though, there’s such a spectrum of sound quality and the variation of sound quality is substantial. If you look at morphosyntax, lexical variation or pragmatic variation, then it would probably be the case of other factors at play: how to deal with the Observer’s Paradox, how to elicit syntactic features in a good way, and so on. However, I think these are issues that you have to deal with in traditional data-collection settings as well. Even if you meet participants face-to-face, I don’t know if that’s very different from the way you’d do it over the phone. I do think though that participants need a lot more supervision. In the case of morphosyntactic variables, for instance, they’re often hard to elicit. You have to formulate specific syntactic questions, if you want to figure out how people in one locality use a different syntax than people in another locality. What we’ve noticed in our study at the moment is that participants need a lot more explanations on our part about what we actually want to collect. But, you have the same thing, I think, in face-to-face situations, when collecting morphosyntactic data. So, you have to instruct participants more anyway, the only difference being that you just do it over Zoom or over the phone.
We now find ourselves in a situation where we are conducting a lot of conversations in a way that we would have thought of as unnatural before the pandemic. However, I think that now this is becoming naturally occurring speech. What do you think would be some areas or questions about language and linguistics that we can learn more about in this way?
That’s a philosophical question. Honestly, the question really is: how does the virtual setting affect our speech patterns? That’s the main question. And that’s something we know absolutely little about as of now. We don’t even know what happens in situations where the connection and the signal are challenging. You and I have good internet at the moment. As soon as somebody starts pixeling up and freezing, that totally affects the whole dynamic of a Zoom call. And we’re at institutions or countries where we have very stable internet. Most countries don’t have that. So, I think there are so many parameters that could be studied, like looking at how that all affects the way we talk and behave virtually. And we don’t know that yet, right? We know how people like to respond to word lists, how they respond to different styles that you elicit, but the whole virtual setting, I think, is completely under-researched. So, I think that that would probably be the hottest topic at the moment, I would assume.
Do you think that there is a difference in how speakers perceive conversations happening via Zoom, for instance? Do they think it’s just a regular conversation, but it’s on Zoom, or do you think that there’s a new genre emerging, where considerations of formality, for instance, might be different? And, would that crucially affect the data we collect?
You should study that. I think that it’s probably the biggest question mark at the moment for sociolinguistic researchers. Because we will have to resort to this way of data collection for the next few months, I would assume. Data collection has to move on, you can’t just stop a project for a year or two. I really think that that’s one of the most pressing questions that has to be answered. Intuitively I would say yes. For instance, people may feel more safe in their homes; they may be like ‘I’m wearing sweatpants, I feel very comfortable.’ And that would probably have an effect on their speech, I would assume. So, I think you’re totally onto something here. I think that’s something that has huge potential for a really good paper: looking at how people respond differently in a virtual setting as opposed to a live setting. Looking at how people behave differently virtually versus face-to-face, and how that affects speech, has huge potential. Definitely. What we found in our study, which was a very interesting result, was that participants actually felt more comfortable taking part in the study via Zoom, if that was the only option. I think that means that if you interview them virtually, they’re pretty satisfied. They perceive it as ‘that’s as good as it gets’. However, if participants do face-to-face recordings as well, they will actually notice that they feel even more comfortable in a face-to-face situation, so they prefer that.
And what was it like to recruit the participants for the study? How did you do that?
I think that’s actually also quite different from face-to-face interviews. Usually when we meet somebody for an interview, you can always ask them at the end of the interview if they can connect you to other potential participants from their region, and it snowballs from there. People feel much less obliged if you ask them over Zoom, I find. They’re like ‘whatever, sure, I’ll tell them this or that name’ but nothing ever happens with most people that we have. They feel much more of a need to provide us with names if you meet them in person, I find. Recruitment is tough, but it’s always tough. And the thing is, with open calls on social media, you only target certain groups, so the selection of participants is super biased. When we do open calls, we mostly get females from urban regions. We want symmetrical designs, but lots of males from rural areas are so hard to target. Technically, there’s polling institutions that you can contact. Some of them will help you out with recruitment, with pay, usually.
How easy would it be to use the software to carry out a similar kind of data-collection cycle if you’re a PhD student working independently and let’s say that you don’t really have the highest level of technical skill necessary?
You do need some coding experience. You can adapt the code, but it really depends on the knowledge you have beforehand. If you’re a linguist who only has linguistics training, it’ll be a long while. If you could outsource it to someone like an IT developer who does the coding for you, it shouldn’t take long. And, there’s a lot you can do with the software. You can use levers that participants can manage, you can play audio, you can play video, you can draw stuff, and so on.
Perhaps you could say something about where this is going in the next ten or twenty years? And do you think there are some ways in which we can still reach those hard-to-reach participants who don’t have access to devices and software and internet?
I mean, eventually, they will also have these devices. It’s just a matter of time, I would say. Smartphone penetration, internet penetration, that will be everywhere in a couple of decades at the latest, I would assume. So, by then, we will be able to do the web-based research there as well. It just takes much longer. What’s really lacking at the moment are papers that actually compare studio quality recordings to smartphone recordings. There are some papers in The Journal of Voice that looked at some phonetic parameters and how they are different from smartphone recording to smartphone recording, and compared to studio recordings. So, we need a lot more data on data validation, and quality validation. I think that would be the next logical step.