Grammar Girl here.
I'm Mignon Fogarty.
And today I'm talking with Charles Kemp, professor of the School of Psychological Sciences at the University of Melbourne.
This was such an interesting discussion about something that people in the northern hemisphere are seeing a lot of right now.
Snow.
Snow.
You've likely heard that Inuit languages have dozens or even hundreds of words for snow.
It's a fun fact that pops up in headlines and coffee shop conversations.
But is it actually true?
Or is it, as one linguist famously called it, the Great Eskimo Vocabulary Hoax?
Well, Professor Kemp has been using big data to find out, not just for snow, but for how our environment shapes the very way we speak.
When we first sat down to talk, he told me that he's always been fascinated by the Nether Conklin principle, the idea that the more important something is to a culture, the more words they'll have to describe it.
He pointed to two specific words that illustrate just how specialized language can get.
Well, I can mention a couple that stuck in my mind.
There's a central Alaska Yupik dictionary which has a word, I think it is a trugta, and it's powdery snow that enters through cracks in a house.
So snow where it isn't meant to be.
Another one I remember is a word, an Inuktitut word.
It is kikalupok and it means noisy walking on hard snow.
So the concept of that just intrigued me.
But Charles says that counting these words is harder than it looks.
Many of these languages are polysynthetic, meaning a single word can represent an entire English sentence, like snow that's falling slowly.
Because of that, linguists have argued for years about whether these word counts are even meaningful.
Some scholars even avoid the topic because they worry it exoticizes other cultures.
Charles and his team, including lead author and poet Tammy Kansu, turned to the HathiTrust Digital Library, a massive database of 18 million volumes, to get past anecdotes and into the data.
Instead of just trying to count words, they looked at the proportion of a dictionary.
For example, not just how many snow words exist, but what percentage of the entire vocabulary is devoted to snow, compared to 600 other languages.
I just found this fascinating.
Dr. Kemp joined me from his lab in Melbourne.
Here's our full discussion where we look at how they actually measured these word counts using English as their anchor.
Charles, welcome to the Grammar Girl podcast.
Now that we have a little background.
How do you know how many times the word snow shows up in a particular dictionary?
Because many of these dictionaries are under copyright, we didn't have access to the full text, but we did have access to frequency information.
So we can say for a particular dictionary how many times does the English word snow appear in that dictionary?
And we use that as our measure of frequency.
How much of the dictionary is devoted to terms about snow?
So among all the terms in the dictionary, how many of them are the English word snow?
And we use that to develop our formal measures.
So we can say things like okay, this Inuktitut dictionary has a much higher percentage of terms related to snow than this dictionary say, for Swahili, some other language.
And you said in your paper that all the dictionaries they're all bilingual dictionaries but English is one of the two languages in all the dictionaries.
How do you think that skews the data or does it?
It definitely does skew the data.
One aspect of that is, in some parts of the world, like Russia and Siberia, many of the interesting dictionaries are dictionaries that map between Russian and some other language.
And in South America, often Spanish is the language that's used to write the bilingual dictionary.
So we don't have good coverage of languages in those areas.
A second way in which English affects our analyses is that we're only able to look at lexical elaboration for concepts that correspond to words in English.
So not every concept corresponds to a single word in English and we just can't look at those concepts.
Yeah.
So if everything were mapped against, say like French, how would you imagine that the data might change?
I think, for the kinds of questions we were focusing on in the paper.
So, for example, what are the languages with the greatest proportion of the dictionary devoted to snow?
And does the climate of the area in which a language is spoken affect terminology for rain and snow-related terms, and so on?
I think it would make very little difference to those sorts of questions.
But when you begin to look at maybe questions that deal with maybe deeper aspects of culture.
So things like emotions or values and things like that.
I could see that the language we're using is the pivot language could have a stronger influence.
Yeah, no, I mean, I want to get away from weather.
But first, like I grew up in Seattle.
And so I'm curious about words for rain and how they are represented in your database and what you found out about words for rain.
That was one of the most interesting things to me, because sometimes people say oh, the words for snow thing.
It's just obvious if you live in a snowy area, you're going to have lots of words for snow.
And I think it's kind of obvious too.
Then what about the case for rain?
My intuition was that languages in the rainiest parts of the world are going to be the ones with the most extensive rain-related vocabularies.
But we didn't really find that to be true.
In fact, some of the languages with...
The most extensive rain vocabularies are languages from places like South Africa.
And so, in retrospect, this makes some sense, I think, because if you live in an area where rainfall is uncertain, then you might need to talk about it and anticipate it in its absence, because rain is critical for survival.
And so I think that's the difference between rain and snow.
If snow is not there, you're just not going to be talking about it.
If rain is not there.
That's a big problem and you will be talking about it and hoping that it is going to arrive soon.
So this to me, the notion of common sense, how the notion of common sense plays out in situations like this isn't obvious, and you need to do the empirical analysis and see how the data and the evidence actually play out.
Yeah, yeah.
In retrospect, it makes so much sense.
Like, if you're in a drought, you're going to talk about rain a lot because you want it to come, even though it's not there.
Something I never experienced in Seattle.
You see words for this.
So words which you know correspond to the idea of eagerly anticipating rain in its absence, or a word that can mean misfortune and it can also mean absence of rain.
So you see kind of these metaphorical connections suggesting that absence of rain is a bad thing.
Oh, that's neat.
Okay, so moving beyond weather, what are some of the other connections that you found?
One of the things that was, I think, a novel finding was that oceanic languages seem to have a relatively high number of smell-related terms.
Oh, why?
Why?
We're not sure why.
Previously, linguists and anthropologists had documented certain cultures have many terms for snow and in fact, there are maps that researchers like Asif and Majid have published, showing the distribution of languages known to have many terms for snow.
And oceanic languages just weren't included in those maps, I think because they hadn't been extensively studied.
Our data, we did find this hotspot.
And so there are languages, like you know, Marsha Lee's has terms like the lingering smell of fish on utensils or clothes, things like that kind of really interesting terms for snow.
I would like to understand better why that's happening.
And I think that's another research project that our findings open up.
Yeah.
And dance, I think dance was another interesting one.
That's right.
We found that languages of North America have extensive dance-related vocabularies, also languages in Africa and some in Australia as well.
So this is kind of moving away from the physical world and the physical environment and moving more towards aspects of culture.
And so I think one of the advantages of our approach is you can look at so many different concepts.
So it's not just physical environment anymore.
It's many different aspects of culture and environment.
That's so interesting.
And so you built this database, is that right?
We compiled this data set, I guess, by selecting, identifying the dictionaries from the Haiti Trust and then going through and doing a lot of pre-processing and analysis.
And I should mention that this was a team effort and the lead author, Tamilin Kishiksura, and she is not only a scholar, but a poet as well.
So kind of the ideal person to be thinking deeply about words in many different languages.
Wonderful.
And now that this resource exists, what other kind of work can be done with it?
Well, we think it provides an opportunity for people who are interested in a certain area to study that with our database.
So just as I said, there was previous work on smell terms.
There's been previous work on many other aspects of vocabulary.
So work on kinship terms, work on terms for emotions.
All of those topics I think can be studied with the database and we made it publicly available so that people can take a look.
We also put an app online so anyone interested can go and use the app and see for a concept they might be interested in, like chocolate, for example.
What are the languages in the dataset which seem to have the most chocolate-related terms?
How fun.
So you started with existing connections that other people had already formed, like the words for snow, like the words for dance and rain.
Other people had done preliminary research on that or different kinds of research to show that some cultures had these and not all of them validated in your database.
So why do you think that there were some that you know?
Your research didn't find the same connection that other researchers found.
Yeah.
Yeah, in most cases, I think that's a limitation of the data that we have.
So one example is because we are working with frequency data only, we can't separate out the parts of a dictionary that are example sentences from the parts of the actual definition.
So I think one example that didn't work out we didn't find that Hindi has a large number of terms for doctors.
And I think that may be because doctor often appears in example sentences.
And so that just adds a lot of noise to to an analysis of terms for doctor.
So I think, yeah, it's not that we're saying previous scholars were wrong.
It's more that the cases that we don't find support for are cases that our methodology really isn't well suited to address.
You know, we talked earlier about English being the, I think you call it the pivot language.
Is that what you called it for all the dictionaries?
And, you know, I was thinking about English.
And could you even do something like this for English?
Because, unlike you, know the Inuit languages that are spoken by a relatively small number of people, English is spoken, you know, all over the world.
Or even if you just look all over the United States, the cultures are so different in different parts of the country.
Do you think that you could even do this kind of research on English?
Or is it just too many different people speak it that you would never find patterns?
I think you can do it, and we can already do that to a limited extent with our data, because the database does include dictionaries for Old English and dictionaries for different regional varieties of English.
And so I remember, for Old English terms like hero and warrior come out in the state of the associated.
If you look at the literature, I think Anna Wierzbiska and colleagues have pointed out that English seems to have an unusually large repertoire of speech act verbs.
So verbs like promise and suggest and recommend and so on.
And this is the kind of thing that you might expect to find if you set up the analysis in a way that could compare English on an equal but on a fair basis with other languages.
That is interesting.
I wonder if the hero stuff from Old English is just because the documents that have survived tend to be those epic poems like Beowulf and things like that.
I wonder if that's influencing those words popping up often.
I think that's right.
That's a reminder that dictionaries don't give you direct insight into languages.
It's actually spoken and used.
It tells you something about language, but not everything.
Right, especially these old, old languages, yeah, where we don't know so much about them.
Well, Charles Kemp, this has been fascinating.
I just think this research is wonderful.
Thank you so much for being here.
Where can people find more about you?
They could go to my website charleskempcom, and that has links to the apps I mentioned that let people explore the data we've been discussing.
Look at where they can find chocolate.
Well, for our Grammarpalusians, we're going to have a bonus segment where we're talking about languages that actually get more complex rather than simpler, over time.
And I think that's going to be fascinating too.
And we'll have Dr. Kemp's book recommendations.
So look for that in your feed if you're one of our supporters.
But for the rest of you, that's all.
Thanks for listening.
