personal pronouns in Brazilian Portuguese20 Feb 2015
While English doesn’t have a singular you (unless you count thou), Portuguese has two: tu and você, with tu being preferred in some regions and você being preferred in others. A while ago an American friend who is learning Portuguese asked me where exactly people used one form or the other and, despite Portuguese being my first language, I couldn’t give her a proper answer. So, to atone for my unhelpfulness (and because I needed an excuse to start playing with D3) I scraped 680,045 geolocated tweets from Brazil and made the map below, where the higher the tu/você ratio the “redder” the state and the lower the tu/você ratio the “bluer” the state. Hover the mouse on the map to see the state names and respective ratios and number of tweets.
As we see, all states prefer você over tu - not a single state has a tu/você ratio higher than 1. And that preference is strong: in all states but one você is at least twice as popular as tu (i.e., the ratios are no higher than 0.5). The exception is Rio Grande do Sul (the purple, southernmost state), where the contest is close, with a tu/você ratio of 0.88. Which fits the popular perception of gaúchos (as natives from Rio Grande do Sul are often called) having a culture that is heavily influenced by the surrounding Spanish-speaking countries (tu is singular you in Spanish as well).
Other than in Rio Grande do Sul tu is most popular in Santa Catarina (tu/você ratio of 0.47), Maranhão (0.42), Amapá (0.35), Pará (0.30), and Roraima (0.30). Santa Catarina is in the South (it borders Rio Grande do Sul) while the other four states are in the North and Northeast. Você, on the other hand, is most popular in Sergipe (0.05), Minas Gerais (0.06), Goiás (0.06), Mato Grosso do Sul (0.08), Mato Grosso (0.08), São Paulo (0.08), and Espírito Santo (0.09) - basically the entire Southeast and Midwest regions.
I got curious to know how exactly people use tu. You see, conjugation is a tricky business in Romance languages. In English I go, you go, he/she/it goes, we go, you go, they go - the verb almost never changes. But in Portuguese eu vou (I go), tu vais (“singular you” go), você vai (“singular you” go), ele/ela vai (he/she/it goes), nós vamos (we go), vós ides (“y’all” go), vocês vão (“y’all” go), eles/elas vão (they go). The verb almost always changes. And as this example shows, tu and você call for different verb inflections: tu vais but você vai.
Or so Portuguese grammar dictates. But grammar rules are one thing and what people actually say are another. At least where I’m currently living (Brasília), when people use tu they often conjugate it as você: tu vai instead of tu vais; tu foi (“singular you” went) instead of tu foste; and so on. So, when people use tu do they conjugate it correctly?
Turns out they don’t. Out of the 680,045 tweets I scraped tu vai appears in 6,559 of them while tu vais appears in a mere 77 - too few to even allow state-wise comparisons.
Which is probably ok. By merging the conjugation rules of tu and você we simplify Portuguese grammar. Old English conjugation was hard: ic hæbbe (I have), þū hæfst (thou hast), hē/hit/hēo hæfð (he/it/she has), wē/gē/hīe habbaþ (we/”y’all”/they have) - and that’s just the nominative case; there were also the accusative, dative, and genitive cases, each calling for different verb inflections. But over the centuries English grammar got simplified and now conjugation is pretty straightforward - and I don’t see anyone complaining about that. Portuguese still needs to undergo a similar process. The less school time we spend learning arbitrary verb inflections the more time we can spend learning useful stuff.
In case this discussion has awaken your inner etymologist you may like to know that tu and thou are cognate - they both come from the same Proto-Indo-European word. And both have declined for the same reason: people were looking for less intimate ways to refer to one another. The English solution was to imitate the French by using the plural pronoun (you). The Portuguese solution was to imitate the Spanish by using Your Mercy - Vuestra Merced in Spanish, Vossa Mercê in Portuguese. Vossa Mercê is a bit cumbersome to say, so eventually people slurred it into você (just like Spanish speakers eventually slurred Vuestra Merced into usted).
(Btw, if you are into etymology you should check Kevin Stroud’s The History of English Podcast. It’s amazing to the point where I don’t even mind being stuck in traffic anymore.)
It would be interesting to know how usage of tu and você varies in other Portuguese-speaking countries, but only Brazil has a population big enough to yield 600k geolocated tweets each containing tu and/or você in under a week. It would also be interesting to know how tu and você usage has varied over time - in particular, when exactly você overtook tu as the most popular choice. But there was no Twitter before the 21st century (and Google Books N-Gram Viewer doesn’t have a Portuguese corpus).
Some caveats are in order. First, there may be intra-state variation. But for most cities there are too few tweets and I’d rather not say anything about how the people from this or that city speak based on only a handful of tweets. So, state capitals and other large cities are overrepresented here. Second, you need internet access and a device of sorts (a cell phone at least) in order to tweet, so the really poor are excluded from this analysis. There is some evidence that usage of tu declines with income, so I’m probably underestimating tu usage here. Third, people don’t necessarily tweet the same way they speak. For instance, perhaps the people of Rio Grande do Sul write tu more often than they speak it.
If you want the gory details, I used Twitter’s sample stream, which is a continuous stream of a small random sample of all tweets, in JSON format. Programming-wise I used Python’s Tweepy. Here’s my code (minus access key, token secret, etc; you can create your own Twitter-scraping credentials here):
This code will simply scrape the tweets and show them on your screen but won’t save them. To save, you need to point the stdout to some output file, as in:
python scrapeTwitter.py > output.csv.
As you can see from the code, I limited my scraping to posts containing tu, você, voce, or vc (a common shorthand for você in Portuguese). The Twitter API doesn’t care about case, so that also yielded Tu, TU, vC, etc.
I also used Python to parse the tweets:
(You may want to comment out line 67 -
print cleaned.encode('utf8'). Otherwise every tweet you scraped will show on your screen as it is parsed. That can make you lose faith in mankind.)
Below is the code I used to produce the choropleth. Here I only have 27 observations, so I didn’t mind hardcoding the data for each of them. But if you have more observations (say, counties) you probably want to automate the process by having a function map tu/você ratios to the RGB color spectrum.
That’s it. More map-based posts should follow.