Hey smart people, Joe here.
In your whole life, how many books have you readed?
Sorry, I mean read. But not red, like the color. Read, like the
past tense of reeead. I just misspeaked! Mispoke. This… reminds me of a poem:
The verbs in English are a fright.How can we learn to read and write?Today we write,
but first we wrote;We bite our tongues, but never bote.This tale I tell; this tale I told;I
smell the flowers, but never smold.If I still do as once I did,Then do cows moo, as they
That was penned by linguist Richard Lederer. And it’s proof that English is…weird.
We can blame all this confusion on irregular verbs.
Most verbs in English are “regular”. We make their past tense by adding a letter or
two on the end. They’re the difference between what happens now and what happened. But irregular
verbs are… well, not regular. Like the difference between what is and what was.
It’s cute when kids say “I breaked my toy.” But why do the rest of us say “broke”?
Because that’s just what everyone else says, right? We say it how it’s always been said.
But if we were thinking scientifically, we’d ask “How did it get this way?”
And I don’t know about you, but I prefer to think scientifically.
A biologist studies how things are by looking at how they used to be. We find fossils. But
how does one go about finding a fossil… of language?
Well luckily, people tend to write language down.
James Joyce’s Ulysses contains 265,222 words I totally counted, and didn’t just google
Of those words the word “time” is the 74th most frequent, used 376 times.
The word “the” is the most frequently used: 14,877 times. We know that thanks to
another type of book: a “concordance”, an index of words that lists every instance
of every word in a written work.
There’s concordances for Thoreau’s Walden. He enjoyed the woods more than the forest.
The poetry of Edgar Allan Poe, where we find the raven more than Eldorado.
The writings of Descartes (in the original French), medieval recipes, even for the Bible.
A linguist named George Kingsley Zipf looked at these ranked lists of written language
and noticed something funny: Not all words are created equal. Some get used a lot, while
most almost never get used. Like how we say “the” all the time, but almost never say
“hallux” –the anatomical name for your big toe.
When it comes to a trait like height, most people are pretty close to average, while
the very tallest people? Are only maybe three times taller than the shortest. We don’t
vary very much. Height is… normal, it’s literally a “normal distribution.”
But Zipf realized words aren’t normal. Only a few words are very common, while most words
are very un-common. For instance, in Ulysses, there are a thousand words used more than
26 times, a hundred words used more than 265 times, but only ten words used more than 2,653
times. Another way to say this: the 10th most frequently used word is ten times more common
than the 100th most used. This peculiar trend is called Zipf’s Law.
I’m Tacky! It looks like you’re talking about Zipf’s Law. Did you know Vsauce already
did a video about that?
Yeah, it’s a great video… it’s actually what got me thinking about this! But I’m
gonna tell them more than just about Zipf's Law. I want to…
Would you like me to help you click over to that video…
No! I want you to watch THIS video. But if you DID watch Michael’s video on Vsauce,
perhaps by clicking a link in the description–LATER–you’d learn that Zipf’s Law applies to tons of
stuff: Like wealth, the population of cities, how long audiences clap, web traffic, the
size of holes in Swiss cheese, and–especially–language.
Wherever people look, newspapers, other languages, even randomly generated words, pretty much
everything in language obeys Zipf’s Law… well, everything except irregular verbs.
The 12 most common verbs in the English language are be, have, do, say, get, make, go, know,
take, see, come, and think. All irregular. But irregulars are a tiny fraction of all
verbs. English only has around 200 irregular verbs, a mere 3 percent of total verbs.
Instead of having a few commonly used irregular verbs, and lots of rare ones, like Zipf’s
Law predicts, almost all irregular verbs are common, and almost none are rare. Irregular
verbs… are a Zipf exception.
Where do irregular verbs come from? They’re the oldest ones we have. Around
four to six thousand years ago, people stretching from Europe to Western Asia spoke an ancient
language known as Proto-Indo-European. A staggering number of modern languages descend from this.
In PIE, the meaning and tense of words could be changed through a system where vowel sounds
were swapped. This system, the ablaut, can still be heard today in irregular verbs: Dig,
dug. Sing, sang, sung.
At the time, it was just one of many competing systems for changing verbs. But a bit later,
people speaking Proto-Germanic, a dialect descended from PIE, began adding verbs to
the language that didn’t fit these old patterns, so they invented a new way of signifying the
past tense by simply adding “-t” or “-ed” sounds to the end. Back then, these new “regular”
verbs were actually the exception.
As English grew from this Proto-Germanic language, newly added words became automatically regular,
they followed this new rule. And many older verbs began to switch from the old way to
the new. Like how long ago, the knight slew the dragon, but Beyoncé slayed at her last
By the time the Old English story of Beowulf was written, three out of every four verbs
had been “regularized.” There /were/ a handful of verbs that moved in the other direction,
going from regular to irregular, but for every havED or makED that was had or made, there
are dozens of verbs like holp that got helped along. Regular was no longer the exception,
it was the rule.
So why did some irregular verbs go extinct, while others have survived? We all know that
language evolves, similar to how living things do, changing slightly over time. Could language
also undergo some kind of natural selection, is there something about a word that decides
whether it’s strong enough to live on?
We can test this! We just need a bigger data set than one book.
Using ancient grammar textbooks along with databases of millions of written words, researchers
tracked the evolution of 177 verbs that were irregular at the time Beowulf was written.
By the time Chaucer wrote Canterbury Tales, 32 of these had become regular. By the time
we hit modern English, 79 had regularized. The trait that predicted whether or not a
verb would become regular was how often we use it.
The most frequently used verbs tend to stay irregular. The most rarely used become regular.
Surprisingly, there was a sort of hidden Zipfian pattern there after all. If a verb is used
100 times less frequently, it will regularize 10 times as fast. If they’re used 10,000
times less frequently, they’ll regularize 100 times as fast.
Researchers were able to estimate the likely lifespan of irregular verbs. A word like “stink”,
that’s used once every 10,000-100,000 words, has a 50% chance of regularizing within 700
years. Drink, a more common word, will take more like 5,000 years. We can find words today
in the process of going extinct. Do you tend to say dived or dove? Now is your last chance
to be newly wed. Pretty soon, you might be newly wedded. “Wed” is the irregular verb
we think will most likely disappear next.
This seems to be natural selection for language. Usage frequency affects a word’s survival,
and this makes sense. Regular verbs follow a rule. When we encounter a word we don’t
know, we can still figure out its past tense, without memorizing each and every one. Irregular
verbs on the other hand, have to be memorized. If we don’t use them, we lose them. As they’re
slowly forgotten, the “regular” rule is used in their place.
In 1980, after thirty years of work, IBM was able to digitize the complete works of Thomas
Aquinas. Today, this is something that you or anyone who knows how to code, can do in
a few minutes, with a few keystrokes. Concordances, the indexes of language that inspired Zipf
and others to ask these questions, no one really writes those anymore. Except… maybe
they do. It’s called “Google”. A search engine is basically a list of words and phrases,
from around the web, and the pages where they appear. Concordances were just analog Google.
The Google Books project now contains 25 million scanned books stretching back more than 500
years. No matter how many books you read, you could never read every book, or even a
fraction of them, in a lifetime. If you tried to read just the English-language books from
the year 2000 in this collection, at a reasonable pace, without stopping, it would take you
But what could we learn if we made computers read for us? The Google Ngram Viewer is a
search tool we can use to study how human culture has changed over the centuries. It
plots the frequency of strings of one or more words, by year, found in those millions of
We can see when people stopped talking about the Great War, and started calling it World
War I instead. “Evolution” was on the decline until “DNA”
came along. Einstein took physics to the next level.
People like pizza more than hamburgers, but less than ice cream.
What’s the most interesting one you can find?
Of course, as much data as we can pull from millions of digitized books, we haven’t
read them. A computer has. And while it gives us access to an immense amount of data, it
doesn’t tell us perhaps the most important part: The story.
If you thought that Ngram was pretty cool … Sarah over at Art Assignment used it to look at how and different artists
got famous… or not. Link description to that one too.