"Fifty-seven years in this business, you learn a few things. You know what words are funny and which words are not funny," says the comedian, played by Walter Matthau. "Alka-Seltzer is funny. You say 'Alka-Seltzer,' you get a laugh... Casey Stengel, that's a funny name. Robert Taylor is not funny. Cupcake is funny. Tomato is not funny... Cleveland is funny. Maryland is not funny. Then there's chicken. Chicken is funny. Pickle is funny."
And it's true! If you need a place name for a punchline, you're guaranteed to kill with Kalamazoo, Schenectady or Rancho Cucamonga. But why? Psychology professor Chris Westbury at the University of Alberta has a fascinating theory, and it's based on perhaps the two unfunniest words in the English language: statistical probability.
Westbury published a paper in October 2018 in the Journal of Experimental Psychology with the first-rate title, "Wriggly, squiffy, lummox, and boobs: What makes some words funny?" In it, he started with a list of the 5,000 English words rated funniest by real humans and constructed a working mathematical model for predicting the laugh factor of nearly every word in the dictionary.
When Westbury applied his model to a dataset of 45,516 English words, it decided that these 10 words were the funniest of all: "upchuck, bubby, boff, wriggly, yaps, giggle, cooch, guffaw, puffball, and jiggly." Runners-up included "squiffy, flappy and bucko" and the perennial favorites of every 8-year-old on the planet: "poop, puke and boobs." On the other end of the spectrum, the word found to be the absolute least funny was "harassment."
In his paper, Westbury explains that philosophers have been trying to unravel the mystery of humor for millennia. Plato and Aristotle weren't big fans of humor, seeing it mostly as a way of denigrating and feeling superior to others. Cicero introduced "incongruity theory," writing that "the most common kind of joke [is when we] expect one thing and another is said; in which case our own disappointed expectation makes us laugh."
While the incongruity theory of comedy makes perfect sense — even orangutans find switcheroo tricks high-larious — Westbury says it's not a true scientific "theory" in that clearly not every incongruous event is as funny as another. A random coughing fit in a crowded movie theater isn't nearly as comical as a random farting fit. (I mean, just try to say "random farting fit" without smiling.) So the goal of Westbury's modeling experiments was to go beyond philosophical theorizing and come up with a truly quantifiable scale of funny.
The Math of Humor
To do it, Westbury analyzed words in two different ways: by their meaning and by their form. For the first analysis, the researchers looked at "semantic predictors" that group words with similar meanings. Using a free tool developed by Google that identifies words that are commonly used for one another (co-occurrence), Westbury mapped out the semantic relationships between 234 of the funniest human-picked words. From this "correlation plot," the researchers identified six different clusters or categories of funny words: insult, sex, party, animal, bodily function and expletive.
Now this is where things get dangerously mathematical. Since many of the words on the human-rated funny list fell into more than one category, the researchers needed a more precise measurement of how a word's meaning translated into comedy. Using the Google tool, they came up with lists of words most closely related to each of the six categories. Then they came up with average values for each of those word categories using something called linear regression analysis. Those average values for each category — insult, sex, expletive, etc. — became known as "category-defining vectors."
When looking specifically at meaning, it turns out that the funniest words don't necessarily fall cleanly into the most categories, but are the words whose mathematical values are the closest, on average, to those six category-defining vectors. Confused? Here's how Westbury summed it up in a press brief: "The average similarity of a word's meaning to these six categories is itself the best measure we found of a word's funniness, especially if the word also has strongly positive emotional connotations."
But meaning is only one type of measurement. Westbury and his team looked at the form of funny words, things like word length or the individual sounds (phonemes) that make up each word. In this second analysis, the data fit nicely with the incongruity theory of humor. It turns out that the fewer times a word or its phonemes appear, the funnier we think they are. That helps explain why there are so many "k" and "oo," sounds in the funny word lists. They're statistically improbable. Words ending in "le" (like waddle" and "wriggle") were another source of fun, suggesting as the study puts it, "repetition, usually with a diminutive aspect."
So, Why Are We Laughing?
Now this is where things get really interesting. The human brain, it seems, is running all of these complex mathematical models all the time without any of us knowing it. As we watch TV and read and talk to people, our brains are constantly parsing language for subtle semantic cross-connections and statistical probabilities. And the result — at least on this basic, one-word level — is what we call humor.