What do words like waddle, slobber, tingle, oink, and zigzag have in common? These words sound funny, but they are also iconic, with forms that resemble aspects of their meanings. In a new paper we investigate the link between funniness and iconicity in 70,000 English words.
“This is play”
The starting point is a theory about metacommunication: some words (or signs) are more striking than others in terms of their form, which means they draw more attention to themselves and signal “this is play”. We think this explains the first finding of the paper: words that people rate as highly funny are also often rated as highly iconic.
To test how general this finding is we developed a way to predict funniness and iconicity ratings for new words. Based on semantic relationships between millions of English words, we trained an algorithm to predict the iconicity (or funniness) of words that have already been rated by people, and then asked that algorithm to predict iconicity (or funniness) for new words.
For example, say the new word is ‘waggle’. First the algorithm learned that ‘waggle’ occurs in similar contexts to ‘wiggle’ and ‘wobble’. Then it learned that ‘wiggle’ and ‘wobble’ were rated as highly iconic by participants. As a result, it predicts that ‘waggle’ will be highly iconic too. Applying this method to ~70,000 words, we find that the relation between funniness and iconicity holds even for predicted ratings.
But what is it about some words that makes them both funny and iconic? Analysing the words that people rated as most funny and iconic, we found a number of recurring features: complex sequences of sounds at the start like str- and cl- or at the end like -nk and -mp, and an ending -le in verbs that contributes an element of movement and playfulness (as in ‘waggle’ and ‘wobble’).
These structural features, we propose, act as metacommunicative signals that help words stand out as playful, performative, and even poetic. They occur disproportionally in highly rated words. When we combined these cues in an overall structural markedness score, we found structural markedness predicts the iconicity and funniness ratings much better than other measures.
So our three main findings are:
- Words that are rated as highly iconic also tend to be rated as highly funny (in the few thousand words for which we have such ratings)
- This relation holds even in for ratings predicted based on semantic relationships (in ~65.000 words for which we have done this)
- The highly rated words tend to have special forms: they sound different from other words, which invites people to treat them as playful and performative
Making sense of apparent exceptions
We also found some other things. First, funniness and iconicity ratings do not always go hand in hand. There are highly iconic words like ‘roar’ and ‘scratch’ that people don’t feel are funny because they have to do with negative events. There are also words that are rated as very funny like ‘blonde’ and ‘buttocks’ mainly because they tend to be used in jokes; these are not rated as iconic and they are not relevant for our theory.
Another thing we found is that human ratings are far from perfect. As it turns out, for the data we used, the people who rated words for how much they “sound like what they mean” gave high ratings to words like ‘whoosh’ (where the sound of the word resembles aspects of its meaning) but also to words like ‘bedroom’ (which are built by combining meaningful parts).
Only words of the first type are really iconic; the others are merely analysable. Our theory holds only for the first, which means that the 10-15% of analysable words with high iconicity ratings are probably diluting the effects we find. Indeed, when we control for this issue by looking only at words of one piece, the relation between iconicity and funniness comes out a little stronger.
We included this analysis not just to show the subtleties of the effects, but also because we believe lexical ratings (whether done by people or by machines) should never be taken at face value. Now that there are so many types of ratings available, it’s tempting to just throw together a bunch of them and have a look at correlations. But to avoid cherry-picking or reporting false positives, it is important to start with a theoretical question, and to always control the findings with other methods.
Having fun with linguistics
While the study is based on English, its questions are inspired by work on ideophones, highly evocative words found in many languages around the world. And the theory put forward in the paper is general enough to help account for many other examples of playful language described in the literature, and to guide future investigations of the relation between playfulness and iconicity in spoken and signed languages.
Our study also contributes to broadening the perspective of linguistics. While anecdotal reports about perceptions of funniness and iconicity abound, our study is the first to investigate this relation on a large scale in English, and perhaps in any language. That this hasn’t been done before is partly because linguistics has long preferred to focus on “serious” matters. However, we argue that there is nothing frivolous about studying playful language.
Cybernetician Gregory Bateson argued that the very notion of play represents a fundamental transition in the evolution of communication. This is because play requires a form of metacommunication, a way of saying “What I do now is special”. Human language has perfected such forms of metacommunication, and in our paper we trace its influence in the very texture of the lexicon.
To enable others to build on our work we’ve made sure it is open science all the way: all primary data as well as our new predicted iconicity and funniness ratings are publicly available. We also share the Python code for our prediction algorithm and the R code for all of the analyses and figures. And last but not least, the paper itself is also published open access.
- All data and code is in our GitHub repository
- Dingemanse, M., & Thompson, B. (2020). Playful iconicity: Structural markedness underlies the relation between funniness and iconicity. Language and Cognition, 1-22. doi:10.1017/langcog.2019.49