Digital, Features, Languages & Literature
Leave a comment

Attention economies and the evolution of American English

Modern communication is evolving as we speak, says Thomas Hills, professor of psychology at Warwick University. The best-selling novels in Japan in 2007 were all written on mobile phones and read more like text messages than the standard chapter format many of us are used to. With text messages demanding brevity over elegance and our increasing tendency to scan headlines instead of read articles, the quality of writing continues to compress information and the ease of communicating it into ever smaller packages. 

A couple of years ago, my colleague Professor James Adelman and I discovered that the written word in American English over the past 200 years has shown a remarkably similar pattern of compression (Recent evolution in the learnability of American English from 1800 to 2000. Cognition, 143, 87–92). In lockstep with immigration (almost entirely European), American English has responded by becoming easier to learn and communicate.

To give you an example, if you compare modern writing with almost anything written more than a hundred years ago, the change in writing style is immediately obvious. In 1850, in The Scarlet Letter, Nathanial Hawthorne wrote this:

‘Children have always a sympathy in the agitations of those connected with them; always, especially, a sense of any trouble or impending revolution, of whatever kind, in domestic circumstances; and therefore Pearl, who was the gem on her mother’s unquiet bosom, betrayed, by the very dance of her spirits, the emotions which none could detect in the marble passiveness of Hester’s brow.’

That’s one sentence. In fact, I recognise it from high school as the exact sentence that made me put the book down.

In 1994, in Barrel Fever, David Sedaris wrote this: ‘If you’re looking for sympathy you’ll find it between shit and syphilis in the dictionary.’

Both of the above examples are about sympathy. But Sedaris has more sympathy for the modern reader than Hawthorne. Then again, Hawthorne’s readers didn’t have status updates and sexting to distract them, so maybe they got their kicks slogging through a bit of high literature. It’s hard to say.

One of the properties that best distinguishes the above two samples of writing is their concreteness. Concreteness is a subjective measure of a words ability to be imagined as a tangible thing. On a scale from 1 to 5, most people will agree that ‘table’ is a fairly concrete word with a rating near 5. ‘Truth’, on the other hand, is more abstract, with a rating nearer 1.  Hawthorne’s description of sympathy is highly abstract, relying on words like ‘sense’, ‘whatever kind’, and ‘domestic circumstances’. Sedaris dispenses with the hedging and instead focuses on words like ‘find’, ‘shit’, and ‘syphilis’.

Concreteness norms exist for 40,000 English words, collected from ratings provided by thousands of people. This allows us to compute the average concreteness for a typical word over large quantities of text. The figure below shows the average concreteness as it changes in American English for each year in books (Google Ngrams), magazines and newspaper articles (the Corpus of Historical American English), and presidential speeches. In each case, we average the concreteness of all words present in the concreteness norms for a given year and plot this across time. The 40,000 words in the norms capture more than half of all words in common usage, so the figure represents the predominant pattern of change in concreteness in American English

Figure 1: Changes in concreteness in American English since 1800. GN = Google Ngrams. 

Also shown are data for the words known by more than 95 percent of those filling out the concreteness norms (omitting words likely to have been lost from modern usage) and the data using only words used in the 1800 (omitting new words). COHA = the Corpus of Historical American English. Inaugural addresses show the pattern in presidential speeches.

This pattern of increasing concreteness is systemic. It is found within nouns, verbs, and prepositions. It is also found in a general tendency to omit function words, like the articles ‘the’ and ‘a’ and the particles ‘if’, ‘then’, and ‘well’. These are the same kinds of words that often get dropped in texts. Indeed, if this trend continues, we can all expect to be speaking Twitterese in another several hundred years. Many people I know already do.

American English is rising in its concreteness in every way we can measure it. But why? As we show in Hills and Adelman (2015), this pattern is not explained by gender, literacy, or a change in the concreteness of individual words. In Hills, Adelman, & Noguchi (Attention economies, information crowding, and language change. In M. Jones (Ed.), Big Data in Cognitive Science) we further show that this pattern is also not consistent with the pattern one would expect with an increase in second language learners.

Perhaps the most favored hypothesis at present is based on what many behavioral scientists call attention economics. We describe this idea in detail in Hills, Adelman, and Noguchi (2017), but the basic principle is simple. As information markets become more crowded, not all the information can survive. This creates a form of cultural selection for information that is more memorable, more easily communicated, and more rapidly understood.

Concrete words have special properties in this competition for cultural attention.  Psychological studies show that they are more rapidly comprehended, they are more interesting, and they are easier to remember.   In a competition amongst information for the minds of the populace, concrete messages have a natural advantage.

As we show in Hills, Adelman, and Noguchi (2017), this result can be easily computed from Shannon’s information entropy, but a similar result can also be found in simulations of language evolution as a function of population size (Simpler grammar, larger vocabulary: How population size affects language, 2018).

In other words, we can see empirical evidence of a selective force on cultural information that we know responds to the size of the culture in which it is created. It is interesting to speculate on the role this may play in the rise of fake news and the apparent dumbing down of political discourse. Though fake news has been around since the evolution of Batesian mimicry – when an edible insect wears the colors of his poisonous neighbors – the capacity for contemporary news to expose itself to cultural selection makes the fake news that rises to the top of this cultural foam all the more stimulating.

What one often sees in the news, and on search engines more generally, is already the result of cultural selection. This information that already received the up-vote (by click or link) of millions of other people. This is not different in kind from the forces that selected the ‘classics’ in literature or have kept the lively work of Berthe Morisot out of the most memorable impressionist painters. But there is a difference in quantity, meaning that the strength of selection is stronger because information is created by more people and passes through more hands before it reaches us. That difference is profound and it continues to shape what and how we communicate.

Thomas Hills is professor of psychology at Warwick University and director of the Bridges-Leverhulme Doctoral Training Centre in Mathematical and Social Sciences. As a behavioural and data scientist, he is interested in quantitative approaches to language, wellbeing, memory, and decision making. His work involves using ‘big data’ to understand psychological change over cultural time; understanding language learning using network analysis; computational modeling of memory representations and age-related cognitive decline; and information search in decision making. He is also a Fellow of the Alan Turing Institute where his research focuses on understanding the psychology of people in the past, such as how happy people were, using machine learning and historical text.

Share this article:

Leave a Reply

Your email address will not be published. Required fields are marked *