The study, from academics at Cardiff University, Loughborough University and the University of Oxford, used computer software to analyse the range of nouns and adjectives used in 33 of his best-selling Discworld novels.
The results show a significant decrease in the diversity of nouns and adjectives in his later works. This shift was particularly marked in the diversity of adjectives, which decreased below a defined threshold approximately ten years before Pratchett's formal diagnosis.
Sir Terry Pratchett died in 2015 at the age of 66. He had posterior cortical atrophy, a rare form of early-onset Alzheimer's disease that primarily affects visual processing.
Study co-author Dr Melody Pattison, based at Cardiff University's School of English, Communication and Philosophy, said: "Our analysis of Sir Terry Pratchett's novels suggests that subtle changes in linguistic patterns, such as decreased lexical diversity, may precede clinical diagnosis of dementia by a considerable margin. In particular we found the richness of descriptive language in his books gradually narrowed."
We would normally expect less lexical diversity as texts get longer, but even after controlling for text length our findings were still significant. This was not something a reader would necessarily notice, but rather a subtle, progressive change. --Dr Melody Pattison
[...] "Research indicates that memory problems may not be the first symptom of dementia. We wanted to explore whether language could be an early warning sign, and to do this, we used Sir Terry Pratchett's books, who himself suffered dementia.
"Our analysis found that Sir Terry's use of language did indeed change during his career. These results suggest that language may be one of the first signs of dementia, and Sir Terry's books reveal a potential new approach for early diagnosis."
Journal Reference: Brain Sci. 2026, 16(1), 94; https://doi.org/10.3390/brainsci16010094
(Score: 5, Funny) by looorg on Thursday February 12, @02:36PM (10 children)
It's very probable. Unfortunately it's not a viable diagnostic method for most people as most of us do not write much text or books amount of text at any given time or even over long periods of time.
The only drawback to this might be that it might also have been altered, and masked, by editors that might have changed his text to there by hide clues.
In hindsight, for him and others, I recon this can be used on a lot of other conditions to such as drug usage, depressions and other things.
But even if this was viable also for say verbal communication most people might not talk enough and they might not speak to the same people.
Just waiting for them to suggest that AI analysis of speech patterns will solve this problem. If the device is just allowed to monitor all our communications all the time.
(Score: 5, Interesting) by ledow on Thursday February 12, @03:29PM (7 children)
The books were also written over DECADES.
Distinguishing that condition against just natural ageing or even just a change of process (e.g. upgrading the word-processor) would be almost impossible without a huge amount of similar author's content and analysis over those decades.
I know for a fact that after 40 I've started doing far more things like substituting future words (I'm typing a sentence. The next sentence I'm planning in my head has a word. Instead of typing a similar word in the first sentence, my fingers subconsciously type the second word that I'm thinking about and I don't notice until I read it all back), etc. than ever before.
It's quite annoying as I'm a bit of a pedant, and I see more and more errors creep into my typing despite not feeling any different mentally.
(Score: 3, Interesting) by looorg on Thursday February 12, @05:24PM (1 child)
Certainly so. Having typed for a living as a researcher for over a decade I do have a certain text mass available. There is a difference between something I typed in the early 2000 vs 2010 vs 2020 vs today. Just from a sort of natural, or unnatural as academic writing is far from normal, progression or development. Words and sentences, or structure, change or become more or less common.
Also as noted tools change. If you write a lot you tend to get some better word processors then microsoft word that as noted count words and frequency, either on single words or sentences, and they suggest replacements so as to not repeat the same words and sentences over and over again.
Then there is that other aspect of collaborative professional writing where you have various guides usually suggesting words and phrases that you should use and when they should be used and such. Once again interrupting your natural flow or structure.
All these things might change the natural flow or evolution of how you write.
Personally I also find myself the older I get the more I look for the little line under the word to see if I misspelled it or not, problem is when you have two words that might only be a letter apart and then the brain just doesn't notice it at a glance. You might catch if when you read or re-read it at a later time. This is the worst. Little errors that creep in to the text that go unnoticed. That is until things are printed or released. Then you notice it, but can't do anything about it. Then it's all you see when you read that text. It's taunting you by it's very existence.
So while an interesting diagnostic technique. I have doubts on its viability for most people. Even in that regard people that write a lot. As the words might not be natural or their own. If you add in some AI writing "help" then it could be even less useful. They have already noticed that how certain words keep becoming more and more popular, that previously was not very common nor popular. But the AI likes then so eventually the typing peons adapt them to.
(Score: 2) by JoeMerchant on Thursday February 12, @06:26PM
> the older I get the more I look for the little line under the word to see if I misspelled it or not
Spellcheck has been around for 25ish years... My boss 22-ish years ago still misspelled several words in every e-mail and even formal documents he wrote. I think he did it on purpose to avoid looking too smart to his bosses. Yes, it was that kind of place: score too high on an IQ test, never get promoted into management.
🌻🌻🌻🌻 [google.com]
(Score: 2) by JoeMerchant on Thursday February 12, @06:23PM (1 child)
Your user number isn't all that much higher than mine - haven't you been "writing for" SN for 10+ years already?
My Reddit account is 13+ years old, they could run this analysis on what I put there - although some those short snippets don't always expose the range of active vocabulary very well.
🌻🌻🌻🌻 [google.com]
(Score: 3, Insightful) by looorg on Thursday February 12, @06:51PM
For me this might work. For a lot of us it might work. It is a fairly substantial and solid body of text collected over a decade or whatnot. From my user number I was probably signed up within a day or less of it going public. I don't recall now but it can't have been to long after.
That said the text here might leave some things to be desired from a professional viewpoint. It might in some regard be a lot more casual writing. I guess most message boards and such postings are. As in there is no large amount of editing involved, for the obviously technical limitations as there is no edit function.
In that regard perhaps that is even better as a diagnostic tool. More of a direct flow from the brain to the keyboard instead of having been edited over and over again.
The other issue, for me at least, is that I'm not a native English speaker. It's more of a third language. But it might be good enough due to my age and having written and spoken it for at least four:ish decades or so.
(Score: 2) by bd on Thursday February 12, @11:26PM (2 children)
You are absolutely correct.
Even more so, the raw data shows a nearly linear decrease over _all_ of his published novels, that - if anything - plateaued out when he reached his mid-fifties.
The "decline during the last 9 years" only appears after very creative application of a "cutoff threshold" that has nothing special about it in the overall curve form.
What is missing is a comparison between a statistically significant amount of prolific authors that did, and did not die from dementia.
They cited works where the same analysis was done on three authors with dementia and a single non-dementia control?!
So they found _one_ author that did not die of dementia and had constant adjective usage well into his 80's...
Why only study the works of a single author?! This would have been a 10 minute google search and some scraping of project Gutenberg at the most to find a few authors of either category.
(Score: 1) by khallow on Friday February 13, @05:35PM (1 child)
Another effect here would be the pulp mill. His first Discworld novel came out in 1983 and it was three years till the second came out. By the early 2000s (which would be where the mid-50s puts him, being born in 1948), he was cranking out almost two books a year. My take is that level of output will decrease the overall quality of the writing. In that light, he only published two books in the last four years of his life (plus whatever else he had been working on) which appears to be a significant decrease in output in addition to the alleged decline in range of language.
(Score: 2) by looorg on Friday February 13, @09:15PM
As far as I can recall from memory now his daughter said they had books, ideas and short stories for about ten or so books worth in storage. But according to his will it was all be destroyed upon his death. They bulldozed the harddrives and backup with said content after his death. He didn't want them to keep publishing his materials after his death. His daughter now runs the Discworld empire, sort of like some of Tolkiens kids does or did for his estate. I'm not sure if they have really found other authors to write new Discworld material after him.
So it could be cause he didn't want them to publish his works after his death or that perhaps there were other reasons.
(Score: 2) by bzipitidoo on Thursday February 12, @04:32PM
Heck of a future there. In addition to toilets that evaluate our stools to determine our physical health, we'll have expert AI analysis of our posts to figure our mental fitness and acuity! Your toilet will tell you that you don't exercise enough and your diet is terrible, and your social media sites will tell you that you don't do enough crossword puzzles! Compared to that, it's child's play to rip off the veil of anonymity that using a handle gives.
(Score: 4, Interesting) by JoeMerchant on Thursday February 12, @06:21PM
I was put through a psych screening as a pre-requisite for consideration for promotion some 20ish years ago - absolutely nutso expense for HR to get datapoints that any reasonably astute interviewer could get by taking the candidate to lunch, but anyway... as part of that evaluation they did vocabulary counts as well as "relatability" of the vocabulary used. On their scale "more words are better" but also controlling for whether or not the target audience is likely to understand what was said.
🌻🌻🌻🌻 [google.com]
(Score: 4, Insightful) by bzipitidoo on Thursday February 12, @04:23PM
Original:
Making your sentences more complicated with longer words to hide your developing dementia is at last a use for AI?
AI Mangled:
Making your sentences more perplexed with more sesquipedal words to obnubilate your developing dementia is at last a utilization for AI?
(Score: 0) by Anonymous Coward on Thursday February 12, @04:36PM (1 child)
This article belongs in the GNU-Terry-Pratchet dept.
(Score: 3, Informative) by janrinok on Thursday February 12, @04:49PM
Done - it might take a few minutes to appear on the site.
Thank you.
[nostyle RIP 06 May 2025]
(Score: 4, Interesting) by spiraldancing on Thursday February 12, @08:17PM
Same results ... they found a significant drop in her vocabulary, increase in repeated terms and phrases, etc. The study particularly saw the changes in one of her last books, ironically titled "Elephants Can Remember", at least hinting that Christie herself may have known she was losing something.
https://en.wikipedia.org/wiki/Elephants_Can_Remember#Literary_significance_and_reception [wikipedia.org]
Lets go exploring.
(Score: 1) by gonemissing on Thursday February 12, @11:17PM (2 children)
Pratchett's very late books became less ... fun. They still dealt with serious social issues as the previous books but were defined the characters far more black and white as good or bad and the ridiculosity disappeared, replaced by basically action novels set on Discworld.
The laugh-out-loud-while-you-read-in-bed-and-annoy-your-wife-while-she's-trying-to-sleep moments stopped happening.
I always put it down to him getting older and more conservative, but maybe he had in fact started losing his capacity for the absurd.
RIP Pterry
(Score: 2) by bd on Friday February 13, @12:08AM (1 child)
Yeah, I especially remember the Moist von Lipwig stories repeating themselves... I thought he was just running out of inspiration.
But this may have also lead to confirmation bias in the authors of the paper.
The "nine years before diagnosis" where they claim to observe a drop in adjective varieties coincides with the stories becoming less complex.
But when you look at the raw data, the biggest drop in word type variety over time actually happened before Guards! Guards!
I would naively assume that the decline should get worse the closer you get to the end, and not plateau out?!
(Score: 2) by suxen on Monday February 16, @03:03PM
> the biggest drop in word type variety over time actually happened before Guards! Guards!
So basically around the time he started getting really good. Could a drop in adjective varieties be a component of an improvement in style?