English scholar turns to neuroscience to detect Alzheimer’s disease by analyzing texts

Some people think an academic career in English means research on literature, but Ian Lancashire, professor emeritus of English, proves that it can take someone in other directions.

Lancashire, who joined the University of Toronto in 1968, started off as a bibliographer with Records of Early English Drama. He developed an interest in the digital humanities and eventually launched an online Early Modern English Dictionaries database with 200,000 word entries. In 2010, his fascination in neuroscience, stylistics and a desire to understand creativity led him to write Forgetful Muses: Reading the Author in the Text. That’s how he came to study Agatha Christie and discovered an effect that he linked to Alzheimer’s disease.

In collaboration with Graeme Hirst, a professor of computer science, he picked up on clues suggesting that the famed mystery novelist suffered from Alzheimer’s-related dementia in her final years – a conclusion that chimes with the view of some of Christie’s biographers. Lancashire and Hirst based their findings on an assessment of 14 of Christie’s novels and using a natural-language analysis system to measure the richness and size of her vocabulary, the increase of repeated words and of indefinite words.

Wajiha Rasul of the department of English spoke to Lancashire about his interest in neuroscience and his collaborative research.

How did you get involved in neuroscience research?

My love of poetry and interest in creativity led me to neuroscience. I educated myself by reading research articles. My full professor status allowed me to conduct long-term research projects. By analyzing Agatha Christie, Iris Murdoch, Ross Macdonald and others, I saw patterns that literary analysis alone could not well explain. Neuroscience provided answers, and in Forgetful Muses published a chapter on Agatha Christie.

In 2015 I published Vocabulary and Dementia in Six Novelists, which confirmed absolutely that longitudinal studies of writers can show Alzheimer’s disease cropping up where writing deteriorates. Vocabulary decreases, repeating phrases increase, and authors resort too much to empty words like ‘thing’ – because their memory of richer content words has been lost.

What led to research collaboration on “Longitudinal Detection of dementia through lexical and syntactic changes in writing: A case study of three British novelists”?

Computational linguist Graeme Hirst heard the first talk I gave about Agatha Christie’s results to a small group in the Bahen Centre in 2008 and volunteered to re-analyze my data with a natural-language processing system. I was very nervous about the entire venture. I am an English professor, untrained in neuroscience, and I did not want to publish unscientific results that could be seen as impugning the reputation of a major English author. The patterns I saw shocked me. I could not believe that longitudinal analysis had not been applied to authorship studies. Naturally, I jumped at Graeme's offer.

I needed colleagues with the qualifications I did not. We also needed help from the medical profession. Regina Jokel, a fine speech-language pathologist and Alzheimer's clinician at Baycrest [and assistant professor at U of T], advised us on the medical aspects of the research. Graeme found some funds at Google and applied himself and his graduate student Xuan Lee to the project. Our team of four worked really well.

The department of English had early on played an important role. They gave me funds to OCR a representative sample of Agatha Christie’s novels. Without the department’s support my data collection would have stalled.

How effective is longitudinal or change-based method in detection of dementia and where do you see its future?

It appears to be reliable when written data over some years is available. Of course, not many people publish novels annually. Yet the famous Nun Study shows that smaller sample sizes, such as diary entries, email, letters, tweets, and even recorded conversations, can be used. A noviciate nun was routinely invited to write a brief autobiography when she entered her order, and 40 to 50 years later that piece was compared to the nun's recent writings. Nuns who lacked linguistic density early on tended to develop Alzheimer’s disease in old age. The study was longitudinal and proved effective.

What are the next steps in using language impairment as an early-warning sign of Alzheimer's disease?

In my 2015 article I analyzed the novels of detective-fiction writer Ross Macdonald, who died with Alzheimer's disease, and found that his vocabulary declined markedly in his later novels. His case is even stronger than Christie's. However, Enid Blyton died with a dementia, yet the language of her children's fiction did not lose words and overly repeat phrases.

We clearly need broad-based studies in how the language of many individuals responds to aging and disease. The highly diverse cultures and languages of Canadians give researchers in the humanities here a good base for study. Longitudinal studies have appeared in Europe, which has a comparable population. Both tend to favour holistic approaches to research topics.

Any advice to English students on interdisciplinary research?

Look for faculty members who are interested in computer and language, take their courses, write a paper and propose developing new tests. My department and the Jackman Humanities Institute have been genuinely encouraging of research in this and other very experimental fields.