It’s long been known that people with Alzheimer’s disease have struggles with language that worsen as the disease progresses. Now a new study shows how subtle written language impairments may begin years before an Alzheimer’s diagnosis becomes likely. The study, from researchers at IBM, found that analysis of writing samples from older men and women with normal memory and thinking skills may help predict who develops Alzheimer’s years later.
The study looked at 80 men and women in their 80s. All were part of the large and ongoing Framingham Heart Study, which has followed the physical and cognitive health of a nationally representative group of Americans for decades. Half had developed Alzheimer’s disease by age 85, while the other half remained cognitively normal.
But seven to eight years earlier, when they were in their 70s, none of them had Alzheimer’s disease. All were cognitively normal. And all had completed what’s known as the cookie-theft picture description task, which is often used by researchers to detect impairments in thinking skills.
The task requires people to describe an illustration of a domestic scene, in which a woman dries dishes at an overflowing kitchen sink, while behind her, unawares, a young boy stands on a toppling stool, reaching for a cookie jar on the top shelf while a young girl stands beside him. (You can find a copy of the picture used in the test here: https://www.researchgate.net/figure/The-standardized-Cookie-Theft-picture-Goodglass-and-Kaplan-1983_fig1_315999221 )
A person with normal cognitive skills might describe the scene this way, using language that is rich and descriptive:
“A young boy is reaching for the cookie jar. He is standing on a stool and is almost falling over. His sister is beside him and talking to him. On the other side of the kitchen their mom is wiping dishes. The water from the faucet is running over to the floor.”
Someone with impaired cognition, and at risk for dementia, would be likely to describe the scene differently, omitting verbs, linguistic articles (like “he” or “she”), capital letters, and punctuation. For example, they might write out something along the lines of:
“Boy taking cookies
Mother washing dishes
water overflowing in sink
girl getting cookie from boy
stool falling over”
Someone even more cognitive impaired, perhaps already in the early stages of dementia, might respond to the task this way, with more misspellings, missing letters and clipped descriptions:
getting cookies out of cookie ja
stoop tipping ove
water running out of sink
Girl reaching for cookie”
Using computer artificial intelligence, the researchers analyzed the participants written responses to the cookie-theft task. They looked for subtle differences in the participants’ responses to the illustrated scene, looking for cues like misspellings, telegraphed language, repetitive word use, inconsistent capitalizations, and omissions of words like “the” and “is.”
The researchers found that those who had the most of these linguistic “mistakes” in their writing samples tended to also the ones who went on to develop Alzheimer’s disease seven to eight years later. The test was 75 percent accurate in predicting who would develop dementia. The findings were published in EClinicalMedicine, a medical journal from The Lancet.
Identifying who is at risk for Alzheimer’s as early as possible, and providing therapies to alter the progression of the disease, has long been a top goal of Alzheimer’s research. Cheap and effective tests that assess language, like the picture task used in this study, could one day be used to identify high-risk populations who might benefit from new therapies. Performing a written test every 10 years or so starting in a person’s 40s might even be better, as it would provide a starting point and some clues about progression as well.
By ALZinfo.org, The Alzheimer’s Information Site. Reviewed by Marc Flajolet, Ph.D., Fisher Center for Alzheimer’s Research Foundation at The Rockefeller University.
Source: Elif Eyigoza, Sachin Mathurb, Mar Santamariab, et al: “Linguistic markers predict onset of Alzheimer’s disease.” EClinical Medicine, October 2020