Word by Word: Enhancing Cognitive Testing in Parkinson’s Disease

Words are more than vehicles of thought—they are markers of brain health. In this perspective, Atlantic Fellow Adolfo García and biomedical engineer Franco Ferrante dissect their research on how verbal fluency tests reveal cognitive changes in Parkinson’s disease.

List as many animals as you can in one minute: “dog, cat, elephant, dolphin…” It sounds simple, almost like a party game. Yet, for scientists, this task opens a window into the brain. The words you choose have measurable information, such as their length, how specific they are, and how they cluster together. Such features can reveal a great deal about your cognitive health.

This is what a new international study has uncovered for people living with Parkinson’s disease (PD). Researchers found that by analyzing the properties of words patients produce in quick verbal tests, they can predict their cognitive status with surprising accuracy, in a fully automated way. This novel framework could change how we detect and monitor cognitive deficits in one of the world’s fastest-growing neurological conditions.

The Mind Beyond Tremors

PD is typified by motor symptoms, like tremors, stiffness, and slowed movement. Yet, for many patients, the disease brings another, less visible challenge: cognitive decline. Attention, memory, and mental flexibility can be affected to varying degrees, with up to 57% of patients developing mild cognitive impairment (MCI)—a state where these domains become significantly affected even though patients have not developed dementia.

Cognitive symptoms can impact daily life, communication, and emotional well-being. They also put financial and caregiving strain on families. Yet, detecting them remains a challenge. Traditional tests are lengthy, subjective, and dependent on highly trained professionals—resources that are scarce in many parts of the world, especially in low-income countries. The question is pressing: How can we spot these changes earlier, faster, and at scale?

A novel study, published in Movement Disorders, turned to a short task already used in neurology clinics: verbal fluency tests. In these, participants are asked to generate as many words as possible in one minute, either within a category (“animals”), a theme (“supermarket items”), or starting with a specific letter (like “p”).

Traditionally, clinicians just count how many valid words someone produces, but the research team saw untapped potential. Using digital tools, they dissected each word across six dimensions:

1. Semantic variability: how much the words shift in meaning from one to the next.
2. Granularity: how specific the word is: “dog” vs. “bulldog”.
3. Concreteness: how tangible the word’s referent is: “apple” vs. “justice”.
4. Frequency: how often the word is used in everyday speech.
5. Phonological neighborhood: how many words sound similar.
6. Length: number of sounds in the word.

By feeding these variables into artificial intelligence algorithms, they could detect patterns invisible to human raters.

Putting Words to the Test

The project involved 464 Spanish-speaking patients with PD, one of the largest samples of its kind. In a first study, the team used word properties from the verbal fluency test responses to predict scores on the Mattis Dementia Rating Scale (MDRS)—a standard clinical test for assessing cognitive function. The results were striking: the automated predictions strongly matched patients’ actual MDRS scores, even after accounting for disease severity, duration, medication, short-term memory, attention and working memory scores.

In a second study, the researchers focused on 247 patients carefully split between those with MCI and those without. Here, statistical analyses distinguished the two groups with great precision. PD patients with MCI tended to produce words that were less varied in meaning, less specific, and less concrete. In other words, their verbal “maps” were flatter and less detailed. Furthermore, artificial intelligence algorithms differentiated both groups with an accuracy that outperformed traditional cognitive tests. Even when applied to new groups—including patients who had undergone brain surgery—the approach held up. This suggests the method is robust and generalizable.  

Why do these word properties reveal so much? The answer lies in the brain’s semantic networks where vocabulary choices are handled. Producing a word requires navigating a mental map of meanings. Healthy individuals can shift flexibly between different “neighborhoods” in this map, producing a mix of general and specific terms. While in PD patients, certain brain circuits that support this navigation—especially those tied to semantic memory—become less efficient. Patients tend to stick to words that are familiar, common, vague, and concrete— arguably to reduce cognitive effort—with the magnitude of these linguistic preferences mirroring the severity of the underlying cognitive decline. 

The approach is fast, scalable and reproducible—pointing toward a future where a simple, smartphone-based test could help detect cognitive decline in Parkinson's patients anywhere in the world.

—Neuroscientist Adolfo García and Biomedical Engineer Franco Ferrante

Why Words Matter 

These findings may sound technical, but their implications are deeply practical. First, the approach is fast. Unlike other neuropsychological evaluations that can take half an hour or more, this method only requires about three minutes of speech. That means less fatigue for patients and more efficiency in crowded clinics. Second, it is scalable. Because the analysis is fully automated, it does not rely on the availability of neuropsychologists to rate, score, or perform analyses—a major bottleneck in many low- and middle-income countries, where most patients live but research investment is scarce. Third, it is reproducible. By focusing on measurable word properties rather than human judgment, the method reduces variability between examiners. Taken together, these strengths point toward a future where a simple, smartphone-based test could help detect cognitive decline in PD patients anywhere in the world. Plus, this approach has the potential to enhance precision medicine—tailoring treatment to each patient’s unique cognitive profile.

In the end, our words are more than vehicles of thought. They are markers of brain health—signals that can guide timely intervention and lighten the burden of disease. For the millions living with PD, and the millions more who will face it in the future, this is not just a scientific breakthrough. It is a message of hope, delivered one word at a time.

Publication

Ferrante, F.J., Escobar Grisales, D., López, M.F., Lopes da Cunha, P., Sterpin, L.F., Vonk, J.M.J., Chaná Cuevas, P., Estienne, C., Hesse, E., Amoruso, L., Orozco Arroyave, J.R. and García, A.M. (2025), Cognitive phenotyping of Parkinson’s disease patients via digital analysis of spoken word properties. Mov Disord. https://doi.org/10.1002/mds.70005