How subtle changes in speech can help track cognitive health over time.
Speech contains subtle signals about how the brain processes information
By regularly analyzing speech, changes over time can become visible
The speech test is based on scientific research and is used as a digital biomarker
The focus is on how someone speaks, not what they say
Within Remind, speech is used for monitoring and context, not for diagnosis
Speech seems obvious, but it is neurologically complex. While speaking, multiple brain areas work together, including those for language processing, planning, memory, and motor skills.
Scientific research shows that with cognitive decline, very subtle changes often occur in speech, long before daily problems become apparent. Think of changes in tempo, pauses, or word usage.
Precisely because speaking is an everyday activity, speech lends itself well to tracking changes over time.
In a speech test, a short (60 seconds) spoken assignment is recorded. That recording is then analyzed using algorithms trained on large datasets.
Among other things, the analysis looks at:
speech rate and variation
pauses and patterns of silence
fluency of sentences
repetitions and self-corrections
These characteristics say something about information processing and cognitive load.
It's important to note that a single measurement tells little. The value lies mainly in repetition and comparing measurements over time.
Research into digital biomarkers shows that this longitudinal nature contributes to the early detection of changes in brain function.
What is assessed:
How someone speaks
Structure and rhythm of speech
Changes compared to previous measurements
What is not assessed:
The content or meaning of what someone says
Emotions, opinions, or intentions
Individual diagnoses or conclusions
The speech test is not a listening tool where conversations are recorded and analyzed. The analysis is entirely focused on patterns in speech structure.
Speech analysis as a digital biomarker has been researched for some time within academic and clinical settings.
Research shows that speech patterns can be associated with cognitive changes. Research into the use of speech as an objective and scalable biomarker is also being conducted at academic centers, including Amsterdam UMC.
A 2025 PhD dissertation by Roseanne van den Berg (Amsterdam UMC) examines digital signals that may contribute to tracking cognitive functions over time.
Moreover, speech analysis aligns with broader developments in brain age and digital biomarkers, where multiple measurement methods are combined to better understand changes.
The Amsterdam UMC has established in a recently published study that speech analysis
is a promising way to detect early signs of cognitive decline.
Behind speech analysis lies a so-called algorithm or model. You can see it as a computational method that learns from examples. Instead of following fixed rules, the model is trained with large amounts of speech recordings from different people, linked to known outcomes from research and care.
With the speech model applied by Remind from Canary Speech, it works as follows: the model analyzes thousands to millions of speech fragments and learns to recognize patterns in, among other things, tempo, pauses, variation, and articulation. These patterns are then compared with data from people with and without cognitive disorders. Thus, the model learns which combinations of characteristics occur more frequently with certain forms of cognitive change.
Importantly, the model does not listen to the content of what someone says, but to how someone speaks. The meaning of words is therefore not relevant, only the way in which speech is produced.
Applied more widely than just dementia and Alzheimer’s
In research and specialized care contexts, this type of model is studied and applied in various neurological and psychiatric disorders, including Alzheimer's, mild cognitive impairment, and Parkinson's. Within Remind, this technology is used exclusively for non-diagnostic monitoring of changes over time.
Applications in clinics and hospitals by Microsoft
This technology is now also used in clinical practice. In the United States, speech analysis is applied within medical documentation, for example via Microsoft DAX Copilot. There, speech technology is used to assist doctors during consultations, with research also investigating how speech characteristics can contribute to recognizing cognitive load and change in patients.
Within Remind, the same scientific foundation is applied, but in a more accessible, repeatable form. By analyzing speech at multiple points in time, insights into changes over time are gained, which aligns with how cognitive decline usually develops: gradually and subtly.
Example of the speech analysis in the Remind app
Remind combines multiple domains to better understand changes in brain health.
Speech analysis is never a standalone signal within Remind. Changes in brain health often manifest on multiple levels simultaneously. That’s why Remind deliberately chooses a multi-modal approach, combining speech with other measurements.
Speech patterns are examined in conjunction with lifestyle, memory tests, and other cognitive signals. This prevents conclusions from being drawn based on a single measurement method and helps to better understand subtle changes. The strength lies not in one test, but in bringing together multiple perspectives over time.
Speech is a natural, daily function in which the brain is continuously active. By analyzing speech in a careful and scientifically substantiated manner, an additional way emerges to monitor changes in brain health over time.
Within Remind, speech is therefore used as a digital biomarker: supportive, contextual, and always in combination with other insights.
Are you interested in the other assessments we offer at Remind? Read more here:
Frequently Asked Questions
