Memory Exam detects cognitive change earlier than clinical-grade written memory exams. It is given at top institutions and neurology offices in the United States. Now, you can take it online.
So, when people wonder how early is early, it's important to know they are really asking two different questions:
How early can Memory Exam detect cognitive change before symptoms appear? And, how does Memory Exam compare with other available tests? These are separate questions we will explore in detail below.
Detecting changes before symptoms
A clinical-grade written memory exam can detect cognitive change eight to 10 years before symptoms appear. Memory Exam goes further. Using patented verbal technology, Memory Exam measures not just what you answer, but how you speak. By tracking subtle vocal nuances, Memory Exam detects the first markers of cognitive decline.
Let’s look at what the research reveals.
First of all, it's difficult to compare studies on how early cognitive changes can be detected. They often involve different people and measure time in various ways. However, one study took a more consistent approach.Researchers studied 128 people with an inherited form of Alzheimer’s disease. In these families, the age at which a parent developed symptoms can help estimate when their children might develop symptoms. The researchers used that estimate as a common timeline and looked at spinal fluid and brain scans, along with thinking and memory in the same group of people.
Working backward from the expected onset of symptoms, they found this sequence:
- About 25 years before symptoms: A change in a marker of amyloid, a protein associated with Alzheimer’s disease, could be detected in spinal fluid.
- About 15 years before: Brain scans could show amyloid buildup, rising levels of tau (another protein associated with Alzheimer’s) and shrinkage of brain tissue.
- About 10 years before: Researchers found changes in the brain’s energy use and in verbal memory, which is the ability to remember information heard and recall it later.
- About 5 years before: A general cognitive score, similar to those used in brief screening tests in doctors’ offices, began to show changes.
There are important limits to these findings. The study captured a snapshot of the participants rather than following them over many years. The researchers estimated when symptoms might appear based on the age of onset in a parent; they did not observe the actual onset in each participant. And because this inherited form of Alzheimer’s is rare, the findings may not apply to the much more common form that develops later in life.
The study offers a useful look at how different signs may emerge at different times. But it does not establish exactly when a particular test will detect changes in an individual, or whether the same timeline applies to most people at risk of Alzheimer’s.
- About 5 years before: A general cognitive score, similar to those used in brief screening tests in doctors’ offices, began to show changes.
- About 10 years before: Researchers found changes in the brain’s energy use and in verbal memory, which is the ability to remember information heard and recall it later.
- About 15 years before: Brain scans could show amyloid buildup, rising levels of tau (another protein associated with Alzheimer’s) and shrinkage of brain tissue.
What additional evidence shows
Two studies of the more common, late-onset form of Alzheimer’s disease point to a similar timeline.
One followed 87 people, 27 of whom later developed mild cognitive impairment. Verbal memory began to decline about eight years before diagnosis. The other tracked 382 people as they progressed toward Alzheimer’s disease and found that learning ability showed the earliest decline, beginning eight to nine years before onset.Together, findings from the three separate groups of people listed above suggest cognitive changes are detectable eight to 10 years before symptoms appear.
How Memory Exam compares to other tests
Memory Exam uses stratified normative data rather than a single pass-or-fail threshold.
In other words, a memory score means little in isolation. To interpret it meaningfully, you need to understand how it compares with the performance of people with similar characteristics.The first is timing. Look again at the last two lines of that list: recall of spoken material and a brief overall cognitive score were measured in the same people on the same schedule, and one moved about 10 years before expected symptoms while the other moved about 5. The study of 87 people shows why. It timed each ability separately, and they did not move together. Verbal memory declined about 8 years before diagnosis, learning a list across repeated attempts and fluency about 4 years, and executive function, processing speed and the point at which a family member noticed something about 2 years. A brief screen samples a little of everything, which is the right design for the job it does, and the reason it cannot be early.
The second part is bigger, and it is about how the score is read. A brief clinic screen is scored against a fixed cut-off, the same number for everybody, with at most a single adjustment for schooling and nothing at all for age or sex. Three findings show what that costs:
- Age and education. When one widely used screen was given to 205 healthy adults aged 70 to 99, 46% of them scored below the standard cut-off. Nearly half of a group with no cognitive diagnosis looked impaired, because the threshold was never built for their age.
- Sex. On a spoken word-list test given to 985 people, scoring against norms that ignore sex produced a 20% error rate: about 10% of women were missed, and about 10% of men were flagged who should not have been.
- How much sex matters. In 4,428 adults aged 30 to 91, sex accounted for 8% to 10% of the variation in word-list scores over and above age. Norms built on age alone flagged low scores too often in men and not often enough in women.
A raw memory score means almost nothing on its own. It only becomes information next to what people of your age, your education and your sex actually achieve, and a single threshold cannot carry that. This is why the exam scores against stratified normative data rather than a pass mark, and it is what the normative study we are currently recruiting for is built to establish. There is a fuller breakdown of how the different kinds of memory test compare, and of what a memory score actually means.
Against an MRI scan: the word list won
210 older adults with no cognitive problems were followed for 7 years to see who would go on to develop mild cognitive impairment. Three things measured at the start were compared on how well they predicted it:
- Immediate recall of a spoken word list sorted people correctly about 71% of the time, where 100% would be perfect and 50% is a coin toss
- The size of the hippocampus on an MRI scan: about 66%
- A visuospatial score: about 57%
Somebody reading you a list and you saying back what you remember beat a brain scan. The researchers made the practical point themselves: the test is short, costs almost nothing, needs no special equipment, and can be delivered digitally, none of which is true of an MRI.
Against a blood test or an amyloid scan: they see it earlier
A page that sells a memory screen has an obvious incentive to leave this part out, so here it is. Amyloid in spinal fluid changes around 25 years before symptoms, and amyloid on a scan around 15. No thinking test comes close to that, and none ever will, because the brain compensates for a long time before performance gives anything away.
Blood tests have also become genuinely good. The leading one was about 85% accurate in general practice at identifying Alzheimer's changes, rising to 92% to 94% if you accept that 12% to 17% of results come back unresolved and need something else to settle them.
What those tests are for is the part that gets lost. They are meant for people who already have symptoms, ordered by a clinician who already suspects something, to work out what is causing it. They are not offered to people with no symptoms, and the people who develop them say so. They tell you about biology; they do not tell you how you are doing.
So on lead time they win, and it does not settle the question, because a test you cannot have is not an option. A recall test answers a different question, and answers it for anyone, at home, repeatedly, without a referral and without needing a reason to suspect anything.
Recognising a word is not the same as recalling it
Every finding on this page rests on one task: hearing something, then saying back what you remember, with nothing in front of you to choose from. That is called free recall, and it is the specific thing that carries the early signal.
In 161 people with no symptoms, free recall tracked the changes linked to abnormal blood markers better than recalling a story or recalling paired words did. So it is not memory testing in general that sees things early. It is unprompted recall.
Most memory tests you can take on a screen work the other way round. They show you something and ask whether you have seen it before, which is recognition, and recognition stays intact long after unprompted recall has started to slip. That is not a flaw in the idea of recognition, which is genuinely useful: our own exam measures it deliberately, because the gap between what you can recognise and what you can recall unaided distinguishes a word you never stored from one you stored and cannot retrieve. The problem is a test that measures only recognition, because it can look reassuring for years. There is a fuller explanation of what a verbal memory exam is and why the format matters as much as the content.
One number we are not using
You will see it written that fluency declines 12 years before diagnosis. We could not trace that figure to a study that reports it, so it does not appear on this page.
The fluency findings we can trace are smaller. In a long-running population study, the way people moved between ideas during a fluency task already differed from healthy volunteers 5 years before dementia began, which was the earliest point that study looked. The 87-person study above puts fluency at about 4 years. A likely source of the confusion is a well-known paper whose title names a 12-year study window, not a 12-year head start.
Fluency is worth measuring and we measure it. It is not the part that sees things first.
What a head start of eight years is for
Knowing earlier is only worth something if there is something to do with the time, and there is.
Up to 45% of dementia risk worldwide is linked to things that can be changed, which is the conclusion of the current international review and of the World Health Organization guidance that followed it. We have written about which risks those are and how much each one carries. Nearly all of them respond better to years of attention than to months.
The other thing time buys is a comparison against yourself. Read against your own result from a few years ago, a score says something no normative table can. That only works if the earlier result exists, which is an argument for testing before you have a reason to.
Two honest limits. A screen is not a diagnosis, and nothing here means a low score predicts dementia for any individual person. And these studies measured groups over years, not one person on one afternoon. What a screen can tell you is whether your recall sits where it should for someone like you, and whether it has moved since last time.
Or try the one-minute demonstration first, to see what the format is like.
Where this information comes from
Clinical and biomarker changes in dominantly inherited Alzheimer's disease. The New England Journal of Medicine, 2012. The 128-person study, and the source of the 25, 15, 10 and 5 year figures.
The 12 Years Preceding Mild Cognitive Impairment Due to Alzheimer's Disease: The Temporal Emergence of Cognitive Decline. Journal of Alzheimer's Disease, 2015. The 87-person study, and the paper whose title is the likely source of the 12-year figure.
Trajectories and timing of accelerated decline in specific memory domains preceding Alzheimer's disease. Frontiers in Aging Neuroscience, 2026.
Beyond brain scans: verbal memory testing as an efficient cognitive biomarker for preclinical Alzheimer's disease. Neuroscience, 2026. The 210-person comparison against hippocampal volume.
Age- and education-adjusted normative data for the Montreal Cognitive Assessment in older adults age 70-99. Aging, Neuropsychology, and Cognition, 2015. The source of the 46% figure.
Sex-specific norms for verbal memory tests may improve diagnostic accuracy of amnestic mild cognitive impairment. Neurology, 2019. The source of the 20% error rate.
Mayo Normative Studies: Regression-Based Normative Data for the Auditory Verbal Learning Test for Ages 30-91 Years and the Importance of Adjusting for Sex. Journal of the International Neuropsychological Society, 2021.
Free Recall Outperforms Story Recall in Associations with Plasma Biomarkers in Preclinical Alzheimer Disease. The Journal of Prevention of Alzheimer's Disease, 2024.
Plasma phospho-tau217 for Alzheimer's disease diagnosis in primary and secondary care using a fully automated platform. Nature Medicine, 2025. The blood test accuracy figures.