Memory and Cognitive Decline
Losing a word is not losing your mind.
Most people who worry about their memory do not have a neurodegenerative disease. A meaningful number of them have something treatable that nobody has looked for. And a smaller number have something serious that would benefit enormously from being found early. Those three groups are separated by pattern, not by how frightening the symptom feels, and this page is about how that separation is actually made.
Begin reading
How a memory is made
These fail differently and the difference matters more than the severity.
The spectrum
Subjective cognitive decline, with normal testing
Mild cognitive impairment, with independence preserved
Dementia, where independence is affected
The line is drawn at daily independence, not at how bad it feels. Progression is not inevitable.
There is a particular kind of fear that comes with forgetting. It is not like other symptoms, because the organ you would use to assess it is the one you no longer trust. That fear stops a lot of people from getting assessed, which is exactly backwards.
Assessment is how the treatable causes get found. Waiting is how they get missed.
This page covers what memory actually is, and how it fails. What ordinary ageing looks like, and what does not. Where the lines sit between noticing a change, mild cognitive impairment and dementia. The causes that are genuinely reversible, and why delirium is an emergency. How the major dementias begin differently. What testing settles, and what the evidence says about lowering risk.
The central idea
Cognitive decline and dementia are not the same word, and the difference is independence.
Which is a clinical judgment, made with a person, not a score on a screening test.
Free guide
Get Brain on Fire
The guide explains how to work out which cognitive function has changed, and what belongs in a proper history. It covers the foundational tests that find the reversible causes, before anyone reaches for anything more specialised.
Brain on Fire
What separates normal ageing from a change worth investigating.
Three steps, three failures
Memory is not one process.
Information has to be attended to and encoded, consolidated into something durable, and then retrieved. Those are separate operations supported by different systems, and which one is failing tells you more than how often it fails. This is the distinction that a fifteen minute appointment most often skips and that changes the interpretation completely.
The three stages
What failure looks like
Encoding: the conversation never went in at all
Consolidation: it went in and did not stick
Retrieval: it is in there and will not come
The test that separates them is whether a cue brings it back. If a prompt produces the memory, that is retrieval, and retrieval failure is common and usually benign. If prompting does not help, that is encoding, and encoding failure is the pattern that changes the questions being asked.
Attention deserves particular emphasis, because it sits upstream of everything else. If attention is compromised by poor sleep, pain, anxiety, medication or simply doing three things at once, encoding fails, and the person experiences that as a memory problem. It is not one. Fixing attention fixes it.
The baseline everyone needs first
What ordinary ageing actually does.
Some cognitive change with age is expected, well described and not a disease. Processing speed slows from early adulthood onward. Retrieval becomes less reliable, which is why names and nouns go missing and then arrive twenty minutes later. Dividing attention between several tasks gets harder. Learning something genuinely new takes more repetitions than it once did.
Some things do not change in ordinary ageing. Holding onto new information once it is learned. Finding your way somewhere familiar. Managing your own affairs, and recognising the people in your life. Vocabulary and accumulated knowledge often keep improving into the seventies.
This distinction is not cosmetic. It is the first sorting step in any competent assessment, and it is the reason a person can be genuinely worried and genuinely fine at the same time.
Slower retrieval is ageing. Information that never went in is a different question.
Expected against concerning
Where the line sits
Expected against concerning
Expected with age
Processing speed slows
Names arrive late, but they arrive
Dividing attention gets harder
Learning new material needs more repetitions
Things get misplaced occasionally
Worth assessing
New information does not stick even with prompting
The same question repeated within one conversation
Getting lost in a familiar place
Tasks that used to be automatic becoming difficult
A change in judgment, personality or social behaviour
Family noticing before the person does belongs firmly in the second list, and it is one of the more useful signals in the whole assessment.
Three labels, one continuum
Where the lines are drawn.
Subjective cognitive decline means you notice a change and formal testing does not detect one. It is not nothing. On average, groups of people with persistent subjective decline progress at a somewhat higher rate than groups without it. That is a statement about groups, not a prediction about you. It is a reason to look for causes, not a reason to assume the worst.
Mild cognitive impairment means testing detects a change beyond what age explains, while daily independence is preserved. This is the category most often misunderstood. A meaningful proportion of people with mild cognitive impairment remain stable for years, and a proportion revert to normal, usually because something reversible was driving it. Others progress. Which of those happens depends heavily on what caused it.
Dementia is not a specific disease. It is a level of severity, defined by cognitive decline in at least one domain that is significant enough to interfere with independent function. The disease causing it is a separate question, and that question has several possible answers.
Mild cognitive impairment is not early dementia. It is a description of where someone is now.
The three tiers
Defined by function
The three tiers
Subjective cognitive decline
You notice a change and formal testing is normal
Mild cognitive impairment
Testing detects a change beyond what age explains
Daily independence is preserved
Dementia
Decline significant enough to affect independent function
Which disease is causing it is a separate question with several possible answers
Movement along this spectrum is not one directional. People revert, particularly when something reversible was driving the impairment in the first place.
The one that is missed most
Delirium is not dementia, and it is urgent.
Delirium is an acute disturbance of attention and awareness that develops over hours to days and fluctuates through the day. It is caused by something: infection, medication, dehydration, metabolic disturbance, pain, alcohol withdrawal, or an underlying illness that has not been found yet. It is common in hospital, common in older adults, and it is regularly written off as confusion or as dementia getting worse.
Delirium against dementia
Delirium
Onset over hours to days
Fluctuates markedly through the day
Attention is the core deficit
Dementia
Onset over months to years
Relatively stable from one day to the next
Memory or another domain leads the picture
Two presentations
Hyperactive: agitated and restless, and usually noticed
Hypoactive: quiet, withdrawn and sleepy, and mistaken for tiredness or low mood
Delirium is a medical emergency. It needs the cause found urgently, and having dementia already does not make new confusion expected. It makes delirium more likely, not less important.
Two points that change outcomes. Delirium often takes weeks to clear fully even after the cause is treated, so someone still not themselves a fortnight later has not necessarily declined permanently. An episode is also associated with worse cognitive trajectories afterward. That is the argument for preventing it. Review medications, treat infection early, protect sleep, and keep glasses and hearing aids in place during illness.
Look here before anywhere else
Causes that are reversible, and causes that are additive.
A minority of cognitive impairment reverses completely when a single cause is corrected, and pages implying otherwise are selling something. What is far more common, and more useful, is that several of these are present at once and each contributes a slice. Correcting three can move a person from struggling to functioning without any one of them having been the answer. This is the part of the workup where thoroughness earns the most, and the part most often abbreviated.
The single highest yield review
Anticholinergic load in particular, which accumulates quietly across bladder medications, older antihistamines, some antidepressants and sleep aids. Benzodiazepines, opioids, sedating over the counter products and polypharmacy in general all matter. Higher cumulative anticholinergic exposure is associated with worse cognition and with dementia risk in large cohorts. Nothing on this list is faster to check.
Common, treatable, routinely missed
Repeated overnight oxygen dips and fragmented sleep impair attention, encoding and executive function. It is associated with cognitive decline, and treating it improves daytime cognition in people who adhere to therapy. Snoring, witnessed pauses, unrefreshing sleep and morning headache should trigger a study, not a supplement.
It impairs cognition directly
Depression slows processing, weakens concentration and disrupts encoding, and in older adults it can be severe enough to look like dementia. That presentation is treatable. Late life depression also raises subsequent dementia risk, so treating it is worth doing on both counts rather than being framed as ruling something out.
And the reference range problem
Deficiency causes cognitive change, and neurological damage can precede any anaemia. Levels in the low normal band can still be functionally deficient, which is where methylmalonic acid and homocysteine earn their place. Metformin and long term acid suppression both lower absorption, and both are common.
Both directions, both correctable
Hypothyroidism slows processing and dampens mood and energy. Thyrotoxicosis in older adults can present as apathy and confusion rather than agitation. One blood test opens the question. Thyroid Health covers what the numbers mean.
The largest single modifiable risk factor
Untreated hearing loss is the strongest of the identified modifiable contributors to dementia risk at a population level. Mechanisms under discussion include cognitive load, reduced input and social withdrawal. A large trial found hearing intervention slowed cognitive decline in participants at higher risk, though not in the lower risk group. Testing hearing is not a niche step.
Dose dependent and often understated
Heavy intake causes direct cognitive impairment and, through thiamine deficiency, can cause an abrupt and partly preventable amnestic syndrome. Alcohol also fragments sleep and interacts with medication. Intake is systematically underreported, so this deserves an actual conversation rather than a tick box.
Cumulative, silent, modifiable
Diabetes, hypertension in midlife, dyslipidaemia and smoking damage small vessels over decades, and vascular contribution is present in a large share of dementia at autopsy. This is slow damage, which means it is also the territory where earlier action changes the most. Blood Sugar and Metabolic Health goes further.
A specific triad worth knowing
Gait disturbance, urinary incontinence and cognitive change, classically in that order, with enlarged ventricles on imaging. It is uncommon, it is often over diagnosed, and outcomes from shunting vary. It still matters, because it is one of the few causes where surgery can help. The gait pattern is the clue that gets it looked for.
The head injury nobody remembered
Blood collecting slowly between the brain and its coverings, often weeks after a fall that seemed minor and sometimes after no remembered injury at all. Risk is higher on anticoagulants and in older adults. Cognitive change with headache, drowsiness or new gait unsteadiness needs imaging rather than watchful waiting.
Ordinary chemistry, real effects
Hypercalcaemia, sodium abnormalities, liver disease, advanced kidney disease and severe deficiency states all impair cognition. These are found on inexpensive tests that are already routine, which is precisely why skipping them in favour of something more specialised is such an expensive mistake.
Attention is the casualty
Persistent pain consumes attentional resources, disrupts sleep and lowers mood, and the medications used for it frequently sedate. The cognitive complaint is real and its origin is upstream. Treating the pain and reviewing what is being taken for it usually moves the cognition.
Two more belong here without needing a card of their own. Undiagnosed coeliac disease and other malabsorption states, which produce deficiencies rather than acting directly. And untreated visual impairment, which reduces input and engagement in much the same way hearing loss does. Both are cheap to check and neither is usually checked.
On order of operations
The question is not only whether something reversible is present. It is how many contributors are present at once.
Which is why finding one and stopping is the most common way a workup fails.
If it is a neurodegenerative process
The major dementias begin differently.
Dementia is a level of severity, not a diagnosis. Underneath it sit distinct diseases with distinct openings, and the opening is diagnostically informative because these conditions become harder to tell apart as they advance. Mixed pathology is common, particularly in older people, and a single clean label is often less accurate than the picture in front of you.
Memory encoding first
The typical opening is failure to form new memories, so recent conversations and events do not stick and prompting does not retrieve them. Repeating questions, misplacing items in odd places and difficulty learning new routines follow. Naming and orientation change later. There are atypical forms that begin with visual or language symptoms instead, and those are frequently misattributed at first.
Executive function and speed first
Slowed processing, difficulty planning and organising, and a stepwise or patchy course rather than a smooth one. Gait change often appears early. It coexists with Alzheimer’s pathology in a large proportion of cases, and the vascular component is the part that responds to blood pressure, glucose, lipids and smoking.
Fluctuation, hallucinations, movement
Marked fluctuation in attention and alertness, well formed visual hallucinations, parkinsonian features, and acting out dreams during sleep, which can precede everything else by years. Severe sensitivity to antipsychotic medication makes correct identification a safety issue rather than an academic one.
Behaviour or language first, and younger
Memory can be relatively preserved early. What changes is behaviour, judgment, empathy and social conduct, or else language production or word comprehension. Onset is often in the fifties or sixties, and the behavioural form is misdiagnosed as a psychiatric or personality problem with painful regularity.
This section exists to make the point that pattern matters, not to enable self diagnosis. Telling these apart requires history from someone close to the person, formal testing, imaging and often specialist assessment. What it should change for a reader is the willingness to describe the actual pattern rather than reporting bad memory and leaving it there.
Why identical pathology produces different lives
Cognitive reserve.
Autopsy studies repeatedly find people who met the pathological criteria for Alzheimer’s disease and had no dementia while alive. That observation is the foundation of reserve: the capacity to sustain damage before function visibly fails, through more efficient networks and the recruitment of alternatives. Reserve does not remove pathology. It changes the point at which pathology becomes disability, and that difference is measured in years.
Reserve is not protection from disease. It is a longer runway before disease becomes disability.
What each test can and cannot settle
The testing, and what it actually answers.
Memory assessment has three separate jobs: measure cognition properly, look for the treatable causes, and only then ask whether a neurodegenerative process is present. Doing the third before the first two is how people end up with a frightening label and an untreated thyroid.
The three layers of testing
Measuring cognition
Screening tools take a few minutes and produce a flag, not a diagnosis. Scores are affected by education, language, hearing, anxiety and sleep.
Formal neuropsychological assessment takes hours, measures each domain separately, and is what actually characterises the pattern.
Finding the treatable causes
B12 and folate, thyroid function, full chemistry, calcium, glucose and A1C, liver and kidney function
A full medication review, a validated mood screen, sleep and apnoea assessment, and formal hearing and vision testing
Asking whether it is neurodegenerative
Structural imaging, to exclude other causes and to show the pattern of tissue loss
Blood and spinal fluid biomarkers, in specialist settings, with real caveats
The third layer answers nothing useful if the second layer was skipped, which is the sequencing error that produces most of the harm in this area.
A word on the newer blood biomarkers, which are being marketed faster than they are being validated. Phosphorylated tau assays can indicate Alzheimer’s pathology with useful accuracy in people who already have symptoms and are being assessed by a specialist. That is a real advance. What they do not do is tell an asymptomatic person what their future holds. They also perform less well outside the populations they were validated in, including in people with kidney disease and higher body mass. Amyloid is also present in people who never develop dementia. Testing without a plan for what the result would change is not information.
What the population evidence supports
Risk is not fixed.
Large scale evidence syntheses now attribute a substantial minority of dementia cases worldwide to modifiable risk factors spread across the life course. That figure is an estimate built on associations, and it should be read as a direction rather than a promise. It still points somewhere useful: the factors are identified, most of them are ordinary, and the ones that matter most are not the ones being sold.
Two features of this evidence matter more than the list itself. The first is timing. Blood pressure and hearing act across midlife, which means the window that changes the most is often decades before anyone is worried about their memory. The second is that these factors are additive, so acting on several imperfectly outperforms perfecting one.
Multidomain trials combining diet, exercise, cognitive activity and vascular risk management have shown modest benefit in at risk older adults in some studies and not in others. Modest and real is a fair summary, and a better basis for action than either dismissal or the claims made by prevention programmes selling packages.
The interventions with the best evidence are unglamorous, cheap, and mostly not for sale.
Modifiable factors
Across the life course
Modifiable factors across the life course
Earlier life
Education, and the cognitive reserve it helps build
Midlife
Hearing loss, treated rather than tolerated
Blood pressure
Head injury, prevented
Alcohol intake, obesity, cholesterol
Later life
Smoking, diabetes, physical inactivity
Depression, social isolation, untreated vision loss
Air pollution exposure
These are associations at population scale, not guarantees for any individual. They are still the best map available, and acting on several of them imperfectly outperforms perfecting any one.
The sequence
How I evaluate a memory concern.
This describes a clinical process carried out with a person, not a checklist to run on yourself, and it frequently ends in a referral. Where I differ from a short appointment is not in believing something exotic. It is in refusing to close the treatable layer early.
Seek urgent assessment
Some memory change is not a slow story.
Everything on this page describes a gradual change assessed over weeks. The following are different, and they are time critical. If any of these is happening, call emergency services or seek same day assessment rather than reading further.
Confusion developing over hours to days
Sudden inability to speak or understand speech
One sided weakness, numbness or facial drooping
Sudden severe headache, unlike any before
Confusion with fever or a stiff neck
A first seizure at any age
Cognitive change after a head injury, even weeks later
New drowsiness or difficulty staying awake
Cognitive decline progressing over weeks rather than years
New psychiatric symptoms with seizures or abnormal movements
Loss of consciousness
Sudden vision loss
Rapid cognitive decline over weeks to a few months deserves particular emphasis. It is not typical of the common neurodegenerative diseases and it raises the possibility of causes that are treatable when found early, including autoimmune encephalitis, infection, structural lesions and severe metabolic disturbance. It warrants urgent neurological assessment rather than a routine referral. If you are having thoughts of suicide or you are in immediate danger, call or text 988 to reach the Suicide and Crisis Lifeline in the United States, or call 911.
Free guide
Work the treatable layer before anything else
Brain on Fire walks through how to describe which function changed, and what belongs in a proper history. Which foundational tests find the reversible causes, and which symptoms need urgent assessment rather than an appointment.
Own your biology
Being assessed is not the same as being diagnosed.
The fear of what an assessment might find is the single most common reason people delay one. That logic runs backwards. Assessment is how the treatable causes are found, and they are found far more often than the untreatable ones.
If the answer turns out to be a neurodegenerative disease, finding it earlier still buys something real. Time to plan. Time to treat what is treatable alongside it. And the chance to make decisions while you are still the one making them.
Ageing from disease Slower retrieval is expected. Information that never went in is a different question.
Severity from cause Dementia describes how much function is lost. It does not name what is causing it.
Additive from singular Several contributors at once is the usual finding, and correcting three imperfectly beats finding one and stopping.
Risk from destiny A substantial share of risk sits with factors that can be acted on, most of them decades before anyone worries.
Bring someone who knows you well.
No pressure, and nothing to buy. Bring what changed and roughly when. Every medication and supplement, including the over the counter ones. How you sleep, and any results you already have. And if you can, a person who has watched the change from the outside. That history is worth more than any single test.
Common questions
Questions about memory and cognitive decline.
Short, plain answers to what people ask most.
How do I know if my memory problems are normal ageing?
Does mild cognitive impairment always progress to dementia?
What causes of memory loss are reversible?
What is the difference between delirium and dementia?
Are the new blood tests for Alzheimer’s worth getting?
Can memory loss be prevented?
Do brain training games or supplements help?
When should I get assessed?
