How to Test Yourself for Brain Fog (and Track Whether It's Getting Worse)

Why asking yourself how you feel does not work

You can test yourself for brain fog, and it does not involve a questionnaire. A usable self-test is a short cognitive task with a numeric output, run at two fixed points in the day and compared against your own earlier scores. The figure that carries information is the gap between your morning and afternoon results, and whether it is widening.

Subjective check-ins fail for three reasons. The first is circularity: you are using the faculty that is impaired to assess the impairment. The second is a missing anchor - by 3pm you have no record of what your 9am thinking felt like. The third is that effort does not fall in step with performance; work often feels harder while output quietly gets worse.

The clearest external demonstration comes from a 2011 analysis in PNAS of more than a thousand parole rulings. The share of favourable decisions fell steadily across each judicial session and rebounded after food breaks. That pattern was recovered from the ruling record, not from anything the judges said about their own state.

There is a physiological reason the afternoon is where this concentrates. Every decision fires glutamate in the lateral prefrontal cortex, and when it accumulates faster than the brain clears it, neural signalling becomes less efficient and decision quality falls - the mechanism set out on the Numin science page. That is a change in the substrate, not in attitude, which is why the failure is invisible from the inside. Fog and fatigue are also not the same thing.

The cognitive tests that actually exist

Cognitive testing is a mature field, and its instruments are narrow, specific and boring by design. None is a brain fog test, because brain fog describes an experience rather than a diagnosis. These tasks measure components - vigilance, working memory, inhibition, set-switching - and fog usually presents as a decline in one or more.

• Psychomotor vigilance - respond as fast as you can each time a stimulus appears, for five to ten minutes. The most sensitive of the group to sleep loss and time on task. Output: median reaction time plus a count of lapses, conventionally responses slower than 500 milliseconds.
• N-back working memory - a stream of items arrives and you report whether the current one matches the item n places back. Accuracy is the output, and it degrades under load faster than reaction time.
• Stroop interference - name the ink colour of a word that spells a different colour. The gap between congruent and incongruent trials indexes inhibitory control.
• Trail making, parts A and B - connect numbered circles in order, then a sequence alternating numbers and letters. Part B minus part A isolates the cost of switching between two rules.
• Sustained attention and go/no-go tasks - long, monotonous, built to catch the lapses short tests miss, and far too long to run twice daily for a month.

Consumer brain fog tests are a different category. Most are lightly repackaged versions of the tasks above, and running one is not a bad idea, but they are screening curiosities rather than diagnostics: no age- or education-matched norms, no control over conditions, and input latency that varies enough between devices to swamp the effect you are hunting for.

A five-minute check you can run tomorrow

The protocol below has three constraints: short enough to run twice a day for a month, workable with nothing but a browser and a sheet of paper, and numeric at every step. None of it is diagnostic.

The three tasks

1. Reaction time, two minutes. Pick one free browser reaction-time test and never change it. Run twenty trials and record the median, not your fastest - the fastest trial is the one measurement guaranteed not to represent your average state.
2. Serial sevens, ninety seconds. Choose a three-digit starting number once and keep it forever. Subtract seven repeatedly, out loud or on paper, for ninety seconds. Record two figures: correct subtractions, and errors. This is the arithmetic used in bedside mental status examinations, and it loads working memory harder than it looks.
3. Alternating sequence, ninety seconds. On blank paper, write A, 1, B, 2, C, 3 and keep alternating letters and numbers until you reach M and 13. Record elapsed seconds and errors. This is trail making part B compressed onto a page: the difficulty is holding both rules and switching between them.

Five numbers per run, about five minutes. The morning session goes within the first hour of work, before your first consequential decision; the afternoon session between 2:30 and 3:30pm. Hold both times to within thirty minutes and hold the conditions constant: same device, same room, phone face down, no music, and a fixed relationship to caffeine.

Log the date, the time, the five numbers, hours slept, caffeine timing and dose, and a subjective clarity rating from one to five. The rating exists so that in three weeks you can see how loosely it tracks the objective numbers.

Your baseline is a week, not a morning. Run the battery twice daily for seven days, then discard the first three sessions, because early scores improve from learning the tasks. What remains is a set of morning scores whose ordinary spread you can see.

Tracking the number over weeks

The quantity of interest is the daily delta: afternoon score minus morning score. One day's delta is close to meaningless; ten paired sessions start to mean something, and a month is where a real pattern separates from noise.

Decide in advance what counts as a real drop, using your own baseline week rather than a threshold from the internet. Take the range of your seven morning reaction-time medians, highest minus lowest. Any afternoon result inside that range is ordinary variation; results outside it on most days are the signal.

"A test run once tells you almost nothing. The same test run forty times tells you exactly when your brain stops keeping up."

Four things will move these numbers more than fog does.

• Sleep - the largest single influence on reaction time, and it drags the morning score down too, contaminating the delta from both ends. Log hours; exclude any night below your normal.
• Caffeine - restores reaction time without necessarily restoring judgment. Taken between your sessions on some days and not others, you are measuring caffeine timing rather than cognitive decline.
• Illness - even a mild viral illness moves cognitive scores, often for longer than the symptoms last. Mark the day and drop it.
• Stress - acute stress can sharpen simple reaction time while degrading switching and inhibition, so a tense afternoon may improve one number and worsen another.

Two procedural rules protect the exercise. Do not change tools mid-experiment: a new device resets your baseline. And do not check the running average daily, because expectation moves subjective ratings far more reliably than measured performance - worth understanding before you evaluate anything you take.

Reading your results, and what to do about them

Four patterns come out of a month of paired sessions.

• Flat delta, steady mornings. No measurable within-day decline; if you still feel foggy, the gap between how you feel and what you score points at sleep, mood or total load.
• Normal mornings, a consistent afternoon drop, recovery by the next day. Within-day depletion, and the most tractable of the four.
• Impaired from waking, every day, for weeks. Not a within-day effect, and it will not respond to within-day fixes.
• Morning scores deteriorating week over week. A trend rather than a cycle.

The last two warrant a clinician, and so does any fog that has persisted for weeks regardless of what your log says. Thyroid disease, anaemia, sleep apnoea, depression, medication side effects, perimenopause and post-viral syndromes all present as cognitive fog, and each is something a doctor can test for and you cannot. Treat your log as a record to bring to an appointment, not a substitute for one, and read up on the range of causes.

The second pattern is the one with a named mechanism behind it. Decisions consume glutamate in the lateral prefrontal cortex, and when it accumulates faster than the brain clears it, signalling efficiency drops and decision quality follows. Numin is a stimulant-free drink built to support that clearance system, designed to prevent a decline from baseline rather than lift performance above it.

We tested it in a randomised, placebo-controlled crossover trial published in Frontiers in Nutrition in 2025. Twenty-three healthy adults completed two 13-hour sessions of sustained cognitive load, seven days apart. On Numin they made 43% fewer decision errors than on placebo and held performance to the end of the session, while the placebo arm declined significantly after about four hours, with no adverse events. The limits: 23 participants, one trial, a competitive-play population under a load longer than most working days. The five ingredients are listed on the science page.

Establish your baseline before you change anything, and your log will tell you whether what you added is doing work. The signature is a narrower afternoon-to-morning gap, not an afternoon score above your morning one.