How We Calculate Your Baseline

Most cognitive assessments compare your score to a population norm — a reference group of people matched by age and education. This tells you roughly where you stand relative to others. It does not tell you whether your cognitive function has changed.

Keel is built on a different principle: your baseline is derived entirely from your own historical data. You are compared to yourself. This is called an n-of-1 or single-subject design in research methodology, and it is the appropriate design for detecting individual change over time — which is exactly what cognitive health monitoring requires.

The practical implication: in your first few weeks of using Keel, you are building your baseline. The more sessions you complete, the more stable and reliable that baseline becomes. Early scores are useful but noisier. After 30 sessions, the statistical model has enough data to detect meaningful deviations from your personal norm.

For each test, Keel fits a rolling statistical model to your historical scores. The model accounts for three things: (1) your average performance level — your central tendency, (2) practice effects — the predictable improvement that occurs as you become more familiar with a task format, and (3) normal variability — the day-to-day noise that is expected and meaningless.

From this model, Keel calculates what your score should be on any given day, given your history. A score significantly above expectation is a good day. A score significantly below — particularly if it persists across multiple sessions — is a deviation from your personal baseline worth paying attention to.

The deviation is expressed in terms of your own standard deviation: how far your current score sits from your personal mean, measured in units of your own typical variability. This is more meaningful than a raw score, because it adjusts for the fact that some people are naturally more variable than others.

When you start a new cognitive task, your scores typically improve for the first several sessions — not because your cognition is improving, but because you are learning the format. This is a well-documented phenomenon in cognitive psychology called the practice effect or test-retest learning effect.

If Keel did not account for practice effects, your early improving scores would artificially inflate your baseline, and a return to your 'true' performance level would look like a decline. Keel's baseline model explicitly fits and removes the practice effect curve, so that your baseline reflects your underlying cognitive capacity rather than your familiarity with the specific test format.

Practice effects are larger for novel, complex tasks (like the spatial memory test) and smaller for simple, familiar-feeling tasks (like reaction time). They typically plateau within 7-10 sessions. Keel weights early sessions appropriately lower while the model is still learning your pattern.

Your Keel baseline cannot tell you whether your absolute cognitive level is high, average, or low compared to the population. It can only tell you whether your performance has changed relative to your own history. A person with high baseline cognitive ability who declines by 15% may still score higher than the population average — but the decline is still real and worth monitoring.

Keel is a personal tracking tool. It is not a clinical diagnostic instrument. Your baseline trend is a signal, not a diagnosis. Any meaningful trend — upward or downward — should be discussed with a healthcare provider, who can order appropriate clinical assessments and account for factors Keel cannot observe.