Norms and percentiles
For 9 tests, ConductCognition compares a participant's main score with a published norm for their age band and reports a z-score and a percentile. Every test reports its scores and quality flags.
A literature norm is the mean and standard deviation of scores that a published study reported for a test, for each age band. ConductCognition places the participant's score on that distribution and reports how many standard deviations it lies from the mean (the z-score) and the matching percentile of a normal distribution.
Scores are signed so that a higher percentile always means better performance. A reaction time faster than the norm's mean gets a percentile above 50.
| Test | Score compared | Published norm | Sample | Ages with a percentile |
|---|---|---|---|---|
| Simple Reaction Time | Mean Reaction Time (ms)Lower is better | Woods et al. (2015) | 1,469 | 18 to 79 |
| Choice Reaction Time | Mean Reaction Time (ms)Lower is better | Woods et al. (2015) | 1,466 | 18 to 79 |
| Go/No-Go Task | d-Prime (d')Higher is better | Votruba & Langenecker (2013) | 276 | 18 to 69 |
| Eriksen Flanker Task | Flanker Effect (ms)Lower is better | Weintraub et al. (2013) | 4,859 | 18 and over |
| Stroop Color-Word Test | Stroop Effect (ms)Lower is better | Van der Elst et al. (2006) | 1,856 | 18 and over |
| Digit Span Test | Forward Span (score)Higher is better | Woods et al. (2011) | 763 | 18 to 79 |
| Trail Making Test A | Completion Time (ms)Lower is better | Tombaugh (2004) | 911 | 18 and over |
| Trail Making Test B | Completion Time (ms)Lower is better | Tombaugh (2004) | 911 | 18 and over |
| Paired Associate Learning | Total Errors (count)Lower is better | Llinares-Benadero et al. (2019) | 1,349 | 18 and over |
Each published study used its own version of the task, so treat a percentile as a reference point for research, not as a clinical score.
When a percentile is withheld, the result page gives the reason, and the session's scores are kept. The Study Quality Monitor shows which sessions had warnings.
A norm is used only when the published study gave the test the same way and scored the same measure. For several tests, the studies we reviewed did not: published verbal fluency norms count words spoken to an examiner, while this test collects typed words, and a published stop-signal norm came from a smartphone and tablet game, while this test uses arrow keys on a computer. Those tests report scores and quality flags, which suit comparisons within a study.
A percentile can also come from ConductCognition's own sessions. The rule is fixed in the scoring code: an age band of a test needs at least 50 valid sessions, run with standard settings on a supported device and scored with the current scoring version, before a platform norm replaces the literature norm for that band. Sessions from studies of a clinical or patient sample do not count. Tests built on words, such as verbal fluency and word-list recall, keep a separate platform norm for each participant language. Each result with a percentile shows the age band and sample size of the norm it was compared with.
9 tests: Simple Reaction Time, Choice Reaction Time, Go/No-Go Task, Eriksen Flanker Task, Stroop Color-Word Test, Digit Span Test, Trail Making Test A, Trail Making Test B and Paired Associate Learning. Each test page states its own percentile status.
One of the conditions was not met: changed settings, no age, a device or quality warning, an age band the norm does not cover, or a free study. The result page states the reason.
Yes. For scores where lower is better, such as reaction time or completion time, a lower score gives a higher percentile, so a higher percentile always means better performance than the norm sample.
Free studies that use only the free tests show scores without percentiles. Paid plans and pay-as-you-go studies show percentiles for the tests listed on this page.
No. A percentile shows where a score falls in a published research sample, given with a different version of the task. Use it as a reference point for research.
A norm built from ConductCognition's own sessions. An age band of a test needs at least 50 valid sessions, run with standard settings on a supported device and scored with the current scoring version, before a platform norm replaces the literature norm for that band. Sessions from clinical or patient samples do not count, and tests built on words keep a separate norm for each participant language.
How each test is scored: automated scoring. What percentiles cost: pricing.
Start free and preview any test. Percentiles come with paid plans and pay-as-you-go studies.