Processing Speed Test — Standard Scores & Reference Values

Processing speed (Gs) is a formal psychometric factor: how quickly you can execute simple, already-learned operations under time pressure. Its classic measure is symbol–digit coding — the best-known clinical example is the Symbol Digit Modalities Test (SDMT): 90 seconds, a key pairing nine symbols with digits 1–9, scored as the number of items completed correctly. Our web test follows that format with an original symbol set (the SDMT symbols are copyrighted); it is a training and self-tracking tool, not the clinical instrument.

The closest fully-normed digital neighbor is the NIH Toolbox Pattern Comparison Processing Speed Test — 90 seconds of same/different picture judgments, normed on a national sample of 4,859 participants aged 3–85 (Carlozzi et al. 2015: https://pmc.ncbi.nlm.nih.gov/articles/PMC4542749/). It proves the format normable at scale — but it is a different task (picture comparison, not symbol coding), so its norms cannot be relabeled as ours.

Your device matters less here than on a reaction-time test. Browser and platform pipelines add measurable noise — Anwyl-Irvine et al. measured end-to-end reaction-time delays of roughly 71–87 ms depending on platform, browser and device (2020, Behavior Research Methods: https://link.springer.com/article/10.3758/s13428-020-01501-5), and Passell et al. 2021 found mobile devices run 50–200 ms slower than desktop on the same tasks (https://link.springer.com/article/10.3758/s13428-021-01597-3). But a coding task aggregates items of 2–4 seconds each, so a ±100 ms device offset is a small fraction of every answer — where it would be a third of a 273 ms reaction time.

Practice effects on symbol coding are large and durable — repeat SDMT studies find gains persisting for up to 18 months (Novakova et al. 2025: https://pmc.ncbi.nlm.nih.gov/articles/PMC12357967/). Our countermeasure is the one clinicians use: alternate forms. The pairing key reshuffles at every session start, so your trend reflects lookup speed rather than a memorized key — though a residual warm-up effect is unavoidable and we will not pretend otherwise.

WRITTEN-VERSION ANCHOR — NOT COMPARABLE WITH THIS WEB TEST. These cutoffs come from the supervised pen-and-paper SDMT (110 items, written answers, clinical administration) reproduced in the NIAAA MATCH manual. Our web task uses an original symbol set, key/button input and self-administration — a web score must never be read against this table. It is shown to give the age gradient an honest, sourced shape, nothing more.

If you want a digitized processing-speed task WITH norms, the honest pointer is the NIH Toolbox Pattern Comparison test: 90 seconds of same/different judgments on pairs of pictures, up to 130 items, normed on 4,859 participants aged 3–85 representative of the US population (Carlozzi et al. 2015: https://pmc.ncbi.nlm.nih.gov/articles/PMC4542749/). It demonstrates that a 90-second format can carry real norms — and it is still not interchangeable with symbol coding, which adds the lookup-and-remember component. That is exactly why we publish our own live community distribution (the strip above) instead of borrowing someone else’s table.

Web timing is noisy: Anwyl-Irvine et al. (2020) measured end-to-end response delays of roughly 71–87 ms across platform/browser/device combinations (https://link.springer.com/article/10.3758/s13428-020-01501-5), and Passell et al. (2021) found mobile devices running 50–200 ms slower than desktop on identical tasks, with Android about 0.56 SD behind Windows after covariates (https://link.springer.com/article/10.3758/s13428-021-01597-3). On a simple reaction test that noise swallows a third of the signal; on a 90-second coding run it rides on items that each take 2–4 seconds, so the device tax shrinks to a few percent of your pace. Still: compare your scores on the same device, and expect a phone to read slightly below a desktop.

Symbol-coding scores rise with repetition and the gain is stubborn: repeat-administration studies of the SDMT find practice effects still measurable up to 18 months later (Novakova et al. 2025: https://pmc.ncbi.nlm.nih.gov/articles/PMC12357967/). The standard clinical countermeasure is the alternate form, and this site builds it in: the symbol–digit key is reshuffled with a seeded random generator at every session start, so memorizing last week’s pairing buys you nothing. Honest residue: the task itself still gets easier with familiarity, so treat your first few sessions as warm-up and your later trend as the signal.

The strongest positive result in this field is the ACTIVE trial: 2,832 older adults randomized to memory, reasoning or speed-of-processing training (JAMA 2002: https://pubmed.ncbi.nlm.nih.gov/12425704/). At the 10-year follow-up, the group receiving the speed-of-processing intervention — a specific, adaptive UFOV-style training protocol — showed a 29% lower dementia risk than controls (Edwards et al. 2017: https://pmc.ncbi.nlm.nih.gov/articles/PMC5700828/). Read that sentence carefully: it is about a particular adaptive training intervention, not about taking a test like this one.

The counterweight: a major review of the brain-training literature found strong evidence that practice improves the trained task and weak evidence for transfer to untrained abilities (Simons et al. 2016: https://gwern.net/doc/dual-n-back/2016-simons.pdf). So our claim is deliberately boring: practicing here improves your score here, visibly, in your local history. We do not claim — and no honest site may claim — that this test prevents dementia or raises intelligence.

Three fixed points to orient a processing-speed score, plus our live one: the paper ceiling (the written SDMT packs at most 110 items into 90 seconds — nobody completes it), the written top-quintile cutoff for young adults (≥ 67 correct, ages 18–24, NIAAA table above), the fully normed digital neighbor (NIH Toolbox Pattern Comparison, n = 4,859 — a different task), and this site’s own live community distribution in the strip under the reference table, which is the only anchor measured on THIS task and THIS input modality.

Run your own 90-second coding round, or print seeded practice sheets for offline use. Open the printable coding sheets →