Why your symbol match score will not match another site's

You got 38 here and something quite different somewhere else. Both can be right — and unlike our other two tasks, this one comes with published measurements of how far apart versions of the same test actually sit.

Version 1.0 · Published 6 September 2026 · Last reviewed 6 September 2026 · About the symbol match test

1. The clock is not the same clock

Our window is 90 seconds. The clinical convention is 120. A digit symbol substitution test conventionally allows two minutes, with the score being the number of correct substitutions made in that time.

That alone makes raw counts incomparable — a third more time is a third more opportunities. But the counts do not scale cleanly either, and we are careful not to claim we know which way they bend. That is exactly why our own test splits its window at 45 seconds and reports both halves separately instead of assuming the rate is flat: if we knew the shape of it, we would not need to measure it. Two thirds of a 120-second score is not a 90-second score, and nobody has published the conversion.

And a shared name does not fix the clock. The Look AHEAD trial's administration form for the WAIS-III Digit Symbol Substitution Test instructs the examiner in as many words: "Allow 90 seconds (1 minute, 30 seconds) for the test." Same instrument, same publisher's materials, a different duration from the convention. If a named clinical test runs on two clocks in two studies, the phrase "digit symbol test" on a website tells you very little about what was measured.

2. How you answer changes the score, and it has been measured

Peer-reviewed journal article · open access · n = 104

You answer ours by pressing a digit. The clinical original is answered by hand-writing a symbol into a box. Other versions use a mouse, a touchscreen, or speaking the answer aloud.

This is not a detail, and it is not our inference. In a 2022 validation study, 104 cognitively unimpaired adults completed three versions of the substitution paradigm in a pseudo-randomised order under supervision. The two computer-based versions — different pictures, same response device — correlated at r = 0.81. A computer-based version against the handwritten WAIS Coding subtest correlated at r = 0.68.

The natural reading, which we flag as a reading rather than a finding: both comparisons change the pictures, and only the second also changes the response from pressing to writing. The step from 0.81 down to 0.68 is what that change added. It is not a controlled manipulation — paper and computer differ in more than one way at once — but the direction is hard to explain otherwise. Writing is slower than pressing, and it is slower by an amount that varies with your handwriting rather than with your processing speed.

The same effect appears in the older literature on the Symbol Digit Modalities Test, whose oral and written forms correlate at .78 — reported in an institutional review of that instrument with the plain conclusion that "the two forms are not interchangeable." Two forms of one test, from one manual, described by the people who use it as not swappable. We have read that review and not the manual it cites, which is why the figure is attributed to the review below rather than presented as the manual's.

3. "Correct" and "completed" are two different scores

Ours counts correct matches. Some versions count items attempted. On a task where a wrong answer still advances you, those two numbers come apart the moment anyone starts rushing — and a timer is an instruction to rush.

Our page reports correct and attempted as two separate counts, never as a percentage on its own, so the gap between them stays visible instead of being averaged away. A site reporting a single number cannot tell you which of the two it is, and for anyone who was hurrying, the difference is not small.

4. Six symbols is not nine symbols

We use six symbol-digit pairs. Both computerised versions in the study above use nine. Fewer pairs is a smaller key, so more of your window goes on matching and less on looking things up — which raises the count without your having got faster at anything.

Two more of our choices belong here, because they change the task rather than its difficulty. Ours regenerates the pairing on every run, so it cannot be carried between attempts. And ours leaves the key on screen throughout, because this is a substitution task, not a recall task. A version that hides the key, or reuses one pairing across sessions, has folded a memory demand into a speed score. That is a legitimate thing to measure. It is not the thing this measures.

What would have to be true for two scores to be comparable

• The same window, or a published method for converting between windows. There is not one.
• The same response mode. Pressing, writing, clicking and speaking are four tasks wearing one name.
• The same number of pairs, and the same rule about whether the key stays visible.
• The same definition of the score — correct, or attempted — stated rather than implied.
• Both scoring rules published, so the arithmetic can be checked instead of trusted.
• Both norm-referenced against a described sample, if either is going to say where you sit relative to other people. Ours is not, and says nothing.

The versions in that study satisfy most of these against each other, under supervision, in one sitting — and still land at 0.68. Almost no pair of web tests satisfies the first four.

So which number should you believe?

Neither, as a statement about you. Ours is a count of correct matches in 90 seconds, on whatever keyboard or screen you had, on one day. That is all it claims and all it can claim — our reference sample is 0 people.

The comparison that survives every objection on this page is the one inside a single run. Our test splits its window at 45 seconds and reports both halves, because both halves are the same length, the same task, the same device and the same person, thirty seconds apart. Nothing on this page can undermine that comparison, because it holds constant everything this page is about. It needs no other people and no other site to mean something.

Sources

• Williamson, M., Maruff, P., Schembri, A., Cummins, H., Bird, L., Rosenich, E., et al. (2022). Validation of a digit symbol substitution test for use in supervised and unsupervised assessment in mild Alzheimer's disease. Journal of Clinical and Experimental Neuropsychology, 44(10), 768–779. tandfonline.com/doi/full/10.1080/13803395.2023.2179977 — open access. The 120-second window, the nine-pair key, and the r = 0.81 / r = 0.68 figures from Study 1 (n = 104).
• WAIS-III Digit Symbol Substitution Test, Look AHEAD study form CF3, version 1 (2011). NIDDK Central Repository. repository.niddk.nih.gov — the "Allow 90 seconds" instruction.
• McDonald, S. (reviewer). Outcome Measure: Symbol Digit Modalities Test (SDMT). UNSW MovingAhead outcome-measure summary. movingahead.psy.unsw.edu.au — the .78 oral–written correlation and the "not interchangeable" conclusion. A secondary review citing Smith (1991). We have read the review and not the manual, and the figure is presented on that basis.