What the Arithmetic Subtest Is
Arithmetic is one of the core subtests of the Wechsler Adult Intelligence Scale. The task is deliberately plain: the examiner reads a word problem aloud, and the candidate answers within a time limit, mentally.
No paper. No calculator. No re-reading. The problem is heard once, held, and worked on in the head.
The mathematics involved is elementary — the kind of proportion, rate, and change problem that appears in school by early secondary level. What makes the subtest difficult is everything around the mathematics.
Where It Sits in the Modern Scale
Arithmetic contributes to the Working Memory Index, alongside Digit Span. That placement is recent, and the history of how it got there is the most interesting thing about it.
Fifty years in the wrong drawer
From the 1939 Wechsler–Bellevue scale through the third edition, Arithmetic sat inside the Verbal half. The reasoning was administrative — the item is delivered in words, so it was filed with the verbal subtests.
The fourth edition in 2008 abolished the verbal/performance split entirely, replacing it with four indexes: Verbal Comprehension, Perceptual Reasoning, Working Memory, and Processing Speed. Arithmetic moved to Working Memory.
Nothing about the subtest changed. What changed was the acceptance that its statistical behaviour had never matched the verbal subtests as well as it matched Digit Span.
Why It Loads on Working Memory
Take a typical item apart and the reason becomes obvious.
- Encode. The problem is spoken, so it must be held as heard — an auditory-verbal load before any arithmetic starts.
- Select. The right operation has to be chosen from the wording, which is a reasoning step.
- Compute. The calculation itself, without external storage.
- Retain. Intermediate results must survive while the next step runs — this is where multi-step items break down.
Only the third step is arithmetic. The other three are memory and control, which is why a person entirely capable of the sum on paper can fail the item in the head.
The Confound Nobody Fully Solved
Arithmetic is the most confounded subtest on the scale, and the manuals have said so for decades.
A low score can mean genuinely limited working memory. It can equally mean weak arithmetic schooling, mathematics anxiety, or a hearing or attention lapse on a single item.
The anxiety problem specifically
Mathematics anxiety consumes working-memory capacity directly — the worrying occupies the same resource the problem needs. On a mentally-computed, time-limited, examiner-observed task, that effect has maximum room to operate.
This is why a competent clinician never interprets Arithmetic alone. It is read against Digit Span, against the other index scores, and against what the person reports about school mathematics.
How It Compares to a Graduate Numerical Test
They measure overlapping things through very different apparatus, and the differences all run in one direction.
- Delivery. Wechsler is spoken and individually administered; a graduate test is read on screen, usually unsupervised.
- Working. Wechsler forbids paper. Most employer tests permit it, and many permit a calculator.
- Data volume. Wechsler items are short and self-contained; employer items sit under tables and charts with more figures than the question needs.
- Purpose. One produces a clinical profile against age norms; the other ranks candidates against an applicant pool.
The practical consequence: Wechsler Arithmetic is nearly a pure working-memory measure with arithmetic as the vehicle, while an employer numerical test dilutes that with reading, extraction, and pacing.
Can an Online Test Approximate It?
Partly, and it is worth being precise about which part.
An online numerical measure can sample the same reasoning content — proportion, percentage change, rate — and can do so reliably, because those items behave well psychometrically regardless of how they are delivered.
What does not survive the translation
The mental-only constraint. Once the candidate can write, the working-memory load that Arithmetic exists to measure largely disappears, and the score becomes a measure of arithmetic reasoning instead.
The standardised individual administration also does not survive — the examiner watching for hesitation, the controlled single reading, the age-normed comparison. Those are what make the clinical score interpretable.
So the honest framing is that an online numerical test overlaps with Arithmetic on content and diverges on construct. It is a useful indicator, not a substitute for an assessment conducted by a qualified professional.
Reading a Low Arithmetic Score
The single most useful comparison is against the rest of the profile.
Low Arithmetic with low Digit Span points at working memory as the common factor, and that pattern has practical consequences — following multi-step spoken instructions, mental calculation in meetings, holding a thread while interrupted.
Low Arithmetic with intact Digit Span points somewhere else entirely: at arithmetic fluency, at anxiety, or at educational history. That version is far more addressable, and it is not a working-memory finding at all.
Either way, one subtest is one data point taken on one day. The scale was built to be read as a profile, and reading a single index score in isolation is the most common misuse of it.