Sample questions, worked

One example from each of the six domains the battery samples, with the answer and the reasoning behind it. These are illustrations written for this page — none of them appear in the scored test.

Six domains, six examples

Verbal reasoning

Ember is to blaze as seed is to …

  • Soil
  • Forest
  • Root
  • Harvest

Answer: Forest

The relation is potential to full expression: an ember is the small form that a blaze grows from. A seed grows into a forest. Soil and root are adjacent to seeds but are not what a seed becomes; harvest is the yield, not the mature form.

Quantitative reasoning

2, 6, 12, 20, 30, …

  • 36
  • 40
  • 42
  • 44

Answer: 42

Differences are 4, 6, 8, 10 — rising by 2 each step, so the next difference is 12. Equivalently each term is n(n+1): 1×2, 2×3, 3×4, 4×5, 5×6, then 6×7 = 42.

Matrix reasoning

In a 3 × 3 grid, each row keeps its shape but rotates it 90° clockwise per column, and the number of dots inside increases by one across the row. The bottom-right cell is missing. Row three holds a square with two dots and a square with three dots.

  • Square, rotated, four dots
  • Square, upright, four dots
  • Circle, rotated, four dots
  • Square, rotated, three dots

Answer: Square, rotated, four dots

Two rules run independently along the row: shape rotates, dot count increments. Both have to be applied. Most wrong answers apply one rule and hold the other constant — that is exactly what the distractors are built to catch.

Spatial processing

A square sheet is folded in half top to bottom, then in half again left to right. A single hole is punched through the folded corner that contains the sheet's original centre. Unfolded, how many holes are there?

  • One
  • Two
  • Four
  • Eight

Answer: Four

Two folds produce four layers, and the punch passes through all of them, so four holes appear — arranged symmetrically around the centre. The common error is answering two, by counting folds rather than layers.

Working memory

Hold the sequence 7 – 2 – 9 – 4, then report it in reverse.

  • 4 – 9 – 2 – 7
  • 7 – 9 – 2 – 4
  • 4 – 2 – 9 – 7
  • 9 – 4 – 7 – 2

Answer: 4 – 9 – 2 – 7

Backward span is harder than forward span because the sequence has to be held and manipulated at the same time rather than just replayed. Real items grow until you fail twice at a length.

Processing speed

Scan the row and mark every instance of the target symbol, as fast as you can, for 45 seconds.

Answer: Accuracy under time pressure

Nothing here is difficult in itself. The measure is throughput: how many correct marks you make before the clock stops, penalised for errors. It is the one section where speed is the point rather than a side effect.

Why the real items are not published

An item only measures anything the first time you meet it. Publishing the scored bank would let anyone memorise the keys, which would break the calibration for everyone else and make the resulting scores meaningless. So the examples above share the format and the logic without sharing the items.

It also means practising on this page will not raise your score. Familiarity with the format removes a little first-contact friction, which is why the landing page offers a warm-up, but the reasoning still has to happen.

Difficulty is not flat

Within every section difficulty rises from near-trivial to genuinely hard. The early items exist to establish a floor, not to challenge you, and the scoring model knows this: solving an easy item earns very little, while a hard item solved moves the estimate a long way. Guessing correctly on a hard item is discounted rather than rewarded in full.

Keep reading

Find out where you actually land.

70 calibrated items, about thirty minutes, no account.

Start the test