Core Module guide

DMAT Latin Squares: A Constraint Method

Applicants who understand the idea of a Latin square but lose time or accuracy when a cell is blank and the options look plausible.

Checked against the site's public facts on 13 November 2026 · 中文版

What this guide gives you

  • Solve the cheapest cell, not the first cell you notice.
  • Use row, column and symbol counts as three independent checks.
  • Never redraw the grid: mark candidates in the blank cell and let constraints remove them.

The rule, stated precisely

A Latin square is a grid in which each symbol appears exactly once in every row and exactly once in every column. The item shows a completed grid with one cell removed and asks which symbol belongs there. For a four-by-four grid with four symbols, each symbol appears four times in the finished square: once per row and once per column. That is the entire rule, and every legitimate solving technique is a consequence of it.

The important consequence is that you never need to solve the whole grid. You only need to prove which symbols cannot occupy the blank cell. If three symbols are eliminated, the fourth is correct by exhaustion. Candidates who try to reconstruct the full square are doing five times the work the item asks for, and they run out of time on a task that should take forty seconds.

One caution about wording: check whether the item asks for the missing symbol in a row, a column, or both. The same grid can support two different questions. If a question says row, do not spend time verifying columns beyond the check that confirms your answer.

Choose the cheapest cell

Scan the row and column that intersect at the blank cell. Every symbol already visible in that row is excluded, and every symbol already visible in that column is excluded. If the row contains three symbols and the column contains the same three, the answer is the fourth symbol immediately. That is the easy case and it is more common than nervous candidates expect.

If two symbols remain possible, the item needs one more constraint. Look for a row or column containing only one instance of a symbol, then check where that symbol can still go. A second-order constraint like this is the normal difficulty level of the task. Write the two candidates as small marks in the blank cell and test them one at a time: assume candidate A is correct, and check whether any other cell becomes impossible. If it does, A is wrong.

The mistake to avoid is starting at the top-left corner because that is where reading starts. Cost is not left-to-right; cost is how few constraints you must combine. Scan the whole grid for a cell with the most filled row and column neighbours, and start there. On a four-by-four grid this scan takes three seconds and routinely saves twenty.

Small grids are where the habit is built. A three-by-three Latin square with three symbols has fewer constraints, so the same rule feels less convincing, and that is exactly why it is a useful training case: you must state the constraint rather than feel your way to an answer. Drill three-by-three grids until you can explain each elimination in words, then move up. Candidates who start at four-by-four often learn the answer pattern for that size rather than the method, and the pattern collapses when the grid grows.

The paired-symbol trap

Many practice grids contain a symmetry that makes two symbols interchangeable in some region. For example, two rows might accept either of two symbols in two cells, so the pair can be swapped without breaking the Latin rule. These ambiguities are harmless in a completed grid but dangerous in a partially solved one, because they let you find a solution that satisfies every constraint you checked yet contradicts a constraint you did not.

The defence is to verify the answer against all three counts: the blank row, the blank column, and the total symbol count across the grid. Each check is cheap. Row and column checks catch local errors; the count check catches a duplicated symbol that slipped into a cell you did not examine. When all three agree, stop checking and move on. Additional verification after three independent checks is not rigour, it is anxiety with a timer attached.

It also helps to notice when the item is deliberately underdetermined. If two symbols both satisfy every visible constraint, the item is probably asking a different question than you assumed, such as which symbol must appear in a particular row somewhere else in the grid. Re-read the prompt before assuming the question is broken.

Solve a generated Latin square

The checker builds a valid four-by-four grid, blanks one cell, and validates your choice against row, column and symbol counts. It also shows the constraint order that solves it.

This guide is independent study material. It is not affiliated with g.a.s.t., APS India, or any German university, and it does not publish official dMAT questions.

Paper discipline for a screen-based test

On a computer-based test you cannot draw on the grid. Use the physical scratch paper the test centre provides, and copy only the blank row and the blank column, not the whole square. Two short lines of symbols are enough to solve most items, and they fit in one corner of the sheet. Copying a full four-by-four grid wastes forty seconds and introduces transcription errors.

If scratch paper is not available for your administration, use the on-screen option labels as your working memory: eliminate options by crossing them out mentally in a fixed order, top to bottom, and state the reason for each elimination. Naming the reason prevents the classic failure where you eliminate an option, forget why, and then re-admit it two steps later.

Practise with the constraint order you will actually use: row symbols, column symbols, then a second-order check. A fixed order removes decisions from the task, and removing decisions is how you buy time for the items that genuinely need thought.

Practice design that builds accuracy

Start with grids where the answer follows from the row and column alone, and get to the point where you can solve them in under a minute without hesitating. Then move to grids that need one second-order constraint. Then practise grids with two blank cells, which appear in harder practice sets and force you to choose a solving order rather than a single cell.

Review by rule, not by score. For each miss, write whether the failure was an unread symbol, a row-and-column swap, an exhausted candidate that was re-admitted, or a misread prompt. Those four causes need four different fixes: slower scanning, a fixed check order, a written elimination list, and a prompt-reading habit respectively. A wrong answer with no diagnosis teaches almost nothing.

Finally, alternate Latin squares with figure sequences in the same session. Both tasks reward constraint reasoning, but they stress different working-memory demands, and alternating them keeps you from over-fitting to one visual format. The transfer is the goal: a small set of rules you can apply to a grid you have never seen.

Official sources

This guide defers to these official pages on format, rules, dates, fees and eligibility. They are external and may change, so check the version that applies to your own administration.

  • Official dMAT websiteTest structure, preparation materials, fees, test centres and the 2027 India test dates.
  • APS India dMAT noticeWho needs the dMAT for APS documentation, exemptions, and how the certificate reaches your APS file.

Questions about this guide

How large are the grids on the real test?

Confirm the current grid size and format on the official dMAT sample material. The method here scales, but the solving time changes with grid size.

Should I fill every cell before answering?

No. Eliminate three symbols for the blank cell and stop. Filling the grid is unnecessary work unless the item asks for more than one cell.

Why does my answer sometimes fail a count check?

Usually a duplicated symbol already in the grid was missed during scanning. Count each symbol once across the whole grid before concluding the grid is valid.

Put the method to work

The free diagnostic shows which of these failure modes is actually costing you marks. Ten questions, fifteen minutes, instant topic feedback, no account.