Army Cognitive Test Sections Explained: What Each One Hides

The Army Cognitive Test (ACT) is the ability test every British Army soldier applicant sits. It replaced the British Army Recruit Battery, the BARB, in 2018, and many recruiters and prep sites still call it by the old name. It has five short sections, and one thing about them matters more than anything else: on every section, the material you are working on disappears before you answer.

The one mechanic all five sections share

The Army's own practice site shows it on the first screen. Error Detection puts two strings in front of you and tells you to study them, then to press Next Step to answer. When you press it, both strings vanish and only the answer buttons remain. Every account of the other four sections describes the same thing: the rules, the calculations, the words or the statements go, and you answer from what you held on to.

That turns each section into two jobs. The first is the actual task, counting the mistakes or ordering the names. The second is carrying one answer across to a screen where you cannot check it. Candidates who treat the answer screen as the place to think are reconstructing something they can no longer see, and on a section with no countdown and no warning before it ends, those seconds are items they never reach.

The five sections, one by one

These are in the order the Army's practice site lists them. No source states the live order, so it is the best guide available.

Error Detection

An original string of letters, numbers and symbols sits above a copy. You count how many characters in the copy are wrong, from 0 to 4, then press Next and tap your count. Every position that differs is one mistake, so two swapped characters are two. Scan once, left to right, keeping a running tally, and remember that zero is a real answer.

Orientation

Two rules describe a pair of arrows stacked one above the other: one about colour, such as “Black ABOVE White”, and one about direction, such as “Up Left BELOW Down Right”. The rules disappear and six pairs appear, of which exactly one obeys both. Convert every BELOW into top-then-bottom before you tap Next, and hold the answer as one picture rather than two facts.

Number Fluency

One short calculation, then a second in its place, then the question: which result was greater, the first, the last, or were they the same? You never see both at once and there is no calculator. Decide as soon as the second sum resolves, and remember that Same is a real answer.

Word Rules

Three categories in order, such as tool, fruit and metal, then three words, then the question: how many words matched the category in their own position? A hammer under “fruit” does not count, even though it is a tool. Count as the words appear; you never need to remember the words themselves.

Deductive Reasoning

Two statements compare three people or things, for example that Leah is faster than Tom and Tom is faster than Priya. They disappear and a question appears with three answers. This is the one section descended directly from BARB's transitive inference test. Build a line as you read, and translate negative wording on sight: “Priya is not as fast as Tom” means Tom is faster.

How long, and how many questions

The Army publishes neither. The only figures anywhere come from preparation sites: five minutes each for Error Detection and Orientation, and seven and a half minutes each for the other three, with typical runs of about 30 to 50 items per section. There is no fixed number of questions, because the items stream until the clock ends. Candidates describe the whole sitting, including briefings and practice items, as taking about an hour.

Screener and Verifier: which sitting counts

You meet the ACT twice. First comes a short online screener through your application portal, which the Army says is practice only and is used to see how well you are likely to do. It will not load properly on a phone, so use a laptop or tablet. Then, at the Assessment Centre, you sit the “Army Cognitive Test (Verifier)” under supervision on a touch screen. That is the sitting that produces your score, and you find out the result before you leave. The Army allows three attempts in total.

What the GTI means for your application

Your result is one number, the General Trainability Index. On the old BARB it was scaled so the average applicant scored about 50, with most applicants between 40 and 60. The Army does not publish the score each role needs, but serving soldiers and recruiters on the Army's own Q&A site have quoted these:

Many preparation sites still print a minimum of 26. That is a BARB-era figure, and an Infantry recruiter gave 34 as the floor in 2023. The Army also changes its thresholds, so ask your recruiter for the score your chosen role needs.

Technical trades: the TST

Some technical trades also need the Technical Selection Test, a maths paper the Army describes as similar to a GCSE textbook: the Army names technical trades in the Royal Engineers, REME and the Royal Corps of Signals, and Ammunition Technician in the Royal Logistic Corps. Since at least 2025 it has not been taken at the Assessment Centre. One preparation site describes 55 multiple-choice questions in 45 minutes with no negative marking; the Army publishes no format.

How to prepare

Practise each section on its own until its rule is automatic, because the time you spend remembering the rule is time the clock charges you for. Then sit all five back to back at least once, because holding your pace across five different kinds of task for the best part of an hour is its own skill. Work quickly: the Army's own practice advice is that you will get a better score by working as fast as you can, and to make your best guess and move on when an item seems too hard.

Our Army Cognitive Test practice runs all five sections with the material hidden before you answer, no clock on screen, and a full sitting that ends with an estimated GTI. If you are applying to the RAF or the Royal Navy rather than the Army, you will sit the Defence Aptitude Assessment instead. And if you have met the Thomas GIA, you have met BARB's other descendant: its Reasoning task is a close cousin of Deductive Reasoning.