Has my TARA score really improved?
Check whether a jump between two practice marks is bigger than normal test noise, using UAT-UK's published measurement error.
TARA Critical Thinking practice marks
Enter two marks out of 22
Every test result wobbles. We use UAT-UK's published measurement error to show how big a change needs to be before it is probably real.
UAT-UK TARA Technical Report, July 2026, Tables 16 and 17
TARA · 2027 entry
Keep the TARA deadline in view.
Save your email for the verified UAT-UK booking reminder and one follow-up to the free TARA reasoning quiz.
We only email you about what you tick, and every email has an unsubscribe link. No account is required.
How we estimate
Scope, source, method, uncertainty, dates and what the answer can and cannot tell you.
How we estimate
Scope, source, method, uncertainty, dates and what the answer can and cannot tell you.
What it covers
Two marks out of 22 on TARA-style practice papers in the same module, or one real TARA score from the 2026 cycle. Measurement error is UAT-UK's published standard error of measurement (SEM) for each module and sitting (Tables 16 and 17).
Sources
- UAT-UK: TARA Technical Report 2025-26, Tables 2-4 and 16-21, pages 8-33. Official.
Method
- We take the largest published SEM for your module: 2.12 raw marks for Critical Thinking and 2.00 for Problem Solving, or 1.24 and 0.98 scaled points. Using the largest keeps the answer cautious.
- The difference between two marks, each with that error, has a spread of the SEM times the square root of 2. A change of 1.96 times that spread or more happens by noise alone less than 5% of the time: 5.9 marks in Critical Thinking and 5.5 in Problem Solving.
- The chance figure is how often noise alone would move a mark at least as far as yours, using a normal curve.
- For one real score we show plus or minus 1.96 SEM around it, capped at 1 and 9.
Uncertainty and limits
- Published SEMs describe live TARA papers. Practice papers may be more or less precise.
- Different papers also differ in difficulty, which adds noise that SEM leaves out, so a real change needs to be even larger across different papers.
- Normal, independent errors and the same error at every score are assumptions that UAT-UK has not validated per score.
Dates and versions
- TARA 2026-entry score distribution, scaling, reliability and timing: data version 2026-09-26.1, checked 26 September 2026.
- Page last changed 27 September 2026.
What the answer means
The answer says whether a change is larger than a 22-question test can reliably tell apart from luck. It cannot say how much you have improved, predict your real score, or compare scores from different years, which sit on different scales.
Questions people ask
How much can a TARA score vary by chance?
UAT-UK's published measurement error for 2026 was about 2 raw marks out of 22 in each module, or roughly 0.9 to 1.2 scaled points. That is one standard error; about 95% of the time the noise stays within twice that.
My mark went up 3 marks. Have I improved?
Maybe, but one jump of 3 is within normal noise on a 22-question paper: luck alone produces a change that size fairly often. Several unseen, timed papers moving the same way is much stronger evidence.
Can I compare my TARA score with last year's?
No. Each cycle's scale is set afresh, anchored to its own October candidates, so scores from different years cannot be subtracted as progress.
Does redoing the same paper count?
Not as evidence of progress. Remembering questions raises the mark without any change in skill. Compare first attempts on papers you have not seen.
Sources
- UAT-UK: TARA Technical Report 2025-26OfficialTables 2-4 and 16-21, pages 8-33. Published 2026-07. Checked 26 September 2026.
TARA 2026-entry score distribution, scaling, reliability and timing: version 2026-09-26.1, First published version.