# TARA score statistics: what the first cycle really looked like

> Official TARA medians and percentiles from UAT-UK’s first report. What counts as a high score, why there is no cut-off, and how CT and PS scale separately.

[Formatted version of this page](https://www.frontiervue.com/exams/tara/guides/score-statistics)

**TARA has no pass mark: across 2,626 first-cycle candidates, official medians were 4.0 for Critical Thinking and 4.7 for Problem Solving, with 90th percentiles of 6.7 and 7.0.** The TARA was first sat in the 2026 admissions cycle across two events, followed by UAT-UK's first technical report on the test. This page sets out the figures that report actually prints, module by module, event by event and group by group, so you can place yourself against the cohort rather than guess, then shows where those figures get misread and which questions the document does not answer. For results day and how to read each module, see [TARA results](/results/tara).

### Quick answer: how to read a TARA score

| Item | Value |
| --- | --- |
| Candidate pool | 976 in October 2025, 1,650 in January 2026 |
| 25th percentile | 2.8 in Critical Thinking, 3.5 in Problem Solving |
| Two separate scores | The two scored modules are reported separately on a 1.0 to 9.0 scale. The report is explicit that they are not comparable with each other and that there is no aggregate or total score. |
| Where 4.5 comes from | The scale was anchored on the October 2025 sitting: median ability fixed at 4.5, the 90th percentile at 7.0. So 4.5 describes the median October candidate. Across the full cycle the medians are 4.0 for Critical Thinking and 4.7 for Problem Solving. |
| First-language gap | The first-language gap lands almost entirely on Critical Thinking: 4.69 for English-first-language candidates against 3.73 for the rest, a difference of 0.95. The same split differs by only 0.11 on Problem Solving, a contrast the report puts down to Critical Thinking's higher language load. |
| Timing | Problem Solving is the speeded module: its median time is exactly the 00:40:00 limit, and 8% of candidates left an item unreached against 1% on Critical Thinking. |

> Every figure here comes from a printed table or sentence in UAT-UK's 2025/26 TARA technical report. Nothing is read off a chart image and nothing is derived across tables.

## Contents
1. The distribution, module by module
2. Where 4.5 comes from: the anchors and the two sittings
3. Group gaps: what the tables show, and how far they carry
4. The clock: Problem Solving is the speeded module
5. What these numbers cannot tell you

## 1. The distribution, module by module

The TARA has three modules and only two of them are scored: Critical Thinking and Problem Solving, 22 multiple-choice items and 40 minutes each. The Writing Task also runs 40 minutes, one prompt chosen from three with a 750-word limit, but it is not placed on the scale and the report prints no score statistics for it. The table below is the part of the report worth committing to memory: both scored modules, read three ways.

- **2,626** TARA candidates in the 2026 admissions cycle, across two events
- **4.0 / 4.7** Whole-cycle median: Critical Thinking / Problem Solving
- **6.7 / 7.0** Whole-cycle 90th percentile: Critical Thinking / Problem Solving
- **1.0 to 9.0** Each scored module reported separately to one decimal place, capped at both ends, with no total

|  | Whole cycle (2,626) | October 2025 (976) | January 2026 (1,650) |
| --- | --- | --- | --- |
| Critical Thinking mean (SD) | 4.21 (1.92) | 4.46 (2.06) | 4.06 (1.82) |
| Critical Thinking 25th / 50th / 75th | 2.8 / 4.0 / 5.4 | 2.9 / 4.5 / 5.8 | 2.8 / 4.0 / 5.2 |
| Critical Thinking 90th percentile | 6.7 | 7.0 | 6.6 |
| Problem Solving mean (SD) | 4.72 (1.77) | 4.66 (1.81) | 4.76 (1.74) |
| Problem Solving 25th / 50th / 75th | 3.5 / 4.7 / 5.8 | 3.5 / 4.5 / 5.8 | 3.5 / 4.8 / 5.7 |
| Problem Solving 90th percentile | 7.0 | 7.0 | 7.0 |
| Observed score range | 1.0 to 9.0 in both modules | 1.0 to 9.0 | 1.0 to 9.0 |

**TARA 2025/26**: Three modules, 40 minutes each

- Critical Thinking: 22 multiple-choice items, scored on the 1.0 to 9.0 scale
  - Whole-cycle mean 4.21
  - Whole-cycle median 4.0
- Problem Solving: 22 multiple-choice items, scored on the 1.0 to 9.0 scale
  - Whole-cycle mean 4.72
  - Whole-cycle median 4.7
- Writing Task: One prompt chosen from three, 750-word limit
  - Not placed on the 1.0 to 9.0 scale
  - No score statistics in the report

## 2. Where 4.5 comes from: the anchors and the two sittings

1. **Calibrate the items** (Rasch + Winsteps)
At the end of each testing window the items are calibrated with a Rasch item-response-theory model in Winsteps, producing a difficulty (b) value for every item. Candidate ability is written as theta.

2. **Turn raw scores into ability** (raw to theta)
The item difficulties generate a raw-score-to-theta table, and the theta value is then scaled to produce the 1.0 to 9.0 reported score.

3. **Anchor two points on October** (4.5 and 7.0)
UAT-UK requires the scale to be fixed to the October ability distribution: median ability maps to 4.5 and the 90th percentile to 7.0. The anchoring abilities are 1.2529 and 2.5610 for Critical Thinking, 0.5029 and 1.8528 for Problem Solving.

4. **Fit the line, cap the ends** (1.0 to 9.0)
A regression line through the two anchors yields the scaling constants: 2.1055 and 1.9112 for Critical Thinking, 3.5686 and 1.8520 for Problem Solving. The transformed scores are capped at 1.0 and 9.0.

5. **Post-equate, share the constants** (No instant score)
Both scored modules are post-equated at the end of the window, which is why no score appears on the day. January reused October's scaling constants so that scaled scores stay comparable across the two events inside one admissions cycle.

Because the anchors were fixed on October, 4.5 describes the median October candidate rather than the median candidate of the cycle. The two sittings did come out differently: the Critical Thinking median was 4.5 in October and 4.0 in January, the larger sitting at 1,650 against 976, and 4.0 across the cycle as a whole. Problem Solving runs the other way, 4.5 in October, 4.8 in January, 4.7 overall. Treat 4.5 as the midpoint only inside the October event.

> Do not read the gap between the two sittings as a difficulty gap. The report is explicit: a raw score depends on both candidate ability and form difficulty, while the scaled score strips form difficulty out and sits on one scale within the cycle, and both events used the same scaling constants. A difference in means between October and January is therefore a difference in who sat the test, not in how hard the paper was. The report's own illustration: a candidate scoring 6.5 in Critical Thinking, in either month, has higher ability than one scoring 4.2. Note too that the TARA was not used by the University of Cambridge this cycle, so candidates could pick either event freely and the two groups were never going to be alike.

## 3. Group gaps: what the tables show, and how far they carry

- **Critical Thinking, English first language:** Mean 4.69 (1,300 candidates, 50%; median 4.7)
- **Critical Thinking, other first language:** Mean 3.73 (1,326 candidates, 50%; median 3.5, a gap of 0.95)
- **Problem Solving, English first language:** Mean 4.78 (Same candidates, median 4.8)
- **Problem Solving, other first language:** Mean 4.67 (Median 4.6, a gap of just 0.11)

The report attributes the Critical Thinking first-language difference to that module's higher language load. If English is not your first language, that one sentence is the most useful line in the document for ordering your preparation. A second set of figures points the same way. Candidates whose country of permanent residence falls under 'Other', meaning outside the UK and the EU, posted the highest Problem Solving mean of the three areas at 4.89 and the lowest Critical Thinking mean at 3.95. Two different cuts of the same cohort, both pointing at the same place to spend time.

### By area of residence

UK 1,189 candidates (45%), Other 1,210 (46%), EU 227 (9%). On Critical Thinking the EU group is strongest at 4.46, with the UK at 4.41 and Other lowest at 3.95. On Problem Solving the order inverts: Other highest at 4.89, UK 4.67, EU lowest at 4.13. The same candidates rank in opposite order on the two modules.

### By gender

62% of candidates identified as Man (1,632) and 37% as Woman (965). The Critical Thinking means are 4.29 and 4.02, a difference of 0.27; on Problem Solving they are 5.06 and 4.13, a difference of 0.94. In the item-level DIF analysis, no Critical Thinking item was flagged by gender, first language, UK school type or UK ethnicity, and only one Problem Solving item (820707) was flagged as C-category DIF, favouring candidates who identified as Man.

### By UK school type (UK candidates, groups of at least 100)

FE or Sixth Form College 512 (43%), Grammar 247 (21%), Private fee-paying 218 (18%), Academy or Secondary 197 (17%). Critical Thinking means run Private 5.13, Grammar 5.00, Academy/Secondary 4.20, FE/Sixth Form 3.89. Problem Solving means run Grammar 5.25, Private 5.08, FE/Sixth Form 4.37, Academy/Secondary 4.25. The report also notes that grammar-school candidates made up 21% of UK TARA candidates against a national figure of around 5% of state-funded secondary pupils.

### By free school meals and parental higher education

13% of UK candidates said they received free school meals (150 people), against almost 25% of pupils nationally who are eligible. That group averaged 3.76 on Critical Thinking and 4.03 on Problem Solving; the 931 candidates (78%) who said no averaged 4.56 and 4.78. On parental higher education, 79% answered yes (2,083 candidates) and averaged 4.33 on Critical Thinking, against 3.74 for the 346 (13%) who answered no. The Problem Solving gap is far smaller: 4.78 against 4.32.

### By ethnicity (UK candidates, groups of at least 50)

Asian 600 (50%), White 277 (23%), Mixed 121 (10%), Black 109 (9%), Other 82 (7%). On Critical Thinking, UK White is strongest at 4.76 with UK Mixed just behind at 4.74, and UK Black is the only group below 4.0 at 3.89. On Problem Solving the order changes: UK Asian leads at 4.84, then Other 4.75, Mixed 4.70 and White 4.63, with UK Black again the only group below 4.0 at 3.73. The orderings do not match across modules, which is itself evidence against treating the two as one ability.

> Two limits when reading group gaps. First, this page repeats the numbers the report prints and does not supply a cause the report itself does not state. Second, a group mean is not a prediction for a person: almost every group spans the full 1.0 to 9.0 range, with the UK Black group's Problem Solving maximum of 7.6 among the few exceptions. No group average should be used to forecast what any individual will score.

## 4. The clock: Problem Solving is the speeded module

- **00:40:00** Median time on Problem Solving, exactly the module limit
- **8%** Share leaving an item unreached on Problem Solving, against 1% on Critical Thinking
- **28%** Share with an unreached or 5-second item on Problem Solving, against 4% on Critical Thinking
- **Over 85%** Share of candidates using between 35 and 40 minutes

|  | Critical Thinking | Problem Solving |
| --- | --- | --- |
| Candidates in the timing analysis | 2,535 | 2,532 |
| Mean time | 00:38:44 | 00:39:43 |
| Median time (limit 00:40:00) | 00:39:49 | 00:40:00 |
| SD of time | 00:02:57 | 00:01:25 |
| Shortest time | 00:15:23 | 00:06:04 |
| Candidates with an unreached item | 33 (1%) | 217 (8%) |
| Unreached items: mean and range | 1.67 items, range 1 to 6 | 2.76 items, range 1 to 12 |
| Candidates with an unreached or 5-second item | 116 (4%) | 742 (28%) |
| Those items: mean and range | 2.23 items, range 1 to 9 | 2.55 items, range 1 to 12 |

The report's own judgement is that Problem Solving is more speeded than Critical Thinking. The two modules carry identical item counts and identical limits, yet the median Critical Thinking time is 00:39:49 while the median Problem Solving time is 00:40:00, the limit exactly. The practical reading is blunt: drill [Problem Solving](/exams/tara/guides/problem-solving) against a hard clock and decide your abandon rule before test day, while the [Critical Thinking](/exams/tara/guides/critical-thinking) bottleneck is more likely to be reading accuracy than the stopwatch.

> One caveat on the timing figures: the report excludes candidates whose module time exceeded 00:40:04, roughly 4% per module, on the assumption that they held an access arrangement. Across the cycle 102 candidates (around 4%) had access arrangements, 101 of them for extra time and/or pause-the-clock; the breakdown is 66 extra time, 28 extra time plus pause-the-clock, 7 pause-the-clock alone, 4 separate room and 7 other. The table above therefore describes candidates working to the standard limit.

## 5. What these numbers cannot tell you

### There is no pass mark and no cut-off

UAT-UK publishes no pass mark and the report contains no threshold of any kind. How each university weighs a TARA score against predicted grades and the rest of an application sits outside its scope. Reading the 75th or 90th percentile as a safe zone is something readers add; the report never says it. The distribution is for locating yourself relative to this cohort, not for finding a line that does not exist.

### Do not compare across cycles

UAT-UK requires the scaling to be fixed to each cycle's October ability distribution, so the transformation is redone every year. The report states plainly that this cycle's scaled scores cannot be compared directly with those from future events. Whenever you meet a figure from another year, establish which cycle it belongs to before you benchmark against it.

### The two modules cannot be added or averaged

The modules are calibrated and scaled separately, and the report states that their scaled scores are not comparable with each other and that no aggregate or total score exists. Universities see two independent numbers, and any attempt to fuse them into one has no basis. The ethnicity and residence tables even reverse their group orderings between modules, which is further evidence that the two are not measuring one thing.

### The Writing Task has no distribution at all

The Writing Task is not reported on the 1.0 to 9.0 scale, and this technical report carries no score statistics for it at all: no mean, no percentiles, no distribution. Any claim about a scoring standard for the TARA writing component does not come from this report.

### A decimal of difference may just be measurement error

Scaled-score reliabilities run 0.61 to 0.77 for Critical Thinking and 0.61 to 0.74 for Problem Solving, with standard errors of measurement between 0.87 and 1.24. Under classical test theory there is roughly a 68% probability that a true score lies within one SEM of the observed score and roughly 95% within two. The report also explains that with only 22 items per module these are short tests whose reliabilities are expected to fall below UAT-UK's other exams, and that anything above 0.70 is satisfactory at this length (the usual rule of thumb is 0.80). A gap between two candidates smaller than one SEM tells you almost nothing.

Turn the figures into a plan. If English is not your first language, put [Critical Thinking](/exams/tara/guides/critical-thinking) first: that is the gap the report ties to language load. Drill [Problem Solving](/exams/tara/guides/problem-solving) against a hard clock, which is what the 28% unreached-or-rushed figure is really telling you. For how the scale itself works, see [the scoring page](/exams/tara/guides/scoring), and for practice material, see [what to use instead of past papers](/exams/tara/guides/no-past-papers). To turn the correct answers from a practice set into an estimated score on this same scale, use the [TARA score calculator](/tools/tara-score-calculator).

> On item quality the report's verdict is positive: 100% of items met the criteria in both modules, well above the 80% target usually set for new items, and every item had a positive point biserial, ranging 0.10 to 0.57 with a mean of 0.38 for Critical Thinking and 0.07 to 0.58 with a mean of 0.36 for Problem Solving. A low score, in other words, cannot be blamed on a faulty paper. Every figure on this page comes from a printed table or sentence in UAT-UK's 2025/26 TARA technical report; nothing is read off a chart image.

## Questions
**What counts as a high TARA score?**

In each module, scores are scaled so that about 10% of candidates score above 7.0 (90th percentile fixed at 7.0) and the median is 4.5 (UAT-UK Explanation of Results, October 2025 and January 2026).
**Is there a TARA cut-off?**

UAT-UK does not provide grade boundaries or a pass mark. Universities use scores alongside the rest of your application.
**Can I add or average my two module scores?**

UAT-UK reports them as two separate scores; each module is equated and scaled separately, and no overall TARA score is issued.
**Is the Writing Task part of my score?**

No. It is unscored by UAT-UK; your response is sent to the universities whose courses require the TARA.
**Are January scores on a different scale?**

No. UAT-UK maps January scores to the same scale as October.

## Official sources
- [UAT-UK TARA Technical Report 2025-26, Table 4: scaled score summary statistics](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161805/UAT-UK-TARA-Technical-Report-2025-26.pdf)
- [Same report: scale construction, anchor points, scaling constants and post-equating](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161805/UAT-UK-TARA-Technical-Report-2025-26.pdf)
- [Same report, Tables 6 to 14: results by gender, residence, first language and school type](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161805/UAT-UK-TARA-Technical-Report-2025-26.pdf)
- [Same report, Tables 16 to 21: reliability, SEM, test time and unreached items](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161805/UAT-UK-TARA-Technical-Report-2025-26.pdf)
- [UAT-UK Annual Reports index (public entry point for all four 2025-26 reports)](https://esat-tmua.ac.uk/about-uat-uk/annual-reports/)
