PISA, TIMSS, PIRLS and ICILS Compared: Purpose, Age Groups and Scoring
PISA, TIMSS, PIRLS and ICILS examine different stages and dimensions of student learning. Their target populations, assessed domains, score scales and reporting cycles are not interchangeable, even when two studies report results near 500 scale points.
Assessment Snapshot
The four studies cover different student populations and different forms of literacy or subject achievement.
Released Assessment Cycles Available for Trend Analysis
Cumulative counts use the official assessment-year sequence for cycles with internationally released results.
Cumulative Released Cycles by Study
PISA and TIMSS currently provide the longest recurring trend series among the four studies.
PISA 2025 and PIRLS 2026 are not added to the released-cycle count because their international results had not yet been published.
Trend analysis
- TIMSS has eight released cycles from 1995 through 2023.
- PISA has eight released cycles from 2000 through 2022; the 2021 cycle moved to 2022.
- PIRLS has five released cycles covering 2001 through 2021.
- ICILS has three released cycles: 2013, 2018 and 2023.
- Trend depth differs by study, so the same start year cannot be assumed across the four series.
Purpose, Population and Reporting Structure
TIMSS is separated by grade and subject, while ICILS reports CIL and optional CT as independent constructs.
| Study component | Organiser | Primary purpose | Target population | Selection basis | Assessed domain | Latest results | Score reporting |
|---|---|---|---|---|---|---|---|
| PISA core domains | OECD | Measure how students apply knowledge and reasoning in varied contexts | Age 15 years 3 months to 16 years 2 months; Grade 7 or higher | Age | Mathematics, reading and science | 2022 | Separate subject scales and domain-specific proficiency levels |
| TIMSS Grade 4 Mathematics | IEA | Measure mathematics achievement related to taught school content | Grade representing four years of schooling from the first ISCED Level 1 year | Grade | Mathematics content and knowing, applying and reasoning | 2023 | Scale points and four international benchmarks |
| TIMSS Grade 4 Science | IEA | Measure primary-level science achievement and reasoning | Grade representing four years of schooling from the first ISCED Level 1 year | Grade | Science content and knowing, applying and reasoning | 2023 | Scale points and four international benchmarks |
| TIMSS Grade 8 Mathematics | IEA | Measure lower-secondary mathematics achievement related to taught content | Grade representing eight years of schooling from the first ISCED Level 1 year | Grade | Mathematics content and knowing, applying and reasoning | 2023 | Scale points and four international benchmarks |
| TIMSS Grade 8 Science | IEA | Measure lower-secondary science achievement and reasoning | Grade representing eight years of schooling from the first ISCED Level 1 year | Grade | Science content and knowing, applying and reasoning | 2023 | Scale points and four international benchmarks |
| PIRLS Reading | IEA | Measure reading comprehension near the transition from learning to read to reading to learn | Grade representing four years of schooling, usually Grade 4 | Grade | Literary reading and reading to acquire and use information | 2021 | Reading scale points and four international benchmarks |
| ICILS CIL | IEA | Measure the use of computers to investigate, create, participate and communicate | Eighth year of schooling, usually Grade 8 | Year of schooling | Computer and information literacy | 2023 | CIL scale points and four proficiency levels |
| ICILS CT | IEA | Measure computational formulation and the development of algorithmic solutions | Eighth year of schooling, usually Grade 8 | Year of schooling | Computational thinking; optional for participating systems | 2023 | Independent CT scale and four proficiency levels |
Table analysis
- PISA is the only study in the comparison that selects students primarily by age.
- TIMSS produces four main achievement combinations: two grades multiplied by two subjects.
- PIRLS focuses on reading comprehension at an earlier stage than PISA reading.
- ICILS CIL and ICILS CT are reported on separate scales.
- CT participation is optional, so its country coverage is narrower than the ICILS CIL coverage.
- The latest released assessment year is not identical across the four programmes.
Age and Grade Sampling Rules
Similar student ages do not make the target populations identical. School-entry age, grade repetition and national grade structures affect coverage.
| Assessment | Main target | Eligibility rule | National variation | Interpretation effect |
|---|---|---|---|---|
| PISA | 15-year-old enrolled students | Approximately 15 years 3 months to 16 years 2 months and enrolled in Grade 7 or higher | Eligible students can be distributed across several grades | Represents an age cohort still enrolled in education rather than one common grade |
| TIMSS Grade 4 | Fourth year of formal primary schooling | Grade representing four years from the first year of ISCED Level 1 | Grade 5 or 6 may be used when needed to match the intended achievement level | Student age can differ across participating systems |
| TIMSS Grade 8 | Eighth year of formal schooling | Grade representing eight years from the first year of ISCED Level 1 | Grade 9 may be used in an approved administration | Not equivalent to the full PISA age-based population |
| PIRLS | Fourth year of formal schooling | Grade representing four years from the first year of ISCED Level 1 | Grade 5 or 6 may be used to better match the intended reading level | The assessed grade is expected to be near the reading-to-learn transition |
| ICILS | Eighth year of schooling | Usually Grade 8 when the average student age is at least 13.5 years | Grade 9 is used when the Grade 8 average age is below 13.5 | The grade rule protects age comparability without becoming an age-based survey |
Assessment analysis
- PISA students can occupy different grades while remaining part of the same age-defined population.
- TIMSS Grade 8 and PISA can overlap in age, but their sampling definitions remain different.
- PIRLS and TIMSS Grade 4 can assess students in the same school year while measuring different domains.
- ICILS uses a minimum average-age rule when determining whether Grade 8 or Grade 9 is appropriate.
- Differences in school-entry age and grade repetition can alter the composition of grade-based and age-based samples.
What Each Assessment Is Designed to Measure
The same subject label can represent different task demands and reporting purposes.
PISA
Central question: Can 15-year-olds use knowledge and reasoning when a task is presented in an unfamiliar or applied context?
Mathematics tasks can require students to formulate a situation mathematically, employ concepts and procedures, and interpret or evaluate the result.
TIMSS
Central question: What mathematics and science knowledge and skills have students developed at defined stages of schooling?
Content domains are combined with the cognitive domains of knowing, applying and reasoning.
PIRLS
Central question: How well can primary-school students construct meaning from literary and informational texts?
Processes include retrieving stated information, making inferences, integrating ideas and evaluating text content or form.
ICILS
Central question: How well can students investigate, evaluate, create and communicate information with digital tools?
The optional CT component examines the formulation of problems and the development or evaluation of algorithmic solutions.
PISA emphasises the application of knowledge among an age cohort. TIMSS places greater weight on mathematics and science content associated with defined school stages. A system can therefore occupy different relative positions without the results being internally inconsistent.
Scoring Systems and Proficiency References
Scale points are model-based reporting units. They are not percentages, examination grades or a common currency across studies.
| Scale | Reference centre | Reference SD | Reported levels | Published cut points | Comparison rule |
|---|---|---|---|---|---|
| PISA mathematics, reading and science | Approximately 500 | Approximately 100 | Domain-specific proficiency levels | Different for mathematics, reading and science | Compare the same domain across participating systems or linked cycles |
| TIMSS mathematics and science | 500 | 100 at scale establishment | Low, Intermediate, High and Advanced international benchmarks | 400, 475, 550 and 625 | Keep grade and subject fixed when comparing results |
| PIRLS reading | 500 | 100 at scale establishment | Low, Intermediate, High and Advanced international benchmarks | 400, 475, 550 and 625 | Compare reading results within PIRLS rather than with TIMSS scores |
| ICILS CIL | 500 | 100 in the 2013 reference metric | Below Level 1 and Levels 1–4 | 407, 492, 577 and 662 | Compare CIL only with the linked CIL scale |
| ICILS CT | 500 | 100 in the 2018 reference metric | Below Level 1 and Levels 1–4 | 330, 440, 550 and 660 | CT and CIL are independent scales and cannot be directly equated |
Score analysis
- 500 is not a pass mark in any of the four studies.
- PISA has no theoretical fixed minimum or maximum score.
- TIMSS and PIRLS use the same numerical benchmark locations, but the benchmark descriptions refer to different subject constructs.
- A TIMSS mathematics score of 550 and a PIRLS reading score of 550 do not describe the same ability.
- ICILS CIL and CT each use a 500-centred reference metric, but the two scales were established with different reference populations.
- Subtracting a PISA score from a TIMSS score does not produce a valid curriculum or application gap.
PISA mathematics 500, TIMSS mathematics 500, PIRLS reading 500 and ICILS CIL 500 are positions on separate reporting scales. The numbers cannot be averaged, subtracted or converted into one shared achievement score.
How International Scale Scores Are Produced
The assessment design measures system-level distributions without requiring every student to answer every item.
| Stage | Process | Function | Interpretation limit |
|---|---|---|---|
| 1 | International item pool | Covers more content and difficulty levels than one student could complete | One student does not receive the entire international assessment |
| 2 | Rotated modules | Assigns different combinations of tasks to sampled students | Raw correct-answer totals are not directly comparable across all module combinations |
| 3 | Item-response modelling | Places item difficulty and population achievement on a linked reporting scale | Scale points are model-based rather than simple percentages |
| 4 | Plausible values | Represent a range of possible achievement values for population analysis | They should not be treated as exact diagnostic scores for individual students |
| 5 | Sampling weights | Adjust estimates so the sample represents the defined student population | Results apply to the covered target population, not automatically to every child |
| 6 | Standard errors and confidence intervals | Quantify uncertainty around estimated means, percentages and differences | Small score or rank differences may not be statistically distinguishable |
Method analysis
- International assessment scores are designed mainly for population and subgroup estimates.
- Rotated modules increase content coverage without extending each student’s testing time to the full item pool.
- Plausible values preserve uncertainty that would be lost by assigning one exact ability estimate to every student.
- Survey weights and replicate methods are required for valid national means, percentages and standard errors.
- A country’s apparent rank should be read with its confidence range rather than as one exact league-table position.
Assessment Cycle History and Current Status
Assessment year and publication year can differ. Active studies should not be reported as released result cycles.
| Study | Released cycle years | Released cycles | Latest released year | Current or next cycle | Publication status | Planned interval |
|---|---|---|---|---|---|---|
| PISA | 2000, 2003, 2006, 2009, 2012, 2015, 2018, 2022 | 8 | 2022 | PISA 2025; PISA 2029 follows | Initial 2025 results scheduled for 8 September 2026; selected additional results follow later | Four years from 2029 |
| TIMSS | 1995, 1999, 2003, 2007, 2011, 2015, 2019, 2023 | 8 | 2023 | TIMSS 2027 | 2023 international results and database released | Four years |
| PIRLS | 2001, 2006, 2011, 2016, 2021 | 5 | 2021 | PIRLS 2026 | Main survey in 2026; international results planned for December 2027 | Five years |
| ICILS | 2013, 2018, 2023 | 3 | 2023 | ICILS 2028 | 2023 international results and database released; 2028 preparation underway | Five years |
Source-year analysis
- PISA moved from 2021 to 2022 after the pandemic-related postponement.
- The PISA sequence changes from a three-year interval through 2025 to a four-year interval beginning with 2029.
- TIMSS provides 28 years between its first 1995 cycle and the 2023 cycle.
- PIRLS 2026 is an active assessment cycle, but 2021 remains the latest year with released international results.
- ICILS 2028 is expected to add greater attention to AI literacy and digital wellbeing while retaining CIL and CT.
- Assessment year should be stored separately from report publication year in comparative databases.
Interpretation Boundaries
Several common comparisons fail because they ignore the target population, domain definition or reporting scale.
Age and grade are not substitutes
PISA’s age cohort can span several grades. TIMSS, PIRLS and ICILS identify a schooling stage, even when the students’ average ages are similar.
500 is not 50 percent correct
A 500-point result reflects a location on a constructed scale. It does not indicate that half of the questions were answered correctly.
One overall education score is not reported
PISA reports mathematics, reading and science separately. TIMSS separates grade and subject. ICILS separates CIL and CT.
Rank differences contain uncertainty
Sample-based means have standard errors. Adjacent positions can fall within overlapping confidence ranges.
System averages hide distributions
Two systems with similar means can have different proportions below basic levels, at advanced levels or across student subgroups.
Association is not causation
Context questionnaires identify relationships between achievement and student, home, teacher or school characteristics. They do not by themselves prove a causal effect.
Source and Methodology
Purpose, domains, target population, cycle sequence and PISA 2025 publication schedule.
Scale construction, approximate 500-point centre, approximate 100-point standard deviation, proficiency levels and ranking uncertainty.
Age eligibility, subject scales and interpretation of PISA score points.
Grade 4 and Grade 8 target definitions, mathematics and science coverage, four-year cycle and curriculum-related design.
Scale centre and international benchmark points of 400, 475, 550 and 625.
Grade 4 reading target, reading purposes, comprehension processes and five-year assessment cycle.
Latest released reading cycle, international benchmarks and participating-system documentation.
Eighth-year target population, CIL definition, optional CT component and study-cycle information.
CIL and CT reference scales, proficiency cut points, participating systems and independent-scale limitations.
PISA 2025, TIMSS 2027, PIRLS 2026 and ICILS 2028 status information is drawn from the responsible OECD or IEA programme.
The comparison covers study purpose, student population, domain structure, reporting scales and cycle history. It does not merge country scores from different assessments.
Cumulative cycle counts in the chart were calculated from the official released assessment-year sequences. Active cycles without published international results were excluded.
All score thresholds are shown in their native assessment scales. No score conversion, cross-assessment ranking or combined education index was calculated. Assessment years refer to the study cycle rather than the date on which the international report or public database became available.
PISA, TIMSS, PIRLS and ICILS results describe sampled populations that meet each study’s eligibility rules. Scale points are not percentages or shared measurement units. Cross-study score subtraction, averaging and direct rank comparison are invalid. Country comparisons should retain the assessment year, domain, grade or age group, standard error, participation status and population-coverage notes.