Comprehensive Junior High School School Ability Test: Replacing the 412-Point Scale with Three Proficiency Levels

In May 2026, 180,000 ninth graders received report cards featuring only "Proficient, Basic, and Needs Improvement" alongside a single essay score. From the 412-point norm-referenced scale to the three-level criterion-referenced system introduced in 2014, the legal basis for this massive national exam is academic assessment; its use in school admissions is merely an added function by local districts. The hidden scores eventually resurface as A++ and cumulative points.

On May 17, 2026, as the bell rang for the final session, ninth graders across Taiwan concluded two days of the Comprehensive Junior High School School Ability Test (CAP) in a single afternoon. Results were mailed on June 5. The National Center for Educational Testing (NCET) at National Taiwan Normal University (NTNU) released statistics simultaneously: 16,293 students nationwide achieved "5A" (Proficient in all five subjects), accounting for 9.04% of all examinees. The most frequent combination was "5B," with 40,598 students, or 22.53%.1

The most counterintuitive aspect of this report card is that it provides almost no numerical scores. Examinees receive a rating for each of the five subjects—"Proficient (A), Basic (B), or Needs Improvement (C)"—plus a six-level score for an essay, with no total score, no percentile, and no ranking.2 This massive national written exam deliberately removes the most convenient number for comparison.

Its legal status is also less widely understood. The CAP functions as an academic proficiency certification mechanism conducted under the Regulations on the Assessment of Learning for Elementary and Junior High School Students. Its use in school admissions stems from ranking rules subsequently added by various school districts.2

I. Two Half-Days, Five Subjects, and One Essay

The 116th year (2027) CAP is scheduled for May 15 and 16; Saturday sessions run until after 4:00 PM, while Sunday sessions conclude in the morning.34 The first morning consists of Social Studies and Mathematics, followed by Chinese and Writing in the afternoon. The second morning covers Natural Science, then English Reading and English Listening. Break times are strictly scheduled: a 40-minute break after Social Studies and a lunch break of one hour and 50 minutes.3

Subject Exam Duration Number of Questions Question Type
Chinese 70 min 38–46 Multiple choice (4 options)
English Reading 60 min 40–45 Multiple choice (4 options)
English Listening 25 min 20–30 Multiple choice (3 options), played twice
Mathematics 80 min 25–30 23–28 multiple choice + 2 non-multiple choice
Social Studies 70 min 50–60 Multiple choice (4 options)
Natural Science 70 min 45–55 Multiple choice (4 options)
Writing Test 50 min 1 One essay

The configuration of subjects and question types reflects how examiners view three years of junior high learning. Aside from the writing test, English listening, and mathematics non-multiple choice questions, all other subjects are strictly four-option multiple choice. In other words, the vast majority of responses in the CAP involve picking one out of four options.3

The English section splits reading and listening into two independent booklets tested in two stages; for listening, each question is played twice with a few seconds of pause between repetitions.3 The rationale for this design is stated in official Q&As: if only reading were assessed, the results "could only provide limited explanation of a student's partial English ability, resulting in seriously insufficient assessment validity."3

II. From 412 Points to Three Proficiency Levels

Before the CAP, junior high students in Taiwan took the Junior High School Basic Ability Test (JHBAT). The difference between the two is more than just a name change. The JHBAT used norm-referenced scoring, with a maximum of 80 points per subject and 12 points for writing, totaling 412 points. The CAP uses criterion-referenced scoring, where grades are divided into three proficiency levels, while writing is scored on a scale of one to six.2

Item Comprehensive Junior High School School Ability Test (CAP) Junior High Basic Ability Test (JHBAT)
Legal Basis Regulations on the Assessment of Learning for Elementary and Junior High School Students Regulations on Diversified Admission for Senior High Schools and Vocational Schools
Primary Function Academic certification for instructional adjustment and personalized guidance Basis for diversified admission channels
Implementation Since 2014, once annually in May Twice annually from 2001 to 2011; changed to once annually in 2012
Basis of Items 9-year curriculum (2014–2021); 12-year basic education curriculum (from 2022) Competency indicators of the 9-year curriculum
Difficulty Setting Moderate Medium-easy
Scoring Method Criterion-referenced Norm-referenced

Since its launch in 2014, the NCET at NTNU has been responsible for item writing, booklet assembly, grading, and scoring. While national exam administration was originally rotated among six special municipalities, the Ministry of Education announced on July 17, 2019, that starting August 1, 2019, NTNU would be commissioned to coordinate all operations.2

The official explanation for this design is straightforward: by reducing grade levels to three, they "appropriately reduce exam pressure" while using criterion-referenced scoring to maintain the function of academic monitoring.2 This represents a recurring tension in Taiwan's educational reform: seeking a balance between "reducing pressure to revitalize learning" and "ensuring quality to maintain competitiveness."2

III. What Exactly is Being Tested?

The principles for item writing are stated in official documents and are quite restrained. Test items are drawn from students' learning and life experiences; they focus only on key learning points assessed through paper-and-pencil tests and require that items do not depend on specific textbook versions, so that any student who has mastered the competency can answer regardless of their textbook.3

The Social Studies section best illustrates this approach. Officials state that questions emphasize understanding, with fewer items testing pure memorization, focusing instead on students' ability to interpret and translate data from charts. Beyond being close to students' lived time and space, items appropriately integrate issues such as gender equality, human rights, the environment, and Indigenous education, aiming to guide students to care about their own society and global issues.3

Natural Science is divided into two layers: "Learning Performance" and "Learning Content." Learning Performance includes scientific cognition, inquiry skills, and scientific attitudes and nature; Learning Content covers three major themes: the composition and characteristics of the natural world, phenomena/laws/functions, and sustainable development. The goal is to test whether students can apply concepts rather than just memorizing terms.3

Mathematics is the only subject with non-multiple choice questions in addition to multiple choice ones; these two non-multiple choice questions account for 15% of the total math score after weighting. Compared to multiple choice, there are fewer non-multiple choice items because the grading process is time-consuming, and officials have noted that "the proportion may be adjusted progressively based on future needs."3

Taken together, these principles show that the CAP does not intend to test entire textbooks. It tests the key learning points of the fourth stage of the curriculum, and examiners admit that as long as students truly master the competencies, questions should not be hindered by any single version of teaching materials.3

IV. Criterion-Referenced: Comparing Against a Threshold, Not Your Neighbor

Criterion-referenced means that a team of subject professors and middle school teachers first sets competency thresholds for each level, then works backward to determine how many questions must be answered correctly for each level. Students are compared against the threshold, not against their fellow examinees.3 The official definitions for the three levels are also clear: "Proficient" means "mastery of the knowledge and abilities learned during the junior high education stage in that subject"; "Basic" means "possessing fundamental academic ability"; and "Needs Improvement" means "not yet possessing such ability."3

However, because three levels appear too coarse for school admissions, additional markers were added. The Proficient (A) and Basic (B) levels are each split in half: the top 25% of those who answer more questions correctly are marked as "++", while the 26th to 50th percentiles are marked as "+". This results in four designations—A++, A+, B++, and B+—while "Needs Improvement" (C) has no such marker.3

Specifically, for the thresholds in 2026 (the 115th year), the numbers were as follows: to reach Proficiency in Chinese, one must answer at least 36 questions correctly; the A++ threshold is 40 questions, meaning a maximum of two errors. The Social Studies proficiency threshold is 48 questions, with A++ at 52. For Natural Science, the proficiency threshold is 43, and A++ is 47.5

English and Mathematics scores are more complex due to weighting. English calculates a weighted score where Reading accounts for 80% and Listening for 20%. To reach Proficiency in Reading, one must answer 38 questions correctly; Listening only distinguishes between Basic and Needs Improvement (the Basic threshold is 13). To achieve overall Proficiency, the weighted score must reach 90.7, while A++ requires 98.14.5 Mathematics weights multiple choice at 85% and non-multiple choice at 15%; the overall Proficiency threshold is 77.10, and the A++ threshold is 91.60.5

Tseng Fen-lan, Deputy Director of the NCET, told CNA at the time that the CAP produces test booklets of similar difficulty every year, so the thresholds for each subject do not vary much annually—"usually just by one question or so."5 This explains why CAP thresholds can be announced annually without drastic fluctuations, and also why a single extra error might cause a student to drop a marker.

V. Why A++ is Not Exactly One-Quarter

The rule for markers states "the top 25%," but the actual published proportions are almost never exactly 25%. The official Q&A uses Natural Science from 2026 as an example: that year, reaching Proficiency required at least 43 correct answers, which 31,473 people met. The number of A++ students should theoretically have been 7,869. However, only 7,194 people answered 48 or more questions correctly—short by 675 people. Consequently, the A++ threshold was lowered to 47. Based on the principle of fairness that "those with the same number of correct answers should have the same marker," all 4,356 people who answered 47 correctly were included in the A++ group, bringing the total to 11,550, or 36.70% of the Proficiency level.3

In other words, the marking method first divides by proficiency level, then adjusts the proportion based on the distribution of correct answers; if it cannot be perfectly balanced, the proportion is allowed to deviate. This technical detail is rarely mentioned in news reports, yet it determines whether many examinees ultimately receive an A++ or an A+.

VI. The Admissions Scene: CAP is Just One Slot in the Ranking

The true source of pressure from the CAP lies not in the exam itself, but in the "excess quota" ranking for non-exam admission. The Ministry of Education's regulations state that whether the CAP is used must be "determined by each district according to local conditions," but it must not be the sole criterion and must be planned alongside other ranking items. If total markers (total cumulative points) are used, they should follow the total score from "Diversified Learning Performance" and precede the individual subject marker rankings.6

In terms of system design, the CAP is deliberately placed after "Diversified Learning Performance," and districts are only permitted to use it as one component. However, how much weight that component carries is decided by each district. After the May 2026 CAP, the cram school industry estimated based on the Taipei-New Taipei-Keelung area ranking rules that the admission threshold for Taipei Municipal Jianguo High School was approximately 34.6 points and for Taipei First Girls' High School was approximately 33.6 points.7 The fact that such estimates are possible demonstrates that in those districts, once CAP markers are converted into points, they effectively carry significant ranking weight.

The overall score distribution also shows where the admission pressure actually lies. In 2026, "5A" accounted for 9.04%, while "5B" was the largest group (22.53%).1 That massive middle block of students who are "all B" is neither part of the discussion regarding elite high schools nor part of the remedial discussions for those needing improvement.

VII. One Essay: From Level Zero to Level Six

The writing test is the only subject in the CAP without multiple-choice questions or a fixed prompt. The 2026 prompt was "Giving things a new appearance," providing three examples: using bartering instead of money at a school fair, turning old wall tiles into coasters, and processing abundant bananas into new products. Examinees were required to combine their own experiences or observations to write about what they would give a new appearance.8

That year, 1,262 people received a level six score, accounting for 0.71%, the highest in nearly three years. Tseng Fen-lan, Deputy Director of the NCET, stated that reaching level four or above meets the standards required by the 108 Curriculum; this year, 78% of examinees met that standard.8 On the other end, 2,187 people received a level zero score, accounting for 1.23%, a significant increase from last year's 753. Lin Hsueh-ling, a teacher at Wenzao Ursuline University of Languages and a grading representative, analyzed that most of these students merely copied or paraphrased the three examples provided in the prompt without writing about their own experiences and feelings.8

The NCET publishes sample essays every year, which to some extent serves as a substitute for "standard answers." One selected sample from 2026 featured a student writing about how, due to cerebral palsy, they had worn pair after pair of corrective shoes since childhood; upon seeing those old, smelly shoes in a storage room as an adult, they initially wanted to throw them away but eventually cleaned and arranged them one by one as glorious marks of their hard work. Another sample described working with a father to transform disposable chopsticks into a Zhuge repeating crossbow; the moment it successfully fired, childhood memories flooded back.8

An essay has only six levels, which may seem more granular than the three proficiency levels for the five subjects, but in reality, it is coarser: it compresses "the ability to write a structurally complete article based on personal experience" into a single number. Yet, this field is also the only place on the entire report card where an examinee's personal story remains visible.

VIII. Who is Left in "Needs Improvement"

The original purpose of the CAP is academic monitoring, so the proportion of "Needs Improvement" has always been its most honest metric. In the 2015 CAP, there were 285,295 examinees. Li Chun-jen, an Associate Professor in the Department of Educational Psychology and Counseling at NTNU, noted in a commentary published by the Taiwan Educational Research Center (TERIC) that the "Needs Improvement" rate for Mathematics reached 33% that year, with 35.8% of non-multiple choice math questions receiving a zero score. He stated bluntly that "multiple choice is a four-option answer," and under loose standards, answering fewer than 6 out of 25 correctly is almost equivalent to pure guessing.9

The fairness of English listening remains another perennial controversy. The official Q&A dedicated an entire section to responding to whether it is "unfair to students in remote areas": the reason given is that the curriculum requires equal emphasis on listening, speaking, reading, and writing; all English teaching materials approved by the Ministry of Education come with listening CDs, and the English listening players used in previous CAP exams are transferred to various junior high schools.3 Whether this defense holds up depends on something that will never be written in statistical figures: whether those players were actually ever turned on.

IX. After the CAP

The schedule for the 116th year (2027) CAP has already been announced; after the two days of testing on May 15 and 16, the next steps are the selection of preferences and distribution for non-exam admission.3 For examinees and parents, there are three other things to understand beyond the report card: how the ranking rules in their specific district are calculated, where last year's admission points for various schools fell, and how much weight the CAP score carries within that specific slot.

The CAP hides scores to prevent junior high teaching from devolving into repetitive drilling. But as long as "excess quota" ranking exists, those hidden scores will regrow in another form, becoming A++, cumulative points, and estimated admission targets. The system's execution hasn't failed; the two goals of "reducing pressure" and "maintaining competitiveness" can only ever be partially achieved simultaneously.

Further Reading

References

  1. CAP Results Released: 712 Students Nationwide Achieve 5A; Writing Level Six Highest — CNA report from June 5, 2026, listing the 2026 figures: 16,293 students (9.04%) achieved 5A, 40,598 (22.53%) achieved 5B, 1,262 reached writing level six, and 2,187 received level zero.↩2
  2. Comprehensive Junior High School School Ability Test: Understanding the CAP — Official explanation of its background and four objectives, including a "Comparison Table between the Educational CAP and the JHBAT," listing differences in legal basis, function, item basis, difficulty, and scoring methods, as well as the transfer of national exam administration to NTNU starting August 1, 2019.↩23456
  3. Q&A for the 116th Year Comprehensive Junior High School School Ability Test — Officially published by NTNU NCET, detailing the history of the CAP, exam times and question counts for each subject, the May 15–16 schedule for 2027, definitions of the three proficiency levels, the marking methods for A++/A+/B++/B+, and a proportional explanation using Natural Science from 2026 as an example.↩2345678910111213141516
  4. Official Website of the Comprehensive Junior High School School Ability Test — The homepage for national exam administration, announcing annual schedules, grade inquiries, level-to-marker correspondence tables, and statistical data download services.↩
  5. CAP Correct Answer Table Released: A++ Thresholds for Each Subject Announced — CNA report from June 2, 2026, listing the 2026 Proficiency and A++ question thresholds for each subject, the 80%/20% weighting for English Reading/Listening, the 15% weight for Math non-multiple choice, and quoting NCET Deputy Director Tseng Fen-lan on booklet difficulty.↩234
  6. Implementation Matters for Senior High School Non-Exam Admission — Ministry of Education regulation, stipulating that whether the CAP is used must be determined by local districts, it cannot be the sole criterion, and regulating the order of total markers and total points in excess quota ranking.↩
  7. Using CAP Scores for Non-Exam Admission: Estimated Jianguo Threshold at 34.6, Taipei First Girls' at 33.6 — CNA report from May 17, 2026, estimating admission points for elite high schools based on the Taipei-New Taipei-Keelung area ranking rules, reflecting the practical ranking weight of CAP markers after conversion.↩
  8. CAP Writing Level Six Highest in 3 Years: Student with Cerebral Palsy Turns Orthopedic Shoes into a Mark of Glory in Selected Sample Essay — CNA report from June 5, 2026, detailing the 2026 writing prompt "Giving things a new appearance" and its three examples, noting 1,262 students (0.71%) reached level six and 2,187 (1.23%) received level zero, quoting Tseng Fen-lan and grader Lin Hsueh-ling on the causes of level zero scores.↩234
  9. Alas, nothing is greater than a dead heart; nothing is more sorrowful in learning than helplessness — Commentary by Li Chun-jen, Associate Professor at NTNU Department of Educational Psychology and Counseling, published by TERIC, discussing the guessing issue in multiple choice using 2015 CAP data (33% "Needs Improvement" in Math and 35.8% zero scores in non-multiple choice).↩
About this article This article was collaboratively written with AI assistance and community review.
Tags
Comprehensive Junior High School School Ability Test CAP non-exam admission 12-year basic education school admissions education
Share

Further Reading

You might also like

Society Evolving · Awaiting fact-check

The College Entrance Examination: A Three-Day Test That Condenses Three Years of High School into a Single Score Sheet

From the 1954 Joint Entrance Examination to the 1994 College Entrance Examination and today’s Subject-Based Test, understanding how Taiwan’s university admissions system balances single-standard metrics with diversified talent selection, and clarifying the distinct paths of the CET, SBAT, and retakes.

閱讀全文
Society Evolving · Awaiting fact-check

Learning Poverty: What We Haven't Learned in Classrooms

The World Bank introduced "Learning Poverty" in 2019 to describe a problem harder to see than school dropout: a 10-year-old who attends school regularly, is compliant, and turns in homework, but cannot read a simple story appropriate for their age. In February 2023, the World Bank updated its global learning poverty rate to remain at 70%. Taiwan's PISA 2022 ranked in the top five globally, with the math scores of the bottom 20% catching up to the OECD average, appearing as a major turnaround. However, the same PISA report showed a gap of 54 points between urban students (571) and rural students (517); a 2019 survey by the Bo-You Foundation across 13 rural junior high schools found that 70% of students in 6 schools scored C in English, and 70% in 5 schools scored C in Math. The Taiwanese version of learning poverty hides within this gap and has not yet been officially measured.

閱讀全文
Lifestyle Evolving · Awaiting fact-check

Taiwan's Education System: An Island Experiment from Exam Hell to a Maze of Diversity

For 30 years, Taiwan tried to dismantle the centralized examination system; the re-examination classes in Nanyang Street shrank from 48 to 3, while private tutoring centers increased by 960 after the new curriculum was implemented. How does an island repeatedly conduct social experiments between academic pressure and educational equity?

閱讀全文
Society Evolving · Awaiting fact-check

Where Did Taiwan's Moral Education Go: A Cancelled Class and Two Decades of Unclaimed Responsibility

Before 1993, every elementary school student in Taiwan took 'Morality and Health' each week, and every junior high student took 'Citizenship and Morality.' The Nine-Year Integrated Curriculum abolished both subjects, 'infusing' moral education into seven major learning areas. The result: in nearly 60 pages of social studies curriculum guidelines, the word 'morality' barely appeared; 78% of teachers said moral education was what most needed strengthening, yet no subject teacher felt it was their responsibility.

閱讀全文