The Harland Review · The Modern Student

High Scores, Hidden Gaps

高分背後,看不見的落差
For
Parents of secondary students bound for university
Reading time
14 minutes
Last updated
August 2026
The Reframe

A grade is a promise. Lately the promise and the skill have come apart.

In November 2025, a working group at the University of California, San Diego published a report on the students arriving in its first-year mathematics courses. One figure is worth sitting with. The students placed into the university's most remedial math course had earned, on average, a high-school math GPA of 3.65. An A-minus record. And yet, on a skills check given to an earlier version of that same course, a quarter of the students could not solve 7 + 2 = ___ + 6. More than sixty percent could not divide a fraction by two.

That is not a story about lazy or unintelligent young people. It is a story about a measurement coming loose from the thing it is supposed to measure. A grade is meant to be a promise: it tells a parent, a university, and the student that a certain body of understanding is in place. When a transcript full of strong grades sits on top of arithmetic a ten-year-old is expected to handle, the promise has quietly stopped being kept.

This is the preparation gap, and it is the subject of this piece. Not the comfortable headline that students today are weaker than students before them, which is an old complaint and a largely mistaken one. The narrower, harder claim is this. The credential a child carries is drifting away from the capability underneath it. The drift is widest at the lower end, and it is growing. And the kind of preparation easiest to buy builds the visible score faster than the durable understanding. We will make the case mostly through mathematics, where the evidence is sharpest and the gap is least possible to hide, and we will be honest about where the argument should stop.

What We See
Two students with the same report card.

Working one student at a time, week after week, you see something a class of thirty hides. Two students can arrive with nearly identical transcripts and be in quite different places.

The first can reproduce a method flawlessly. Given a problem that matches a worked example, the answer comes out clean and correct. Move one step off the script, ask why the method works, or pose the same idea inside an unfamiliar problem, and the floor gives way. The understanding was never there. What was there was a well-drilled procedure, and procedures hold right up until the moment they are asked to bend.

The second student is sometimes slower and sometimes carries a slightly lower grade. But this student can explain why a step is taken, can catch an answer that is obviously wrong, and can carry a familiar idea into an unfamiliar setting. When the next year's material arrives, this is the student who is ready for it.

The report cards do not separate these two. The first child's parents are often the more confident of the two, because the grades say everything is fine. The work tells a different story, and the work is what the next course, and the university after it, will demand.

At a Glance

What the numbers show

A set of figures families ask about, each drawn from the awarding body or study named. Read together, they describe a credential and a capability moving apart.

US grades vs. skill
3.02 → 3.32
Average high-school math GPA rose this much, 2010–2022, while test scores did not, per ACT
US seniors proficient
22%
Twelfth-graders at or above proficient in math, 2024, among the lowest in 20+ years, per NAEP
Taiwan, world rank
3rd of 80
Chinese Taipei mean mathematics score, PISA 2022, per the OECD
Taiwan, internal spread
294 points
Gap between Taiwan's top and bottom math performers, the widest of any PISA 2022 system, per the OECD
Confidence, worldwide
55%
Eighth-graders reporting they are "not confident" in math, TIMSS 2023, per the IEA
Taiwan, productive skill
34% vs. 56%
GEPT High-Intermediate pass rate, speaking and writing vs. listening and reading, 2023, per the LTTC
US college math readiness
~30%
Tested 2024 graduates meeting the ACT math benchmark, per ACT
Taiwan tutoring sector
17,710
Registered cram schools in 2025, a five-year high, 88% focused on test prep, per Ministry of Education data
The Picture That Matters

A world-class average can hide the widest gap on earth

Taiwan's mathematics results are, by any fair reading, excellent. In the 2022 round of the OECD's PISA assessment, fifteen-year-olds in Chinese Taipei posted the third-highest mean math score of any of the eighty systems tested. Nearly a third of students reached the top two proficiency bands, against an international average closer to one in ten. The floor is strong too: the share of students below basic proficiency is among the lowest in the world. None of what follows is an argument that Taiwanese students are weak. They are not.

But an average is a single number laid over a crowd, and it can hide how spread out the crowd is. On the same 2022 assessment, the distance between Taiwan's strongest math performers and its weakest was the widest of any participating system in the world. A national headline of excellence and the widest internal gap on earth are both true at once, and they are describing the same students.

Mathematics performance, PISA 2022 · score range from lowest to highest decile
OECD averagea typical system
mean 472
Chinese Taipei3rd in the world
mean 547
lower scores294-point spread, Taiwan: widest of 80 systemshigher scores
547
Taiwan mean math score, 3rd of 80 systems
294
Points between top and bottom deciles, the widest gap of any PISA 2022 system
32%
Top performers, against an OECD average of 9%
An average tells you how a system does. The spread tells you how different two children inside it can be, even when their report cards look the same.

The OECD noted something further. Between 2018 and 2022, that gap in Taiwan grew wider, and it grew wider in a specific way: the strongest students pulled further ahead while the weakest did not move. A rising average does not mean a rising floor. It can mean the top is climbing while a tail of students is left where it was, carried along on grades that keep promising more than they hold.

Why Mathematics Tells On Us

How a high score sits on a shallow base

To see how a strong grade and a weak foundation can coexist, it helps to be precise about what understanding mathematics involves. The clearest account remains the one set out in Adding It Up, the National Research Council's 2001 study of how children learn mathematics. It describes proficiency as five strands woven together: conceptual understanding, procedural fluency, strategic competence, adaptive reasoning, and a productive disposition toward the subject. The strands are meant to develop in concert and to support one another.

The study's central observation is that they often do not. Students reliably become fluent in procedures faster than they grow to understand them. You can learn the steps of an algorithm and execute them perfectly without grasping why they work, and for a while nothing exposes the difference. The report gives the texture of it directly: roughly half of thirteen-year-olds could complete a straightforward computation, but far fewer could reason about the numbers underneath. Only about a third, asked for a number between 0.03 and 0.04, could name one.

A test built to reward procedure will happily certify the student who has the steps and not the understanding. And here mathematics is unlike most subjects, because it is cumulative. Each year's material is built on the last. Weak fractions make algebra harder than it should be. Weak algebra makes calculus nearly impossible. A conceptual gap does not announce itself. It sits quietly under a column of good grades until a later, harder course finally asks the student to use the understanding that was never built, and the whole structure comes down at once. This is precisely the sequence the university instructors describe. The bill for a shallow foundation does not arrive when the foundation is poured. It arrives years later, in a first-year lecture hall.

The Scoreboard Rises On Its Own

When the grade climbs and the skill does not

If grades and understanding were tracking each other, rising grades would be cause for celebration. The evidence is that they are not tracking. Analyzing students between 2010 and 2022, ACT found that the average high-school math GPA in the United States rose from 3.02 to 3.32. That was the largest increase of any subject. Over the same span, performance on its own standardized math measure did not improve. The grade went up. The skill it was supposed to represent stayed where it was. ACT's chief executive put it plainly: rising grade-point averages have not corresponded with improvements in other measures of achievement, particularly in mathematics.

The independent yardsticks tell the harder story. On the 2024 National Assessment of Educational Progress, only twenty-two percent of US twelfth-graders reached proficiency in mathematics, among the lowest shares in more than two decades. Around thirty percent of tested 2024 graduates met ACT's college-readiness benchmark in math. The grades say one thing. The external measures say another. Where the two disagree, it is the grades that have moved.

This is what makes the University of California faculty worth listening to. In 2026, more than 1,800 mathematics and science instructors across the system signed an open letter reporting that first-year students were increasingly arriving without the basic skills their courses assume. At Berkeley, the letter noted, twenty to thirty percent of students in an early calculus course were showing severe gaps in preparation, and instructors were finding themselves reteaching mathematics from years earlier. These are not commentators with a thesis to sell. They are the people standing at the far end of the pipeline, receiving what twelve years of grades produced, and reporting that the grades had promised more than the students could deliver.

The Taiwan Version

A system very good at producing scores

Taiwan's version of the gap has a particular shape, and it is not the American one. The headline numbers are strong, as we have said. But two patterns inside them are worth a parent's attention.

The first is confidence. Across the high-achieving East Asian systems, a striking pattern recurs in the international data: students post superb mathematics scores while reporting unusually low confidence, low enjoyment, and high anxiety about the subject. Worldwide, the 2023 TIMSS study found that more than half of eighth-graders described themselves as not confident in mathematics. High marks and a shaky relationship with the subject can live in the same student. A score measures what a child can produce under exam conditions on a given day. It does not measure whether the understanding is secure enough to build on, or durable enough to enjoy.

The second pattern is the gap between recognizing and producing. The General English Proficiency Test, Taiwan's own widely-sat English examination, reports its results in a way that makes this visible. At every level, students pass the receptive components, listening and reading, at markedly higher rates than the productive ones, speaking and writing. At the High-Intermediate level in 2023, the pass rate was fifty-six percent for listening and reading but thirty-four percent for speaking and writing. Recognizing the right answer and generating one yourself are different skills, and preparation aimed at recognition builds the first far faster than the second.

It would be easy, and wrong, to name a villain here. The honest target is structural, not any particular institution. Taiwan supports an enormous and sophisticated tutoring sector. As of 2025 the Ministry of Education recorded 17,710 registered cram schools, a five-year high even as the birthrate falls, with around 88 percent focused on test preparation.

A system organized to optimize a measured score will, rationally, optimize the measured score. It gets good at the part that is tested and recognized, and it has less reason to build the part that is harder to see. That is not a moral failing of any tutor or family. It is what the incentive rewards. The chair of one national parents' organization, quoted in the coverage of those 2025 figures, suggested that nurturing self-directed learning might serve children better than endless tutoring. The point is not that preparation is bad. It is that preparation aimed at the scoreboard and preparation aimed at understanding are not the same project, and only one of them is what the next stage rewards.

What The Evidence Will and Will Not Bear

The honest limits of this argument

An argument is only as trustworthy as its willingness to say where it stops. Here is what the evidence supports, and what it does not.

What the evidence supports

Grades have drifted from skill. Multiple independent measures, ACT, NAEP, and university placement data, show grades rising or holding while external assessments do not follow. This is well documented.

The gap is widest at the bottom. Both US and Taiwanese data show the divergence concentrated among lower performers, and in Taiwan's case growing while the top pulls away.

The mechanism is real. That procedural fluency outruns conceptual understanding, and that cumulative subjects expose the difference later, is established in the mathematics-education research.

Where it should stop

This is not "students are getting dumber." Adults have lamented declining youth since at least 624 BCE. A 2019 study in Science Advances showed the belief is partly an illusion of memory and perspective. We are not making that claim.

Some of the gap is widened access, not decline. When more students from under-resourced schools reach a selective university, the entering class changes composition. That is partly a success of access, not simply a fall in ability.

The gap is addressable. When Tennessee paired remedial students with college-level math and real support instead of holding them back, the share passing college math rose from 12 to 61 percent. A gap is not a destiny.

Hold those together and the claim that remains is narrow and sturdy. Not that this generation is failing, but that a credential and a capability have come apart. The gap falls hardest on the students least equipped to absorb it. And the preparation easiest to purchase tends to close the visible gap, not the real one. That is a claim a parent can act on.

Common Situations

How to read your own child's readiness

Not prescriptions. Starting points for noticing whether the grade and the understanding are in the same place.

The grades are strong, but you are not sure they are solid.
The most useful test is not another test. Ask your child to explain why a method works, or to teach you a recent topic. A student who understands can explain it simply and can say what would happen if one part changed. A student who has only memorized will reproduce steps but stall at "why." The grade will not tell you which you have. The explanation will.
A new course suddenly feels much harder than the last.
In a cumulative subject this is usually a signal, not a slump. The difficulty often is not the new material. It is an earlier foundation that was never secure being asked to carry weight for the first time. The instinct is to drill the current topic harder. The more durable fix is to find and rebuild the gap a year or two below, where the structure cracked.
Your child can pass the test but cannot use the idea.
This is the recognizing-versus-producing gap, and it is common in heavily test-prepared students. They select the right answer but cannot generate one, or freeze when a familiar idea appears in an unfamiliar problem. The work here is to move from recognition to production: open problems, explanation, applying an idea somewhere it was not taught. Slower, and the thing universities ask for.
University is coming and you want to know they are truly ready.
Readiness is not the transcript. It is whether your child can learn something hard without being walked through it, recover from being stuck, and reason about unfamiliar material. Those capacities are visible long before university, in how a child handles a problem with no worked example in front of them. That is worth watching for now, while there is still time to build it.
Why We Can Write This Honestly
One student's work, seen week to week.

The public conversation about this gap runs to two poles. At one end, alarm: a generation in decline, standards collapsing. At the other, denial: the numbers are fine, the worry is the same fretting every generation has aimed at the next. We think both miss it. The alarm overstates a fall that the long evidence does not support. The denial overlooks a real and specific divergence between what a grade certifies and what a student can do.

We can sit in that middle because of where we work from. Harland teaches one student at a time. We are not reading a class of thirty through a single average, and we are not selling a score. We see the individual gap between the grade on the page and the understanding in the room, every week, because that gap is the whole of what one-on-one teaching is for. A student who can produce the right answer but cannot explain it is not a statistic to us. He is Tuesday afternoon.

That vantage also obliges us to be honest about the limits. We cannot see a national system from a single desk, and we have been careful above to separate what the data establishes from what it does not. What we can see, more clearly than almost any other vantage, is whether one child's foundation is truly built. That is the thing report cards have stopped reliably reporting, and it is the thing worth knowing.

How Harland Helps

Building the understanding, not just the score.

Our pedagogy is content-based learning: real subject content is the vehicle through which understanding is built, not a set of test patterns drilled in isolation. Working one student at a time in mathematics and the sciences, that approach is built to close the gap this piece describes, the distance between what a child can produce and what a child understands.

01
We find where the foundation cracked.
When a course suddenly feels too hard, the cause is often a gap a year or two below. Teaching one student at a time, we can locate the specific place the understanding broke, rather than drilling the current topic harder and leaving the real fault untouched. The fix is built where the structure failed.
02
We teach for explanation, not just the right answer.
A student who understands the work can say why a method holds, can sense when a result is off, and can take an idea into a problem it was never taught in. We build that capacity deliberately, because it is what the next course and the university after it will demand, and it is exactly what a score alone never certifies.
03
We will not sell you a hollow score.
We can prepare a student for an exam, and we do. But we will not optimize a number while leaving the understanding beneath it empty, because that gap only moves downstream to a harder course where it costs more to fix. If a foundation needs rebuilding, we will tell you, even when that is not the faster answer.

Is the grade telling you the whole story?

If you are not sure whether your child's strong marks rest on a foundation that will hold, that is exactly the question worth asking out loud. Tell us where your child is, and we will help you read what the report card cannot show you.

Start the conversation
PUBLISHED 12 August 2026 · LAST UPDATED 12 August 2026 · All figures verified against the awarding body or study named at the date of publication.
Sources

Where these figures come from

The remedial-cohort average high-school math GPA of 3.65, and the skills-check findings on basic arithmetic. The report also documents the rise in students placing into remedial mathematics.
University of California faculty open letter (2026)
More than 1,800 mathematics and science instructors on first-year preparation, and the Berkeley early-calculus figures. Reported by EdSource, The Economist, and others; signatory count as of late June 2026.
Mean mathematics score of 547 (3rd of 80 systems); the 294-point gap between top and bottom deciles, the widest of any participating system; the 32% share of top performers; and the 2018–2022 widening of the gap.
The 22% of US twelfth-graders at or above proficient in mathematics, among the lowest shares in over two decades.
The rise in average high-school math GPA from 3.02 to 3.32 (2010–2022) against flat test scores; and the share of 2024 graduates meeting the ACT math readiness benchmark.
The finding that more than half of eighth-graders worldwide reported being "not confident" in mathematics, alongside high achievement in the East Asian systems.
Pass rates by level showing productive components (speaking and writing) trailing receptive ones (listening and reading); High-Intermediate 34% versus 56% in 2023.
National Research Council, Adding It Up: Helping Children Learn Mathematics (2001)
The five strands of mathematical proficiency, the finding that procedural fluency outruns conceptual understanding, and the cited figures on computation versus number reasoning among thirteen-year-olds.
Taiwan Ministry of Education cram-school registry, reported July 2025
The 17,710 registered cram schools in 2025, a five-year high, with around 88% focused on test preparation. Market-size estimates are press-reported private estimates, not official figures.
Protzko & Schooler, "Kids these days," Science Advances (2019)
Evidence that the belief in declining youth is perennial and partly an illusion of memory and perspective. The counter-evidence the argument concedes.