The Harland Review
The Modern Student

The Trophy and the Conversation: What Taiwan's English Market Sells

獎盃會說話,孩子卻不會
For
Parents choosing English education for their child
Reading time
~13 minutes
Last updated
August 2026
The Key Insight

Your child's English scores can climb for years while the one thing you want, the ability to speak, never arrives.

Taiwan's measured English is not collapsing. By some yardsticks it is at a record. The country's average TOEIC score in 2024 was the highest since the test arrived here in 2001, and Taiwan's speaking and writing scores rank near the top of Asia. So this is not a story about a nation that cannot learn English. It is a story about a gap that sits underneath the good numbers, and most parents only meet it by accident.

A parent notices it the first time a stranger asks their child a simple question in English and the child, who holds certificates and has stood on a competition stage, cannot answer. Both facts are real at the same moment: the certificate is real, and the child cannot speak. The after-school market produces and displays progress markers, the trophies, the level badges, the polished performances, because those are what a parent can see and what sells a course. Real speaking ability grows slowly, shows up nowhere on a shelf, and resists packaging. So the market optimizes the scoreboard, and over years the scoreboard and the skill drift apart.

That leaves you with a problem the market will not solve for you. The market hands you signals and asks you to trust them, with no honest instrument to tell a child who can speak from a child whom someone has coached to look as though they can. This piece is that instrument. It lays out what each signal truly measures, what it quietly leaves out, and the one question a parent can ask that cuts straight through to the truth.

What We See
A child who wins prizes and cannot order a meal in English. We see it often, and the family is always the last to know.

An elementary-school child came to us recently for an assessment. On paper she was a success. She had attended a well-known English chain since kindergarten, she had won at its national speech competitions, and her school had just chosen her to represent it at a city-level contest. Then we asked her to talk with us, away from any script, and discovered she could barely speak English at all. ChatGPT, it turned out, had written the speech that won her those stages. She had memorized it word for word and understood almost none of it.

We are not telling this story to single out one family. We tell it because it is ordinary. The same shape recurs across the children who come to us from the mainstream-school-plus-buxiban path. An impressive record of visible achievement rests on a thin foundation of real ability, and no one along the way notices the difference.

Her father was relieved, and also shaken, that it surfaced now. Had it not, his daughter could have reached eighteen still carrying an English level far below her grade and far below anything real study or work would ask of her. A shelf of evidence would have told everyone, including herself, that she was fine. What we see, more than any single weak skill, is that the system she was in held no honest mirror. This piece is about why.

At a Glance

The numbers underneath the trophies

A few figures that frame the gap, drawn from the test makers themselves, from Taiwan's own statistics, and from the global research, with sources listed at the end.

Taiwan's weakest skill
Speaking
On the 2024 TOEFL, Taiwan's speaking score (20) was its lowest of the four skills, three points below listening (23). Test-takers here read and listen better than they speak. (ETS)
The test maker's own finding
~10%
Among weaker GEPT candidates, the share who pass the second stage, which tests speaking and writing. Taiwan's own exam board states plainly that speaking and writing lag behind reading and listening. (LTTC)
The scores are not falling
581
Taiwan's average TOEIC score in 2024, its highest since the test launched here in 2001. The gap this piece describes sits underneath rising numbers, not falling ones. (Chun Shin / Focus Taiwan)
The market's size
17,710
Registered cram schools in Taiwan as of mid-2025, a five-year high. Of these, 15,599 focus on test preparation. (Ministry of Education, via Taipei Times)
Where the money is
NT$150bn
The estimated annual value of cram schools serving school-age children, more than NT$100 billion of it from test-prep cram schools, the activity furthest from spontaneous speech. (Taipei Times)
How common the path is
~70%
The share of Taiwanese schoolchildren who attend cram schools, with elementary-age children the single largest market. The model in this piece is the mainstream experience, not a fringe one. (CommonWealth; TEPS)
The Instrument

Four signals a parent reads, and what each one truly measures.

Every signal below is real. A child did win the trophy, did reach the level, did produce the essay. The problem is not that the signals are fake. It is that each one measures something narrower than the thing you care about, and the market presents them as if they measured the whole. Here is what each signal proves, what it leaves out, and the honest test that closes the gap.

Signal One

The competition trophy

A speech or recitation prize is the market's favorite proof. It photographs well and feels objective: a panel of judges chose this child over others.

What it shows
That the child can memorize a fixed passage, pronounce it clearly, and perform it with poise. These are real accomplishments.
What it leaves out
Everything unscripted. A memorized speech is a fixed block of language. It tells you nothing about whether the child can form a sentence no one wrote for them, which is what speaking means.
The honest test
Step off the script. Ask the child an ordinary question they did not prepare for and listen to whether an answer comes.
Signal Two

The level certificate

Chains run students up a ladder of branded levels, each with its own certificate and small tests. The ladder feels like measured, visible progress, one rung at a time.

What it shows
That the child completed a course and passed its internal checks, most of which reward vocabulary recall and grammar drills.
What it leaves out
Whether any of it transfers to use. The levels are the chain's own, set by the chain that profits from selling the next one, and they measure progress through a curriculum, not the ability to communicate.
The honest test
Ask what the child can do, not what level they reached. Can they explain their day, argue a small point, ask a question back?
Signal Three

The polished piece of writing

An essay or a speech comes home looking far stronger than you expected, clean and fluent and well above what the child says aloud. It reads like progress.

What it shows
Less than it used to. A polished text once implied a capable writer behind it. It no longer reliably does.
What it leaves out
Who wrote it. A child can now generate an essay well beyond their ability in seconds, and the tools that claim to detect this are unreliable and biased against non-native writers. The fluency on the page may belong to a machine, not the child.
The honest test
Watch the child write, or have them talk through what they wrote. Ability that lives only on the finished page is not the child's ability.
Signal Four

The advertising

Billboards, glossy brochures, and a polished brand do real work on a parent's confidence. A school that looks successful feels like a school that produces success.

What it shows
That the provider has a marketing budget. Nothing more. Advertising spend and teaching quality are unrelated, and can pull against each other when the budget comes from the fees.
What it leaves out
Any evidence of outcomes. No billboard reports what its students can do. That leaves parents reading marketing as a stand-in for results, because the market offers no transparent measure to read instead.
The honest test
Ignore the brand and ask for the thing itself: a sample of the teaching, and an honest account of where your child stands.
The One Question That Cuts Through
Every honest test above is the same test. Can the child say something in English that no one wrote down for them first?

A trophy, a certificate, a clean essay, a trusted brand: each can sit on top of a child who cannot do this, and none of them will tell you. A short, unscripted exchange will, in about a minute. It is the one measure the market does not sell, because it is the one a coached performance cannot fake. If you take a single thing from this piece, take that question, and use it the next time someone shows you proof of your child's progress.

Why the Scores Rise and the Speaking Does Not

The numbers and the silence are both true, and the reason is the same.

It would be simpler if Taiwan's English were just failing. It is not, and pretending otherwise would be dishonest. The harder, truer picture is that the things tests reward and the thing parents want have come apart, and you can see the split in the data itself.

The gap in the data

Taiwan reads and listens far better than it speaks

Recall that TOEFL profile: speaking is Taiwan's weakest skill, below its own reading and listening. That is not our reading of the data, it is the test makers' own, and Taiwan's exam board says the same in plain language: learners here develop listening and reading faster and find speaking and writing harder. Its data go further. Among candidates whose reading and listening are still weak, only about one in ten passes the stage that tests speaking and writing. The skills that surface first on a worksheet are the skills the system builds. The skills that need a live human on the other side lag behind.

The reason underneath

You can memorize for a test. You cannot memorize a conversation.

Language researchers draw a line between two things that look alike and are not. One is studying a language as a subject, memorizing its words and rules, which a child can do well enough to pass tests. The other is acquiring it, building the underground ability to generate sentences no one supplied, which is what speaking demands and what tests struggle to measure. Drilling and recitation build the first and barely touch the second. A memorized speech is the clearest case: the child produces a perfect block of English and owns none of the machinery that made it, so the moment the script ends, the English ends with it.

The missing ingredient

The one thing that builds speaking is the one thing the model skips

Speaking grows through real use: forming your own sentences, landing or failing to land, and adjusting. Taiwan is not unusual here. It shares the pattern with its neighbors, and the test makers point to the cause: across East Asia, many English learners report using the language in only one to ten percent of daily life. A child can spend years in classes and bank almost no live practice, because the model fills the hours with material to absorb instead of conversations to hold. No amount of drilling replaces the missing ingredient, which is why the scores can rise, year after year, while the speaking stays silent.

What the Market Is Built to Sell

The signals are not an accident. They are the product.

A market this large does what it is built to do. Taiwan's cram schools number in the tens of thousands, and the bulk of the money sits in test preparation. To see why the model produces signals rather than speakers, look at what the model rewards.

The structure

A franchise runs on things it can standardize and show

The large chains grow by repeating themselves: branded curricula, fixed levels, the same materials in every branch, and frequent small tests to mark the way. That structure excels at producing consistent, visible markers of progress, the certificates and the rungs, because those scale across hundreds of locations. It is poor at the thing that does not scale: patient, individual, two-way practice with a teacher who notices what one particular child cannot yet do. In a large class moving through fixed material, the quiet child at the back who has stopped understanding is the easiest thing to miss, and the structure is not built to catch them.

The information problem

The market asks you to judge what you have no way to see

Choosing an English course may be the purchase a parent is least equipped to evaluate. You often cannot check the teachers' real qualifications, you cannot watch the daily teaching before you commit, and no transparent, comparable measure exists of what a provider's students can do. Into that vacuum the market pours the only things it can show: the trophy, the level, the brand, the billboard. Parents are not foolish to read these signals. They are reading the only signals on offer, which is the problem itself. When the market gives you no honest measure, marketing becomes the measure.

The new variable

AI has made the signals easier to fake and the gap easier to hide

The newest pressure on all of this is the one that wrote our opening child's speech. Tools like ChatGPT now let a student produce written English far above their real level in seconds, and in one 2023 survey more than half of students said they had used it to write an essay. The detectors that claim to catch this are unreliable, and they misjudge non-native writers most of all. The effect is to widen the distance between what a child can show and what a child can do, the exact distance this piece traces.

We have written separately about what AI use means for a child's learning, and the short version applies here. Used as a thinking partner it can help. Used as an answer machine it hollows out the very ability it appears to demonstrate, and a speech written by a machine and memorized by a child is that hollowing made visible.

Common Situations

Four places parents find themselves.

The conversations we have most often, and how we would suggest thinking about each. These are starting points, not verdicts on any one child or any one provider.

Situation 01

Your child has the trophies and the certificates, and you have never heard them speak

This is the most common case and the most invisible, because nothing looks wrong. The record is full and the child seems to be thriving. Before you assume the foundation matches the record, run the one test: a short, friendly, unscripted exchange in English, ideally with someone who is not their regular teacher. If the answers come, the signals were true. If they do not, you have learned something far cheaper to fix at ten than at eighteen. Either way you will know, which is more than the certificates can tell you.

Situation 02

You are choosing a first English program for a young child

The instinct to start early is sound. The instinct to choose by brand and billboard is the one to resist. At the point of choosing, ask the provider for two things the signals never include. First, a sample of the real teaching, so you see how a class runs, not how it advertises. Second, an honest account of how they will show you where your child stands and whether it is working. A provider confident in its teaching will show you both without pressure. A provider that answers with its trophy wall and its level system is telling you what it sells, which may not be what you came for.

Situation 03

Your child has studied for years and you cannot tell if it is working

That uncertainty is not a failing on your part. It is the predictable result of a system that hands you levels and scores instead of a clear answer. Stop reading the signals and measure the thing directly. Pick a real task slightly beyond what a script would cover, telling a short story, describing a picture, holding a few minutes of back-and-forth, and see how your child does. Real ability shows up in unscripted use and nowhere else. If years of classes have produced a child who passes tests but cannot yet do this, the classes built the wrong thing, and knowing that is the first step to fixing it.

Situation 04

Your child handed in work that looks far above their level

Treat it as a question, not an accusation. The most useful thing you can do is ask the child to talk you through what they wrote, in their own words, out loud. A child who understands their own work can explain it. A child who generated it and memorized it cannot, and the gap will be obvious and gentle to surface. The aim is not to catch them but to see clearly. A polished page the child cannot stand behind is not progress, however good it looks, and pretending it is lets the real gap grow quietly underneath.

Why We Can Write This Honestly

We see the gap because our work begins by looking for it, and we have no signal to sell you.

A market built on signals does not like to be measured, so it is fair to ask why a tutoring academy gets to describe the problem honestly. The answer is in how we work. We teach 1-on-1, and every engagement begins with an honest assessment of where a child stands. Our work starts at exactly the point the rest of the system avoids: a real look at what the child can and cannot do. The child in our opening did not arrive flagged as struggling. She arrived as a prize-winner, and the gap only appeared because someone finally sat with her, off the script, and looked.

Her father said something we hear often from families who find us, that he was surprised we spend nothing on advertising. We do not, and the reason connects to everything above. We are not trying to sell a signal, so we do not buy billboards to project one. What we offer instead is the opposite of opacity: every lesson produces a written record the parents can read, so a family can see, week by week, what was taught and how their child is doing. The progress is visible because it is real, not because someone packaged it to look that way.

We should be plain about our own limits too. We are a tutoring academy, not a miracle, and building real fundamentals is slower than selling the appearance of them. We cannot promise a trophy by spring. What we can promise is simpler. A family will always know the truth about where their child stands, with no pressure and no bait-and-switch, because a parent who can see clearly is a parent who can stop worrying. For the child in our opening, seeing clearly at ten changed the whole arc of what comes next. That is the work.

How Harland Helps

We build the skill, not the scoreboard.

Harland's Academic English program exists for exactly the gap this piece describes. It is built to grow real, usable English, and to show a family the truth about their child's progress at every step.

01
An honest assessment before anything else
Every student begins with a working session that shows where they truly stand, the unscripted look the rest of the system skips. We would rather a family hear an uncomfortable truth at the start than a comfortable signal that costs them years. There is no entry sales pitch, only a clear picture of the child in front of us and what they need.
02
Real fundamentals, taught to be used
Our teaching is content-based: a child develops English by engaging with real material and producing their own language, not by reciting passages or running vocabulary drills. The aim is the ability the trophies leave out, the capacity to form sentences no one wrote first. It is slower than the appearance of progress, and it is the thing that lasts.
03
Progress a parent can see
Every lesson produces a written record the family can read, so you are never guessing what a level or a certificate means. You see what was taught, how your child responded, and where they are heading, in plain terms. This is the honest instrument the market does not provide, built into the work rather than sold as an extra.

If you want to know the truth about where your child's English stands, that is where we start.

No trophy wall, no level system, no pressure. A conversation with our Student Coordinator and an honest assessment will tell you more about your child's real ability than years of certificates have. From there, if it helps, we build.

Start the conversation
PUBLISHED June 16, 2026  ·  LAST UPDATED June 16, 2026  ·  Figures verified against ETS, the LTTC, Taiwan's Ministry of Education, and published research current to mid-2026
Sources

Sources and references for this editorial

ETS, TOEFL iBT Test and Score Data Summary 2024
Country and region score tables, including section means for Taiwan. Source for Taiwan's skill profile: speaking (20) as the weakest of the four skills, three points below listening (23). ETS cautions that ranking countries on these scores is a misuse of the data, so the piece cites the within-profile gap, not a national ranking.
Language Training and Testing Center (LTTC), GEPT
The General English Proficiency Test, run by the LTTC under Ministry of Education commission. Published scoring information and test-taker analysis. Source for the exam board's own statement that listening and reading develop faster than speaking and writing for Taiwanese learners, and that weaker candidates pass the speaking-and-writing stage only about one time in ten.
Chun Shin Limited, via Focus Taiwan and the Taipei Times
Annual report on TOEIC performance in Taiwan, 2024 results. Source for the record average TOEIC score (581), the highest since the test arrived in Taiwan in 2001, and for the strong regional standing of Taiwan's speaking and writing scores. The evidence that the gap sits underneath rising numbers.
Ministry of Education, via the Taipei Times
Reporting on Taiwan's registered cram-school sector, July 2025. Source for the count of registered cram schools (17,710 as of mid-2025, a five-year high), the number focused on test preparation (15,599), and the sector's estimated annual value (around NT$150 billion, more than NT$100 billion of it from test-prep cram schools).
CommonWealth Magazine and the Taiwan Education Panel Survey
Reporting and panel-survey analysis of cram-school attendance among Taiwanese students. Source for the share of schoolchildren who attend cram schools (around 70 percent) and for elementary-age children as the largest single market, establishing the model as the mainstream experience rather than a fringe one.
Second-language-acquisition research
The established distinction between conscious language learning and subconscious acquisition, and the limits of memorized formulaic language for producing novel speech. Source for why a memorized speech demonstrates recall and pronunciation but not generative ability, and why drilling and recitation build test performance far more than they build the capacity to converse.
ETS, TOEIC reports on test-takers worldwide
Survey data accompanying the TOEIC program on test-takers' use of English in daily life. Source for the finding that English learners across East Asia, Taiwan included, report using the language in only a small fraction of daily life, identifying the lack of authentic use as the structural reason speaking lags.
Study.com survey on student use of AI
Survey of students and educators on the use of ChatGPT for schoolwork, 2023. Source for the finding that a majority of students have used ChatGPT for assignments, including having it write essays, supporting the point that polished written work no longer reliably evidences a child's own ability.
Research on AI-text detection
Peer-reviewed study on detecting AI-generated essays and the bias such detectors show against non-native English writers, with OpenAI's own withdrawal of its detection tool as unreliable. Source for the claim that tools attempting to identify AI-written work are unreliable and disproportionately misjudge non-native writers, so a polished page cannot be trusted as proof of authorship.
US International Trade Administration and Taiwan news reporting
Briefings on Taiwan's Bilingual 2030 policy, its budget, and its implementation challenges, including the 2024 adjustments. Background for the national ambition and spending context, including the teacher-supply shortfalls and the policy's revision amid criticism.