DSE Listening Research Report — Volume 1: Diagnosis

Part One: Diagnosis

Under the label “listening” sits a combined-skills paper, built for the city Hong Kong actually is.

Of the four HKDSE English papers, Paper 3 — Listening and Integrated Skills — is the one most candidates fear, and the one they feel least sure about. Chinese-medium students feel it most. Teachers often describe it as a combined-skills monster wearing a listening costume. The name is listening. The task is high-speed decoding of speech, a large amount of reading across texts, sorting administrative detail, and rewriting into more than one style, all under exam conditions.

The design makes sense once you look at the city. Hong Kong is an international financial centre and a hub for professional services. In law, in multinationals, and in government, the working language is English. A normal day can mean taking in speech in several accents while reading a pile of written material — a clause, a commercial report — and turning that mix, quickly, into a formal public statement or an administrative memo.

That is the ability the HKEAA built this paper to test: whether a school leaver can use English the way that workplace does.

This report draws on HKEAA reports, classroom data, and candidate performance from the first HKDSE in 2012 through to 2026. The finding it will not soften is this: most secondary preparation for Paper 3 has fallen into drilling that breaks the way languages are learned, and that sits far above what students can currently do. It feels like work. It does not teach.

What this report covers

1. How hard the listening really is, set beside other exams. This chapter stays on listening and on spelling while you listen. It compares DSE Paper 3, especially Part A, with IELTS, iGCSE ESL and International A-Level ESL, and with academic exams sat by native speakers: the American SAT and ACT, and British GCE A-Level. The point is stark. Without the immersion of a British or American school exam, and without the multiple-choice prompts those international tests often give, DSE Paper 3 asks candidates to hear advanced vocabulary and spell it, accurately, by hand, with nothing on the page to lean on. At this level, that demand is rare.

2. What the results keep showing: two gaps, then a pile-up. In Part A, Chinese-medium students are hit first by missing background knowledge — the scenes are not theirs — and by a thin store of words they can actually produce. Somewhere between 44 and 55 blanks and short answers, with no options, and working memory collapses because they are listening and spelling at once. The next answer evaporates, and the marks fall in a row. In Part B they locate words and copy them. They do not reshape them. Tone and genre come apart.

3. Why schools drill anyway. Under public-exam pressure, many schools replace teaching with testing: a past paper or a mock, a timer, then the key. That breaks the order in which a second language develops. It never builds automatic recall of words, or the ability to decode speech. It retests failure until the failure feels normal. The cost is energy, not learning.

4. A way through: low-pressure immersion. The report’s practical half is a route that respects how memory and motivation work. Parents and students are caught in three binds at once. The priority underneath everything else is to repair schema — the background of the scenes — and the core listening vocabulary. Using the listening system developed by EHLA’s Oxford team, the report shows how three modules, and a choice between a zero-fatigue walkthrough and a full simulation, turn a long, punishing Paper 3 session into a few minutes of low load. The nightmare eases. Motivation can return. Small sessions accumulate.

This is a technical reading of a public exam, and it is also a diagnosis of how integrated English is taught in Hong Kong. The hope is that schools will reconsider the dogma of drilling harder, and will face the gap that the city’s working language and the exam’s design have opened together.

1. Paper 3 listening, set beside exams at a similar level

Put Paper 3’s listening next to other academic and language exams and one difference dominates: DSE demands far more language the candidate has to produce, not just recognise.

The comparison includes exams for native speakers (SAT, ACT, GCE A-Level English) and exams built for non-native speakers (IELTS, iGCSE ESL, International A-Level ESL).

Does the exam test listening at all? HKDSE Paper 3 Part A: yes. It is compulsory, and the topics run through academic, workplace, and social scenes. iGCSE ESL / International A-Level ESL: yes, often as its own paper. Scenes are school life, interviews, broadcasts. IELTS: yes, as its own module, from everyday survival to basic academic contexts. SAT / ACT / GCE A-Level English: no listening test at all.

What do you have to do? HKDSE: mostly gap-fill, notes in tables, and open short answers. iGCSE / iAL ESL: multiple choice, matching, and a smaller amount of word or phrase fill. IELTS: multiple choice, matching, labelling, and short fills, in roughly equal measure. SAT / ACT / A-Level: not applicable.

Is there anything on the page to help? HKDSE: nothing. A blank line or an empty cell. iGCSE / iAL: options and matches are given; fills often have a synonym or a strong clue in the sentence. IELTS: options, diagrams, or the surrounding sentence. SAT / ACT / A-Level: not applicable.

How hard is the spelling? HKDSE: very. Advanced academic and workplace words have to be spelled exactly, from memory, as productive spelling. iGCSE / iAL: moderate to low. The words are more basic, and minor spelling slips are tolerated more than on the DSE. IELTS: moderate. Fills are mostly everyday words, not advanced academic ones. SAT / ACT / A-Level: not applicable.

What that difference does to the candidate:

Against school-leaving exams in Britain and the United States, the contrast is structural. SAT, ACT, and GCE A-Level English do not test spelling-from-listening at all.

Against IELTS, iGCSE ESL, and International A-Level, DSE Part A is still unusually severe. Those exams lean on multiple choice and matching, which hand the candidate a word to recognise and lighten the load. Part A is a blank. Their dictated words sit in daily life or basic academic nouns. Part A asks for infrastructure or rejuvenation, with no visual help, and it is unforgiving about a missing -s or -ed.

Asking for words to come out automatically, with no options, is a severe test for Chinese-medium students who do not live in English.

2. What the yearly results show

HKEAA English reports and classroom data show the same collapse in Part A, year after year. Four tasks, roughly 44 to 55 blanks or short-answer points, and no options anywhere. In about 30 minutes of recording, that is an accurate answer every 30 to 40 seconds.

The failures are not random. From the deepest cause outward, they fall into three layers.

2.1 The floor: four gaps that break the listening before technique starts

This is where Chinese-medium candidates lose marks before any exam trick or grammar point is in play. Four missing abilities decide it.

1. The scenes are not in their background knowledge. Listening depends on schema. Part A assumes an international, middle-class, administrative, or academic life: wildlife work, overseas volunteering, green buildings, city planning, arts festivals. Students raised in Chinese do not share that life. EMI students have already met many of these ideas in English, in geography, history, or science. A Chinese-medium student has to understand an unfamiliar situation and decode unfamiliar words at the same time. The overload is what teachers hear as “they just switched off”.

2. The listening vocabulary is not there. Too many of the words are unknown, so the point of the question and the gist of the recording never arrive.

3. Recognising a word is not the same as writing it. Many students “know” a word as a Chinese gloss on a page. Part A wants the word produced, by hand, as the sound passes. They hear it, they know what it means, and they cannot get it onto the paper.

4. Spelling is not automatic. Attention goes into forming the letters and never returns to the recording. Speech moves in one direction and does not wait. Spelling takes time. A word like recommendation can cost two or three seconds. Working memory is occupied by the hand, the next target is missed, and the marks fall in a row. That is the domino.

2.2 Three further cracks in the basics

On top of those four gaps, ordinary technical holes make the listening stall.

1. Words are not known finely enough. A word changes meaning with the situation, and similar sounds and similar spellings get swapped.

2. Connected speech was never taught. Linking, weak forms, and dropped sounds turn a string of ordinary words into something that sounds like one unknown word.

3. Grammar fails at the last step. The keyword was heard, then tense, agreement, plural, or word class is wrong, and the mark is gone.

2.3 The marking standard, and the domino in the exam hall

Those gaps meet a very strict mark scheme, and the result is not recoverable in the room.

1. Spelling tolerance is tiny. IELTS still offers a share of multiple choice and matching, so recognition can score. Part A is almost all blank fills. In the moment the sound passes, the candidate is expected to write infrastructure, nutritionist, entrepreneur. Miss the word, or know it and misspell it, and the mark is zero. Get the root right and miss -s, -ed, or a noun ending such as -ment, and the mark is still zero. “I heard it, I understood it, I couldn’t write it” is normal among students who have never spelled these words often enough for the hand to know them.

2. The domino follows at once. While the hand is still on the previous long word, the recording has moved. The next point is missed. Then the one after that.

A note on the tidy-up minute. Teachers often tell students to jot a first letter and complete it in the final minute. Short-term memory fades in 15 to 30 seconds. Twenty minutes later, under exam nerves, a fragment such as “ac…” no longer distinguishes accommodation from academic. The strategy fails in the room, whatever it looks like in a calm classroom.

2.4 Part B: copying instead of rewriting

Part A tests whether sound becomes spelling. Part B tests whether a candidate can rebuild meaning across texts. The marks cluster in two places: what is done with the content, and whether the tone fits.

1. They can find the sentence. They cannot change it. HKEAA reports have criticised verbatim copying for years. Weaker candidates can locate a line with a keyword. They cannot digest it, convert it, or paraphrase it. Pronouns and tenses come across unchanged. A Data File line in the first person — “I think we should delay the project” — is pasted into a third-person report. At the moment of writing, the candidate is moving text, not writing.

2. Register and tone split apart. Appropriacy is a real part of the mark, because the paper is imitating work. The Data File mixes WhatsApp, an angry complaint, and rough meeting notes. Students who cannot hear a shift in tone paste informal or emotional wording into a formal report or proposal. Content and language both suffer, and the writing shows that workplace English is unfamiliar.

3. Linkers become decoration. To cope with the volume, candidates memorise openers and sentence-initial connectives. Furthermore, Moreover, and In addition appear where there is no cause, contrast, or concession to mark. A polished memorised opening, followed by error-strewn lines copied from the file, is obvious to a marker. The gap in style is itself evidence that the language is not at the level claimed.

Taken together, speed and question-familiarity will not fix Paper 3. Part A shows a base that cannot produce words or decode speech. Part B shows a base that cannot rebuild logic or judge tone. If teaching stays on the current drill, it will not produce a different result. The next chapter is about that drill.

Part Two: Breakthrough

Chapter 3. Why the school drill fails, and why teachers are stuck inside it.

The standard senior-secondary answer to Part A is a pair: teach a few local tricks, then run a lot of papers. The evidence says this does not raise the real listening of mid and lower candidates — most of the entry, and most Chinese-medium students. It breaks the order of second-language learning, and it leaves teachers with no good move.

3.1 What schools usually do

Two tracks, used together.

Strategy, taught at the surface. Predicting during the reading time: circle a keyword, guess whether the blank wants a noun, a number, an adjective, a place, a job. Shorthand, because the recording is fast: w/o for without, Govt for government. Distractors: listen for wait, actually, however, because the speaker is about to reverse the answer.

Then testing in place of teaching. As the exam approaches, past papers and publisher mocks are played under a timer and marked immediately. The lesson is the paper.

3.2 Why that drill does not teach

For a strong EMI student the tricks can tidy an ability that is already there. For the majority, forced practice works against them.

1. The paper is too far above the student, so the recording is noise. The weakness is not one skill. Schema is missing, listening vocabulary is thin, speech will not decode, spelling and grammar are unsteady. A timed paper at that point violates the rule of comprehensible input. Nothing just beyond the student is on offer. Only a repeated experience of failing, which is a very efficient way to kill motivation.

2. Exam tricks are a match in fog. When decoding has already failed, working memory fills with anxiety. Prediction, shorthand, and “listen for the turn” were memorised in class and cannot be called up in the hall.

3. Long, dense listening becomes something to avoid. Part A mocks are long, heavy, and humiliating. Students dodge listening outside class. Contact with the language drops. The listening gets worse. The drill caused the avoidance it was meant to cure.

3.3 Teachers are not the cause. The structure is.

This is not a verdict on teachers’ competence or effort. They are inside a bind made of student level and exam pressure.

Students arrive from junior secondary already far behind. In a crowded senior timetable, repairing the base and “covering the exam” cannot both be done properly.

Listening materials a teacher can actually assign are narrow. There is no high-motivation, low-threshold system students will use on their own.

So teaching becomes defensive. Drop the papers and spend the lessons on vocabulary and sound, and the mock scores will not move quickly, while management and parents still want an account. Keep teaching tricks and running papers, and if the public exam is still poor, the procedure at least shows that the teacher tried. The student’s foundation takes the blame.

That defence copies itself. Breaking it means different materials, a different kind of practice, and a different idea of what the assessment is for.

Chapter 4. A way through: immersion that does not hurt

Extra lessons and extra papers have already failed. The route has to sit outside the school timetable, and it has to be something parents and students will actually accept.

This chapter names the three binds, then uses EHLA’s listening system as the worked example: DSE preparation that is short, light, and still the real paper.

4.1 Three binds that suffocate practice at home

Teachers and parents have tried subscribing to listening mocks. Most of those attempts die, because three demands contradict each other.

Relevance. If the material is easy, or does not look like a DSE paper, families decide it will not help the exam and stop.

Load. The moment the material matches DSE difficulty and length, the school nightmare switches on. Spelling under time is exhausting, and students refuse it on reflex.

Exhaustion. A traditional Part A sitting asks for 30 to 40 minutes of unbroken focus. For a weak student that is punishing. One attempt, then a month of nothing. A habit never forms.

Schools will not redesign themselves this year. Parents will not drop the wish for something that helps the exam soon. The only remaining move is in the design of the learning itself: fully DSE, and also short enough that it does not hurt.

4.2 What to repair first: the scenes, then the words

Of the four gaps under Part A, two have to come first. Students do not know the scenes, and they do not know the words.

Listening is schema from the top and vocabulary from the bottom, working at the same time. Part A assumes green architecture, international volunteering, festival planning, and a great deal else in that world. If those ideas are not in the student’s head, and the words are new, prediction and “listen for the turn” are a building with no foundation.

Repair the scenes and the core listening vocabulary first. That is the only base on which the rest can stand.

4.3 Ten design rules that make the “impossible” usable

A system that respects both the exam and the student has to leave ordinary drilling, and ordinary audio played straight through.

1. Sit outside the usual lesson and homework shape. 2. Cover the spread of DSE scenes, not a sample of them. 3. Stay 100% DSE in question type, pace, and logic, so families do not suspect a watered-down substitute. 4. Keep each session to a few minutes. It should be possible anywhere, without a headache. 5. Put a hint one tap away, so “I didn’t catch that” is not a long hunt. 6. Show the whole solution, so the thinking can be watched with the pressure off. 7. Offer a zero-fatigue mode: listen and watch, no writing, when energy is gone. 8. Make the pieces small enough that listening stops feeling like a threat. 9. Turn abstract tricks into steps on the screen. 10. Track the words the ear met, and follow them with spelling the hand has to produce.

4.4 What EHLA built: three modules, and a mode that does not exhaust

After several years of work, EHLA’s Oxford team built a wide bank of Part A scenes, then three modules and a mode switch.

The switch is the part that changes the experience.

Zero-Fatigue Mode is for the fear of writing and the fear of falling behind. The student does not answer. A teacher walks the task: prediction, shorthand, cutting distractors, then the answer. Bilingual subtitles can be on. An unknown word is one tap, and the system remembers what was looked up. Afterwards it pulls the answer words into spelling practice, so what the ear heard becomes something the hand can write. The domino — miss a spelling, miss the next point — never starts. Schema gets built while the affective filter is low, and listening can feel manageable.

Full simulation is the check. No subtitles, timed, the real attempt, once confidence and background knowledge are actually there.

The third module is a short vocabulary bank aimed at the spellings Part A punishes: necessary words, words that sound alike, and words students chronically misspell. A few minutes turns recognition into words the student can produce.

4.5 Who it is for: from avoidance to a habit

The two modes are not a remedial track with a prestige track beside it.

For weaker candidates, and for many Chinese-medium students, the spelling burden and the clock are lifted. They watch and listen, and the scenes and the sounds start to exist. Four to eight minutes is short enough that practice stops being an animal they refuse to approach. Small sessions can accumulate.

For mid-level and strong students, the walkthrough is an expert’s thinking made visible. They see the small leaks in their own technique, and they can sweep hard spellings in the gaps of a day.

Zero-Fatigue Mode is the part that holds the system together. It drops the rule that listening is either a test or it is nothing. When a student is tired, or the base is thin, watching and staying inside the task is itself high-value learning. For a struggling student it is a lifeline. For a strong student it is a fast audit.

Public-exam preparation does not have to be a grind that teaches nothing. Schema, a lower cognitive load, and a system designed around both of those can repair the base and still respect why a student would bother. That is a concrete pattern other DSE teaching can copy.

Who wrote this: EHLA

Rooted in Oxford.

EHLA studies what actually changes English learning, then turns that research into programmes schools and students can run. We work with individual learners and with mainstream Hong Kong schools. More than 300 schools have taken part in EHLA pilot programmes.

“My colleagues at the University of Oxford and I met EHLA in Oxford, and I was completely fascinated by what EHLA is doing…” — Prof. Furlong, former Dean of the Department of Education, University of Oxford