🔁 1. Ôn tập – 10 phút
🔁 Ôn thẻ từ vựng đến hạn
🧠 Nhớ lại buổi trước (không xem tài liệu)
- Buổi 31: “corroborate” nghĩa là gì?
- Buổi 75: vì sao nhóm 20 người khó kết luận?
- “He must have left” và “He might have left” khác nhau thế nào?
Trả lời to hoặc viết ra giấy, rồi mới mở lại buổi trước để kiểm tra.
🎧 2. Đầu vào – 25 phút
Động từ khiếm khuyết suy luận về quá khứ
| Mức chắc chắn | Cấu trúc | Ví dụ |
|---|---|---|
| Gần như chắc chắn đã | must have + V3 | The data must have been entered incorrectly. |
| Gần như chắc chắn không | can’t / couldn’t have + V3 | The result can’t have been due to chance alone. |
| Có thể | might / may / could have + V3 | Something else might have caused the effect. |
| Khả năng khá cao | might well have + V3 | The sample might well have been too small. |
🎧 Bài nghe
A lecturer in the philosophy of science explains how science corrects its own mistakes (giả định)
Nghe tối đa 2 lần trước khi xem lời thoại.
📄 Xem lời thoại (sau khi làm bài)
- Lecturer
- Science is often described as self-correcting. Today I want to look at what that actually means, how the process works in practice, and where it sometimes falls short.
- Lecturer
- Let's begin with a hypothetical example. A research team reports that a particular vitamin supplement improves memory in older adults. The study is published in a respected journal after peer review – that is, after other experts in the field have assessed the paper before publication. Many people assume that peer review guarantees a result is correct. It doesn't. Reviewers usually see only the written report, not the raw data, so they can check whether the methods and conclusions are reasonable, but they can't detect every error.
- Lecturer
- The more powerful check comes afterwards, through replication – other teams repeating the study to see whether they get the same result. In our example, suppose two other groups try to replicate the finding and find no effect at all. What can we conclude? We can't say for certain that the original team made a mistake. The first result might have been a chance finding, especially if the sample was small. The original team might well have tested many outcomes and reported only the one that looked impressive. Or the groups might have used slightly different supplements.
- Lecturer
- Now suppose a fourth team looks at the original data and discovers that some participants were counted twice. At that point, the error must have occurred in the original analysis, and the journal may decide to issue a retraction – formally withdrawing the paper. Retractions are relatively rare, and they're often seen as embarrassing. But in a sense they're evidence that the system is working.
- Lecturer
- So why does self-correction sometimes fail? One reason is that journals have traditionally been more interested in publishing new, positive findings than replications or studies that find nothing. As a result, researchers have had little incentive to repeat other people's work. Another reason is time: a widely reported claim can influence public opinion for years before it's corrected, and the correction rarely receives the same attention.
- Lecturer
- Several reforms aim to address these problems. One is pre-registration, in which researchers publish their plans – including exactly which outcomes they'll measure – before collecting any data. This makes it much harder to report only the impressive results. Another is data sharing, which allows others to check the analysis directly. Neither is a complete solution, but together they make errors more likely to be caught early.
- Lecturer
- The key point for non-scientists is this: a single study, however impressive, should be treated as preliminary. Confidence should grow only when several independent studies point in the same direction – which is what we mean by a scientific consensus.
Giọng đọc là giọng máy của trình duyệt; giọng và tốc độ khác nhau tùy thiết bị. Bài thi thật dùng giọng người với nhiều giọng địa phương.
Câu 1–10: Complete the notes. ONE WORD ONLY.
- 1. Peer review: experts assess a paper before .
- 2. Reviewers usually cannot see the raw .
- 3. Stronger check: replication – other teams the study.
- 4. A failed replication may mean the first result was a finding.
- 5. Risk: reporting only one outcome that looked .
- 6. Retraction = formally a paper.
- 7. Journals prefer new, positive .
- 8. Corrections rarely get the same as the original claim.
- 9. Pre-registration: researchers publish their before collecting data.
- 10. A single study should be treated as .
🧩 3. Ngôn ngữ trọng tâm – 10 phút: mức chắc chắn trong bài giảng
✅ Giảng viên chắc chắn đến mức nào?
1. “The first result might have been a chance finding.”
2. “The original team might well have tested many outcomes.”
3. “The error must have occurred in the original analysis.”
4. “We can't say for certain that the original team made a mistake.”
Suy luận với modal
- The figures don't add up – someone have entered them twice.
- He was in Hanoi all week, so he have attended the meeting in Da Nang.
- The delay might have been caused by the storm – that's the most likely explanation.
- The sample may have too small to show an effect.
🎧 4. Thực hành – 20 phút
- Nghe lại, mở lời thoại: đánh dấu mọi modal suy luận và ghi mức chắc chắn (7 phút).
- Giải thích bằng 45 giây: vì sao một nghiên cứu được bình duyệt vẫn có thể sai (6 phút).
- Sổ cụm từ band 7: falls short, point in the same direction, however impressive, in a sense (7 phút).
💼 Dùng ngay – trình bày phát hiện kèm bằng chứng khách quan (thêm 5–10 phút)
🗣️ 🗣️ Tách điều chắc chắn với điều suy luận
1) Bấm 🔊 Nghe. 2) Nói theo. 3) Bấm 🎙 Ghi âm, nói, rồi ⏹ Dừng. 4) Bấm ▶ Nghe mình để so sánh.
What we know for certain is that three batches failed the test last week.
Điều chúng ta biết chắc là tuần trước ba lô hàng không đạt kiểm tra.
All three were made on Line 2, so the problem must have started there.
Cả ba đều sản xuất trên dây chuyền 2, nên vấn đề hẳn bắt đầu từ đó.
The new supplier's material might well have contributed, but we haven't confirmed it yet.
Nguyên liệu của nhà cung cấp mới có khả năng đã góp phần, nhưng chưa xác nhận.
It can't have been the operator, because the same person ran Line 1 without problems.
Không thể do người vận hành, vì cùng người đó chạy dây chuyền 1 không có vấn đề.
We'll know more once the lab results come back on Friday.
Chúng ta sẽ biết thêm khi có kết quả phòng thí nghiệm vào thứ Sáu.
🔒 Bản ghi âm chỉ nằm trên máy của bạn, không gửi lên mạng, và mất khi tải lại trang.
Trình duyệt này chưa hỗ trợ ghi âm. Bạn vẫn có thể nghe và nói theo.
📚 Từ vựng của buổi
- hypothesis – giả thuyết
The study tested a new hypothesis.
- replicate – lặp lại (nghiên cứu)
Other teams failed to replicate the result.
- replication – sự lặp lại nghiên cứu
Replication is a key check in science.
- peer review – bình duyệt
Peer review does not guarantee accuracy.
- retraction – sự rút bài
The journal issued a retraction.
- sample size – cỡ mẫu
A small sample size limits confidence.
- bias – thiên lệch
Reporting only positive results creates bias.
- self-correcting – tự điều chỉnh, tự sửa sai
Science is described as self-correcting.
- preliminary – sơ bộ
These results are preliminary.
- consensus – sự đồng thuận
A scientific consensus takes time to form.
✅ 5. Tự kiểm tra – 10 phút
✅ Tự kiểm tra
Nghe lại bản ghi của mình và tích những điểm bạn đã làm được. Ô tích chỉ lưu trên máy này.