Đề hoàn chỉnh — làm một mạch 70 phút. Phần Nghe (câu 1–20): bấm nút Nghe ở trên, mỗi phần chỉ nghe MỘT lần. Phần Đọc (câu 21–40): dùng ba bài đọc bên dưới. Chỉ mở lời thoại sau khi đã nộp bài.

PHẦN ĐỌC — Bài 1 (câu 21–27)

The Return of the Night Train

For thirty years the sleeper train looked like a museum piece. Budget airlines undercut it on price, high-speed daytime services undercut it on time, and across Europe operators withdrew route after route. Between 2000 and 2019 the continent lost the majority of its international night services. Then, quite suddenly, the direction reversed. New routes have opened between Vienna and Paris, Berlin and Brussels, Stockholm and Hamburg, and operators report occupancy rates they had not expected.

The obvious explanation is environmental. A night train produces a small fraction of the emissions of the equivalent flight, and the growth coincided with a rise in public concern about aviation. But the operators themselves give a more prosaic account. What changed, they argue, was the value of a hotel night. A sleeper is not competing with a flight alone; it is competing with a flight plus a night's accommodation plus two airport transfers. Once city-centre hotel prices rose above a certain point, the arithmetic moved.

The economics remain difficult. Sleeper carriages are expensive to build, carry far fewer passengers per metre than a seated coach, and spend the daylight hours idle. A seat on a high-speed train can be sold three times in a day; a berth is sold once. Staffing costs are higher, and track access charges in some countries are levied at rates designed for freight and daytime passenger traffic rather than for a service that runs slowly through the night precisely so that it arrives at a civilised hour.

This last point has attracted attention from regulators. A night train is in the odd position of wanting to travel slowly, since arriving at three in the morning helps nobody, yet it is charged as though speed were the service it was selling. Several governments have begun adjusting these charges, and Austria has gone further, treating its night network as public infrastructure rather than as a commercial operation expected to pay for itself.

Whether the revival lasts is genuinely uncertain. The new routes have so far attracted a particular kind of passenger: someone who could afford to fly and is choosing not to. That is a real market, but it is not obviously a large one, and it may be sensitive to the price of the alternative in a way that mass markets are not.

PHẦN ĐỌC — Bài 2 (câu 28–34)

Why Teams Choose the Wrong Idea

Research on group decision-making has repeatedly found an uncomfortable pattern: groups spend most of their discussion on information that everyone already possesses. In the classic experimental design, participants are each given a unique piece of evidence along with a set of shared facts. The unique pieces, taken together, point clearly towards one option. The shared facts point towards another. Groups overwhelmingly choose the second.

The phenomenon is called the shared information bias, and its cause appears to be social rather than cognitive. Mentioning a fact that others already know produces immediate agreement, which is rewarding. Mentioning something nobody else knows produces silence, questions, and the risk of being wrong in public. Participants who do raise unique information are, in follow-up ratings, judged less competent by their colleagues than those who repeat what is already common ground — even when their information turns out to have been decisive.

Several remedies have been tested with varying success. Simply instructing groups to share everything has little effect. Assigning explicit roles works better: when one member is designated as responsible for a particular domain, their contribution in that domain is expected rather than volunteered, and the social cost disappears. Longer discussions help somewhat, since unique information tends to surface late, but only if the group has not already converged on an answer.

The most effective intervention found so far is also the simplest: having each member write down their view privately before any discussion begins. This prevents the earliest speakers from setting an anchor, and it means that a person holding unique evidence has already committed to its implications before discovering that nobody else agrees. Organisations that have adopted the practice report that meetings become slower and, by their own account, noticeably less pleasant.

That last observation deserves attention. The bias is not a malfunction; it is the by-product of a group doing something valuable, namely establishing common ground and cohesion. Interventions that improve decision quality tend to work by making the group less comfortable, and a team that is never uncomfortable is probably not surfacing much that is new.

PHẦN ĐỌC — Bài 3 (câu 35–40)

Mapping What Was Never Written Down

Historians of ordinary life face a structural problem: the records that survive were made by institutions, and institutions recorded what concerned them. Tax rolls, court proceedings, parish registers and estate accounts tell us a great deal about who owed money, who broke the law and who was baptised. They tell us very little about what most people did on most days.

One response has been to read institutional records against their purpose. A court case about a disputed boundary describes, incidentally, how a field was worked. An inventory drawn up for probate lists the contents of a room in an order that reveals how the room was used. The historian is not interested in the dispute or the valuation but in the detail that had to be included for the document to do its job.

This technique has limits that its practitioners acknowledge. Incidental detail is not random: it appears when it is relevant to the institution's concern, so the resulting picture is skewed towards whatever generated disputes. Households that never went to law, and there were many, leave no such traces. There is a real danger of reconstructing the life of the litigious and calling it the life of the ordinary.

Material evidence offers a partial correction, since a broken pot cares nothing for what a clerk thought worth recording. Archaeology has revised several long-standing accounts of diet and of trade, in each case by finding that what people evidently ate and used did not match what the documents implied. But material evidence has its own bias: it preserves the durable and loses the perishable, which means it over-represents ceramics and metal and under-represents textiles, wood and food.

The methodological consensus, such as it is, holds that no single class of evidence can be trusted alone, and that the useful work happens where two biased sources disagree. A discrepancy between what the accounts record and what the middens contain is not an inconvenience to be resolved in favour of one or the other; it is frequently the most informative thing available, since it marks the place where an institution's interest and a household's practice came apart.