IELTS Reading: True / False / Not Given (Practice with Answers)

Before electricity reached the coast, keeping a lighthouse lit was manual labour without interruption. Keepers trimmed wicks every four hours through the night, hauled oil up spiral stairs, and polished the lens after every storm. The British service logged each watch in a leather book, and inspectors arrived without warning to check that the pages were genuine.

Trinity House began experimenting with automation in the 1880s, when a clockwork mechanism allowed the Daunt Rock light in Ireland to burn unattended for eight hours at a stretch. Progress was slow because the Board of Trade judged that a light without a keeper was a light without a witness. If a flame failed at sea, no one ashore would know. That objection held for decades, and keepers continued to climb.

The economics of the 1930s settled the argument. A keeper's station cost around 1,400 pounds a year to run, most of it wages and supply boats, while an automated light on the same rock needed a third of that. One by one the stations emptied. The last manned lighthouse in Britain, North Foreland in Kent, closed its final watch in November 1998, and the keepers left carrying their belongings in grocery crates.

What remained on the rocks was not simply machinery. Engineers had learned that automation works best when a building is designed around it from the start, so the new lights were rebuilt with flat roofs for helicopter access and tanks of propane sized to last a full winter without a visit. The older towers, with their narrow stairs and rooms shaped for human bodies, proved expensive to convert, and several were simply capped and darkened. Historian Bella Bathurst later observed that the keepers had been part of the apparatus itself: remove the man and the machine had to be redesigned to notice what he once noticed.

In 2011 a psychology professor named Daryl Bem published a set of experiments appearing to show that people could feel the future. The studies used standard methods, passed standard review, and appeared in a leading journal. Their impossibility triggered a question that had been waiting decades to be asked: if the machinery of psychology research could certify results like these, what else had it certified?

The answer arrived through an act of mass self-examination. A consortium of nearly two hundred researchers agreed to repeat one hundred published psychology experiments, using the original materials wherever possible and registering their methods in advance. Roughly a third of the replications produced statistically significant results. The figure was read two ways: as a catastrophe, by those who noted that two-thirds of the field's flagship findings had failed to reproduce, or as reassurance, by those who pointed out that a third surviving is roughly what statistics predicts when original studies were borderline.

The causes dug deeper than fraud, which remains rare. Small samples exaggerate chance effects. Flexible analysis lets researchers test many specifications and publish the one that worked. Journals prefer tidy, surprising findings over careful null results, so the literature fills with the winner's curse: the luckiest estimate of each effect, not the truest. None of this requires dishonesty, only incentives.

Reform has followed the diagnosis. Preregistration, where analysts commit to a method before seeing data, spread from clinical trials into laboratory fields. Journals began accepting studies on the quality of their design rather than the direction of their results, and several funders now require that grantees share raw data and code. Culture moves slower than policy. A survey in 2021 found that most researchers still could not name a single study they had tried and failed to replicate themselves, not because they had none, but because trying was never part of the routine.

In 1901, sponge divers off Antikythera found a corroded lump of bronze in a Roman-era shipwreck and delivered it, among the statues and jars, to an Athens museum. Months later a museum archaeologist noticed a gear wheel embedded in the crust, and the lump began a century of reinterpretation.

X-ray imaging in the 1970s revealed the truth: at least thirty interlocking gears, dials and inscriptions, a mechanical computer that modelled the movements of the Sun and Moon, predicted eclipses and tracked a calendar of athletic games. Nothing of comparable complexity survives from the ancient world, and nothing approaching it appears again in the archaeological record for over a thousand years.

That silence is the mechanism's second story. The device implies a tradition: workshops designing geared astronomy, craftsmen able to cut teeth into bronze wheels to tolerances later associated with clocks. Greek texts mention such machines, but historians had treated the passages as literary exaggeration, since no physical evidence existed. One object revised the estimate of what ancient technology could do and, by implication, how much of it has simply not survived.

The wreck itself dates the machine loosely, to perhaps 150 to 60 BC, and its maker remains unknown. Inscriptions suggest repairs by more than one hand, and the design incorporates lunar theory associated with astronomers of Rhodes, which sharpened an old suspicion that the island's workshops were turning out instruments, not just ideas. Whatever its origin, the mechanism ended its working life as cargo, crated for a Roman buyer who never received it, and so became the only ancient computer we have, by accident of a storm.