Glasshouse: IELTS Academic Reading practice test

Glass is what happens when certain molten materials are cooled too quickly to arrange themselves into crystals. Sand, which is mostly silicon dioxide, will do it if you can get it hot enough, and adding soda lowers the temperature at which it melts to something a furnace can reach, while lime stops the result dissolving in water. The atoms in the finished material are disordered in the way the atoms of a liquid are disordered, which is the origin of the persistent claim that the glass in old windows is thicker at the bottom because it has flowed downward over the centuries. It has not. Early window glass was uneven when it was made, and glaziers sensibly set the thick edge at the bottom.

For its first fifteen hundred years glass was a luxury. It was worked by wrapping molten material around a shaped core of clay and dung, or by casting it in moulds and grinding the result, both of them slow methods that produced small, heavy and expensive objects. The transformation came in the first century before our era, somewhere on the eastern Mediterranean coast, when somebody discovered that a blob of molten glass on the end of a hollow iron rod could be inflated. Blowing is fast, it uses very little material, and it requires no mould. Within a century glass vessels had gone from treasure to the ordinary packaging of the Roman world.

The Romans also produced the first window glass, at first cast in shallow trays and later blown, and the chemistry of what they made tells its own story. Their glass was fluxed with natron, a mineral soda gathered from dry lake beds in Egypt, and the composition is so consistent across the empire that it is clear the raw glass was made at a small number of sites and shipped everywhere else to be worked. When that supply failed, some time around the ninth century, European glassmakers turned to the ash of burnt plants instead, which is why medieval northern glass has a quite different composition and a greener colour.

Making flat glass remained the central difficulty for another thousand years. One method was to blow a bubble, open it out and spin it rapidly until centrifugal force flattened it into a disc, from which panes were cut; the thick lump where the rod had been attached is the bullseye that still survives in imitation in pub windows. Another was to blow a long cylinder, cut it lengthwise and flatten it in a kiln. Both leave surfaces that are wavy enough to distort what is seen through them. The alternative was to cast a plate and then grind and polish both faces, which produces excellent glass and consumes so much labour that plate glass was a material for palaces.

The problem was solved in 1959, and the solution is worth describing because it is so unlike the methods it replaced. Molten glass is poured continuously onto the surface of a bath of molten tin. Glass floats on tin, tin is a liquid and therefore perfectly flat, and the two do not react, so the glass spreads out under its own weight to a uniform thickness with two fire-polished surfaces and is drawn off the far end as a solid ribbon. Nothing is ground and nothing is polished. The company that developed it spent seven years and a very large sum before a single saleable metre emerged, and within a decade almost all flat glass in the world was made this way.

What transparency made possible is easy to underestimate because it is everywhere. A glass vessel allows a reaction to be watched while it happens, which is the precondition for chemistry as an experimental science. Lenses gave the microscope and the telescope, and within a few decades of each other the seventeenth century acquired the cell and the moons of Jupiter. Spectacles extended the working life of every literate craftsman in Europe by perhaps twenty years. The mirror, which is glass with a metal backing, changed portraiture and, some historians argue, the sense of oneself as an object that others see.

The most recent chapter required glass of a purity nobody had previously attempted. In the 1960s it was proposed that light could carry telephone traffic along a glass fibre, provided the fibre absorbed little enough: the target set was a loss of no more than twenty decibels per kilometre, at a time when the best optical glass lost that much in twenty metres. The obstacle turned out to be impurities, chiefly iron, and it was overcome within four years. Modern fibre is transparent enough that a block of it several kilometres thick would be about as clear as a window pane, and essentially every intercontinental data connection now runs through it.

There is a pattern in all of this worth noticing. Glass has been improved, repeatedly, not by making it stronger or cheaper but by making it more perfectly what it already was: flatter, clearer, purer. Most materials are valued for what they do. This one is valued for what it fails to do, which is to interfere with light, and every advance in three thousand years has consisted of failing to interfere a little more completely.

The word multitasking was borrowed from computing, where it never meant what people take it to mean. A processor with a single core does not execute two programs simultaneously; it runs a little of one, saves its state, loads the other and runs a little of that, fast enough that a human cannot detect the alternation. The borrowed term has carried the misunderstanding with it. When a person appears to be writing a report and following a conversation, they are not doing both. They are alternating, and the alternation has a cost that the processor's does not, because a human state is not saved and reloaded cleanly.

The cost has been measured in a great many laboratories in much the same way. A participant is given two simple tasks, deciding whether a number is odd or even and whether a letter is a vowel or a consonant, and is asked to alternate between them. The response is slower and less accurate on the trials immediately after a switch than on trials that continue the same task, and the difference is large enough to be obvious. What makes the result interesting is that the cost does not disappear when the switch is announced in advance and the participant is given time to prepare. Preparation reduces it. Something always remains, which suggests that part of what has to happen cannot be done until the new task actually begins.

The applied case that has been studied most is driving. It is widely assumed that the danger of a telephone lies in holding it, and that a hands-free arrangement removes the problem. The evidence does not support this. Drivers in conversation on a hands-free device show delayed braking, reduced scanning of the road and a striking pattern in which they look directly at a hazard and fail to register it; in experiments using eye tracking, they fixate on objects they later cannot report having seen. Conversation with a passenger is measurably less disruptive, apparently because a passenger can see the road and falls silent when the situation demands it, which a caller cannot do.

The most uncomfortable finding in the field concerns who is doing it. A study that divided people by how much they habitually used several media at once expected the heavy group to be better at managing divided attention. They were worse: worse at ignoring irrelevant information, worse at switching between tasks, and worse on tests of working memory. Whether the habit causes the deficit or the deficit encourages the habit is not settled. A separate line of work found that the people who rate their own ability to multitask most highly perform worst on objective measures of it, and that the correlation runs through impulsiveness and a taste for novelty rather than through any capacity to divide attention.

It would be wrong to conclude that nothing can be done at the same time as anything else. A skill that has been practised to the point of automaticity makes almost no demand on the limited resource, which is why an experienced driver on an empty motorway can hold a conversation and a novice cannot, and why walking and talking coexist perfectly well until the pavement becomes complicated. The predictions that hold up concern competition for specific resources rather than for attention in general: two tasks interfere most when they use the same sense, the same kind of response and the same form of internal representation. Listening while reading is hard because both are verbal. Listening while walking is not.

Priya Raghunathan, who studies interruption in workplaces, thinks the laboratory picture has been carried too far into ordinary life. The switching tasks used in experiments are arbitrary and unrelated to one another, whereas a person moving between a document and an email about the same project is switching between things that share context, and the cost of that is much smaller. She also questions the widely quoted figure for how long it takes to return to a task after an interruption, which she traces to a small observational study and considers an artefact of how the researchers defined resumption. Her own field data suggest recovery is often a matter of a minute or two rather than the twenty or more that circulate.

She is right about the number and, I think, wrong about what follows from it. Field studies of knowledge work, including her own, consistently find that people switch activity every few minutes and that a substantial proportion of switches are self-initiated rather than caused by anybody else. Even if each resumption is cheap, the pattern leaves very little uninterrupted time, and the tasks that suffer are the ones requiring a structure to be held in mind while it is worked on. Laboratory measures of speed and error are poor instruments for that kind of damage, and they understate it, because the work they use is work that does not require anything to be held.

The practical conclusion is not a moral one, and the framing of attention as self-discipline has done a good deal of harm. Almost nobody resists an available interruption reliably, and treating the failure as a character defect produces guilt rather than concentration. What works is arrangement: making the interruption unavailable for a period, agreeing that a response is not expected within minutes, and separating the hours in which one is reachable from the hours in which one is not. Organisations that require constant availability and then complain that nobody thinks deeply are not describing a failure of their staff. They are describing the system they have built.