IELTS Academic Reading Practice Test 207

The word 'microphone' was coined in 1827 by the British scientist Charles Wheatstone, but it did not describe something to speak into. Wheatstone used it for an instrument designed to detect and amplify very faint sounds, a device closer to a stethoscope than to a telephone.

For the next forty years microphones remained laboratory curiosities, of interest to physicists studying sound and to a handful of inventors trying to send speech along a wire. The difficulty was fundamental: a microphone must convert the minute pressure changes of a sound wave into an electrical signal that varies in exactly the same way, and the components needed to do this reliably did not yet exist.

In 1861 Johann Philipp Reis, a German schoolteacher, built a transmitter in which a thin membrane vibrated against a metal contact. When the membrane touched the contact a circuit closed; when it sprang back the circuit broke. The device could reproduce the pitch of a note and the rhythm of a tune, and it is said to have carried a word or two intelligibly. But speech is a continuously varying waveform rather than a series of makes and breaks, and the interrupted current that Reis produced was far too crude to carry it. Reis himself doubted whether the design could ever work well, and he described his apparatus as a toy.

The first person to send recognisable speech was Alexander Graham Bell, whose family background lay in elocution and whose wife and mother were deaf. Working in Boston, Bell tried several transmitters. The most successful, described in the patent granted to him in March 1876, was a liquid device: a diaphragm with a needle attached rested above a cup of water made conducting with acid. As sound made the diaphragm vibrate, the needle moved up and down, altering the depth of immersion and therefore the resistance of the path through the liquid. The current varied in step with the sound, and the speech that reached the receiver was distorted but intelligible.

The device that made the telephone practical came from several inventors almost at once. In 1877 Thomas Edison developed a transmitter in which a button of carbon was pressed between two metal plates; in 1878, in London, the inventor David Edward Hughes demonstrated a microphone consisting of a small box of loose carbon granules traversed by a needle. Both exploited the same property of the material: when carbon granules are packed together, their resistance falls as the pressure on them increases. The vibration of a diaphragm squeezed the granules and modulated a current far larger than any liquid cell could carry, so the signal needed no amplifier.

Carbon microphones were crude and noisy, and their performance drifted as the granules settled, yet they were cheap and produced a strong signal. For the next hundred years almost every telephone in the world used one, engineers preferring a component they understood to the cleaner but quieter alternatives being developed in laboratories. The carbon microphone also made broadcasting possible when radio stations opened in the early 1920s, although its background hiss was one reason why studio engineers soon began to look elsewhere.

The alternative they wanted was the capacitor, or condenser, microphone. In 1916 Edward Wente of the Bell Telephone Laboratories in New York built one in which a very thin metal diaphragm formed one plate of a capacitor and a fixed backplate the other. Sound moved the diaphragm, the changing gap altered the capacitance, and because the capacitor carried a constant charge the variation appeared as a change in voltage. The moving part was almost weightless by the standards of the time, which gave a flatter and more faithful response, but the microphone required a polarising voltage and an amplifier built into the same housing.

Two further designs followed in the 1920s and 1930s. The ribbon microphone, patented by Walter Schottky and Erwin Gerlach in 1924 and manufactured by RCA from the early 1930s, suspended a corrugated metal ribbon in a magnetic field. The ribbon acted as both diaphragm and conductor, and because it was open to the air on both sides it was equally sensitive to sound arriving from front or back. The moving-coil, or dynamic, microphone used a small coil attached to a diaphragm and placed in the field of a permanent magnet. Rugged and needing no power supply, it became the standard for live performance.

The next revolution was chemical as much as electrical. In 1962 Gerhard Sessler and James West, working at Bell Laboratories, showed that a plastic film could hold a permanent electric charge. Such an electret film could take the place of the polarising supply, which meant that a good condenser microphone could be made small, cheap and battery-powered. Electret microphones were soon fitted to cassette recorders and hearing aids, and the microphone inside a modern smartphone belongs to the same family.

Since the 2000s most microphones have been made not by assembling parts but by etching silicon in the same factories that produce computer chips. These micro-electro-mechanical systems, or MEMS, microphones are a few millimetres across and cost cents to produce, which is why a phone can hold several of them and use the differences between their signals to suppress background noise. The journey from Wheatstone's acoustic lever to a chip the size of a grain of rice took nearly two centuries. The microphone has become so common that most people notice it only when it fails.

Speech synthesis is the production of intelligible speech by machine, and its purposes have always been twofold. The first is to read aloud text that exists only as characters on a page or a screen, from a railway announcement to a novel. The second is to give a voice to people who cannot speak for themselves. These two aims make different demands on the technology, since a listener will tolerate a mechanical voice reading a timetable but expects something warmer from a book, and something personal from a device that speaks for a person.

The earliest attempts were mechanical. In 1779 the Danish physicist Christian Kratzenstein won a prize offered by the Academy of Sciences in St Petersburg for resonating chambers that produced the vowel sounds of five languages. A few years later the Hungarian inventor Wolfgang von Kempelen, better known for a chess-playing automaton, published a description of a speaking machine that used a bellows for lungs, a vibrating reed for vocal cords and a leather resonator shaped by the hand for the mouth. It produced recognisable words and short phrases, and its failures were as instructive as its successes.

In the nineteenth century the work moved from imitation to analysis. Alexander Graham Bell, whose father had devised a notation showing how speech sounds are formed, spent years trying to build a machine that could speak, and what he learned about sound led him towards the telephone. The telephone was in one sense a by-product of speech research, though it carried a voice rather than creating one, and for the next sixty years the two fields developed separately.

Electric synthesis began at Bell Telephone Laboratories. In 1939 Homer Dudley completed the vocoder, which analysed speech into a set of frequency bands and a signal representing the pitch of the voice, and could reconstruct it from those measurements. Its descendant, the Voder, was demonstrated at the World's Fair in New York that year. The Voder was not automatic: a trained operator, using a keyboard and a foot pedal, had to shape the sounds in real time to produce sentences. Audiences were astonished, but the machine was a musical instrument rather than a reading device.

The first computer to sing arrived in 1961, when John Kelly and Louis Gerstman used an IBM 704 at Bell Laboratories to synthesize the song Daisy Bell. In 1968 a team in Japan produced the first system capable of converting unrestricted English text into speech in software, and through the 1970s Dennis Klatt at the Massachusetts Institute of Technology mapped in detail the acoustic cues on which listeners rely. The work reached commercial form in the DECtalk device of 1984, and a closely related synthesiser based on Klatt's research gave the physicist Stephen Hawking the voice by which he became known.

A modern system is conveniently divided into a front end and a back end. The front end deals with text, which is far messier than it looks: it must expand 'Dr' and '1990', decide whether 'lead' rhymes with 'feed' or 'bed', and recognise that a telephone number is not a quantity to be read out. It then converts each word into a sequence of phonemes, the units of sound that distinguish one word from another. Finally it predicts prosody - where the voice should pause, which syllable should be stressed and how the pitch should move. The back end turns this plan into an audio waveform.

For two decades the dominant back end was concatenative synthesis. A voice actor was recorded reading thousands of carefully chosen sentences, and the recording was cut into short units, often single phonemes or pairs of them. To say something new, the system searched its collection for the units that would join most smoothly and selected the best sequence. In 1996 Andrew Hunt and Alan Black showed that choosing longer stretches, sometimes whole phrases, produced far more natural results. Their weakness was rigidity: each voice required its own expensive database, and no amount of editing could make it say a word the actor had not recorded.

The most recent transformation has come from machine learning. In 2016 researchers at DeepMind in London described WaveNet, a system that generated sound one audio sample at a time, predicting each sample from those before it, rather than joining recorded pieces. Within a year, systems such as Tacotron had learned to convert text directly into an intermediate picture of the sound, which a second network turned into audio. Because these models learn from large collections of recordings, they acquire intonation and timing that earlier engineers had to specify by rule, and a model can be trained on one speaker until it reproduces that voice convincingly.

Impressive as they are, neural systems have not solved everything. English spelling supplies a steady stream of traps, and the names of people and places are frequently outside the vocabulary the system has seen. Expressive speech - anger, warmth, irony - remains difficult to specify and to evaluate. Researchers are also working on languages with few written resources, for which there is little recorded speech to learn from. And the ease of copying a voice has raised questions about consent, fraud and the status of recordings as evidence, questions that technology alone cannot answer.

Almost every large technology company now calls the people in its customers' contact lists friends, and the word has been stretched to cover a colleague in another country, a stranger met once at a conference and an aunt who posts photographs of her garden. When a term is used so widely, it is tempting to conclude that the argument about online friendship is merely a quarrel about vocabulary. It is not. Behind the confusion lie real questions about what friendship is, what it asks of us and whether a screen can supply those things.

The most useful starting point remains Aristotle's account in the Nicomachean Ethics, written in the fourth century BC. He distinguished friendships based on usefulness, on pleasure and on goodness of character. The first two are easily formed and easily dissolved, since they last only as long as the benefit or the enjoyment. The third, which he considered the highest, requires time, familiarity and a degree of shared life; two people cannot know whether they are good for each other, he argued, until they have eaten together and lived alongside one another. Sceptics about online friendship quote this passage with satisfaction.

The most influential modern statement of that scepticism comes from Sherry Turkle, a psychologist at the Massachusetts Institute of Technology. In Alone Together, published in 2011, she argued that digital communication offers the feeling of company without its obligations. Messages can be edited, delayed or abandoned; a conversation can be left without excuse. What looks like intimacy, she suggests, may be a performance for an audience, and the constant availability of a device can crowd out the slower, less rewarding work of being with someone in person.

There is something in this, and anyone who ignores it is being sentimental. But two qualifications are needed. The first is that online contact mostly supplements existing friendships rather than replacing them: messages exchanged by people who also meet for coffee are a continuation of the relationship by other means. The second is that for some people the alternative is not rich face-to-face friendship but none at all. Someone housebound by illness, living in a remote village or belonging to a small and scattered minority may find online the only community in which friendship of any kind is possible.

Robin Dunbar, whose research on the brain size of primates produced the widely quoted figure of about 150 as the number of people with whom a person can sustain a stable relationship, has argued that this figure sets a limit which social media cannot lift. His surveys suggest that those who use such sites do not have larger networks of genuinely close friends, only a wider audience of weak acquaintances. This is an important corrective. It measures, however, how many relationships a person has, not how good any of them is, and a count of friendships cannot answer a question about their quality.

Friendship has survived other media. For two centuries the Republic of Letters, a network of scholars and enthusiasts who corresponded across Europe, allowed people who never met to argue, console and advise one another over many years, and anyone who has read such correspondence knows how much warmth can travel in ink. Writing is slower than speech, and slowness has advantages: a considered reply, written the next morning, may be more honest than a remark made in the noise of a room. Many people, including those who find face-to-face conversation difficult, say that they disclose more of themselves in text.

The medium nevertheless has costs of its own. A message can be ignored at no immediate price, and friendship is partly made of the things people do when it would be easier to do nothing. Public posting, too, encourages broadcasting rather than exchange: a photograph sent to a hundred people is not a conversation but a performance, and the response it invites is applause. It is significant that the most painful experiences reported in research on social media are rarely violent disagreements; they are the quiet ones - the reply that never came, the group chat from which one was silently left out.

So which is it? I would suggest three tests, none of them new. Is the feeling reciprocal, in the sense that both people give as well as receive? Does the friendship last long enough to survive disagreement and neglect? And does it survive a moment when something is genuinely demanded of one of the parties - time, money, attention, an inconvenient truth? Friendships that pass these tests are not made less real by the fact that they were conducted through a screen.

My conclusion, then, is that online friendships can be real friendships, and sometimes are. But that possibility should not be turned into a general claim. A friendship sustained entirely online is thinner, in most cases, than one that also involves presence, and it would be dishonest to pretend that a screen is a substitute for being there. The medium alone cannot settle what a relationship is worth; conduct can. What is needed is not less technology but more caution in the use of the word friend, and more willingness to be one.