After only a few short chapters, I had my first uncomfortable moment: "Reviewers  have customarily declared in order to praise a translation to the skies that it sounds as if it had been written in English. This is hollow praise ..." At which point I became all too conscious that only last week I had praised Sophie Lewis's translation of Marcel Aymé's The Man Who Walked Through Walls in almost exactly that fashion. "Where," asks Bellos, "is the bonus in having a French detective novel for bedtime reading unless there is something French about it?" I then remembered that I had also said that Lewis's translation retained the Gallic flavour, so consider myself mostly off the hook. Towards the end of the book, Bellos gives a scornful list of the meaningless adjectives used by book reviewers to describe translations they think are good: fluent, racy, stylish ... I have used "damned fine", which doesn't feature in the list. Is that OK?

Bellos has used this book, in part, as a means of demolishing received ideas about translation. I am all in favour of demolishing received ideas but, as Gloria Steinem said, the truth will set you free, but first it will piss you off. I would have lazily assented to the proposition that a translation is no substitute for the original, but this, as Bellos points out, is a stupid thing to say when you consider that, in fact, a substitute for the original is exactly what a translation is. And if we didn't have translations, then we would, as he points out, have no knowledge of the Bible, the works of Tolstoy, or Planet of the Apes.

People have always been saying daft things about translation. José Ortega y Gasset said "almost all translations done until now have been bad ones", which Bellos demonstrates is ludicrous by experimenting with replacements, eg "Almost all firefighters up till now have been bad ones." Referring to the "extravagant" amount of attention that has been paid by scholars to the story of the tower of Babel  our search for an original, unitary human tongue being its testament  he says "it is far from obvious that their time is well spent." (If anything, the idea of there being one original language and now many is exactly the wrong way round.)

He dismisses as sexist nonsense the old adage, first expressed in French, that translations are like women  if they're beautiful, they're not faithful, and if they're faithful, they're not beautiful; although he doesn't mention the even stupider and more sexist nonsense that Nabokov spouted when he said that women don't make good translators. One of the earliest descriptions of translators is a rather complex-looking collection of cuneiform wedges which means, in Sumerian, "turner"; this Bellos likes, and his meditation on the various words languages use for the concept of translation is itself illuminating.

There is no reason not to trust Bellos completely. Not only have other translators, such as Michael Hofmann, raved about this book: Bellos has the chops for this, as he has translated Georges Perec, whose Oulipian linguistic games are challenging enough to read as it is (not to mention purposefully challenging to write). Although even he balked at Perec's La disparition, the novel famously written without the letter "e"; Gilbert Adair managed that, and Bellos gives him his due.

It is also engagingly written, not to mention fascinating throughout, and any initial misgivings I might have had about the viability of a popular, almost 400-page work about translation are completely vanquished. And although about 10 times as many books get translated out of English than into it  we are an insular linguistic community  we must salute the ingenuity of the translator's art. This is as good a way of doing it as any. What reaction can you have, save awe and pity, when you try to translate "white as snow" into a language whose speakers have never seen snow, or come across a fact like this one: "There is no fully adequate Russian translation of something as apparently simple as the word 'cheese'"?

A frolicsome cover, and a title and subtitle that perform in two different registers of cool, mask a disquisition of remarkable freshness on language, speech and translation. In short, punchy, instructive chapters that take in such things as linguistics, philosophy, dictionaries, machine translation, Bible translations, international law, the Nuremberg trials, the European Union and the rise of simultaneous interpreting ("the Soviet delegate has just made a joke"), David Bellos, Princeton professor and translator of Georges Perec, Ismail Kadare and others, makes a maximalist case for translation as perhaps the definitive human activity.

The great Australian poet Les Murray says simply: "We are a language species." There are some 7,000 languages currently in use in the world. Were there to be parity among them  which of course there isn't  that would give rise to 25 million different pairs of languages, and therefore twice as many "potentially separate translation practices"; French into English not being the same as English into French. Suddenly, a new 49 million-ply international industry looms. In fact, though, there are only around 50 languages that participate significantly in the give and take of translation  resulting in a paltry 2,500 language relationships (thus winnowing the field by a factor of a handy 20,000)  and those not particularly equally.

There is a top table of a dozen or so languages (Chinese, Hindi, Arabic, Spanish, and so on) which will obtain access to most of the world for you, and then there is English, which is the "inter-language" of choice. (Bellos has a faible for technical terms, and luckily another one for explaining them.) For some reason  Hollywood, science, military alliances, tourism, pop music, the web  it's most people's favourite alternative language. "The reasons why English has made a clean sweep [...] are not straightforward. Among them we cannot possibly include the unfortunate but widespread idea that English is simpler than other languages." It's a very typical Bellos notion, that of the effective wrong idea.

Of the million translated books listed by Unesco since its inception, fully one tenth are translated into English, and a staggering two thirds from English. This means (at least) two things: books written in English are at an enormous advantage against books written in any other language; and "whatever language you write in, the translation that counts is the English one". English is the floor, the language of international competition, the language of visibility. Sans English, you do not pass go. Sans English, there is no go.

It also means that in the field of translations, English is in play fully three-quarters of the time. Who knew? Really, it ought to mean that we have a keen sense of our good fortune; that (as befits a trading nation and an island state) we are comfortable with the idea of translation, knowing as we do that our books have an absurdly easy time of it abroad, and that we are kept tolerably well supplied from other languages; that we devote a respectable amount of thought to the commerce of books and ideas through a jostle of languages. That, to adapt one of Bellos's eccentric and wonderful heroes, a Francophone writer with the nom de plume of Antoine Volodine, we revere English not as the language of Byron and Shakespeare and JK Rowling, but as the ultimate repository of, say, Eco and Remarque and Kawabata. Instead of which, translation is a dirty secret, and the level of thinking about it ought to be an embarrassment. (That's not Bellos, by the way, it's me.)

Because Bellos seems to have that rare and wonderful thing, a sunny, Scotch (or Scotch-educated) temperament. There may be no particular reason for things to be the way they are  "the solar structure of the global book world wasn't designed by anyone"; he points it out, and moves on. He doesn't accuse, doesn't lament, doesn't gripe. The terrain is fraught with misconceptions, many of them ancient and indurated. He clears away a few: "The Great Eskimo Vocabulary Hoax" (that's the "hundred words for snow" idea to you); the canard of fidelity; the "no substitute for the original" prejudice against translation (mostly  though you don't find him saying so  from people who have never learned a language or crossed a street to read an author); the idea of languages as settled, finite things whose natural habitat is dictionaries (as if wild animals came from zoos); the idea of language as separate from speech (this too he has a label for: "scriptism"); the idea that Robert Frost ever said anything like "poetry is what gets lost in translation" (it was news to me too that he hadn't).

He doesn't make translating appear any harder than it is, or better than it is. He describes his work as rewriting foreign books into something he engagingly calls "English-minus"  English with the least possible amount of local or regional or national variations, so that it can be read all over the (English-speaking) world. Literary translators "don't have too hard a time", he says. "Using one word for another isn't special, it's what [people] do all the time." And then, for the T-shirt: "Translators just do it in two languages."

When Bellos finally settles on a thing for a translation to be, it doesn't come trailing philosophical clouds of glory and ingenuity. In that empirical, British, stone-kicking way, finding a dauntingly simple word for something almost indescribable, he asks that it be "a match". I could say anyone with an interest in translation should read Is That a Fish, but there wouldn't be very much point; instead, anyone with no interest in translation, please read David Bellos's brilliant book.

Michael Hofmann's Selected Poems is published by Faber.