There are two ways in which we can think of literary translation: as reproduction, and as recreation. If we think of translation as reproduction, it is a safe and harmless enough business: the translator is a literature processor into which the text to be translated is inserted and out of which it ought to emerge identical, but in another language. But unfortunately the human mind is an imperfect machine, and the goal of precise interlinguistic message transference is never achieved: so the translator offers humble apologies for being capable of producing only a pale shadow of the original. Since all he is doing is copying another's meanings from one language to another, he removes himself from sight so that the writer's genius can shine as brightly as may be. To do this, he uses a neutral, conventionally literary language which ensures that the result will indeed be a pale shadow, in which it is impossible for anybody's genius to shine. Readers also regard the translator as a neutral meaning-conveyor, then attribute the mediocrity of the translation to the original author. Martin Amis, for example, declares that Don Quixote is unreadable. without stopping to think about the consequences of the fact that what he has read or not read is what a translator wrote, not what Cervantes wrote. If we regard literary translation like this, as message transference, we have to conclude that before very long it will be carried out perfectly well by computers. There are many pressures encouraging translators to accept this description of their work, apart from the fact that it is a scientific description and therefore must be right. Tradition is one such additional encouragement, because meaning-transference has been the dominant philosophy and manner of literar3 translation into English for at least three hundred years. The large publishing houses provide further encouragement, since they also expect the translator to be a literature-processor, who not only copies texts but simplifies them as well, eliminating troublesome complexities and manufacturing a readily consumable product for the marketplace. But there is another way in which we can think of literary translation. We can regard the translator not as a passive reproducer of meanings but as an active reader first, and then a creative rewriter of what he has read. This description has the advantages of being more interesting and of corresponding more closely to reality, because a pile of sheets of paper with little squiggly lines on them, glued together along one side. only becomes a work of literature when somebody reads it, and reading is not just a logical process but one involving the whole being: the feelings and the intuitions and the memory and the creative imagination and the whole life experience of the reader. Computers cannot read, they can only scan. And since the combination of all those human components is unique in each person, there are as many Don Quixotes as there are readers of Don Quixote, as Jorge Luis Borges once declared. Any translation of this novel is the translator's account of his reading of it, rather than some inevitably pale shadow of what Cervantes wrote. It will only be a pale shadow if the translator is a dull reader, perhaps as a result of accepting the preconditioning that goes with the role of literature-processor. You may object that what l am advocating is extreme chaotic subjectivism, leading to the conclusion that anything goes, in reading and therefore in translation; but it is not, because reading is guided by its own conventions, the interpersonal roles of the literary game that we internalise as we acquire literary experience. By reference to these, we can agree, by reasoned argument, that some readings are more appropriate than others, and therefore that some translations are better than others. Which of the fo