Few questions in the study of language have generated as much heat, or as little consensus, as the one in this essay's title. The intuition behind it is seductive. Anyone who has learnt a second language knows the vertiginous sense that a familiar thought has to be rebuilt from different materials, and that something – a nuance, a joke, a shade of feeling – does not quite survive the journey. From there it is a short step to a grander claim: that the language we happen to speak quietly dictates what we are able to think.
That claim is usually associated with two American scholars of the early twentieth century, Edward Sapir and his student Benjamin Lee Whorf. Whorf, who worked by day as a fire-prevention inspector for an insurance company, developed his ideas in the evenings, and they were bold. Studying the Hopi language of the American South-West, he argued that its speakers did not think of time as a series of countable units flowing from past to future, as speakers of European languages do, and that their grammar both reflected and entrenched this difference. The idea that would later be called 'linguistic relativity' was born, and it travelled far beyond the seminar room.
It also met with a formidable backlash. Later fieldwork suggested that Hopi has perfectly serviceable ways of talking about time, and that Whorf had overstated his case. Meanwhile, from the 1950s onwards, an influential school of linguistics emphasised what all human languages share rather than how they differ, and the notion that a grammar could imprison a mind began to look not merely wrong but faintly embarrassing. The strong version of the hypothesis – that language determines thought, so that a concept with no word is literally unthinkable – has few defenders today. It falls at the first hurdle: people routinely grasp ideas for which they have no word, which is precisely why languages keep coining new ones.
Yet the story did not end there, because a weaker and more interesting claim survived. Perhaps language does not build the walls of thought but merely nudges the traffic within them. Since the 1990s, psychologists have tested this idea with considerable ingenuity. Russian, for instance, has two basic colour words where English has one: a word for lighter blues and a separate one for darker blues. In laboratory tasks, Russian speakers were found to be slightly quicker at telling apart two shades when they fell on either side of that boundary. The effect is small – measured in fractions of a second – and English speakers can, of course, see the difference perfectly well. But it suggests that habitual categories leave a faint imprint on perception.
More striking still is research among speakers of certain Aboriginal Australian languages, which describe space not with words like 'left' and 'right' but with the equivalents of north, south, east and west. Speakers of such languages say, in effect, that there is an ant on your south-west leg. To talk like this, one must know where north is at every moment, and they do: tested in unfamiliar surroundings, they can point to the cardinal directions with an accuracy that astonishes visitors. Here, plausibly, the demands of a language have trained a skill. The language does not make other ways of thinking impossible; it makes one particular way indispensable.
Sceptics urge caution, and with reason. Correlation is not causation: a community that lives by landmarks and horizons may develop both its spatial vocabulary and its sense of direction for the same environmental reasons, without one causing the other. There is also publication bias to reckon with, since a study that finds a striking difference between speakers is more likely to be noticed than one that finds none. And effects that appear in carefully controlled experiments can evaporate in the messiness of ordinary life. Some of the most quoted findings have fared badly. A widely reported experiment suggested that people described objects in ways that echoed the grammatical gender of their names in their own language – a bridge, say, as 'elegant' or 'sturdy' depending on whether the noun was feminine or masculine. Attempts to repeat it have produced weaker or inconsistent results.
Where does this leave us? Somewhere less dramatic than Whorf hoped, but more interesting than his critics allowed. Language is probably not a prison, nor even a pair of tinted spectacles that we can never remove. It is better thought of as a set of well-worn paths across a landscape: we can walk anywhere we like, but we tend to go where the ground is already trodden. The thoughts we find easiest to have are, to some modest degree, the ones our language has made easy to express. That is not a cage; it is a habit. And habits, unlike cages, can be noticed – which may be the best argument of all for learning another language.