Two Languages, One Brain

AThe belief that a single mind can accommodate only so much language has proved unusually durable, outliving both the collapse of the studies that first appeared to support it and a century of evidence pointing the other way. Its everyday form is familiar enough. Someone who speaks two languages is assumed to know each of them a little less well than a person who speaks one, and to pay for the second in hesitation, in fruitless searching for names, and in that momentary blankness which specialists call a tip-of-the-tongue state. The assumption is not pure superstition. On a standard naming test, in which participants see line drawings and say aloud what each one shows as fast as they can, bilinguals are reliably slower, by something in the region of 40 to 60 milliseconds, and get through fewer pictures in a minute. Where the field divides is over what that delay is a measure of. A gap in knowledge, a difficulty in reaching what is known, and an artefact of how vocabulary happens to be totalled are three quite different propositions, and the last thirty years have been spent trying to tell them apart.

BElsa Varga, who directs the language sciences programme at the Doornik Centre, has argued since 2009 that most of the apparent deficit dissolves as soon as the arithmetic is put right. Vocabulary tests, she points out, are given in one language at a time, so a bilingual who can name a hammer in Portuguese but not in French is recorded as not knowing hammers at all. When her group re-tested 1,240 adults in 2017 and credited an item as known if it was available in either language, the distance between the two groups almost vanished: monolingual participants recognised 88 per cent of the list, bilinguals 87. Scored one language at a time, the very same bilinguals reached 62 per cent in their weaker language. Nothing had changed but the bookkeeping. Varga is unusually careful about what this does and does not establish. Her figures speak to how much is stored, she concedes, and say nothing about how quickly any of it can be reached, which is precisely what a naming test is built to capture.

CTomás Herrera, a psycholinguist at Ravensmoor, is sceptical that the slowness is an artefact of anything. Naming pictures in a booth is artificial, he accepts, yet the same hesitations turn up in recordings of unremarkable conversation, where nobody is being timed and no drawings are involved. Analysing 51 hours of speech gathered in a hospital whose staff addressed patients in one language and colleagues in another, his team found that bilingual speakers broke off in mid-clause about a third more often than monolingual colleagues, and abandoned a chosen word and began again roughly twice as frequently. Retrieval, he maintains, is genuinely harder when two labels compete for a single concept, and no amount of generous scoring will make that competition disappear. The weakness of his evidence is one he states himself. The recordings covered eleven speakers over three weeks, every one of them an adult who had taken up the second language after the age of twenty, and no pause could be tied with confidence to a search for a word rather than to the interruptions of a busy ward.

DA third position holds that both camps have been asking a question pitched at the wrong level. On the account developed at the Institute of Mind and Language in Trondheim, what matters is not bilingualism in itself but the conditions under which a speaker's two languages are habitually used. Where the languages are kept apart, one at home and one at work, the speaker must repeatedly hold down whichever language does not belong, and the bill arrives at the moment of switching; laboratory tasks put that charge at roughly 190 milliseconds. Where the two are interleaved freely inside a single conversation, as they are among many bilingual communities, suppression is unnecessary and the switching charge falls close to zero, although speakers of this kind do slightly worse when they are later obliged to keep the languages apart. A survey of 190 secondary teachers carried out between 2014 and 2019 located the pattern in an unexpected place. Those who taught in one language and lived in another showed the effect plainly, whereas colleagues who mixed both in the staffroom did not show it at all.

ENone of this settles the older question of whether two languages are a burden or a gift, and the honest answer may be that the question was badly put. The costs that survive scrutiny are narrow and modest: a delay of a few tens of milliseconds in producing one word, detectable only in a quiet room and invisible in conversation, together with a thinner store of words in whichever language is used less. Several of the benefits claimed for bilingualism have shrunk in much the same way as the samples testing them have grown. Neither camp now expects a single answer that will hold for every speaker in every setting. What has replaced the old dispute is a more modest and more useful picture, in which one arrangement, two systems permanently available and one of them selected, yields different consequences according to how often the selection must be made. Varga and Herrera agree on almost nothing else. Both now insist that the effects are conditional, which is not the conclusion the twentieth century expected but is the one the measurements will bear.