The Language That Comes Home at Three

AA language nest is a preschool in which nothing is spoken but the ancestral language, and in which the adults doing the speaking are usually old enough to have learned it at home. The first opened at Torvik in 1982 with eleven children and three grandmothers; by 2019 the model had been copied into two hundred and fourteen centres across seven countries and had acquired the standing of a default remedy. The reasoning behind it is plain. A language that has stopped passing from parent to child will not be rescued by lessons at sixteen, so the handover must be reinstated at three. What remains unsettled is the size of the debt a nest is being asked to clear in a morning. Whether the fluency measured at the end of a nest year belongs to the classroom, or to the small number of households that would have raised speakers in any case, has divided the field for three decades.

BMarit Ekholm, a sociolinguist at the Institute for Language Ecology in Torvik, holds that the mechanism is real but that the accounting built on it has been careless. Between 1994 and 2022 her group followed six hundred and twelve children through forty-eight nests, logging every hour of contact with the language and testing unscripted production each spring. Below roughly fourteen hundred cumulative hours almost no child produced a sentence that had not been rehearsed; above two thousand, four in five did. Her quarrel is not with that finding but with what was laid on top of it. The number of centres in the region more than doubled after 2005, a rise that ministry summaries present as recovery, while the average child's yearly contact slid from six hundred and ten hours to four hundred and forty as sessions were shortened to admit more families. The expansion, in short, was paid for out of the very quantity her data had identified as decisive.

CThe account makes a prediction, and Ekholm has tested it. If exposure sets the ceiling, children in nests running full days ought to outperform those attending mornings only, and they did: at seven the full-day group reached a median of sixty-eight on her production index against thirty-one. The advantage was confined, however, to children with at least one adult at home who answered in the language, however clumsily. Where no adult answered, the full-day children finished at thirty-five and the morning children at twenty-nine, a difference her own statistics call noise. Hours, she concludes, buy comprehension reliably and production only conditionally, and she has taken to describing the nest as an amplifier rather than a source, on the grounds that it multiplies whatever the household supplies, including nothing.

DAna Belen Cortes, who studies language socialisation at the Centre for Minority Sociolinguistics in Oviedo, replies that hours are the wrong currency. What decides whether a child keeps a language, she argues, is the number of separate people it can be used with, since a tongue heard from a single voice is learned as a subject rather than as a way of living among others. Across thirty-three communities her group counted the adults outside the nest who used the language with children unprompted. Where a child could find at least five such adults, fifty-two per cent were still using the language weekly at fifteen; where fewer than three could be found, the figure was nine per cent. Session length, by contrast, explained almost none of the variation once that count was held constant. She concedes two things. All thirty-three communities lie within a single state whose schools already teach the language two hours a week, and the counts of adult speakers rested on parents' recall rather than on observation.

EThe dispute reached the funding forms in 2018, when the regional authority began to require any nest applying for money to show that it also ran a weekly class for the parents, and applications without one were turned down. Cortes points out, awkwardly for her own position, that an authority can pay for a class but cannot pay for the willingness to use a language in a shop or a kitchen, so the density she argues for is not a thing the instrument can manufacture. Ekholm answers that this appetite for whatever can be entered on a form is precisely why session hours went on being cut while centre counts rose, since an hour taken out of a timetable leaves no trace an auditor is looking for. Both regret one habit above the rest: that the field still reports children enrolled rather than households converted, a figure that flatters every authority quoting it.

FA settlement of sorts is emerging in which the two accounts are divided between community types rather than ranked against one another. Where a generation of home speakers survives, the nest's hours decide what a cohort gains and session length is worth arguing over; where the last household speaker has gone, enrolment secures a curriculum without producing the everyday company that would carry the language past the school gate. The unresolved question is whether a language rebuilt through nests keeps the grammar of the one it is replacing, since the earliest nest cohorts are only now old enough to raise children of their own, and no series yet runs long enough to answer it. What both researchers reject is the assumption behind the first enrolment targets, namely that a language returned to a timetable has been returned to a community.