Emergent classroom sign language

How can a schoolyard of deaf children invent a grammar that no adult ever taught them?

In 1977 a school for special education opened in Managua, and for the first time deaf Nicaraguan children were gathered in large numbers rather than raised in isolated hearing families. The teaching was oral — lip-reading and Spanish speech — and it largely failed. No signed language was taught, because at that point there was none in the country to teach.

Within about a decade the children were signing to each other in something with consistent word order, verb agreement, and grammatical devices nobody had shown them. The obvious explanation is that an adult must have brought a language in. The historical record does not support that, so the question is genuinely live: where does a grammar come from when there is no one to copy?

Start with what the children actually brought. Most had homesign — gestures worked out with their own families, expressive enough for daily life but private to each household, and lacking the systematic machinery linguists call grammar. Pool a few hundred of those in a playground and what should you expect? Something like a contact system: shared vocabulary negotiated on the spot, with word order varying and each signer improvising. That is roughly what the first arrivals produced, and it is not yet a language.

So the interesting event is what happened next — and this is the step that decides the whole explanation. If a grammar simply accumulated with practice, the strongest signers of that first cohort should have been the most systematic. They were not. Ann Senghas and colleagues, studying the community across cohorts, found the regularities arriving with the younger children who joined later — children whose input was the messy, variable system of the older ones, and whose output was more regular than what they had received.

Sit with how strange that is. The output exceeded the input in orderliness. Whatever produced the grammar was therefore not transmission; a copy cannot be cleaner than its original. Something in the young learner is imposing structure that was not there to be learned.

What kind of structure? The clearest documented case concerns how motion is expressed. Where a hearing Nicaraguan gestures a rolling descent as one continuous movement, and where the earliest signers largely did the same, later cohorts split it into two separate signs — the manner of movement and the path of movement — and produced them in sequence. That is a real grammatical move: an analogue gesture has been broken into discrete recombinable units. Discreteness is a design feature of every human language, and here it can be watched appearing, cohort by cohort, in a community whose age is known.

Now push on the mechanism, honestly. Why would a child regularize? A reasonable account is that young learners are not recording their input but building a system from it, and system-building imposes consistency — a rule that applies sometimes gets applied always. This is consistent with what is seen elsewhere: young children learning from inconsistent input tend to over-regularize it, while adults reproduce the inconsistency more faithfully. And it matches the creolization pattern that Derek Bickerton argued for, in which a variable pidgin becomes a fully grammatical creole in the hands of the generation that acquires it as a first language.

I would flag where this is less settled than the story usually sounds. How much of the structure follows from an innate language faculty, how much from general learning biases toward regularity, and how much from the ordinary pressure to communicate efficiently in a group is actively contested. Some scholars have also questioned the sharpness of the "no prior language" framing, noting that deaf Nicaraguans had contact and homesign traditions before 1977. What the case establishes firmly is the shape of the emergence — successive cohorts, each systematizing further — rather than a settled account of its cause.

One condition deserves emphasis, because it explains why this had not happened before. None of it required teaching, but all of it required a community. Homesign had existed in Nicaragua for as long as deaf children had. What 1977 supplied was density: enough peers, in daily contact, over enough years for a new cohort to arrive and find something to systematize. A school built to teach speech accidentally supplied the one ingredient a language needs.

Three refinements.

First, "invented" overstates the agency. No one sat down and devised a grammar. Each child did the ordinary thing a child does with linguistic input — extract a system from it — and the aggregate of those acts, repeated across cohorts, is what the grammar is. The result looks designed and no one designed it.

Second, this is not evidence that teaching is unnecessary. The community produced its own language because the education system failed to provide one, and the human cost of that failure — years of impoverished schooling for the very children involved — was real. The finding is about language acquisition, not a case for withholding instruction.

Third, the direction of the effect is the counterintuitive part and worth holding onto: the youngest and least experienced members were the innovators. In most skills, expertise flows downward from those who have practised longest. In language emergence it runs the other way, because the relevant capacity peaks early and then fades.

• Al-Sayyid Bedouin Sign Language, a village signed language that emerged under quite different conditions.
• What homesign shows about which structures a single isolated child can build alone.
• The critical-period evidence for why late learners regularize less.

Every term the collection defines is gathered in the glossary.