Nine Words

Nine.

That's the number you landed on — at the last checkup, or counting on your fingers one evening, or comparing against a chart someone shared in a group chat. Mama. Dada. More. Dog. Ball. Then the uncertain ones: does the sound she makes exclusively for bananas count? What about the grunt that might be "more"?

The chart says fifty by eighteen months. Three hundred by two. Somewhere between now and then, the parenting literature promises, comes the vocabulary explosion — a term that implies something sudden and passive, as if your child will wake up one morning fluent in nouns. It's a comforting metaphor. It's also wrong.

The Spread

Start with this: the number nine means almost nothing on its own.

When the MacArthur-Bates Communicative Development Inventory was re-normed in 2023 with over 6,500 children — weighted, for the first time, to U.S. Census demographics — the new scores came in lower than previous norms, especially below the 90th percentile. Earlier editions had been biased toward educated, English-speaking families. The corrected picture is wider and messier.

At sixteen months, some babies produce zero words. Others produce well over a hundred. Both fall within the normal distribution. By twenty-four months, the standard deviation in vocabulary size — 175 words — is more than half the median of 319. The range of normal at any given age is so wide that no single number can locate your child meaningfully within it.

And the number depends on who's measuring. Weaver and Saffran (2026) found that parent reports on the CDI frequently failed to predict how the same babies performed in looking-while-listening experiments — the two gold-standard measures of early word knowledge often don't agree. Your count of nine isn't wrong. It's one angle on something that resists being counted.

What's Compounding

Between fifteen and eighteen months, four things happen simultaneously. None is the cause. All are necessary.

A word-learning shortcut emerges. At fourteen months, when your baby hears a new word in the presence of a familiar object and an unfamiliar one, she looks at the familiar object — the thing she already knows. By seventeen months, she does the opposite: she infers that the new word must name the unfamiliar thing. This is mutual exclusivity, and it transforms word learning from effortful mapping to rapid inference. At sixteen months, your baby is at the crossover — the shortcut almost online. But it isn't a rule she applies. Brody and colleagues (2024) showed that the shortcut only works when the new word is spoken with prosodic emphasis — the slight stress you naturally put on a new label. She's reading your communicative intent, not following a constraint.

Known words begin teaching new ones. At fifteen months — but not at twelve — babies can use words they already know to infer the meaning of words they've never encountered, for objects they've never seen. Luchkina and Waxman (2025) showed that hearing "apple, banana, grapes" followed by a novel word allows fifteen-month-olds to guess it probably means a food. The semantic network itself becomes a learning tool. Each word makes the next one easier to acquire — not because of some cognitive switch, but because the web of known words provides scaffolding for the unknown.

Walking changed the conversation. Walle and Campos (2013) found that walking onset predicts non-linear acceleration in both receptive and productive language, independent of age. Not because walking is linguistic, but because it restructured what happens between you and your child. The baby who walks approaches objects, holds them up, brings them to you. Your speech shifted — more labels, more questions, more narration of what she's doing rather than what she's looking at. The input changed when the legs did.

The renovation enables retrieval. The prefrontal cortex renovation I wrote about in "Nothing Is Happening" — the same circuits being pruned and myelinated at fifteen months — feeds directly into word learning. Thériault-Couture and colleagues (2025) tracked 145 children and found that vocabulary and executive function develop bidirectionally: better inhibitory control means sharper focus on word-learning cues, and more words provide labels that support cognitive flexibility. A positive feedback loop, powered by the neural infrastructure that was silently reorganizing while nothing seemed to be happening.

Not an Explosion

The word explosion is the problem. It implies the vocabulary appears — rapidly, mysteriously, as a developmental event. Gómez Díaz and colleagues (2024) tested the two competing theories directly: the "naming insight" model (babies suddenly realize that things have names) versus the accumulator model (vocabulary grows as experience compounds). Dense longitudinal data from monolingual and bilingual infants. The accumulator model won. There is no switch. There is no insight moment. There is only the compounding of experience — every pointing episode, every time you followed her gaze to the dog and said dog, every third reading of Goodnight Moon.

The largest study of infant language input to date — Bergelson and colleagues (2024), one thousand children across twelve countries and forty-three languages — found three things that predicted how much babies vocalized: age, clinical diagnoses, and the amount of adult talk they heard. Not gender. Not socioeconomic status. Not whether the family was bilingual. The quantity of conversation directed at the child.

And it compounds forward. Ferjan Ramírez and colleagues (2024) followed families from infancy to kindergarten: consistent parentese — that high-pitched, melodic, vowel-stretched speech you probably use without thinking about it — predicted lexical diversity, sentence length, and conversational turn frequency at age five. The investment you're making at sixteen months shows up in measurable school readiness four years later.

What Nine Becomes

One more thing the word count doesn't capture. It's not just how many words — it's what kind. Perry and colleagues (2023) found that the composition of a toddler's vocabulary at sixteen to twenty-four months — specifically, whether it's rich in shape-based nouns like cup, ball, and spoon — predicted language outcomes four to seven years later. The shape bias, a word-learning shortcut where children generalize new labels by shape, reflects healthy accumulation. Its absence isn't a count problem. It's a process problem.

So the parent counting words is reading the wrong metric — but not for the reason they fear. The number isn't too low or too high. The number is the wrong unit. What matters is the trajectory: whether each word is making the next one easier. And the composition: whether the words reflect the kind of embodied, interactive experience that builds.

Your baby has nine words, or twelve, or six. By winter she'll have hundreds. The distance between here and there will look, in retrospect, like something that happened suddenly. It won't have been sudden. It will have been every conversation you've already had — the ones where she said nothing back, and the ones where she said bah, and the ones where you understood anyway. All of it, compounding quietly, until the day you realize you've stopped counting.