There’s a number that seems to have shaped early-years policy for thirty years.

By age three, the number says, children growing up in higher-income families have heard, on average, thirty million more words than children growing up in lower-income families. Thirty million. The number is so large that it’s fairly easy to remember. It has appeared in TED talks, government white papers, parenting books, charity fundraising decks, and more than one podcast episode I half-listened to and made my ears perk up. It is the number behind lots and lots of interventions. It is, for many, the foundational fact about early language.

It is also, depending on whom you ask, slightly true, badly framed, statistically fragile, methodologically narrow, politically loaded, and, when used in its purest, decontextualised form, wrong.

I came across this study while reading up about language acquisition in bilingual children, which this study is NOT about, but my (recovering marketer’s) mind found it really interesting and decided to dig a bit deeper into what the science of language acquisition actually tells us, what gets lost when a finding becomes a slogan, and what any of it means for the choices a parent makes on any normal, uneventful day.

But let’s go back to the source. The original study, published by Betty Hart and Todd Risley in 1995, was a small longitudinal investigation. They recruited 42 families in Kansas - a mix of professional, working-class, and households on welfare subsidies, and recorded an hour of household conversation each month for two and a half years, beginning when each child was around seven months old.

What they found, when they tallied the words, was striking to them. Children in professional families heard, on average, around 11 million words per year. Children in working-class families heard around 6.5 million. Children in welfare-recipient families heard around 3 million. By age three, the cumulative gap between the highest and lowest input groups was roughly thirty million words.

More importantly, when Hart and Risley followed up with a subset of the children at age nine, the vocabulary measured at age three predicted academic outcomes six years later. The shape of the early input, was the argument, had cast a long shadow.

This was a remarkable piece of longitudinal data. It still is. The methodology involving recording naturalistic input over years, in actual homes is slow and patient observational work that doesn’t get done very often. Whatever else is true about the controversy that followed, Hart and Risley did real work, and the data they collected is real.

The problem is what happened next.

The “thirty million word gap” became a catchphrase. By the early 2010s, it had hardened into a kind of conventional wisdom about why some children fall behind in school. Providence, Rhode Island won a $5 million Bloomberg Mayors Challenge grant in 2013 to launch Providence Talks, a scheme that gave low-income parents wearable devices to count the words they said to their babies. The Clinton Foundation partnered with Univision on Talking is Teaching. At the University of Chicago, the paediatric surgeon Dana Suskind turned the number into a book, a TED talk, and eventually a whole curriculum for parents. The implication, always present even when not stated it seems to me, was that low-income parents talk less to their children, and that’s possibly why their children don’t do as well in school.

This was where the trouble started.

In 2019, the linguist Douglas Sperry and his colleagues published a re-examination of the original methodology. Using five different datasets, a much larger sample, and a way of counting that captured language input from sources beyond the primary caregiver (think extended family, neighbours, ambient adult conversation, siblings) they redid the maths with a broader definition of “input”. The gap shrank dramatically. In some samples it disappeared entirely. In others, the rank order of socioeconomic groups was actually reversed.

Their argument, in essence, was this: Hart and Risley counted only certain kinds of speech, that is child-directed speech from the primary caregiver, and treated it as the totality of a child’s linguistic input. But in many communities, particularly working-class and minority communities, children are surrounded by ambient adult speech that wasn’t being captured. They listen to grandparents. They listen to neighbours on the porch. They listen to the radio playing in the kitchen. None of this was in Hart and Risley’s count.

By treating one culturally-specific style of speaking to children as universal, the original study had inadvertently “pathologized” the speech communities of families who simply do childhood differently - and is nothing to do with income levels.

This resonates with me especially strongly as the Italian culture I grew up in was rich in ambient speech, and I notice it also when my daughter visits Italy: there is always one (or multiple, which can be a bit of a headache sometimes) conversation in the background, someone talking loudly in a neighbouring home with their window open, the chitchat in the queue at the chemist, or the little interactions while buying bread at the panetteria.

This is, to me, the part of the story that matters. It doesn’t make the original finding wrong, the original finding still has value, but it shows what happens when a piece of social-science research gets taken up by the quest for a simple story about why some children do well and others don’t.

There is a particular danger to research that produces a clean number, a temptation that needs to be questioned, always.

Thirty million is so easy to remember, and almost irresistible. It is concrete, quantifiable, that it makes the problem of early language inequality look like a matter of maths.

If the gap is thirty million words, you can fix it by adding more words. Tempting. You can print posters that say “talk to your baby lots.” You can launch an initiative that promises to close the gap.

What you cannot do, at least, not as easily, is sit with the harder question, which is: what kind of conversation, from whom, in what kind of relationship, makes a difference to the kind of life a child eventually lives?

As one might suspect, the answer to that question, drawn from the broader literature on early language acquisition, is messier and less tractable. It involves the quality of conversation, not its quantity. It involves whether a child’s vocalisations are responded to. Whether stories are read, and how: straight through, or with pauses for prediction and questions. Whether feelings are named in the household, or only commands issued. Whether the child has the experience of being a participant in conversation, or only the recipient of instruction.

These factors are quantity-independent. A family that speaks to its child in only seven million words a year, but does so responsively, by answering vocalisations, reading dialogically, naming emotions, helps children do better than a family that speaks twelve million words a year mostly as background noise.

The research that supports this - think of work on dialogic reading, on serve-and-return interaction, on caregiver responsiveness, on emotional vocabulary, is more robust than the original Hart and Risley study. It has been replicated across more populations, more languages, more cultural contexts. It is also, unfortunately for the slogan lovers, much harder to fit on a poster.

I think about all of this most when I’m tired.

In the evening, the bath running, a two-year-old trying to climb into the tub fully clothed or begging me to ‘no pelli’ (which means ‘no capelli’, her plea to not wash her very very curly hair, as washing involves brushing which is not her favourite sport!), and a part of my brain wanting to put on a podcast and let the bath happen quietly. I am, in those moments, the person the research is about: the parent whose choices in the next twenty minutes constitute, in some small way, the linguistic environment of her child.

What the research tells me, both the original 1995 finding and its 2019 re-examination, is that the question is not how many words am I producing in this bath. The question is whether I am responsive. Whether, when she points at the tap and says acqua, I respond — sì, l’acqua, è troppo calda o va bene? — instead of nodding and looking at something else (possibly my puffy eyes in the mirror by this point of the day). Is the bath a small conversation, or a small monologue? Or in our case, a sung monologue with audience commentary (i.e. mine!).

I get this right about half the time. The other half, I look at the tap and zone out a bit. I do not say this to be self-flagellating. But because the research, properly read, takes the pressure off in some directions and applies it in others. It does not require me to fill the bath with words. It requires me to be there.

This is, I think, what gets lost when thirty million words becomes the slogan.

There is another piece of this that’s been spinning in my mind, even though it is not strictly about linguistics - but it’s more reflective of my international development studies and experience.

The deficit framing of the original study had political consequences. By treating the speech of low-income, working-class, and minority families as less (insert what you wish: less sophisticated, less language-rich, less developmentally generative) the early-years sector ended up implicitly arguing that the solution to language inequality was to make poor families talk more, in the manner of professional families. This was a strange argument, when you stop to look at it. It located the problem inside the family rather than inside the structural conditions that shaped the family.

A child whose parent works two shifts cannot be talked at for the same number of hours per day as a child whose parent works one. A grandmother who speaks a different language and is told her input “doesn’t count” because the school operates in English does count, in the home, in very real ways. A community that does childhood through aunties and cousins and neighbours rather than through one-on-one parental speech is not a deficient community. It is a different, often richer, community.

None of this means we should stop thinking about language inequality. It means we should think about it less as a matter of word count and more as a matter of structural support: for the parents who are tired, for the heritage languages that are at risk, for the books that are not in the household, for the child who needs more than one adult holding their attention.

So what do I take from all of this, on any boring, uneventful day, deciding what to do with the next twenty minutes?

I take, first, that the broad shape of the original finding is probably right: language exposure in the first three years matters, a lot, and the kind of language a child is surrounded by shapes the kind of language they grow up to use. This is intuitively obvious and it is also empirically supported. Hooray, as I am a real chatterbox so this makes me feel a little better.

I take, second, that the content of that exposure matters more than the count. Responsiveness over volume. Quality of attention over number of words. A bath that becomes a conversation over a bath that becomes a monologue with no audience. This is also a lesson for me in listening, slowing down, and responding. As a bonus, in those moments, when my daughter is in equal parts tired and relaxed in the bath tub, the funniest stories from her day come to the surface, and her little narration, full of incomprehensible sounds, a mix of languages and beautiful vivid images fills the room: I’d give anything to bottle it up and never forget.

I take, third, that the deficit framing is wrong, and that whatever else I am doing in our household, I want to model for my daughter a way of being among words that doesn’t exist in contrast to other people’s. Italian is one room. English is another. The Greek her friend at nursery hears at home is a third. The Polish her key worker switches into when she calls her mother is a fourth. The household next to ours speaks French; the cafe down the road plays Brazilian music. The world is a chorus, and thirty million is just the part of it that happened to get counted.

What I am trying to do here through this Substack, through the picture books I am making, through the slow conversations at home, is not to add to a child’s word count. It’s to support the kind of home (mine, in primis) where conversation is something we do, in any language we have, in the small daily scenes that won’t make it to a slogan.

The thirty million word gap is real. Ok. But the more important number, the one that doesn’t get a campaign, is the one between the words and the silence. The small, responsive yes, I see you that turns input into language and language into a relationship.

That number is impossible to measure. Which is why it is so easily forgotten. And why I felt the need to write about it.

Sowing Tales is a publication on slow, screen-free, multilingual childhood. Letters about once a fortnight, free.

Sources

The original finding

• Hart, B. & Risley, T.R. (1995). Meaningful Differences in the Everyday Experience of Young American Children. Paul H. Brookes Publishing. The longitudinal book

• Hart, B. & Risley, T.R. (2003). “The Early Catastrophe: The 30 Million Word Gap by Age 3.” American Educator, Spring 2003, 4–9. The shorter, widely-circulated article that turned the finding into a slogan.

The re-examination

• Sperry, D.E., Sperry, L.L., & Miller, P.J. (2019). “Reexamining the Verbal Environments of Children From Different Socioeconomic Backgrounds.” Child Development, 90(4), 1303–1318. The paper that broadened the definition of “input” and found the gap much smaller.

• Golinkoff, R.M., Hoff, E., Rowe, M.L., Tamis-LeMonda, C.S., & Hirsh-Pasek, K. (2019). “Language Matters: Denying the Existence of the 30-Million-Word Gap Has Serious Consequences.” Child Development, 90(3), 985–992. A formal defence of Hart & Risley from five leading language-development researchers, published in the same journal. Sperry then replied to the reply. The debate is real.

The interventions the finding inspired

• Providence Talks (Rhode Island, launched 2013). Bloomberg Mayors Challenge grant winner; distributed wearable word-counting devices to low-income families.

• Too Small to Fail. Clinton Foundation initiative (2013), partnered with Univision on the Talking is Teaching campaign.

• Suskind, D. (2015). Thirty Million Words: Building a Child’s Brain. Dutton.

• The TMW Center for Early Learning + Public Health at the University of Chicago (founded 2010 as the Thirty Million Words Initiative).

On responsiveness over volume

• Weisleder, A. & Fernald, A. (2013). “Talking to Children Matters: Early Language Experience Strengthens Processing and Builds Vocabulary.” Psychological Science, 24(11), 2143–2152.

• Rowe, M.L. (2012). “A Longitudinal Investigation of the Role of Quantity and Quality of Child-Directed Speech in Vocabulary Development.” Child Development, 83(5), 1762–1774.

• Romeo, R.R. et al. (2018). “Beyond the 30-Million-Word Gap: Children’s Conversational Exposure Is Associated with Language-Related Brain Function.” Psychological Science, 29(5), 700–710. MIT/Harvard neuroimaging work finding that the number of conversational turns predicts brain activity in language areas more strongly than raw word count does.

I don’t remember the exact day, but I remember the exact sweet moment. She is on the carpet, blabbering to herself and pretending to give a small doll a nappy change. It’s autumn so she must be 18 months old or thereabouts. I am doing something mundane, folding clothes, or maybe sneaking away for a tea, while singing one of her absolute favourites — Wind the Bobbin (the soundtrack of my life oscillates between Sleepy Bunnies, Nella Vecchia Fattoria, Baba Black Sheep and Stella Stellina depending on the time of the day, but I digress). She announces, without looking up, ”Mamma, [insert noise that sounds a bit like ‘kata’] Nonno.”

It takes me a moment to twig what she is saying. We have recently come back from Italy where we spent a bit of time with my parents, her grandparents, her Nonno e Nonna, and Nonno does have a big booming voice and likes to sing made-up songs, partly because he doesn’t know the lyrics. So it does take me a moment but I then get it. She wants me to sing (canta) in Nonno’s voice.

So I do. I clear my throat, drop my voice an octave or two lower than feels natural, puff my cheeks, and begin: Wind, wind, wind the bobbin up. Wind, wind, wind the bobbin up. It is a children’s song about factory work invented in nineteenth-century Lancashire (I had to look this up). I am singing it in the voice of an Italian grandfather, in a kitchen in London, to an 18 month old who has not noticed anything strange about this arrangement at all. She nods along, satisfied. I continue.

When the song ends she says, ”Altra, Nonna.” which in her world of word associations I take to mean ’once more, like Nonna’. My voice changes again — thinner now, gentler, a little quavering, the way her Italian grandmother actually sounds on the phone. Wind, wind, wind the bobbin up. Same song, second voice. She listens like a small conductor. When I finish she says, ”ancora” again, and we go around once more.

This is a thing that happens in our house now. It happens several times a day. Sing it like Nonno. Sing it like Nonna. It applies to every song in her repertoire — the English ones, the Italian ones, the Sinhala lullaby her Seeya, her paternal granddad, sings her. All of them now come in three versions: my voice, Nonno’s voice, Nonna’s voice.

She has begun to sing them in those voices herself. She announces it, like a tiny presenter: Adesso canto come Nonno. Now I sing like Nonno. The cheeks puff. The voice descends. The wheels on the bus, Giro Giro Tondo, Tikiri Tikiri Tikiri Liya (ටිකිරි ටිකිරි ටිකිරිලියා) — all of them, sometimes, are sung in the deep booming voice of a man neither she nor I see in person more than once every few months.

I am a linguist by training, and I cannot help asking what this is, exactly.

Three languages live in our house. Italian with me, English with her father and pretty much everywhere else — the nursery, the playground, the world — some Sinhala from her father’s heritage, a sprinkling of German from my years abroad. Two sets of grandparent voices live in our house too. The Italian ones — my parents — who we video call once or twice a week from a small town outside Milan. The Sri Lankan ones — her papi’s parents — who we call in Australia where they live, and read picture books over a slightly grainy connection. She loves all four of them. She has limited time with any of them — both sets live across a sea — and what she has of them, she takes very fondly, and seriously. And here let me pause a moment to acknowledge how lucky she is to have four wonderful grandparents. As someone who grew up with only two Nonnas, one of whom lived with us, I delight in the fact that she has them all, alive and well.

But the voices she imitates are specifically Italian. Nonno. Nonna. Not — for example — the voices of seeya and Ba, or technically Achie, her Sri Lankan grandfather and grandmother. She adores them too. They sing to her too, when we visit. She does not, however, do their voices.

I will come back to that. The Italian ones, first, arrive in our house as voices.

I think this is mostly about how the languages themselves carry voice.

Italian is, of all her languages, the most prosodic. It is what linguists call a syllable-timed language — every syllable holds roughly equal duration, and the music of the sentence comes from pitch and timbre and stress rather than from the rhythmic compression that English does. You can sing Italian in a way you cannot quite sing English. You can also tell whose voice an Italian sentence is in, even if the words don’t change. The pitch range carries more identifying information than it does in English.

Anne Fernald’s cross-linguistic work on infant-directed speech — the high-pitched, sing-song register adults use with babies the world over — showed that the language behaves differently in this register across cultures. Italian motherese carries a noticeably wider pitch range than English motherese. Italian mothers, technically, do not speak to their babies. They sing to them (and btw, I take this to another level as I literally sing through our day: it was from very young the only thing that seemed to settle my daughter, which is ironic if you knew me and my shocking singing voice).

What Italian exaggerates, a small Italian-listening child notices. She is sorting voices the way another child might sort colours: Nonno is one voice, Nonna is another. They are categorically different in a way that her English-speaking grandparents — for whom the pitch range is narrower, the cadence less punctuated — are not, to her ear, quite as different from each other.

So when she wants to play with voice, she reaches for the Italian set. They are the most legible voice-pair she owns.

Beneath this, something else is happening that I find remarkable.

Around the age of two, children begin to develop what linguists call register variation: the conscious ability to use their voice to indicate something about the speaker, the role, the situation. The classic example is a toddler addressing a baby in a higher-pitched, gentler voice than the one she uses with her father. It is a sign of an emerging theory of mind — the slowly-arriving recognition that other people have inner lives different from her own, and that you can hold them in mind well enough to imitate them.

Bilingual children tend to develop register variation a little earlier than monolingual children, because they are already, by necessity, practising language-switching from the moment they have language at all. They have learned that the same person can be addressed in different ways and the same chair called something different depending on who is in the room. That awareness — the I can switch — gets internalised early, and it doesn’t only switch languages. It switches voices, roles, characters.

When she sings Wind the Bobbin Up in Nonno’s voice, she isn’t mimicking, rather it is metalinguistic awareness, in toddler form. She is holding a voice in her head as a separate object from the language it speaks, and she is applying it to whatever song she likes. She has discovered that voice and language are decoupled.

So I guess this is a developmental win; and also, frankly, the cutest thing to watch.

It is no accident, I think, that the test material she has chosen is nursery rhymes.

Nursery rhymes are built for prosody. Wind the Bobbin Up, Giro Giro Tondo, Tikiri Tikiri Tikiri Liya — or any of the dozen others, are all constructed of rhythm, rhyme and repetition. They don’t have fantastically memorable lyrics, they are not even particularly good songs (if you ask me). They have survived and keep going strong in my view because they tune the ear and give a child a framework on which to hang the strange and irregular shapes of her language. They are, in a sense, language-training disguised as fun.

Layering voice-play on top of nursery rhymes is double work. You are practising the rhythm and rhyme and stress patterns of the song and you are practising the timbre and pitch of a particular speaker. And it seems to me that for a multilingual child, this is exactly the kind of layered exercise she is built to enjoy. Of course she does it. Of course she asks for it. Of course she announces Adesso come Nonno. It’s her gym, and she is loving doing reps.

The other part that really touched me, and I think the essay is actually about — is what happens when I notice that the Italian grandparents arrive in our house through voice and the Sri Lankan grandparents arrive in our house through story.

Her Seeya and Ba are present in her life in a completely different form. During our last visit, they read my daughter many books, and a couple really stuck with her, to the point we still read one of them every night before bed, the story of a cheeky black cat called Slinky Malinki (or Slinky Inky, if you asked her). She does things with her Sri Lankan grandparents that she does not with her Italian ones. She knows what their garden smells like. She knows what their favourite food smells like. She does not do their voices, but she can tell how Ba likes to water her plants, or Seeya likes to prepare his Weetabix for breakfast.

And I love noticing these subtle differences.

They tell me that absent grandparents arrive in a child’s life through whatever channel best matches them. The channel is partly a function of language access, partly a function of geography, partly a function of who the grandparents themselves are. The Italian grandparents are vocally expressive (aka loud and with constant chattering) in person and video, they sing along to nursery rhymes even when they don’t know the words; their voices are in the air. The Sri Lankan grandparents love to tend to the garden, to feed birds and observe other small creatures, to cook traditional foods and to read lovely books with my daughter; they are, in a way, on the page.

Both are real, precious forms of presence-making across distance.

What it teaches me — and might be useful to other parents — is that the trick of staying close to relatives who live far away is not necessarily to spend more time, which is often impossible. The trick is to find the channel that fits the person. Some grandparents are voices. Some are stories. Some are recipes — the smell of a kitchen well lived. Some are objects — a single wooden spoon that arrives in the post. Some are practices that a child learns from afar — a way of dipping the tea bag in the mug, of talking to the birds, of singing softly to herself in the bath.

The child will pick the perfect channel for each of them — and it’s a delight to witness.

A few small offerings, if any of this is useful to you.

For the parents of multilingual children with relatives — grandparents, aunts, uncles, godparents, family friends who function like family — who live a long way away:

Send the voice. Not just video calls, where the connection is patchy and the child is shy. Send recorded songs. Send lullabies sung straight into the phone, badly, with no pretence. Ask a grandparent to record themselves singing the thing they sang to your partner as a child. Ask an aunt to record the song she used to fall asleep to, or to read a short book you both have. Play them at bath time. The voice is a kind of presence the algorithm cannot dilute.

Send the story: buy picture books from the heritage country, even if you can’t yet read them aloud yourself. Ask the relative to record themselves reading one in their own language. Ask them to write down — in their own language, no translation needed — the bedtime story they remember from their own childhood. Slip the words into your child’s life.

Send the object: a small thing in the post that has a connection to a shared experience, however small. A spoon, a doll, a piece of cloth, a dried flower. Things have a stubborn way of carrying the person who owned them.

Send the practice: a way of doing something. A song before food. A morning prayer. A pose, a gesture, a habit. Children pick these up and turn them into their own without knowing they are doing it.

Don’t impose ‘the channel’: watch what your child does spontaneously, and trust it. If she wants to sing in Nonno’s voice and not in Seeya’s, that is information about how she is processing two heritage lines. The most trusting response is to widen each channel — not to redirect it. If Seeya reaches her through story, send more stories. If Nonno reaches her through song, send more songs.

Encourage voice-play of any kind: it is a developmental win, and it is free.

It is Saturday morning now. She is in the garden. She is singing, quietly, to herself, Tikiri Tikiri Tikiri Liya. The likely mispronounced Sinhala syllables float across the space and out into the day. After a few bars she pauses, looks at me, and says, ”Adesso come Nonno.”

So she sings in a deep voice, cheeks puffed, lips closed together with the same seriousness with which an Italian grandfather might announce, on a hot August afternoon, that it’s now time for a stroll and a gelato. She finishes. She nods, satisfied. She begins again, come Nonna this time, very softly, almost mouthing the words.

Three languages, two voices, one small body on a London patio. Somewhere outside Milan, a Nonno is reading the daily paper finishing his espresso. Somewhere outside Sydney, a Seeya is engrossed in a police thriller book. None of them know what she is doing. All of them are present.

This is the long embrace of grandparents who live overseas. A child puts her arms around them instinctively. The grandparents barely have to do a thing.

— Paola

---

Sowing Tales is a publication on slow, screen-free, multilingual childhood. Letters about once a fortnight, free. If this resonated, subscribe at the top of the page.

---

Sources

These are the sources that back the linguistic and developmental claims I lean on in the essay, mapped to the line of argument it supports, with a note where the literature is more specific than I am (and how fascinating it is to come back to this way of writing and researching, after a good few years).

”Italian is a syllable-timed language… you can sing Italian in a way you cannot quite sing English.”

- Pike, K. L. (1945). The Intonation of American English. University of Michigan Press. The original distinction between stress-timed and syllable-timed languages.

- Ramus, F., Nespor, M., & Mehler, J. (1999). Correlates of linguistic rhythm in the speech signal. Cognition, 73(3), 265–292. The empirical basis for grouping Italian and Spanish as syllable-timed against the stress-timed Germanic languages including English. https://www.sciencedirect.com/science/article/abs/pii/S0010027799000581

”Italian motherese carries a noticeably wider pitch range than English motherese.”

- Fernald, A., Taeschner, T., Dunn, J., Papousek, M., de Boysson-Bardies, B., & Fukui, I. (1989). A cross-language study of prosodic modifications in mothers’ and fathers’ speech to preverbal infants. Journal of Child Language, 16(3), 477–501. The foundational cross-linguistic comparison. Italian — alongside other Romance languages — shows the widest pitch excursion in mothers’ speech to infants. https://www.cambridge.org/core/journals/journal-of-child-language/article/abs/crosslanguage-study-of-prosodic-modifications-in-mothers-and-fathers-speech-to-preverbal-infants/

- Fernald, A. (1989). Intonation and communicative intent in mothers’ speech to infants: Is the melody the message? Child Development, 60(6), 1497–1510. The companion paper on why this pitch-range exaggeration matters for the infant ear.

”Around the age of two, children begin to develop register variation… It is a sign of an emerging theory of mind.”

- Andersen, E. S. (1990). Speaking with Style: The Sociolinguistic Skills of Children. Routledge. The classic study of how young children adjust voice and register to mark speaker, role, and situation.

- Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of-mind development: The truth about false belief. Child Development, 72(3), 655–684. The standard meta-analysis on theory-of-mind emergence — strictly speaking the *false-belief* shift happens later (around 3–4), but the precursor capacities, including the kind of speaker-as-other modelling that register-switching requires, are present from around two. https://srcd.onlinelibrary.wiley.com/doi/10.1111/1467-8624.00304

”Bilingual children tend to develop register variation a little earlier than monolingual children… metalinguistic awareness, in toddler form.”

- Bialystok, E. (2001). Bilingualism in Development: Language, Literacy, and Cognition. Cambridge University Press. The standard reference on bilingual cognitive flexibility and earlier metalinguistic awareness.

- Bialystok, E. (2009). Bilingualism: The good, the bad, and the indifferent. Bilingualism: Language and Cognition, 12(1), 3–11. A more accessible review that includes the register-switching advantage. https://www.cambridge.org/core/journals/bilingualism-language-and-cognition/article/abs/bilingualism-the-good-the-bad-and-the-indifferent/

”Nursery rhymes are built for prosody… they tune the ear.”

- Trehub, S. E. (2003). The developmental origins of musicality. Nature Neuroscience, 6(7), 669–673. Trehub at the University of Toronto on infant musicality and the overlap between music and language processing in the developing brain. https://www.nature.com/articles/nn1084

- Patel, A. D. (2008). Music, Language, and the Brain. Oxford University Press. The book-length treatment of the music-and-language overlap that underlies why rhythmic, rhyming, repetitive material lands so well for young children.

Honesty flag: the “bilingual register-switching advantage” line is supported by the literature but the size of the advantage is contested in the broader bilingual-cognitive-advantage debate (see Paap, Johnson & Sawi 2015 for the sceptical reading). I’ve kept the claim to an observable minimum — that bilingual children practise switching from earlier — which is hard to dispute.

Hello, and welcome. I’m Paola.

Sowing Tales is a publication about slow, screen-free, multilingual childhood — the early years especially, and the small daily choices that shape a child before they have the words for them. I write from inside it: as the mother of a two-year-old, a linguist returning to research, raising my daughter across Italian, English, with a little Sinhala for heritage and some German from my au-pair days.

If you’ve just arrived, here are three good places to begin:

- Mama. Mamma. Mamá — what a baby’s first word actually means, and why almost every language in the world landed on the same one.

- The 30-million-word gap (coming mid-June) — the famous study about how much we talk to our children, what it gets right, and what it quietly gets wrong, IMO.

- The sound before the word (coming end June)— on babbling, motherese, and everything that happens before language properly arrives.

What to expect if you stay: a long-ish essay every two to three weeks, some of it in Italian, the odd shorter Note in between. Never spam, never sponsored, always something I’d want to read myself. And I might share the odd DIY bits and bobs I am using with my daughter - home made books, picture cards and other DIY bits and bobs.

For the full practical companion — from picture-book guides to reading lists — there’s My Mini Linguist, the website cousin to this Substack.

And if something here resonates, I’d love you to subscribe — it’s free — and to write back. I read everything.

— Paola

---

Twelve months and a few days old. A Tuesday morning in April. The kitchen still half-dark, the kettle just clicked off, my daughter standing on the chair helping me pour oats in the pan to make porridge — and the word came out as if it had always been there.

Mamma.

She said it once, looking down at the pan. Then she lifted her head, looked directly at me, and said it again — slower, deliberately — mamma — with the cadence she’d heard a thousand times in the way I responded to her cry, the way I called from the next room, the way Italian carries the word like a question even when it isn’t one.

I didn’t cry, which surprised me.

I picked her up, which surprised neither of us.

I am a linguist by training and I have read more materials on early word acquisition than any single person probably needs to. I knew this would happen, more or less, around now. I knew the first word would likely be a bilabial consonant (she had been calling her paternal Grandma ‘Ba’ since she was 6 months old, but to me that sort of didn’t feel yet like talking). I knew, statistically, it would probably be a word for me or her Papi (dad). None of this prepared me for the actual sound of it. The theory had been one thing. The Tuesday morning was another.

---

Across nearly every language family in the world, the word for mother starts with the syllable /ma/.

Italian mamma. English mama. German Mama. Spanish mamá. French maman. Russian мама. Mandarin māma. Welsh mam. Even where the letters follow what might feel a slightly reverse order, like Sinhala, the language in which she calles her paternal grandparents, which uses අම්මා ammā. Even in languages whose paths haven’t crossed in eight thousand years, the word arrives in roughly the same shape — a soft consonant made with closed lips, opening into the most open vowel a human voice can produce.

The reason is not poetry, though it is sometimes mistaken for poetry. The reason is anatomy.

Roman Jakobson, the great Russian-American linguist, proposed in 1962 that the consonant /m/ and the vowel /a/ are the most natural sounds a human infant can produce. The /m/ is the sound that emerges with the lips closed and the mouth at rest — which is to say, the position the lips are in for most of the first months of life. The /a/ is what happens when the mouth opens. Babies feeding at the breast or the bottle close their lips, hum slightly, and as the latch breaks, the mouth opens onto the air. Mmmmm-aaaa.

Adults, hearing this sound, mapped it to the closest social referent — mother — and the convention rippled outward through every culture that ever depended on a mother to feed an infant. Which is all of them.

When a Tuesday-morning toddler says *mamma* for the first time, what she is doing — at the level of pure phonology — is naming the sound her own body has been making since before she had words at all. The word is, in a sense, the sound of feeding. Of being held. Of the first thing she ever did with her mouth.

I find this devastating, in the gentle sense of the word. There is a way of hearing this fact that strips it of romance, and I sometimes try to hear it that way out of intellectual honesty: it’s just bilabials, just feeding mechanics, just convention. But the romance leaks back in around the edges anyway. Mamma is not arbitrary. Mamma is what the body says when the body is fed.

---

The other thing I knew, in theory, was that first words are not random.

Look at any list of children’s first fifty words across any language and you will find roughly the same categories: agents (mama, dada, baby), things that move and have agency (cat, dog, ball, car), things that appear and disappear (more, gone, up, down, all done), things you eat or drink (milk, water, bread), things you can do with your body (kiss, jump, fall).

You will usually not find house. You will not find lamp. You will not find Tuesday.

A small child does not say house first because the house does not move. The house does not love them back. The house is simply there. What a child wants a name for, before anything else, is the people and things that act on her, and the things she can act on. The first word is a cognitive choice. A child saying mamma is saying: this person is the person I most want a name for. The first word is the first time a child says here, look. It is the beginning of a long lifetime of pointing.

---

I am raising my daughter bilingual with snippets in two additional languages, which sounds harder than it is and is also exactly as hard as it is.

Italian with me. English with her dad and pretty much everywhere else — the nursery, the playground, the world; some German from my years living there and some Sinhala which is the ancestral language of her dad (plus a bit of Spanish with close friends, which I have to say is coming along nicely).

The order maps neatly to time spent. At the moment (she is two now), Italian came first because Italian is not only what I speak with her when we are together, it’s also what I default to when I am tired and unguarded; mamma is the word I use for my own mother, the word I called for in the dark as a child myself. English is multiplying every week now that nursery has started. German is on its way; she has nine or ten German words, mostly short songs, and a good ear for the consonant clusters her English-speaking peers find difficult. Sinhala is important for the cultural connection, for foods, festivals we don’t have translations for. And of course for calling her family members like her dad would have called them growing up. Spanish is for now the language of fun and play, as she is particularly fond of her little friend and his Spanish speaking mama.

Code-switching is already happening, the way it always happens, the way every bilingual or multilingual child does it. She says acqua to me and water to her father in the same minute, mostly choosing the language that fits the listener. She also This is not a sign of confusion. This is a sign of a small linguistic cartographer who has already mapped which language belongs in which room.

The texture of bilingual life is not symmetric. Italian is for emotional moments — vieni qui, amore, non piangere. English is for transactional moments — let’s get your shoes on, do you want a snack, careful with that. The peppering of Sinhala words, Telefoniamo a seeya? (shall we call grandad?), and the occasional Spanish joyful explosion, like Hola amor!

I notice these patterns and file them away as a researcher would, which is the only way I know how to live inside what I’m watching. But the moments that arrive are also the moments that surprise me most. She fell off a step last week and cried mamma — not mummy — and the choice was not random. Pain went through Italian first, because pain reaches for whoever held the body when it was new.

---

I almost let it slip.

I want to say this clearly because the research often makes heritage-language transmission sound like a choice you make once, calmly, with intention. It isn’t. It’s a daily choice you make a hundred times a day, and many of those choices fail because you are tired, because the English word came first, because you were in a hurry, because you couldn’t remember the Italian for bib in the moment.

There were several months — she was eight, nine, ten months old — when I noticed I was speaking to her in English by default. Soft English. Tired-evening English. Italian was something I’d reach for at bedtime, when I had the energy for one Stella stellina, and not much more.

Then my brother in law in Milan, who speaks good English but is not as confident, asked me Parlale in italiano, altrimenti come lo impara? In that same trip I saw my mum as she held my daughter on her lap and sang to her in Italian, and at some point my daughter said nonna, clearly, while reaching for her face — and something in me arranged itself. The word had come from somewhere. Italian was already a room she lived in, and I had been letting the door swing shut behind me when I left it.

I started again, more deliberately. Conscious Italian in the morning, before the day turned English. Italian at bath time. Italian when I sang her to sleep. The book in Italian before the book in English. None of it heroic. All of it sustainable.

Now, two years in, Italian is solidly hers. She has more Italian words than I had at her age, partly because I am paying closer attention than my parents could afford to.

---

The research on heritage language is settled enough that I can summarise it without nuance: maintaining a child’s heritage language correlates with stronger family cohesion, better cross-generational communication, more secure cultural identity, lower rates of adolescent isolation, and no worse outcomes in the majority language. The fear that bilingual children fall behind in their dominant language is not borne out by any longitudinal data I have seen.

What is also settled, less happily, is that without deliberate effort, heritage languages typically disappear by the third generation. The pattern is consistent across migration histories: grandparents speak Italian, parents speak Italian and English, children understand Italian but answer in English, grandchildren do not speak Italian at all.

I am the third generation in my own family, more or less. My grandparents spoke Italian. My parents speak Italian and English (mum) or French (dad). I speak both. My daughter is the inflection point. Without conscious work, she would be the generation that loses it.

This is the inheritance — not the bel paese version, the romantic Italy of summer holidays and pasta sauces (more on my research on this another time). The other inheritance: the ability to call her granddad and have a conversation. The ability to read a Topipittori book in the language it was written in. The felt sense that Italian is hers, not foreign — a room in the house she grew up in, not a museum she occasionally visits.

---

What I want for her is not fluency. Fluency is a bonus. What I want is belonging in two languages. (Note: there is also belonging in three cultures, and two ethnicities, as a mixed race child, but again, more on this in the future).

There is a specific kind of bilingual person — the unsettled bilingual, the one who is fluent in two languages but at home in neither — and the difference between belonging and unsettlement is largely made in early childhood. It’s made in whether the heritage language was emotional or only instructional. Whether it was the language of bedtime or only the language of grandparents. Whether the child’s body learned its first words and its first songs and its first prayers in that language, or only learned its grammar.

I want my daughter to fall in love in Italian someday, if she chooses. I want her to dream in Italian. I want her to have a registro — a register, a tonal range — in Italian that goes from playful to angry to tender, the full instrument, not the simplified tourist version. None of these things will happen unless Italian is in her body before she knows what language is.

Which is why we sing, badly, every night — first in Italian, then in English. *Stella stellina, la notte si avvicina. Twinkle, twinkle, little star*

Two languages in the same fifteen minutes. Rooms in the house she lives in, lit one after the other.

---

The Tuesday morning she said *mamma* for the first time, I didn’t write it down. I had a fiercely independent daughter making her own porridge and a kettle that had clicked off three minutes ago. I think I’ll remember the space around us. I think the word will travel with me anyway, with or without the notebook.

What I know now that I didn’t know then is that mamma was not, in fact, her first word. Her first word was the mmmmm she’d been making in her sleep at four months old. Her first word was the aaaa she opened her mouth into when I lifted her out of her Moses basket. Her first word was the held breath before nursing, the closed-lips hum, the slow open vowel afterwards.

Mamma is what a baby’s body has been saying since before there were words. We just didn’t have a way to write it down until she did it on a Tuesday, stirring in the oats, looking up at me as if to check whether the word had landed where she meant it to.

It had.

---

*Sowing Tales is a publication on slow, screen-free, multilingual childhood. Letters about once a fortnight, free. If this resonated, you can subscribe at the top of the page.*

Sources:

• Oh & Fuligni (2010), Social Development

• Cummins (1979), Review of Educational Research

• Bilingual lexical-deficit meta-analysis (2022), Psychonomic Bulletin & Review

• Heritage language & mental health benefits (2023), PMC

• Meta-analysis: heritage language & ethnic identity

• Only English by the Third Generation? (2002), Demography