Language — built in turns, not volume

Research on language acquisition over the last two decades has changed our understanding of how the linguistic brain is built. Three central findings stand out, and together reshape practical recommendations:

1. Babies start discriminating sounds before they speak — and lose that capacity through use, not age.
2. Word volume matters less than conversational turns.
3. Screens and electronic toys don't replace the human voice — on the contrary, they often interfere.

This pillar lays out the science behind each finding and what to do concretely at each phase.

1. From gestation to first sounds — perceptual window

The baby hears inside the womb. The auditory system is functional by week 24-26, and she recognizes the maternal voice at birth — preferring it to any other. Classic studies show that Brazilian and French newborns already discriminate Portuguese rhythm from French.

Werker and Tees, in their seminal 1984 study, showed that 6-month-old babies discriminate sounds from any human language, but lose that capacity by 10-12 months if not exposed^{Werker & Tees 1984}. It's the prototype of a critical period in humans: the brain specializes by pruning unused capacities.

This has practical consequence: rich and responsive exposure in the first months isn't luxury — it's architecture construction. And exposure to multiple languages in this window keeps discriminative capacities that will be useful later.

2. Parentese — not cute, structured

"Parentese" (or infant-directed speech) is that higher-pitched, exaggerated speech, with elongated vowels and broad prosodic contours that adults use intuitively with babies. For decades it was treated as mere affective behavior. Patricia Kuhl's work at I-LABS (University of Washington) showed that parentese optimizes phoneme discrimination by the baby and predicts vocabulary at 18-24 months^{Kuhl et al. 2014}.

More than that: Ramírez et al.'s (2017) randomized trial showed that teaching parents to use parentese (format, contours, questions inviting response) improves language outcomes at 14 months^{Ramírez et al. 2017}. It's not just descriptive — it's causal and modifiable.

Use parentese without guilt. It doesn't infantilize, doesn't delay "adult" speech. It's the native language of someone learning to be linguistic.

3. The word gap revisited — turns > words

Hart and Risley (1995) published the book that popularized the "30 million word gap": children from professional families heard ~3x more words by age 3 than children in poverty, and that difference correlated with vocabulary and school performance^{Hart & Risley 1995}. For years, interventions tried to increase word quantity.

Romeo et al. (2018) at MIT, using neuroimaging (fMRI), refined the story: the number of conversational turns (back-and-forth), not total word count, predicts brain activation in language areas (superior temporal gyrus, Broca's area) and language abilities at 4-6 years^{Romeo et al. 2018}. The effect was independent of socioeconomic status.

Fernald et al. (2013) added another angle: differences in language processing speed are already detectable at 18 months^{Fernald et al. 2013}, that is, before expressive vocabulary clearly diverges.

Practical implication: you don't need to "talk more". You need to converse — ask questions, wait (even if the answer is just a sound), expand her response. Five minutes of genuine turns are worth more than an hour of monologue.

4. Shared reading — the "turbo charger"

Hutton et al. (2015) showed in fMRI that preschoolers with rich reading environments have greater activation in brain networks for language and mental imagery processing, especially in the left parietal-occipital cortex^{Hutton et al. 2015}. Shared reading isn't pastime — it's direct training of neural networks.

And it's not just any reading. Whitehurst et al. (1988) developed dialogic reading, a technique where the adult doesn't just read but converses about the book with the child. The PEER acronym summarizes:

• Prompt — ask a question about what's on the page.
• Evaluate — assess the child's response.
• Expand — expand the response with 1-2 additional words.
• Repeat — repeat the expanded utterance for the child to absorb.

Studies show children receiving dialogic reading advance months in language development in just weeks^{Whitehurst et al. 1988}.

Start early — even before she "understands". At 6-9 months, books with contrasting images, flaps, and textures. At 12-18 months, simple books with few words per page where she points. At 2-3 years, narratives she starts to tell back.

5. Screens, electronic toys, and the "video deficit"

The AAP, based on decades of converging research, recommends zero screens before age 2 (except video calls with family) and maximum 1h/day of high-quality co-viewed content between 2-5 years^{AAP 2016}. Why so strict?

• Video deficit: babies under 2 can't learn from screens with the same efficiency as from a real person. Even "educational" content is processed incompletely.
• Substitution: every minute on screen is a minute less of conversational turns.
• Contingency: humans respond to the baby at the right moment — screens are fixed contingency.

Sosa (2016) tested three toy types during parent-baby interaction: electronic toys (lights, recorded sounds), books, and traditional toys (blocks, pots). Electronic toys drastically reduced the quantity and quality of parental speech, words directed to the baby, and conversational turns^{Sosa 2016}.

Educational electronic toy isn't educational — it's a distractor of interaction.

6. Bilingualism — open window, not delay

Research on early bilingualism has dispelled several popular myths. Bosch and Sebastián-Gallés (2001) showed that babies exposed to two languages from birth clearly discriminate both within a few months^{Bosch & Sebastián-Gallés 2001}. De Houwer (2007) consolidated decades of evidence showing that the OPOL strategy (one parent, one language) performs best in keeping both languages active, as long as input in each language is consistent (at least 20-30% of directed time)^{De Houwer 2007}.

Bilinguals may show apparent "delay" until 24-30 months (vocabulary in each language individually smaller than monolinguals), but in total vocabulary there's no difference, and from that age on they catch up. No cognitive harm. Possible benefit in executive function and theory of mind.

If more than one language is native to a caregiver, it's a gift — not a delay.

7. Milestones by age (overview)

The normal range is wide. See reference tables for details.

8. Practical synthesis

1. Speak parentese to her from the womb. Sing, read, narrate the everyday.
2. Pause. Wait for the response — sound, gesture, word. Respond as if it were full speech. Volume doesn't replace turns.
3. Daily dialogic reading. Use PEER. Start with picture books, evolve with age.
4. No screens before age 2. No electronic toys. Block and pot beat any lit-up "educational" item.
5. Bilingualism is an open window — if applicable to your family, OPOL is the best path.