In summary, by leveraging naturalistic neural recordings and LLMs, this study identifies a previously unrecognized information-making process in the speaker’s brain. We would like to thank our co-authors: Haocheng Wang, @TomSheffer17807, @DariaLioub, @SchainMariano@HassonLab 🙏
If you saw our modeling paper, this is the infrastructure behind it. New preprint from the First 1,000 Days (1kD) Project. It took 5 years to build what this paper describes. Worth it.
Dense measurement doesn't just give you more data. It tells you what's universal, what's household-specific, and when aggregation obscures more than it reveals.
Deeply grateful to our partners, collaborators, and the families who made this possible. We'd love to hear your thoughts. Preprint: biorxiv.org/content/10.64898…
There is no “average family”. Every home has its own lexical signature — the words that fill one child's day look meaningfully different from another's. Averaging across families doesn't reveal structure. It flattens it.
Scale is not enough. Science requires a system: longitudinal design, behavioral measures, data infrastructure, scalable AI-based analysis, and a feature table that links recordings and annotations back to individual children over time.
Dense measurement doesn't just give you more data. It tells you what's universal, what's household-specific, and when aggregation obscures more than it reveals.
Deeply grateful to our partners, collaborators, and the families who made this possible. We'd love to hear your thoughts. Preprint: biorxiv.org/content/10.64898…
Learning depends not only on rich everyday input, but also on replay of past experience, suggesting that cycles of experience and consolidation are critical for early language development.
Excited to share this work, and deeply grateful to our collaborators and partners who made it possible. We’d love to hear your thoughts.
Preprint: biorxiv.org/content/10.64898…