The Company a Word Keeps
Why the words around a word may matter as much as the word itself
By Eyal Rav-Noy, Co-Founder, Capit Learning
There is a quiet finding buried in the science of how children learn language, and it has stayed with me since I first read it.
A team of researchers at the Hebrew University of Jerusalem recently asked a deceptively simple question: when a child learns a new word, does it matter which words it shows up next to? Not the word itself. Not how often the child hears it. The company it keeps.
The answer, it turns out, is yes — and the effect lasts for decades. It's a small window into how children learn words, and a reminder that the process is stranger and more relational than we tend to assume.
The Anchoring Index
The researchers built something they called an "anchoring index." For thousands of words in child-directed speech, they measured how often each word appeared right after a high-frequency word — words like the, is, a, in. The idea is that these common, well-known words act as anchors: stable, familiar reference points that help a child's brain segment and lock down the less familiar word sitting beside them. Then they did something clever. They took those childhood anchoring scores and tested them against how quickly adults recognized those same words in the lab.
Make it stand out
Whatever it is, the way you tell your story online can make all the difference.
The words that had been well-anchored in childhood — the ones that kept good company early — were processed faster in adulthood. Not a little faster in the moment. Faster years later, as a permanent property of the mental representation. A word learned next to a familiar anchor forms a sturdier memory than the same word learned in isolation or next to other strangers.
I find this beautiful, and I find it useful, and I want to be careful to separate the two.
Why Language Looks the Way It Does
The beautiful part is what it tells us about why language looks the way it does. Languages are Zipfian: a handful of words appear constantly, and the vast majority are rare. That lopsidedness has always seemed like an accident or an inefficiency. The anchoring research suggests it might be a feature. The common words are scaffolding. They give every rare word a familiar neighbor to lean on. A child wading into the flood of speech is never far from solid ground.
The useful part is quieter, and I want to resist overstating it, because the temptation in education is always to take an elegant finding and turn it into a mandate.
What It Means for Teaching Reading
This research is not a reading method. It says nothing about how to teach a child the relationship between sounds and print, which remains, in my view, the foundational work, and the work most reading instruction still gets backward. Anchoring operates at a different layer. It is about sequence: not which words you teach, but the order and company in which you place them once a child can read them.
And at that layer, it offers something practical. When we build the sentences and short passages a child reads to practice a newly learned spelling pattern, we have a choice about arrangement. We can string the hard new words together and hope they stick. Or we can seat each one beside a familiar anchor — the nut, it is on the mat — so the known word steadies the new one. Same words. Same sounds. Different company. The research suggests the second version doesn't just feel easier; it builds a more durable trace.
It is a small refinement. But the best instruction is often an accumulation of small refinements, each grounded in how the mind actually works rather than in how we assume it should.
I keep returning to the larger lesson. We tend to think learning is mostly about the thing being learned — the word, the rule, the fact. This research is a reminder that learning is relational. A new word is easier to hold when it arrives in good company.
The children who learn to read most easily are not necessarily the ones who are taught the most. They are the ones for whom each new thing arrives anchored to something they already know.
That is a design principle. And it is one worth building around.
The research described here is drawn from Matas, Arnon, and Siegelman, Anchoring and the learnability advantage of Zipfian distributions (Hebrew University of Jerusalem). The interpretations and applications are my own.