tl;dr-ELT

How do you imagine the brain goes about constructing a sentence? You’d think it grabs a pronoun, looks around for a verb, then starts arranging everything into a tidy hierarchy -what linguists would call a syntax tree (or parse tree), with words stacked in layers of structure. However, a new study by Yngwie A. Nielsen (Aarhus University) & Morten H. Christiansen (Cornell University), published in Nature Human Behaviour, suggests something far simpler may be going on: our brains rely heavily on linear, reusable sequences, not just hierarchical grammatical structures. The idea challenges a 70‑year tradition in linguistics that treats hierarchical syntax as the defining feature of human language.

The study

The researchers combined three strands of evidence:

• Eye‑tracking experiments showing that readers process familiar multi‑word sequences faster, even when those sequences don’t form grammatical constituents.
• Analysis of spontaneous phone conversations, revealing that many of the most frequent 3–4 word sequences in English are non‑constituents such as “can I have a” or “it was in the”.
• Priming tasks, demonstrating that once people encounter a sequence, they process it more quickly the next time -evidence that these sequences are mentally represented as units.

Although the data come from English, the authors argue the implications extend to language evolution, child language development & adult L2 learning.

The findings

The central claim: language is not built solely from hierarchical structures. Instead, speakers draw on a repertoire of short, linear sequences of word classes -the linguistic equivalent of LEGO pieces- which can be combined rapidly in real time.

To understand the contrast:

• Hierarchical structures (the traditional view) assume that sentences are built from nested constituents:
  “the cake” → noun phrase → joins with “ate” → verb phrase → joins with “she” → full sentence→ full sentence → “she ate the cake”
• Linear sequences (the study’s view) are frequent, ready‑made patterns like “I don’t know if…”, “at the end of the…”, or “do you want to…”, which may not form neat constituents but are processed as meaningful units.

Key findings:

• Many high‑frequency sequences do not fit hierarchical grammar at all yet are central to how we process language.
• These sequences are stored, recognised & primed, meaning they behave like cognitive building blocks rather than on‑the‑fly constructions.
• If language relies less on hierarchical syntax, the gap between human language & animal communication may be smaller than traditionally assumed.

We could see it like this. A learner may store “I was wondering if…” as a single processing unit long before they can explain complement clauses. This matches what teachers often see: fluency grows from familiarity with patterns, not mastery of rules.

Teacher Takeaways? Once again, this study isn’t designed to directly inform classroom practice, but it offers ideas that can sharpen how we think about patterns, processing & fluency.

• Rebalance grammar & patterning. Hierarchical explanations (e.g. clause structure, VP/NP boundaries) are still useful, but learners may rely more on linear predictability -the high‑frequency word‑order patterns that help them anticipate what comes next- than on abstract rules. Teaching should reflect both.
• Elevate non‑constituent chunks -frequent word sequences that aren’t full grammatical units but are processed as familiar patterns. “do you want to”, “I don’t think so”, or “in the middle of the” are cognitively real, even if they don’t neat grammatical units. Treat them as legitimate teaching targets, not exceptions (since they don’t form proper constituents in a syntax tree).
• Use priming deliberately. Repeated exposure to the same sequence across tasks (listening → speaking → reading → writing) can accelerate processing. It’s a means to build fluency through pattern recycling rather than rule rehearsal.

Why this matters

This study doesn’t kill grammar -but it does suggest that our mental grammar is flatter, more sequential & more experience‑driven than the classic syntax‑tree model. For the classroom, it reinforces something teachers have long intuited: learners often speak fluently by leaning on familiar sequences, not by assembling sentences from hierarchical rules.

How do you help learners spot patterns?