tl;dr-ELT

Ever wondered why languages organise themselves the way they do –how they arrange sounds, words & phrases into patterns that feel natural to speakers? A new paper by Richard Futrell (UC Irvine) & Michael Hahn (Saarland University), published in Nature Human Behaviour, offers a fresh angle: perhaps the structure of language didn’t emerge from cultural design or biological specialisation, but from the simple fact that humans process a continuous stream of linguistic units -sounds, syllables & words arriving one after another in time- with limited memory & predictive capacity.

The study

Futrell & Hahn model language as a one‑dimensional sequence of symbols, essentially a single line of sounds or words that the brain processes one after another in real time, rather than as a branching structure we can see all at once. They argue that this linear stream is handled under tight cognitive constraints, meaning our working memory can only hold a small amount of linguistic information at any moment & our ability to predict what comes next is limited. These bottlenecks, they suggest, are powerful enough to shape the very structure of language.

They combine simulations of artificial “languages” with different mappings between meanings & forms, & cross‑linguistic corpus analyses spanning phonology, morphology, syntax & semantics. In each case, they compare real languages comparing real languages with counterfactual baselines -altered versions of the same data where word order, affixes or sound patterns are deliberately scrambled to break the natural structure. This lets the authors test whether real languages genuinely minimise processing load compared with versions where structure has been removed.

The findings

Systematic structure emerges naturally: When the system tries to keep processing effort low, it automatically starts grouping bits of meaning into small, reusable word‑like pieces rather than expressing whole meanings in one big chunk.
Languages that try to pack everything into a single form (holistic codes) end up being harder to process because the brain has to predict too much at once.

Locality matters: Languages work better when related words stay close together.
If you scramble the order so that connected elements are far apart, the processing load shoots up.
Learners struggle when words that belong together are too far apart, because the brain has to hold information for longer & the sentence becomes harder to process.

Hierarchy falls out of the maths: When meanings have nested or layered relationships, the most efficient systems naturally produce well‑nested phrase structures -the kind of hierarchical patterns we see in real syntax.
In other words, you don’t need to design hierarchy; it appears automatically when the system tries to minimise processing effort.

Cross‑linguistic evidence supports the model

The authors also checked whether real languages behave the way the model predicts. They looked at data from many different languages & compared each one with a scrambled version of itself. Across the board, the real languages were easier to process.

Here’s what they found:

• Sounds: Real words follow sound patterns that make them easier to predict than versions where the sounds are shuffled.
• Word building: Real affixes (like plural endings or verb endings) make processing easier than artificial versions, even in languages with complex systems like Arabic.
• Phrases: Real adjective–noun pairs (like big house) are easier to process than versions where the order is randomly changed.
• Word order: The noun‑phrase orders that are most common across the world are also the ones that minimise processing effort.
• Meaning: Features that don’t strongly go together (like number) tend to be expressed separately & systematically (e.g., plural –s), while features that naturally cluster tend to appear together inside a single word.

Why this matters for ELT

For teachers, a lot of this will feel surprisingly familiar.

• Learners struggle when words that belong together (often called “dependencies”) are separated by long distances, because the brain has to hold information for longer & the processing load increases.
• Chunking isn’t just a learning strategy; it reflects how language itself is structured to reduce cognitive load. Morphological & syntactic “irregularities” often mirror deeper statistical patterns in meaning.

Teacher Takeaways?

• Keep connected words close. Show learners that English usually puts words that belong together side by side, & that sentences get harder when those words are far apart.
• Teach chunks. Phrases like at the moment, I don’t think so or do you want to… match how the brain naturally handles language.
• Play with word order. Give learners scrambled examples so they can feel how much harder a sentence becomes when the natural order is broken.

How do you help learners make sense of longer or more complex sentences in your classroom?