Scientists Explain Why Human Language Never Evolved Like Computer Code
Human language is extraordinarily rich and detailed. From the perspective of information theory, however, many of the same ideas could, at least in principle, be communicated in a much more compressed form. This raises an intriguing question: if computers can efficiently transmit information using ones and zeros, why do humans not communicate through a similarly compact digital code?
Linguist Michael Hahn, based in Saarbrücken, explored this puzzle together with Richard Futrell from the University of California, Irvine. The researchers developed a model that helps explain why human language evolved the way it did. Their findings were published in Nature Human Behaviour.
Why Language Isn’t “Efficient” In The Computer Sense
Around 7,000 languages are spoken across the world today. Some are used by only a few communities, while others—including Chinese, English, Spanish, and Hindi—are spoken by billions of people. Despite major differences in vocabulary, sound systems, and grammar, all human languages share the same basic purpose: expressing meaning by combining words into phrases and arranging them into sentences.
At first glance, this may seem unnecessarily complicated. If nature generally favors efficiency and resource conservation, why would the human brain rely on such an elaborate communication system instead of something more compact and mathematically efficient?
In theory, a binary-style system could compress information more effectively than spoken language. However, Hahn and Futrell argue that human language was not optimized to minimize bits of information. Instead, it evolved to minimize mental effort during real-time communication.
Language Is Grounded In Shared Experience
According to Hahn, language is shaped by ordinary human experience. If someone invented a completely abstract word for something nobody had ever encountered—for example, “gol,” defined as half cat and half dog—listeners would have no reliable way to interpret it because the concept would not connect to anything familiar.
Similarly, combining recognizable words into a meaningless jumble does not help communication. A scrambled word such as “gadcot” contains letters from “cat” and “dog,” but it does not clearly map onto understandable concepts. By contrast, the phrase “cat and dog” works instantly because it connects directly to ideas people already recognize.
The Brain Prefers Predictable Patterns Over Maximum Compression
Hahn summarizes the broader principle simply: what appears mathematically “inefficient” may actually be easier for the brain to process. Human language constantly relies on background knowledge, shared context, and learned expectations about what is likely to come next.
He compares this process to taking a familiar route to work. Even if the usual path is not the shortest one, it often feels easier because it is predictable and requires less active attention. A shorter but unfamiliar route may feel more mentally exhausting because it demands continuous focus and decision-making.
Similarly, a fully compressed digital communication system might transmit information quickly, but it would likely require far more cognitive effort for humans to produce and interpret. Both speaking and understanding would become mentally demanding tasks.
How Prediction Helps The Brain Understand Speech
The researchers emphasize that the human brain constantly predicts which words and sentence structures are likely to appear next. These expectations are built through years of exposure to a native language, making common linguistic patterns deeply ingrained.
Hahn illustrates this idea using German. The phrase “Die fünf grünen Autos” (“the five green cars”) is immediately understandable to German speakers, while a rearranged version such as “Grünen fünf die Autos” (“green five the cars”) feels unnatural and far more difficult to process.
In the normal sentence structure, each word gradually narrows the range of possibilities. “Die” signals certain grammatical expectations, “fünf” suggests a countable noun, and “grünen” indicates a plural object associated with a color. By the time the final noun appears, the brain has already reduced uncertainty and can easily integrate the final piece of meaning.
When the order is scrambled, however, those predictive cues arrive in unexpected positions, making it harder for the brain to construct meaning step by step.
What This Could Mean For AI
Hahn and Futrell also demonstrated these effects mathematically, showing that human language appears optimized for reducing cognitive load rather than maximizing compression efficiency.
The findings suggest that language evolved specifically for human brains operating under real-time constraints—not for the most compact possible encoding of information.
The research may also have implications for artificial intelligence systems, including large language models such as ChatGPT and Microsoft Copilot. Understanding how humans rely on prediction, familiarity, and gradual interpretation may help engineers design AI systems that communicate in ways that feel more intuitive and mentally effortless for users.
