Why Do Cuisines Worldwide Obey the Same Mathematical Laws?
> *"Imagine you're an alien who just arrived on Earth. You walk into a Beijing hutong, a Parisian street corner, a Mexican market, a spice shop in Delhi. What you see confuses you: humans combine ingredients in thousands of different ways, creating seemingly endless varieties of food. But when you feed these recipes into a computer, a stunning pattern emerges—no matter the country or culture, all cuisine obeys the same deep mathematical laws."*
---
1. The Code of the Kitchen
In 1935, American linguist George Kingsley Zipf discovered something strange while studying English texts: if you rank all words by frequency of use, the 1st-ranked word appears roughly twice as often as the 2nd, three times as often as the 3rd, four times as often as the 4th...
This became known as Zipf's law, one of the most mysterious statistical regularities ever discovered. It applies not only to English but to Chinese, Swahili, Inuit—virtually all human languages. Strangely, it also applies to city populations (the largest city is about twice the size of the second largest), web links, and even animal calls.
A natural question follows: if language obeys Zipf's law, do other symbolic systems created by humans obey it too?
In April 2026, a team led by Ganesh Bagler at IIIT Delhi gave a resounding answer in a large-scale study spanning global cuisines: yes, cooking does too.
They analyzed tens of thousands of traditional recipes, using named-entity recognition algorithms to annotate each dish's ingredients, cooking techniques, utensils, and cultural attributes. Then they asked a simple question: do recipes, as a "symbolic system," follow the same statistical laws as natural language?
The results were striking. Not only does Zipf's law apply to ingredient usage—three other equally profound statistical laws apply as well, drawn from information retrieval, quantitative linguistics, and statistical physics.
---
2. Four Universal Laws Governing the Kitchen
Law 1: Zipf's Law — A Few Ingredients Rule the World
The Bagler team found that ranking all ingredients in global recipes by frequency of use yields a near-perfect power-law relationship—just like words in language.
What does this mean? Open any world cookbook at random. No matter which country's chapter you land on, you'll find: a small number of ingredients dominate the vast majority of dishes, while most ingredients appear in only a few recipes.
This is no coincidence. In Bagler's dataset, foundational ingredients like onion, garlic, salt, and pepper function like "the," "of," and "and" in language—they're everywhere. Meanwhile, "rare words" like saffron, truffle, and tamarind appear only in specific contexts.
Zipf's law shows a three-part structure across 50 languages: a stable high-frequency segment (function words / staple ingredients), a smoothly declining mid-frequency segment (common content words / main ingredients), and a sharply bending low-frequency segment (rare words / specialty ingredients). Bagler found that the Zipf curve for ingredient use is structurally remarkably similar to that of language.
Linguist Sander Lestrade (Radboud University, 2017) proposed that Zipf's law can be explained by the interaction of syntax and semantics: function words (like articles) are few but essential, while content words (like nouns) are numerous with highly varied frequencies. Does this explanation extend to cooking? Perhaps: base seasonings are like "function words"—needed in every dish, but limited in variety—while main ingredients are like "content words"—rich in variety with enormous frequency differences.
Law 2: Heaps' Law — The Ceiling on Culinary Diversity
In 1978, information retrieval specialist Harold Stanley Heaps proposed a rule about vocabulary growth in texts: as you read more text, the number of new words you encounter grows sublinearly.
Specifically, for a text of N words, the number of distinct words V(N) ≈ k × N^β, where β is typically between 0.4 and 0.6. This means: after reading 1 million words, reading another million will yield far fewer new words than the first million.
The Bagler team applied this to cooking and found an equally deep pattern: as the recipe corpus grows, the rate at which new ingredients and techniques appear declines.
Imagine collecting 100 Italian recipes. The first 100 dishes might contain 50 distinct ingredients. But expanding to 1,000 dishes might only raise the count to 120—not 500. By 10,000 dishes, the increase slows further.
The intuition here is profound: each dish is a "sentence," and its ingredients are the "vocabulary." Like natural language, cooking exhibits lexical saturation—within a given cultural and technological framework, the number of possible innovations is finite.
A 2023 study found that text generated by the GPT-Neo model also follows Heaps' law, with larger models producing vocabulary growth patterns closer to human text. This hints at something deeper: Heaps' law may be a universal feature of any system generated from finite elements via combination rules—whether human language, AI text, or human cooking.
Law 3: The Menzerath–Altmann Law — The Complexity Trade-off
In 1954, German phonetician Paul Menzerath observed while analyzing German syllables that the longer a word, the shorter its syllables on average. He summarized it as: "the larger the whole, the smaller its parts."
In 1980, quantitative linguistics pioneer Gabriel Altmann formalized this as the Menzerath–Altmann law (MAL), with an elegant mathematical form:
y = a × x^b × e^(-c×x)
where x is the number of constituent units and y is their average length.
The law has been found to apply at nearly every level of linguistics: phoneme–syllable, syllable–word, word–clause, clause–sentence. More surprisingly, it also applies to genomics (the more exons in a gene, the shorter the average exon), music, and even monkey calls.
The Bagler team's third major finding: recipes also follow a Menzerath–Altmann-type relationship.
Specifically, the number of ingredients in a dish (x) and the "average information" per ingredient (y) follow the relationship MAL predicts. More ingredients means each contributes less information to the whole—like a long sentence where each word carries less average information.
A 2021 study in *PLOS ONE* revealed a deeper implication: a simple "monkey-at-the-typewriter" random model can produce a rough Menzerath correlation, but cannot produce the precise "inversion region" predicted by the full Menzerath–Altmann law. MAL is not merely a statistical artifact—it is a genuine marker of complexity distinguishing complex systems from random ones.
In cooking, what does the existence of this "inversion region" imply? Perhaps it marks the transition from "simple home dishes" to "complex banquet dishes"—the point where adding more ingredients no longer lowers the average information per ingredient, and may even increase it through refined pairings.
Law 4: Log-Normal Distribution — The Statistical Signature of Nutrition
The team's fourth finding was even more unexpected: the concentrations of macronutrients (protein, fat, carbohydrates) in recipes follow a log-normal distribution.
The log-normal distribution is extremely common in nature. It describes the outcomes of random processes that "multiply, then take the logarithm." For example, adult heights are roughly normally distributed, but adult incomes are log-normal—because income grows multiplicatively (salary doubling), not additively.
Why should macronutrient concentrations in recipes be log-normal? The Bagler team notes this is consistent with observations in packaged foods. One possible explanation: a recipe's nutritional content is the product of many independent factors (ingredient choice, proportions, cooking method), each affecting the outcome multiplicatively.
This finding links cooking to a vast range of biological phenomena—from bacterial growth to neuronal firing rates, from species abundance to earthquake magnitudes, log-normal distributions are everywhere.
---
3. Generative Models: Why These Laws Emerge
Discovering statistical regularities is one thing; understanding why they exist is another. The Bagler team not only discovered the four laws but also proposed a set of minimal generative models to explain them.
Mechanism 1: Preferential Reuse
The name borrows directly from "preferential attachment" in complex network science—the core of the Barabási–Albert model, which explains why internet links, scientific citations, and social networks all follow power laws.
In cooking, preferential reuse means: widely used ingredients are more likely to be used again. Salt, onion, and garlic are ubiquitous not only because they're versatile, but because cooks tend to reuse reliable, proven ingredients.
This mechanism directly produces Zipf's law: a few "star ingredients" accrue disproportionate usage frequency while the vast majority are marginalized.
Mechanism 2: Constrained Sampling
Cooking is not free creation. A cook's choices face multiple constraints:
- Geographic: inland regions lack access to seafood
- Seasonal: winter ingredients differ from summer's
- Cultural: religious taboos and traditions restrict available ingredients
- Technological: some ingredients require specific techniques
Mechanism 3: Incremental Modification
Recipes are rarely "invented from scratch." More often, a cook starts from a known recipe and makes small changes—a different spice, adjusted heat, a swapped ingredient. This incremental evolution resembles gradual mutation in biological evolution.
Incremental modification explains why "distances" between recipes are usually small, and why culinary innovation is bounded in rate (consistent with Heaps' law).
The team showed that the interaction of these three simple mechanisms reproduces all four statistical laws. This means: the deep structure of global cuisine was not planned by any central designer—it emerges through self-organization from these simple local rules.
---
4. From Kitchen to Cosmos: The Shared Grammar of Complex Systems
The deepest implication of Bagler's work is that cooking, language, cities, genomes, and neural networks—all these seemingly unrelated systems—share the same deep mathematical grammar.
This is not a metaphor but an empirical fact. Power-law distributions, log-normal distributions, sublinear growth—these patterns recur throughout nature, from earthquake frequencies to mass extinctions, from wealth distributions to neuronal firing.
Why?
Physicist Per Bak's 1996 theory of self-organized criticality offers one explanation: many complex systems naturally evolve toward a "critical point" where small events can trigger large consequences, and large events follow power laws. The sandpile model is the classic example: as grains fall, the pile self-organizes to a critical slope where one more grain may cause nothing—or an avalanche—and avalanche sizes follow a power law.
Another explanation comes from information theory. In 2025, Łukasz Dębowski of the Polish Academy of Sciences demonstrated a rigorous mathematical derivation chain from Zipf's law through Heaps' law and Hilberg's hypothesis to neural scaling laws. The chain starts with the statistical structure of natural language and ends at the performance curves of modern large language models. Dębowski's proof suggests: these statistical laws may be properties of information itself, not of any particular system.
What does this mean for cooking? Perhaps when humans create recipes, they unknowingly follow an optimal strategy of information efficiency—expressing the richest "semantics" (flavor and culture) with the fewest "words" (ingredients). This echoes the "principle of least effort" Zipf himself proposed for language.
---
5. Conclusion
When George Kingsley Zipf died in 1949, his law was still viewed by many as a statistical artifact—an interesting numerical game with no deep meaning. Nearly 80 years later, we find it governs not only human language but the human kitchen.
The Bagler team's work tells us: cuisine is not merely an expression of culture—it is also a creation of mathematics. From Beijing's zhajiang noodles to Parisian croissants, from Delhi's butter chicken to Mexican tacos—all these seemingly arbitrary creations obey the same deep laws.
This does not diminish the artistry of cooking. Quite the opposite—it makes the art more wondrous. Within the framework of these universal laws, human cooks still possess boundless room for innovation. Just as poets create infinite meaning from a finite vocabulary, cooks create infinite flavors from finite ingredients.
The laws set the boundaries; the art dances within them.
---
References
1. Bagler, G., Tewari, G.K., Yadav, A.R. et al. *Universal statistical laws governing culinary design.* arXiv:2604.28021 [physics.soc-ph] (2026). 2. Zipf, G.K. *Human Behavior and the Principle of Least Effort.* Addison-Wesley (1949). 3. Heaps, H.S. *Information Retrieval: Computational and Theoretical Aspects.* Academic Press (1978). 4. Menzerath, P. *Die Architektonik des deutschen Wortschatzes.* Dümmler (1954). 5. Altmann, G. *Prolegomena to Menzerath's law.* *Glottometrika* 2, 1-10 (1980). 6. Ahn, Y.-Y., Ahnert, S.E., Bagrow, J.P. & Barabási, A.-L. *Flavor network and the principles of food pairing.* *Sci. Rep.* 1, 196 (2011). 7. Torre, I.G., Dębowski, Ł. & Hernández-Fernández, A. *Can Menzerath's law be a criterion of complexity in communication?* *PLOS ONE* 16, e0256133 (2021). 8. Dębowski, Ł. *From Zipf's Law to Neural Scaling through Heaps' Law and Hilberg's Hypothesis.* arXiv:2512.13491 (2025). 9. Yu, S., Xu, C. & Li, H. *Zipf's law in 50 languages: its structural pattern, linguistic interpretation, and cognitive motivation.* arXiv:1807.01855 (2018). 10. Caprioli, C. et al. *The networks of ingredient combinations as culinary fingerprints.* *npj Science of Food* 9, 5 (2025).