Abstract
The subject-verb-object (SVO) word order prevalent in English is shared by about 42% of world languages. Another 45% follow the SOV order, 9% the VSO order, and fewer languages use the three remaining permutations. None of the many extant explanations of this phenomenon take into account the difficulty of implementing these permutations in the brain. We propose a plausible model of sentence generation inspired by the recently proposed Assembly Calculus framework of brain function. Our model results in a natural explanation of the uneven frequencies. Estimating the parameters of this model yields predictions of the relative difficulty of dis-inhibiting one brain area from another. Our model is based on the standard syntax tree, a simple binary tree with three leaves. Each leaf corresponds to one of the three parts of a basic sentence. The leaves can be activated through lock and unlock operations and the sequence of activation of the leaves implements a specific word order. More generally, we also formulate and algorithmically solve the problems of implementing a permutation of the leaves of any binary tree, and of selecting the permutation that is easiest to implement on a given binary tree.
Introduction
Linguists have long observed a striking imbalance in the distribution of basic word orders across the world's languages. While there are six possible permutations of Subject (S), Verb (V), and Object (O), nearly 90% of languages utilize either SOV (e.g., Japanese, Hindi) or SVO (e.g., English, Mandarin). Rare orders like VOS, OVS, and OSV account for only a tiny fraction of linguistic diversity.
Traditional theories attribute these patterns to communication efficiency or cognitive ease of learning. However, this research explores a fundamental "missing link": the neural implementation cost. We hypothesize that the physical constraints of the human brain—specifically the metabolic and computational difficulty of generating sequences from hierarchical structures—dictate which word orders become dominant.
Neural Implementation Model
Our model is built upon the Assembly Calculus (AC), a framework designed to bridge high-level cognition with low-level neural dynamics. We represent a basic sentence using a standard binary syntax tree where the "Sentence" node branches into a "Subject" and a "Verb Phrase" (VP), with the VP further splitting into "Verb" and "Object."
To transform this hierarchy into sequential speech, the brain must activate the leaves (S, V, O) in a specific order. We propose two neural primitives for this process:
- Lock (Inhibition): A higher-level node prevents a child leaf from firing prematurely.
- Unlock (Dis-inhibition): Once a leaf fires, it "unlocks" another leaf or node that was previously inhibited.
As shown in the figure above, the SVO order follows a natural progression through the tree structure. The "energy" or metabolic cost of these operations determines the likelihood of a word order's emergence, modeled using a Boltzmann distribution.
Linguistic Frequencies and Implementation Cost
The core insight of this model is that the dominance of SVO and SOV is a direct consequence of the syntax tree's inherent asymmetry. Because the Verb and Object are grouped together within the Verb Phrase, they are computationally "closer" to each other than to the Subject.
In our "Basic Model," where all neural operations have equal cost, SVO and SOV are the only permutations that can be implemented with a single lock and a single unlock operation. The other four permutations require significantly more complex coordination, involving multiple inhibitory and dis-inhibitory steps.
By comparing the articulation paths (see Figure above), it becomes clear why orders like OSV or VOS are rare: they are metabolically "expensive" for the brain's syntactic hardware to execute.
Key Results and Predictions
The model provides both qualitative and quantitative matches to real-world linguistic data:
- Frequency Matching: The Basic Model predicts SVO and SOV frequencies at approximately 39% each, remarkably close to the observed 42-45%. An "Extended Model" can perfectly fit the precise frequencies of all six orders by adjusting the metabolic cost parameters for specific inter-regional brain signaling.
- Neural Connectivity: Fitting the model to data allows us to predict the relative difficulty of neural signaling between brain regions. For instance, the model suggests that dis-inhibiting the Verb area from the Subject area is more costly than the reverse, providing a testable hypothesis for systems neuroscience.
- Algorithmic Generalization: We formally defined the "Leaf Order Problem"—finding the easiest permutation to implement on any given binary tree. This problem is solvable in $O(n \log n)$ time for uniform costs but becomes NP-hard with variable costs.
Discussion and Limitations
While the model provides a compelling neural explanation for word order frequencies, it currently focuses on simple three-word sentences. Real-world language involves complex nested structures, adjectives, and morphological features like "case" markings, which might mitigate the neural cost of otherwise "expensive" orders. Future work will aim to integrate these neural implementation costs with existing theories of communication efficiency to provide a holistic account of linguistic evolution.
Summary updated